ISLAB/CAISR - User contributions [en]

Online Continual Learning for Time-Series Forecasting

2026-02-06T10:48:27Z

Islab: Created page with "{{StudentProjectTemplate |Summary=This project explores online continual learning to train models that adapt to changing environments and new tasks while retaining previously..."

{{StudentProjectTemplate
|Summary=This project explores online continual learning to train models that adapt to changing environments and new tasks while retaining previously learned knowledge.
|Keywords=Online Continual Learning; Time-series Forecasting; concept drift
|Supervisor=Nuwan Gunasekara, Yuantao Fan
|Status=Ongoing
}}
This project explores how online continual learning methods can be designed and employed for time-series forecasting so that models can be adapted rapidly to evolving environments and new tasks without catastrophic forgetting.

Data-Driven Activity Recognition and Energy Consumption Forecasting for Heavy-Duty Vehicles

2025-11-10T14:48:56Z

Islab:

{{StudentProjectTemplate
|Summary=develop a machine learning framework for activity recognition and energy consumption forecasting, in collaboration with Volvo Group
|TimeFrame=Fall 2025 or Spring 2026
|References=[1] Liu, J., Liu, Y., Zhu, W., Zhu, X., & Song, L. (2023). Distributional and spatial-temporal robust representation learning for transportation activity recognition. Pattern Recognition, 140, 109568.

[2] Khaertdinov, B., Ghaleb, E., & Asteriadis, S. (2021, August). Contrastive self-supervised learning for sensor-based human activity recognition. In 2021 IEEE International Joint Conference on Biometrics (IJCB) (pp. 1-8). IEEE.

[3] Zhang, J., Zi, L., Hou, Y., Wang, M., Jiang, W., & Deng, D. (2020). A deep learning‐based approach to enable action recognition for construction equipment. Advances in Civil Engineering, 2020(1), 8812928.

[4] Xing, Y., Lv, C., Cao, D., & Lu, C. (2020). Energy oriented driving behavior analysis and personalized prediction of vehicle states with joint time series modeling. Applied Energy, 261, 114471.

[6] Axelsson, H., & Wass, D. (2019). Machine Learning for Activity Recognition of Dumpers.

[7] Zhang, C., Shang, F., Liu, H., Wan, L., & Feng, W. (2025). FedAGC: Federated Continual Learning with Asymmetric Gradient Correction. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 3841-3850).

[8] Wang, Z., Fan, Y., Ydreskog, H., and Nowaczyk, S. (2025). Investigation on machine learning models for forecasting auxiliary energy consumption of HD-BEVs. In The 38th International Electric Vehicle Symposium & Exhibition. Electric Vehicle Symposium and Exhibition
|Supervisor=Yuantao Fan, TBD
|Level=Master
|Status=Open
}}
This thesis focuses on developing an AI-driven approach for activity recognition and energy consumption forecasting in heavy-duty vehicles. You will work with multivariate time-series data collected from on-board sensors in heavy-duty vehicles, creating compact and structured data representations through dimensionality reduction, feature extraction, and representation learning techniques to derive meaningful embeddings that capture the contextual dynamics of vehicle operations.

Potential research directions include: i) Multi-modal representation learning, integrating heterogeneous data sources (e.g., telemetry, environmental, and operational data) to enhance model robustness; ii) Contrastive learning, employing data augmentation strategies to pre-train models for improved generalization; iii) Federated learning, enabling decentralized and privacy-preserving model training across distributed edge devices; and iv) Domain adaptation and transfer learning, ensuring robustness across diverse operational contexts and vehicle configurations; v) Online learning with computationally efficient models for real-time training and inferences. Furthermore, achieving a certain degree of model interpretability is essential for understanding decision mechanisms and building trust in model outputs. This may involve clustering and visualizing embeddings to analyze operational states and transitions. Finally, a computational efficiency analysis shall be conducted to balance model performance with scalability and feasibility for real-world deployment.

You will work with the Advanced Analytics Team (Volvo Group), and collaborate with domain experts, stakeholders in different tech streams.

Please contact Yuantao for more details.

Please apply via the following link (Volvo portal):

https://jobs.volvogroup.com/job-invite/26402/

Anomaly Detection for Heavy-duty Vehicles

2025-11-10T14:48:16Z

Islab:

{{StudentProjectTemplate
|Summary=develop contextual and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles; in collaboration with Volvo Group
|TimeFrame=Fall 2025 or Spring 2026
|References=Li, Z., Zhu, Y., & Van Leeuwen, M. (2023). A survey on explainable anomaly detection. ACM Transactions on Knowledge Discovery from Data, 18(1), 1-54.

Li, Z., & Van Leeuwen, M. (2023). Explainable contextual anomaly detection using quantile regression forests. Data Mining and Knowledge Discovery, 37(6), 2517-2563.

Pasini, K., Khouadjia, M., Same, A., Trépanier, M., & Oukhellou, L. (2022). Contextual anomaly detection on time series: a case study of metro ridership analysis. Neural Computing and Applications, 34(2), 1483-1507.

Han, X., Zhang, L., Wu, Y., & Yuan, S. (2023, October). On root cause localization and anomaly mitigation through causal inference. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management (pp. 699-708).

Carvalho, J., Zhang, M., Geyer, R., Cotrini, C., & Buhmann, J. M. (2024). Invariant anomaly detection under distribution shifts: a causal perspective. Advances in Neural Information Processing Systems, 36.

Fan, Y., Nowaczyk, S., & Antonelo, E. A. (2016, July). Predicting air compressor failures with echo state networks. In PHM Society European Conference (Vol. 3, No. 1).

Gallicchio, C. (2024). Euler state networks: Non-dissipative reservoir computing. Neurocomputing, 579, 127411.
|Supervisor=Yuantao Fan, TBD
|Level=Master
|Status=Open
}}
Ensuring operational safety is crucial for commercial heavy-duty vehicles. The aim of this project is to design and develop contextual and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles. Beyond safety considerations, cost-efficient and reliable maintenance strategies are essential for operating commercial vehicle fleets, where operational availability and total maintenance cost directly impact profitability.

By applying anomaly detection techniques to identify early signs of component degradation or failure, this project seeks to enable proactive and predictive maintenance. Such a system would minimize unplanned downtime, extend component lifespan, and reduce overall maintenance costs, thereby enhancing the safety, reliability, and economic performance of heavy-duty vehicle fleets. A promising research direction involves investigating representation learning techniques (such as contrastive learning, continual learning, federated learning etc.) to capture and encode key characteristics of time series data for anomaly detection. For instance, many state-of-the-art time methods utilize autoencoders, using learned embeddings in the latent features or reconstruction errors to compute anomaly scores. In addition, methods that are inherently explainable (e.g. causal graph/relations learned via causal inferences), that can be learned in an incremental setting (e.g. decision tree-based approaches) or computationally efficient (e.g. echo state network via reservoir computing), are of great interest as well. The developed approach will be evaluated on a real-world dataset collected from commercial vehicle fleets.

You will work with the Advanced Analytics Team (Volvo Group) and collaborate with domain experts, stakeholders in different tech streams.

Please contact Yuantao for more details.

Please apply via the following link (in Volvo portal):

https://jobs.volvogroup.com/job-invite/26403/

Data-Driven Activity Recognition and Energy Consumption Forecasting for Heavy-Duty Vehicles

2025-11-10T13:37:41Z

Islab:

Anomaly Detection for Heavy-duty Vehicles

2025-11-10T13:36:14Z

Islab:

Anomaly Detection for Heavy-duty Vehicles

2025-11-10T13:35:25Z

Islab:

{{StudentProjectTemplate
|Summary=develop contextual and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles; in collaboration with Volvo Group
|TimeFrame=Fall 2025 or Spring 2026
|References=Li, Z., Zhu, Y., & Van Leeuwen, M. (2023). A survey on explainable anomaly detection. ACM Transactions on Knowledge Discovery from Data, 18(1), 1-54.

Li, Z., & Van Leeuwen, M. (2023). Explainable contextual anomaly detection using quantile regression forests. Data Mining and Knowledge Discovery, 37(6), 2517-2563.

Pasini, K., Khouadjia, M., Same, A., Trépanier, M., & Oukhellou, L. (2022). Contextual anomaly detection on time series: a case study of metro ridership analysis. Neural Computing and Applications, 34(2), 1483-1507.

Han, X., Zhang, L., Wu, Y., & Yuan, S. (2023, October). On root cause localization and anomaly mitigation through causal inference. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management (pp. 699-708).

Carvalho, J., Zhang, M., Geyer, R., Cotrini, C., & Buhmann, J. M. (2024). Invariant anomaly detection under distribution shifts: a causal perspective. Advances in Neural Information Processing Systems, 36.

Fan, Y., Nowaczyk, S., & Antonelo, E. A. (2016, July). Predicting air compressor failures with echo state networks. In PHM Society European Conference (Vol. 3, No. 1).

Gallicchio, C. (2024). Euler state networks: Non-dissipative reservoir computing. Neurocomputing, 579, 127411.
|Supervisor=Yuantao Fan, TBD
|Level=Master
|Status=Open
}}
Ensuring operational safety is crucial for commercial heavy-duty vehicles. The aim of this project is to design and develop contextual and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles. Beyond safety considerations, cost-efficient and reliable maintenance strategies are essential for operating commercial vehicle fleets, where operational availability and total maintenance cost directly impact profitability.

By applying anomaly detection techniques to identify early signs of component degradation or failure, this project seeks to enable proactive and predictive maintenance. Such a system would minimize unplanned downtime, extend component lifespan, and reduce overall maintenance costs, thereby enhancing the safety, reliability, and economic performance of heavy-duty vehicle fleets. A promising research direction involves investigating representation learning techniques (such as contrastive learning, continual learning, federated learning etc.) to capture and encode key characteristics of time series data for anomaly detection. For instance, many state-of-the-art time methods utilize autoencoders, using learned embeddings in the latent features or reconstruction errors to compute anomaly scores. In addition, methods that are inherently explainable (e.g. causal graph/relations learned via causal inferences), that can be learned in an incremental setting (e.g. decision tree-based approaches) or computationally efficient (e.g. echo state network via reservoir computing), are of great interest as well. The developed approach will be evaluated on a real-world dataset collected from commercial vehicle fleets.

You will work with the Advanced Analytics Team (Volvo Group) and collaborate with domain experts, stakeholders in different tech streams.

Please contact Yuantao for more details.

Please apply via the following link:

https://jobs.volvogroup.com/job/G%C3%B6teborg-Master-Thesis-Anomaly-detection-in-time-series-for-heavy-duty-BEVs-417-15/1330250455/

Tabular Health Data Under Attack: Benchmarking Privacy Risks and Defenses

2025-11-03T08:11:18Z

Islab: Created page with "{{StudentProjectTemplate |Summary=This thesis aims to investigate privacy attacks and defenses in tabular health data. |TimeFrame=Autumn25-Spring26 |References=[1] He, Z., Ouy..."

{{StudentProjectTemplate
|Summary=This thesis aims to investigate privacy attacks and defenses in tabular health data.
|TimeFrame=Autumn25-Spring26
|References=[1] He, Z., Ouyang, C., Wen, L., Liu, C. and Moreira, C., 2025. TabAttackBench: A Benchmark for Adversarial Attacks on Tabular Data. arXiv preprint arXiv:2505.21027.
[2] Alshantti, A., Rasheed, A. and Westad, F., 2025. Privacy Re‐Identification Attacks on Tabular GANs. Security and Privacy, 8(1), p.e469.
|Supervisor=Jens Lundström, Eric Järpe, Atiye Sadat Hashemi
|Level=Master
|Status=Open
}}
This thesis aims to investigate privacy attacks and defenses in tabular health data, focusing on understanding how sensitive information can be inferred from structured datasets and how modern privacy-preserving techniques can mitigate these risks. The project will involve studying and implementing state-of-the-art attack methods (e.g., membership and attribute inference) and defense mechanisms (e.g., differential privacy and adversarial noise injection) on benchmark datasets such as MIMIC-III, IV, which are commonly used in healthcare research. The goal is to provide a comprehensive evaluation framework for measuring privacy–utility trade-offs and to propose or refine novel defense approaches that enhance protection while maintaining analytical value in health-related tabular data.

Anomaly Detection for Heavy-duty Vehicles

2025-10-28T23:10:47Z

Islab:

Data-Driven Activity Recognition and Energy Consumption Forecasting for Heavy-Duty Vehicles

2025-10-28T23:10:25Z

Islab:

Anomaly Detection for Heavy-duty Vehicles

2025-10-28T23:08:46Z

Islab:

{{StudentProjectTemplate
|Summary=develop contextual and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles; in collaboration with Volvo Group
|TimeFrame=Fall 2025 or Spring 2026
|References=Li, Z., Zhu, Y., & Van Leeuwen, M. (2023). A survey on explainable anomaly detection. ACM Transactions on Knowledge Discovery from Data, 18(1), 1-54.

Li, Z., & Van Leeuwen, M. (2023). Explainable contextual anomaly detection using quantile regression forests. Data Mining and Knowledge Discovery, 37(6), 2517-2563.

Pasini, K., Khouadjia, M., Same, A., Trépanier, M., & Oukhellou, L. (2022). Contextual anomaly detection on time series: a case study of metro ridership analysis. Neural Computing and Applications, 34(2), 1483-1507.

Han, X., Zhang, L., Wu, Y., & Yuan, S. (2023, October). On root cause localization and anomaly mitigation through causal inference. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management (pp. 699-708).

Carvalho, J., Zhang, M., Geyer, R., Cotrini, C., & Buhmann, J. M. (2024). Invariant anomaly detection under distribution shifts: a causal perspective. Advances in Neural Information Processing Systems, 36.

Fan, Y., Nowaczyk, S., & Antonelo, E. A. (2016, July). Predicting air compressor failures with echo state networks. In PHM Society European Conference (Vol. 3, No. 1).

Gallicchio, C. (2024). Euler state networks: Non-dissipative reservoir computing. Neurocomputing, 579, 127411.
|Supervisor=Yuantao Fan, TBD
|Level=Master
|Status=Open
}}
Ensuring operational safety is crucial for commercial heavy-duty vehicles. The aim of this project is to design and develop contextual and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles. Beyond safety considerations, cost-efficient and reliable maintenance strategies are essential for operating commercial vehicle fleets, where operational availability and total maintenance cost directly impact profitability.

By applying anomaly detection techniques to identify early signs of component degradation or failure, this project seeks to enable proactive and predictive maintenance. Such a system would minimize unplanned downtime, extend component lifespan, and reduce overall maintenance costs, thereby enhancing the safety, reliability, and economic performance of heavy-duty vehicle fleets. A promising research direction involves investigating representation learning techniques (such as contrastive learning, continual learning, federated learning etc.) to capture and encode key characteristics of time series data for anomaly detection. For instance, many state-of-the-art time methods utilize autoencoders, using learned embeddings in the latent features or reconstruction errors to compute anomaly scores. In addition, methods that are inherently explainable (e.g. causal graph/relations learned via causal inferences), that can be learned in an incremental setting (e.g. decision tree-based approaches) or computationally efficient (e.g. echo state network via reservoir computing), are of great interest as well. The developed approach will be evaluated on a real-world dataset collected from commercial vehicle fleets.

You will work with the Advanced Analytics Team (Volvo Group Technology) and collaborate with domain experts, stakeholders in different tech streams.

Please contact Yuantao for more details.

Anomaly Detection for Heavy-duty Vehicles

2025-10-28T23:06:58Z

Islab:

{{StudentProjectTemplate
|Summary=develop contextual and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles; in collaboration with Volvo Group
|TimeFrame=Fall 2025 or Spring 2026
|References=Li, Z., Zhu, Y., & Van Leeuwen, M. (2023). A survey on explainable anomaly detection. ACM Transactions on Knowledge Discovery from Data, 18(1), 1-54.

Li, Z., & Van Leeuwen, M. (2023). Explainable contextual anomaly detection using quantile regression forests. Data Mining and Knowledge Discovery, 37(6), 2517-2563.

Pasini, K., Khouadjia, M., Same, A., Trépanier, M., & Oukhellou, L. (2022). Contextual anomaly detection on time series: a case study of metro ridership analysis. Neural Computing and Applications, 34(2), 1483-1507.

Han, X., Zhang, L., Wu, Y., & Yuan, S. (2023, October). On root cause localization and anomaly mitigation through causal inference. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management (pp. 699-708).

Carvalho, J., Zhang, M., Geyer, R., Cotrini, C., & Buhmann, J. M. (2024). Invariant anomaly detection under distribution shifts: a causal perspective. Advances in Neural Information Processing Systems, 36.

Fan, Y., Nowaczyk, S., & Antonelo, E. A. (2016, July). Predicting air compressor failures with echo state networks. In PHM Society European Conference (Vol. 3, No. 1).

Gallicchio, C. (2024). Euler state networks: Non-dissipative reservoir computing. Neurocomputing, 579, 127411.
|Supervisor=Yuantao Fan, TBD
|Level=Master
|Status=Open
}}
Ensuring operational safety is crucial for commercial heavy-duty vehicles. The aim of this project is to design and develop contextual and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles. Beyond safety considerations, cost-efficient and reliable maintenance strategies are essential for operating commercial vehicle fleets, where operational availability and total maintenance cost directly impact profitability.

By applying anomaly detection techniques to identify early signs of component degradation or failure, this project seeks to enable proactive and predictive maintenance. Such a system would minimize unplanned downtime, extend component lifespan, and reduce overall maintenance costs, thereby enhancing the safety, reliability, and economic performance of heavy-duty vehicle fleets. A promising research direction involves investigating representation learning techniques (such as contrastive learning, continual learning, federated learning etc.) to capture and encode key characteristics of time series data for anomaly detection. For instance, many state-of-the-art time methods utilize autoencoders, using learned embeddings in the latent features or reconstruction errors to compute anomaly scores. In addition, methods that are inherently explainable (e.g. causal graph/relations learned via causal inferences), that can be learned in an incremental setting (e.g. decision tree-based approaches) or computationally efficient (e.g. echo state network via reservoir computing), are of great interest as well. The developed approach will be evaluated on a real-world dataset collected from commercial vehicle fleets.

You will work with the Advanced Analytics Team, at Volvo Group Technology, and collaborate with domain experts, stakeholders in different tech streams.

Please contact Yuantao for more details.

Yuantao Fan

2025-10-28T15:10:18Z

Islab:

{{Person
|Family Name=Fan
|Given Name=Yuantao
|Title=Associate Senior Lecturer
|Position=PhD
|Email=yuantao.fan@hh.se
|Image=Yuantao_Fan_img2.jpg
|Office=E513
|Subject=Learning Systems
}}
{{AssignProjects}}
{{AssignProjects
|project=In4Uptime
}}
{{AssignSubjectAreas
|SubjectArea=Vehicle Diagnostics
}}
{{AssignSubjectAreas
|SubjectArea=Data Mining
}}
{{AssignApplicationAreas
|ApplicationArea=Intelligent Vehicles
}}

__NOTOC__
{{ShowPerson}}
{{InsertSubjAreas}}
{{InsertProjects}}
{{PublicationsList}}
[[Category:staff]]

Yuantao Fan

2025-10-28T15:07:55Z

Islab:

{{Person
|Family Name=Fan
|Given Name=Yuantao
|Title=Assistant Professor
|Position=PhD
|Email=yuantao.fan@hh.se
|Image=Yuantao_Fan_img2.jpg
|Office=E513
|Subject=Learning Systems
}}
{{AssignProjects}}
{{AssignProjects
|project=In4Uptime
}}
{{AssignSubjectAreas
|SubjectArea=Vehicle Diagnostics
}}
{{AssignSubjectAreas
|SubjectArea=Data Mining
}}
{{AssignApplicationAreas
|ApplicationArea=Intelligent Vehicles
}}

__NOTOC__
{{ShowPerson}}
{{InsertSubjAreas}}
{{InsertProjects}}
{{PublicationsList}}
[[Category:staff]]

Data-Driven Activity Recognition and Energy Consumption Forecasting for Heavy-Duty Vehicles

2025-10-28T14:52:44Z

Islab:

{{StudentProjectTemplate
|Summary=develop a machine learning framework for activity recognition and energy consumption forecasting, in collaboration with Volvo Group
|TimeFrame=Fall 2025 or Spring 2026
|References=[1] Liu, J., Liu, Y., Zhu, W., Zhu, X., & Song, L. (2023). Distributional and spatial-temporal robust representation learning for transportation activity recognition. Pattern Recognition, 140, 109568.

[2] Khaertdinov, B., Ghaleb, E., & Asteriadis, S. (2021, August). Contrastive self-supervised learning for sensor-based human activity recognition. In 2021 IEEE International Joint Conference on Biometrics (IJCB) (pp. 1-8). IEEE.

[3] Zhang, J., Zi, L., Hou, Y., Wang, M., Jiang, W., & Deng, D. (2020). A deep learning‐based approach to enable action recognition for construction equipment. Advances in Civil Engineering, 2020(1), 8812928.

[4] Xing, Y., Lv, C., Cao, D., & Lu, C. (2020). Energy oriented driving behavior analysis and personalized prediction of vehicle states with joint time series modeling. Applied Energy, 261, 114471.

[6] Axelsson, H., & Wass, D. (2019). Machine Learning for Activity Recognition of Dumpers.

[7] Zhang, C., Shang, F., Liu, H., Wan, L., & Feng, W. (2025). FedAGC: Federated Continual Learning with Asymmetric Gradient Correction. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 3841-3850).

[8] Wang, Z., Fan, Y., Ydreskog, H., and Nowaczyk, S. (2025). Investigation on machine learning models for forecasting auxiliary energy consumption of HD-BEVs. In The 38th International Electric Vehicle Symposium & Exhibition. Electric Vehicle Symposium and Exhibition
|Supervisor=Yuantao Fan, TBD
|Level=Master
|Status=Open
}}
This thesis focuses on developing an AI-driven approach for activity recognition and energy consumption forecasting in heavy-duty vehicles. You will work with multivariate time-series data collected from on-board sensors in heavy-duty vehicles, creating compact and structured data representations through dimensionality reduction, feature extraction, and representation learning techniques to derive meaningful embeddings that capture the contextual dynamics of vehicle operations.

Potential research directions include: i) Multi-modal representation learning, integrating heterogeneous data sources (e.g., telemetry, environmental, and operational data) to enhance model robustness; ii) Contrastive learning, employing data augmentation strategies to pre-train models for improved generalization; iii) Federated learning, enabling decentralized and privacy-preserving model training across distributed edge devices; and iv) Domain adaptation and transfer learning, ensuring robustness across diverse operational contexts and vehicle configurations; v) Online learning with computationally efficient models for real-time training and inferences. Furthermore, achieving a certain degree of model interpretability is essential for understanding decision mechanisms and building trust in model outputs. This may involve clustering and visualizing embeddings to analyze operational states and transitions. Finally, a computational efficiency analysis shall be conducted to balance model performance with scalability and feasibility for real-world deployment.

You will work with the Advanced Analytics Team, at Volvo Group Technology, and collaborate with domain experts, stakeholders in different tech streams.

Please contact Yuantao for more details.

Anomaly Detection for Heavy-duty Vehicles

2025-10-28T14:51:50Z

Islab:

{{StudentProjectTemplate
|Summary=develop context-aware and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles; in collaboration with Volvo Group
|TimeFrame=Fall 2025 or Spring 2026
|References=Li, Z., Zhu, Y., & Van Leeuwen, M. (2023). A survey on explainable anomaly detection. ACM Transactions on Knowledge Discovery from Data, 18(1), 1-54.

Li, Z., & Van Leeuwen, M. (2023). Explainable contextual anomaly detection using quantile regression forests. Data Mining and Knowledge Discovery, 37(6), 2517-2563.

Pasini, K., Khouadjia, M., Same, A., Trépanier, M., & Oukhellou, L. (2022). Contextual anomaly detection on time series: a case study of metro ridership analysis. Neural Computing and Applications, 34(2), 1483-1507.

Han, X., Zhang, L., Wu, Y., & Yuan, S. (2023, October). On root cause localization and anomaly mitigation through causal inference. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management (pp. 699-708).

Carvalho, J., Zhang, M., Geyer, R., Cotrini, C., & Buhmann, J. M. (2024). Invariant anomaly detection under distribution shifts: a causal perspective. Advances in Neural Information Processing Systems, 36.

Fan, Y., Nowaczyk, S., & Antonelo, E. A. (2016, July). Predicting air compressor failures with echo state networks. In PHM Society European Conference (Vol. 3, No. 1).

Gallicchio, C. (2024). Euler state networks: Non-dissipative reservoir computing. Neurocomputing, 579, 127411.
|Supervisor=Yuantao Fan, TBD
|Level=Master
|Status=Open
}}
Ensuring operational safety is crucial for commercial heavy-duty vehicles. The aim of this project is to design and develop context-aware and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles. Beyond safety considerations, cost-efficient and reliable maintenance strategies are essential for operating commercial vehicle fleets, where operational availability and total maintenance cost directly impact profitability.

By applying anomaly detection techniques to identify early signs of component degradation or failure, this project seeks to enable proactive and predictive maintenance. Such a system would minimize unplanned downtime, extend component lifespan, and reduce overall maintenance costs, thereby enhancing the safety, reliability, and economic performance of heavy-duty vehicle fleets. A promising research direction involves investigating representation learning techniques (such as contrastive learning, continual learning, federated learning etc.) to capture and encode key characteristics of time series data for anomaly detection. For instance, many state-of-the-art time methods utilize autoencoders, using learned embeddings in the latent features or reconstruction errors to compute anomaly scores. In addition, methods that are inherently explainable (e.g. causal graph/relations learned via causal inferences), that can be learned in an incremental setting (e.g. decision tree-based approaches) or computationally efficient (e.g. echo state network via reservoir computing), are of great interest as well. The developed approach will be evaluated on a real-world dataset collected from commercial vehicle fleets.

You will work with the Advanced Analytics Team, at Volvo Group Technology, and collaborate with domain experts, stakeholders in different tech streams.

Please contact Yuantao for more details.

Data-Driven Activity Recognition and Energy Consumption Forecasting for Heavy-Duty Vehicles

2025-10-28T14:48:41Z

Islab:

{{StudentProjectTemplate
|Summary=develop a machine learning framework for activity recognition and energy consumption forecasting; collaboration with Volvo
|TimeFrame=Fall 2025 or Spring 2026
|References=[1] Liu, J., Liu, Y., Zhu, W., Zhu, X., & Song, L. (2023). Distributional and spatial-temporal robust representation learning for transportation activity recognition. Pattern Recognition, 140, 109568.

[2] Khaertdinov, B., Ghaleb, E., & Asteriadis, S. (2021, August). Contrastive self-supervised learning for sensor-based human activity recognition. In 2021 IEEE International Joint Conference on Biometrics (IJCB) (pp. 1-8). IEEE.

[3] Zhang, J., Zi, L., Hou, Y., Wang, M., Jiang, W., & Deng, D. (2020). A deep learning‐based approach to enable action recognition for construction equipment. Advances in Civil Engineering, 2020(1), 8812928.

[4] Xing, Y., Lv, C., Cao, D., & Lu, C. (2020). Energy oriented driving behavior analysis and personalized prediction of vehicle states with joint time series modeling. Applied Energy, 261, 114471.

[6] Axelsson, H., & Wass, D. (2019). Machine Learning for Activity Recognition of Dumpers.

[7] Zhang, C., Shang, F., Liu, H., Wan, L., & Feng, W. (2025). FedAGC: Federated Continual Learning with Asymmetric Gradient Correction. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 3841-3850).

[8] Wang, Z., Fan, Y., Ydreskog, H., and Nowaczyk, S. (2025). Investigation on machine learning models for forecasting auxiliary energy consumption of HD-BEVs. In The 38th International Electric Vehicle Symposium & Exhibition. Electric Vehicle Symposium and Exhibition
|Supervisor=Yuantao Fan, TBD
|Level=Master
|Status=Open
}}
This thesis focuses on developing an AI-driven approach for activity recognition and energy consumption forecasting in heavy-duty vehicles. You will work with multivariate time-series data collected from on-board sensors in heavy-duty vehicles, creating compact and structured data representations through dimensionality reduction, feature extraction, and representation learning techniques to derive meaningful embeddings that capture the contextual dynamics of vehicle operations.

Potential research directions include: i) Multi-modal representation learning, integrating heterogeneous data sources (e.g., telemetry, environmental, and operational data) to enhance model robustness; ii) Contrastive learning, employing data augmentation strategies to pre-train models for improved generalization; iii) Federated learning, enabling decentralized and privacy-preserving model training across distributed edge devices; and iv) Domain adaptation and transfer learning, ensuring robustness across diverse operational contexts and vehicle configurations; v) Online learning with computationally efficient models for real-time training and inferences. Furthermore, achieving a certain degree of model interpretability is essential for understanding decision mechanisms and building trust in model outputs. This may involve clustering and visualizing embeddings to analyze operational states and transitions. Finally, a computational efficiency analysis shall be conducted to balance model performance with scalability and feasibility for real-world deployment.

You will work with the Advanced Analytics Team, at Volvo Group Technology, and collaborate with domain experts, stakeholders in different tech streams.

Please contact Yuantao for more details.

Anomaly Detection for Heavy-duty Vehicles

2025-10-28T14:39:33Z

Islab:

{{StudentProjectTemplate
|Summary=develop context-aware and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles; collaboration with Volvo
|TimeFrame=Fall 2025 or Spring 2026
|References=Li, Z., Zhu, Y., & Van Leeuwen, M. (2023). A survey on explainable anomaly detection. ACM Transactions on Knowledge Discovery from Data, 18(1), 1-54.

Li, Z., & Van Leeuwen, M. (2023). Explainable contextual anomaly detection using quantile regression forests. Data Mining and Knowledge Discovery, 37(6), 2517-2563.

Pasini, K., Khouadjia, M., Same, A., Trépanier, M., & Oukhellou, L. (2022). Contextual anomaly detection on time series: a case study of metro ridership analysis. Neural Computing and Applications, 34(2), 1483-1507.

Han, X., Zhang, L., Wu, Y., & Yuan, S. (2023, October). On root cause localization and anomaly mitigation through causal inference. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management (pp. 699-708).

Carvalho, J., Zhang, M., Geyer, R., Cotrini, C., & Buhmann, J. M. (2024). Invariant anomaly detection under distribution shifts: a causal perspective. Advances in Neural Information Processing Systems, 36.

Fan, Y., Nowaczyk, S., & Antonelo, E. A. (2016, July). Predicting air compressor failures with echo state networks. In PHM Society European Conference (Vol. 3, No. 1).

Gallicchio, C. (2024). Euler state networks: Non-dissipative reservoir computing. Neurocomputing, 579, 127411.
|Supervisor=Yuantao Fan, TBD
|Level=Master
|Status=Open
}}
Ensuring operational safety is crucial for commercial heavy-duty vehicles. The aim of this project is to design and develop context-aware and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles. Beyond safety considerations, cost-efficient and reliable maintenance strategies are essential for operating commercial vehicle fleets, where operational availability and total maintenance cost directly impact profitability.

By applying anomaly detection techniques to identify early signs of component degradation or failure, this project seeks to enable proactive and predictive maintenance. Such a system would minimize unplanned downtime, extend component lifespan, and reduce overall maintenance costs, thereby enhancing the safety, reliability, and economic performance of heavy-duty vehicle fleets. A promising research direction involves investigating representation learning techniques (such as contrastive learning, continual learning, federated learning etc.) to capture and encode key characteristics of time series data for anomaly detection. For instance, many state-of-the-art time methods utilize autoencoders, using learned embeddings in the latent features or reconstruction errors to compute anomaly scores. In addition, methods that are inherently explainable (e.g. causal graph/relations learned via causal inferences), that can be learned in an incremental setting (e.g. decision tree-based approaches) or computationally efficient (e.g. echo state network via reservoir computing), are of great interest as well. The developed approach will be evaluated on a real-world dataset collected from commercial vehicle fleets.

You will work with the Advanced Analytics Team, at Volvo Group Technology, and collaborate with domain experts, stakeholders in different tech streams.

Please contact Yuantao for more details.

Anomaly Detection for Heavy-duty Vehicles

2025-10-28T14:39:04Z

Islab:

{{StudentProjectTemplate
|Summary=develop context-aware and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles
|TimeFrame=Fall 2025 or Spring 2026
|References=Li, Z., Zhu, Y., & Van Leeuwen, M. (2023). A survey on explainable anomaly detection. ACM Transactions on Knowledge Discovery from Data, 18(1), 1-54.

Li, Z., & Van Leeuwen, M. (2023). Explainable contextual anomaly detection using quantile regression forests. Data Mining and Knowledge Discovery, 37(6), 2517-2563.

Pasini, K., Khouadjia, M., Same, A., Trépanier, M., & Oukhellou, L. (2022). Contextual anomaly detection on time series: a case study of metro ridership analysis. Neural Computing and Applications, 34(2), 1483-1507.

Han, X., Zhang, L., Wu, Y., & Yuan, S. (2023, October). On root cause localization and anomaly mitigation through causal inference. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management (pp. 699-708).

Carvalho, J., Zhang, M., Geyer, R., Cotrini, C., & Buhmann, J. M. (2024). Invariant anomaly detection under distribution shifts: a causal perspective. Advances in Neural Information Processing Systems, 36.

Fan, Y., Nowaczyk, S., & Antonelo, E. A. (2016, July). Predicting air compressor failures with echo state networks. In PHM Society European Conference (Vol. 3, No. 1).

Gallicchio, C. (2024). Euler state networks: Non-dissipative reservoir computing. Neurocomputing, 579, 127411.
|Supervisor=Yuantao Fan, TBD
|Level=Master
|Status=Open
}}
Ensuring operational safety is crucial for commercial heavy-duty vehicles. The aim of this project is to design and develop context-aware and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles. Beyond safety considerations, cost-efficient and reliable maintenance strategies are essential for operating commercial vehicle fleets, where operational availability and total maintenance cost directly impact profitability.

By applying anomaly detection techniques to identify early signs of component degradation or failure, this project seeks to enable proactive and predictive maintenance. Such a system would minimize unplanned downtime, extend component lifespan, and reduce overall maintenance costs, thereby enhancing the safety, reliability, and economic performance of heavy-duty vehicle fleets. A promising research direction involves investigating representation learning techniques (such as contrastive learning, continual learning, federated learning etc.) to capture and encode key characteristics of time series data for anomaly detection. For instance, many state-of-the-art time methods utilize autoencoders, using learned embeddings in the latent features or reconstruction errors to compute anomaly scores. In addition, methods that are inherently explainable (e.g. causal graph/relations learned via causal inferences), that can be learned in an incremental setting (e.g. decision tree-based approaches) or computationally efficient (e.g. echo state network via reservoir computing), are of great interest as well. The developed approach will be evaluated on a real-world dataset collected from commercial vehicle fleets.

You will work with the Advanced Analytics Team, at Volvo Group Technology, and collaborate with domain experts, stakeholders in different tech streams.

Please contact Yuantao for more details.

Anomaly Detection for Heavy-duty Vehicles

2025-10-28T12:56:02Z

Islab:

{{StudentProjectTemplate
|Summary=develop context-aware and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles
|TimeFrame=Fall 2025 or Spring 2026
|References=Li, Z., Zhu, Y., & Van Leeuwen, M. (2023). A survey on explainable anomaly detection. ACM Transactions on Knowledge Discovery from Data, 18(1), 1-54.

Li, Z., & Van Leeuwen, M. (2023). Explainable contextual anomaly detection using quantile regression forests. Data Mining and Knowledge Discovery, 37(6), 2517-2563.

Pasini, K., Khouadjia, M., Same, A., Trépanier, M., & Oukhellou, L. (2022). Contextual anomaly detection on time series: a case study of metro ridership analysis. Neural Computing and Applications, 34(2), 1483-1507.

Han, X., Zhang, L., Wu, Y., & Yuan, S. (2023, October). On root cause localization and anomaly mitigation through causal inference. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management (pp. 699-708).

Carvalho, J., Zhang, M., Geyer, R., Cotrini, C., & Buhmann, J. M. (2024). Invariant anomaly detection under distribution shifts: a causal perspective. Advances in Neural Information Processing Systems, 36.

Fan, Y., Nowaczyk, S., & Antonelo, E. A. (2016, July). Predicting air compressor failures with echo state networks. In PHM Society European Conference (Vol. 3, No. 1).

Gallicchio, C. (2024). Euler state networks: Non-dissipative reservoir computing. Neurocomputing, 579, 127411.
|Supervisor=Yuantao Fan, TBD
|Level=Master
|Status=Open
}}
Ensuring operational safety is important for commercial heavy-duty vehicles. The aim of this project is to design and develop context-aware and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles. Beyond safety considerations, cost-efficient and reliable maintenance strategies are critical for the effective operation of commercial vehicle fleets, where operational availability and total maintenance cost directly influence overall profitability.

By applying anomaly detection techniques to identify early signs of component degradation or failure, this project seeks to enable proactive and predictive maintenance. Such a system would minimize unplanned downtime, extend component lifespan, and reduce overall maintenance costs, thereby enhancing the safety, reliability, and economic performance of heavy-duty vehicle fleets.

A promising research direction is to explore various approaches for learning useful representations (e.g. contrastive learning, continual learning, federated learning etc.) that can capture and encode key characteristics of time series data for anomaly detection. For instance, autoencoders can be trained, and learned embeddings in the latent features or reconstruction errors can be used to compute the anomaly score. Methods that are inherently explainable (e.g. causal relations learned via causal inferences), can be learned in an incremental setting (e.g. decision tree-based approaches) or computationally efficient (e.g. echo state network via reservoir computing), are of interests as well. The developed approach will be evaluated and compared with sota methods on a real-world dataset collected from commercial heavy-duty vehicles.

The plan is to work with the Advanced Analytics Team (Volvo Group Technology) and have the opportunity to collaborate with domain experts and stakeholders.

Please contact Yuantao for more details.

Asynchronous Federated Learning for Commercial Vehicle Fleets

2025-10-28T11:32:07Z

Islab: Created page with "{{StudentProjectTemplate |Summary=explore and design Asynchronous Federated Learning strategies for commercial vehicle fleets in AI-driven digital services |TimeFrame=Fall 202..."

{{StudentProjectTemplate
|Summary=explore and design Asynchronous Federated Learning strategies for commercial vehicle fleets in AI-driven digital services
|TimeFrame=Fall 2025 or Spring 2026
|References=[1] Xu, C., Qu, Y., Xiang, Y., & Gao, L. (2023). Asynchronous federated learning on heterogeneous devices: A survey. Computer Science Review, 50, 100595.

[2] Chen, Z., Liao, W., Hua, K., Lu, C., & Yu, W. (2021). Towards asynchronous federated
learning for heterogeneous edge-powered internet of things. Digital Communications and Networks, 7(3),
317-326.

[3] Imteaj, A., & Amini, M. H. (2020, December). Fedar: Activity and resource-aware federated learning model for distributed mobile robots. In 2020 19th IEEE International Conference on Machine Learning and Applications (ICMLA) (pp. 1153-1160). IEEE.

[4] Wu, W., He, L., Lin, W., Mao, R., Maple, C., & Jarvis, S. (2020). SAFA: A semi-asynchronous protocol for fast federated learning with low overhead. IEEE Transactions on Computers, 70(5), 655-668.

[5] Chen, Y., Sun, X., & Jin, Y. (2019). Communication-efficient federated deep learning with layerwise asynchronous model update and temporally weighted aggregation. IEEE transactions on neural networks and learning systems, 31(10), 4229-4238.

[6] Xie, C., Koyejo, S., & Gupta, I. (2019). Asynchronous federated optimization. arXiv preprint arXiv:1903.03934.

[7] Sun, W., Lei, S., Wang, L., Liu, Z., & Zhang, Y. (2020). Adaptive federated learning and digital twin for industrial internet of things. IEEE Transactions on Industrial Informatics, 17(8), 5605-5614.

[8] Zhang, Y., Duan, M., Liu, D., Li, L., Ren, A., Chen, X., ... & Wang, C. (2021). CSAFL: A clustered semi-asynchronous federated learning framework. IJCNN, pp. 1-10.

|Supervisor=Yuantao Fan, Zahra Taghiyarrenani
|Level=Master
|Status=Open
}}
Traditional machine learning workflows often depend on centralized data aggregation, which creates high communication overhead and raises privacy concerns. While federated learning (FL) seeks to address these issues, it faces scaling difficulties in large, heterogeneous fleets due to differences in device capabilities and connectivity. Synchronization requirements in standard FL can cause idle waiting for slower devices, reducing overall efficiency. In addition, repetitive computations on previously learned data lead to unnecessary resource
consumption in large-scale deployments. These obstacles emphasize the need for more flexible, asynchronous approaches that accommodate changing device connectivity, varying computation resources and data availability, and diverse fleet configurations.

This project explores and aims to design scalable Asynchronous Federated Learning (AFL) solutions that account for fleet heterogeneity, storage efficiency, and computational constraints. By utilising multiple aggregated models and personalized learning, the research aims to enhance model accuracy, adaptability and efficiency for AI-driven digital services in commercial vehicle fleets. Target application areas include energy consumption forecasting, representation learning, anomaly detection, and related intelligent mobility tasks.

Existing research on AFL spans several key areas: i) node selection: while classical FL selects nodes based primarily on their data volume, AFL prioritizes nodes with heightened resilience, network and computational capacity. For example, work presented in [2] employs a heuristic greedy node selection strategy to iteratively involve nodes in global learning based on local computing and communication resources, whereas other works use trust scores [3], or crashing probability [4] to guide the selection process; ii) weighted aggregation: increasing the weight of recently updated local models [5] or adjusting a hyperparameter to balance convergence speed and variance due to staleness [6]; iii) gradient compression, to reduce communication expenses. Fourth, employing semi-asynchronous FL as a hybrid approach, aggregating local models that arrive early, and involving slow devices based on the magnitude of stabledness; iv) cluster-based FL, which groups nodes based on aggregation frequency [7], gradient direction, or latency [8] to optimize overall performance. These research directions address key challenges related to heterogeneity, communication constraints, and model convergence, and are of interest to explore.

Anomaly Detection for Heavy-duty Vehicles

2025-10-28T10:58:35Z

Islab:

{{StudentProjectTemplate
|Summary=develop context-aware and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles
|TimeFrame=Fall 2025 or Spring 2026
|References=Li, Z., Zhu, Y., & Van Leeuwen, M. (2023). A survey on explainable anomaly detection. ACM Transactions on Knowledge Discovery from Data, 18(1), 1-54.

Li, Z., & Van Leeuwen, M. (2023). Explainable contextual anomaly detection using quantile regression forests. Data Mining and Knowledge Discovery, 37(6), 2517-2563.

Pasini, K., Khouadjia, M., Same, A., Trépanier, M., & Oukhellou, L. (2022). Contextual anomaly detection on time series: a case study of metro ridership analysis. Neural Computing and Applications, 34(2), 1483-1507.

Han, X., Zhang, L., Wu, Y., & Yuan, S. (2023, October). On root cause localization and anomaly mitigation through causal inference. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management (pp. 699-708).

Carvalho, J., Zhang, M., Geyer, R., Cotrini, C., & Buhmann, J. M. (2024). Invariant anomaly detection under distribution shifts: a causal perspective. Advances in Neural Information Processing Systems, 36.

Fan, Y., Nowaczyk, S., & Antonelo, E. A. (2016, July). Predicting air compressor failures with echo state networks. In PHM Society European Conference (Vol. 3, No. 1).

Gallicchio, C. (2024). Euler state networks: Non-dissipative reservoir computing. Neurocomputing, 579, 127411.
|Supervisor=Yuantao Fan, TBD
|Level=Master
|Status=Open
}}
Ensuring operational safety is important for commercial heavy-duty vehicles. The aim of this project is to design and develop context-aware and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles. Beyond safety considerations, cost-efficient and reliable maintenance strategies are critical for the effective operation of commercial vehicle fleets, where operational availability and total maintenance cost directly influence overall profitability.

By applying anomaly detection techniques to identify early signs of component degradation or failure, this project seeks to enable proactive and predictive maintenance. Such a system would minimize unplanned downtime, extend component lifespan, and reduce overall maintenance costs, thereby enhancing the safety, reliability, and economic performance of heavy-duty vehicle fleets.

A promising research direction is to explore the use of representation learning methods, e.g. deep learning-based approaches (including time series embedding methods), that can capture and encode key characteristics of time series data for anomaly detection. For instance, autoencoders can be trained, and learned embeddings in the latent features or reconstruction errors can be used to compute the anomaly score. Methods that are inherently explainable (e.g. causal relations learned via causal inferences), can be learned in an incremental setting (e.g. decision tree-based approaches) or computationally efficient (e.g. echo state network via reservoir computing), are of interests as well. The developed approach will be evaluated and compared with sota methods on a real-world dataset collected from commercial heavy-duty vehicles.

The plan is to work with the Advanced Analytics Team (Volvo Group Technology) and have the opportunity to collaborate with domain experts and stakeholders.

Please contact Yuantao for more details.

Anomaly Detection for Heavy-duty Vehicles

2025-10-28T10:58:16Z

Islab: Created page with "{{StudentProjectTemplate |Summary=develop context-aware and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty ve..."

{{StudentProjectTemplate
|Summary=develop context-aware and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles
|TimeFrame=Fall 2025 or Spring 2026
|References=Li, Z., Zhu, Y., & Van Leeuwen, M. (2023). A survey on explainable anomaly detection. ACM Transactions on Knowledge Discovery from Data, 18(1), 1-54.

Li, Z., & Van Leeuwen, M. (2023). Explainable contextual anomaly detection using quantile regression forests. Data Mining and Knowledge Discovery, 37(6), 2517-2563.

Pasini, K., Khouadjia, M., Same, A., Trépanier, M., & Oukhellou, L. (2022). Contextual anomaly detection on time series: a case study of metro ridership analysis. Neural Computing and Applications, 34(2), 1483-1507.

Han, X., Zhang, L., Wu, Y., & Yuan, S. (2023, October). On root cause localization and anomaly mitigation through causal inference. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management (pp. 699-708).

Carvalho, J., Zhang, M., Geyer, R., Cotrini, C., & Buhmann, J. M. (2024). Invariant anomaly detection under distribution shifts: a causal perspective. Advances in Neural Information Processing Systems, 36.

Fan, Y., Nowaczyk, S., & Antonelo, E. A. (2016, July). Predicting air compressor failures with echo state networks. In PHM Society European Conference (Vol. 3, No. 1).

Gallicchio, C. (2024). Euler state networks: Non-dissipative reservoir computing. Neurocomputing, 579, 127411.
|Supervisor=Yuantao Fan, TBD
|Level=Master
}}
Ensuring operational safety is important for commercial heavy-duty vehicles. The aim of this project is to design and develop context-aware and explainable anomaly detection algorithms for monitoring critical components and their efficiencies in heavy-duty vehicles. Beyond safety considerations, cost-efficient and reliable maintenance strategies are critical for the effective operation of commercial vehicle fleets, where operational availability and total maintenance cost directly influence overall profitability.

By applying anomaly detection techniques to identify early signs of component degradation or failure, this project seeks to enable proactive and predictive maintenance. Such a system would minimize unplanned downtime, extend component lifespan, and reduce overall maintenance costs, thereby enhancing the safety, reliability, and economic performance of heavy-duty vehicle fleets.

A promising research direction is to explore the use of representation learning methods, e.g. deep learning-based approaches (including time series embedding methods), that can capture and encode key characteristics of time series data for anomaly detection. For instance, autoencoders can be trained, and learned embeddings in the latent features or reconstruction errors can be used to compute the anomaly score. Methods that are inherently explainable (e.g. causal relations learned via causal inferences), can be learned in an incremental setting (e.g. decision tree-based approaches) or computationally efficient (e.g. echo state network via reservoir computing), are of interests as well. The developed approach will be evaluated and compared with sota methods on a real-world dataset collected from commercial heavy-duty vehicles.

The plan is to work with the Advanced Analytics Team (Volvo Group Technology) and have the opportunity to collaborate with domain experts and stakeholders.

Please contact Yuantao for more details.

Data-Driven Activity Recognition and Energy Consumption Forecasting for Heavy-Duty Vehicles

2025-10-28T09:35:37Z

Islab:

{{StudentProjectTemplate
|Summary=Developing a machine learning framework for activity recognition and energy consumption forecasting, in collaboration with Volvo
|TimeFrame=Fall 2025 or Spring 2026
|References=

[1] Liu, J., Liu, Y., Zhu, W., Zhu, X., & Song, L. (2023). Distributional and spatial-temporal robust representation learning for transportation activity recognition. Pattern Recognition, 140, 109568.

[2] Khaertdinov, B., Ghaleb, E., & Asteriadis, S. (2021, August). Contrastive self-supervised learning for sensor-based human activity recognition. In 2021 IEEE International Joint Conference on Biometrics (IJCB) (pp. 1-8). IEEE.

[3] Zhang, J., Zi, L., Hou, Y., Wang, M., Jiang, W., & Deng, D. (2020). A deep learning‐based approach to enable action recognition for construction equipment. Advances in Civil Engineering, 2020(1), 8812928.

[4] Xing, Y., Lv, C., Cao, D., & Lu, C. (2020). Energy oriented driving behavior analysis and personalized prediction of vehicle states with joint time series modeling. Applied Energy, 261, 114471.

[6] Axelsson, H., & Wass, D. (2019). Machine Learning for Activity Recognition of Dumpers.

[7] Zhang, C., Shang, F., Liu, H., Wan, L., & Feng, W. (2025). FedAGC: Federated Continual Learning with Asymmetric Gradient Correction. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 3841-3850).

[8] Wang, Z., Fan, Y., Ydreskog, H., and Nowaczyk, S. (2025). Investigation on machine learning models for forecasting auxiliary energy consumption of HD-BEVs. In The 38th International Electric Vehicle Symposium & Exhibition. Electric Vehicle Symposium and Exhibition
|Supervisor=Yuantao Fan, TBD
|Level=Master
|Status=Open
}}
This thesis focuses on developing a machine learning–based approach for activity recognition and energy consumption forecasting in heavy-duty vehicles. You will work with multivariate time-series data collected from on-board sensors in heavy-duty vehicles. The research will emphasize creating compact and structured data representations through dimensionality reduction, feature extraction, and representation learning techniques to derive meaningful embeddings that capture the contextual dynamics of vehicle operations.

Potential research directions include: i) Multi-modal representation learning, integrating heterogeneous data sources (e.g., telemetry, environmental, and operational data) to enhance model robustness; ii) Contrastive learning, employing data augmentation strategies to pre-train models for improved generalization; iii) Federated learning, enabling decentralized and privacy-preserving model training across distributed edge devices; and iv) Domain adaptation and transfer learning, ensuring robustness across diverse operational contexts and vehicle configurations; v) Online learning with computationally efficient models for real-time training and inferences. Furthermore, achieving a certain degree of model interpretability is essential for understanding decision mechanisms and building trust in model outputs. This may involve clustering and visualizing embeddings to analyze operational states and transitions. Finally, a computational efficiency analysis shall be conducted to balance model performance with scalability and feasibility for real-world deployment.

The plan is to work with the Advanced Analytics Team (Volvo Group Technology) and have the opportunity to collaborate with domain experts and stakeholders.

Please contact Yuantao for more details.

Data-Driven Activity Recognition and Energy Consumption Forecasting for Heavy-Duty Vehicles

2025-10-28T09:35:21Z

Islab:

{{StudentProjectTemplate
|Summary=Developing a machine learning framework for activity recognition and energy consumption forecasting, in collaboration with Volvo
|TimeFrame=Fall 2025 or Spring 2026
|References=[1] Liu, J., Liu, Y., Zhu, W., Zhu, X., & Song, L. (2023). Distributional and spatial-temporal robust representation learning for transportation activity recognition. Pattern Recognition, 140, 109568.

[2] Khaertdinov, B., Ghaleb, E., & Asteriadis, S. (2021, August). Contrastive self-supervised learning for sensor-based human activity recognition. In 2021 IEEE International Joint Conference on Biometrics (IJCB) (pp. 1-8). IEEE.

[3] Zhang, J., Zi, L., Hou, Y., Wang, M., Jiang, W., & Deng, D. (2020). A deep learning‐based approach to enable action recognition for construction equipment. Advances in Civil Engineering, 2020(1), 8812928.

[4] Xing, Y., Lv, C., Cao, D., & Lu, C. (2020). Energy oriented driving behavior analysis and personalized prediction of vehicle states with joint time series modeling. Applied Energy, 261, 114471.

[6] Axelsson, H., & Wass, D. (2019). Machine Learning for Activity Recognition of Dumpers.

[7] Zhang, C., Shang, F., Liu, H., Wan, L., & Feng, W. (2025). FedAGC: Federated Continual Learning with Asymmetric Gradient Correction. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 3841-3850).

[8] Wang, Z., Fan, Y., Ydreskog, H., and Nowaczyk, S. (2025). Investigation on machine learning models for forecasting auxiliary energy consumption of HD-BEVs. In The 38th International Electric Vehicle Symposium & Exhibition. Electric Vehicle Symposium and Exhibition
|Supervisor=Yuantao Fan, TBD
|Level=Master
|Status=Open
}}
This thesis focuses on developing a machine learning–based approach for activity recognition and energy consumption forecasting in heavy-duty vehicles. You will work with multivariate time-series data collected from on-board sensors in heavy-duty vehicles. The research will emphasize creating compact and structured data representations through dimensionality reduction, feature extraction, and representation learning techniques to derive meaningful embeddings that capture the contextual dynamics of vehicle operations.

Potential research directions include: i) Multi-modal representation learning, integrating heterogeneous data sources (e.g., telemetry, environmental, and operational data) to enhance model robustness; ii) Contrastive learning, employing data augmentation strategies to pre-train models for improved generalization; iii) Federated learning, enabling decentralized and privacy-preserving model training across distributed edge devices; and iv) Domain adaptation and transfer learning, ensuring robustness across diverse operational contexts and vehicle configurations; v) Online learning with computationally efficient models for real-time training and inferences. Furthermore, achieving a certain degree of model interpretability is essential for understanding decision mechanisms and building trust in model outputs. This may involve clustering and visualizing embeddings to analyze operational states and transitions. Finally, a computational efficiency analysis shall be conducted to balance model performance with scalability and feasibility for real-world deployment.

The plan is to work with the Advanced Analytics Team (Volvo Group Technology) and have the opportunity to collaborate with domain experts and stakeholders.

Please contact Yuantao for more details.

Data-Driven Activity Recognition and Energy Consumption Forecasting for Heavy-Duty Vehicles

2025-10-28T09:28:33Z

Islab: Created page with "{{StudentProjectTemplate |Summary=Developing a machine learning framework for activity recognition and energy consumption forecasting, in collaboration with Volvo |TimeFrame=F..."

{{StudentProjectTemplate
|Summary=Developing a machine learning framework for activity recognition and energy consumption forecasting, in collaboration with Volvo
|TimeFrame=Fall 2025 or Spring 2026
|References=[1] Liu, J., Liu, Y., Zhu, W., Zhu, X., & Song, L. (2023). Distributional and spatial-temporal robust representation learning for transportation activity recognition. Pattern Recognition, 140, 109568.

[2] Khaertdinov, B., Ghaleb, E., & Asteriadis, S. (2021, August). Contrastive self-supervised learning for sensor-based human activity recognition. In 2021 IEEE International Joint Conference on Biometrics (IJCB) (pp. 1-8). IEEE.

[3] Zhang, J., Zi, L., Hou, Y., Wang, M., Jiang, W., & Deng, D. (2020). A deep learning‐based approach to enable action recognition for construction equipment. Advances in Civil Engineering, 2020(1), 8812928.

[4] Xing, Y., Lv, C., Cao, D., & Lu, C. (2020). Energy oriented driving behavior analysis and personalized prediction of vehicle states with joint time series modeling. Applied Energy, 261, 114471.

[6] Axelsson, H., & Wass, D. (2019). Machine Learning for Activity Recognition of Dumpers.

[7] Zhang, C., Shang, F., Liu, H., Wan, L., & Feng, W. (2025). FedAGC: Federated Continual Learning with Asymmetric Gradient Correction. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 3841-3850).

[8] Wang, Z., Fan, Y., Ydreskog, H., and Nowaczyk, S. (2025). Investigation on machine learning models for forecasting auxiliary energy consumption of HD-BEVs. In The 38th International Electric Vehicle Symposium & Exhibition. Electric Vehicle Symposium and Exhibition
|Supervisor=Yuantao Fan, TBD
|Level=Master
|Status=Open
}}
This thesis focuses on developing a machine learning–based approach for activity recognition and energy consumption forecasting in heavy-duty vehicles. You will work with multivariate time-series data collected from on-board sensors in heavy-duty vehicles. The research will emphasize creating compact and structured data representations through dimensionality reduction, feature extraction, and representation learning techniques to derive meaningful embeddings that capture the contextual dynamics of vehicle operations.

Potential research directions include: i) Multi-modal representation learning, integrating heterogeneous data sources (e.g., telemetry, environmental, and operational data) to enhance model robustness; ii) Contrastive learning, employing data augmentation strategies to pre-train models for improved generalization; iii) Federated learning, enabling decentralized and privacy-preserving model training across distributed edge devices; and iv) Domain adaptation and transfer learning, ensuring robustness across diverse operational contexts and vehicle configurations; v) Online learning with computationally efficient models for real-time training and inferences. Furthermore, achieving a certain degree of model interpretability is essential for understanding decision mechanisms and building trust in model outputs. This may involve clustering and visualizing embeddings to analyze operational states and transitions. Finally, a computational efficiency analysis shall be conducted to balance model performance with scalability and feasibility for real-world deployment.

Interactive XAI using LLMs

2025-10-28T08:39:47Z

Islab: Created page with "{{StudentProjectTemplate |Summary=XAI methods would benefit greatly from becoming more interactive: this project aims to explore the use of LLMs for this purpose |TimeFrame=Fa..."

{{StudentProjectTemplate
|Summary=XAI methods would benefit greatly from becoming more interactive: this project aims to explore the use of LLMs for this purpose
|TimeFrame=Fall 2025
|Supervisor=Slawomir Nowaczyk & TBD
|Level=Master
|Status=Open
}}
Many XAI methods have been proposed. Several can, in principle, be used interactively.
However, evaluating them is typically done through experiments with humans, which is expensive and slow.

In this thesis, we'd like to (as part of research project KEEPER) explore the use of LLMs to develop and evaluate interactive XAI methods.

We expect this thesis to result in a scientific paper.

Captioning Engine for AD/ADAS data using Multi-Modal Large Language Models

2025-10-24T14:15:04Z

Islab: Created page with "{{StudentProjectTemplate |Summary=Use advanced LLMs to describe and interpret sensor data from autonomous vehicles. |TimeFrame=Spring 2026 |Supervisor=Felix Rosberg, Cristofer..."

{{StudentProjectTemplate
|Summary=Use advanced LLMs to describe and interpret sensor data from autonomous vehicles.
|TimeFrame=Spring 2026
|Supervisor=Felix Rosberg, Cristofer Englund
|Level=Master
|Status=Open
}}
Captioning Engine for AD/ADAS data using Multi-Modal Large Language Models – Use advanced LLMs to describe and interpret sensor data from autonomous vehicles. Apply latest October 31

How to Apply?

Submit your CV, motivation letter, and grade transcripts.

Applying as a pair? Include your partner’s name in the application.
Planned start: January 2026 (flexible)
Application deadline: October 31, 2025 (applications reviewed continuously)

Read more and apply at Zenseacts webpage: https://career.zenseact.com/jobs/6537044-master-thesis-projects-in-sensing-perception

Cristofer Englund

2025-10-24T14:09:01Z

Islab:

{{Person
|Family Name=Englund
|Given Name=Cristofer
|Title=Professor
|Cell Phone=+46 708 560 227
|Position=Dean of School of Information Technology
|Email=cristofer.englund@hh.se
|Image=CEnglund800.jpg
|Street Address=https://www.hh.se/information/sok-personal.html?person=688095C0-1FFA-446E-9C94-73EC5E8DA29E
|Office=F505
|url=https://scholar.google.com/citations?user=xstRD04AAAAJ&hl=sv
|Given_Name=Cristofer
|Family_name=Englund
}}
{{AssignApplicationAreas
|ApplicationArea=Intelligent Vehicles
}}
{{ShowPerson}}
{{PublicationsList}}

[[Category:Staff]]

Cristofer Englund

2025-10-24T14:07:56Z

Islab:

{{Person
|Family Name=Englund
|Given Name=Cristofer
|Title=Professor
|Cell Phone=+46 708 560 227
|Position=Dean of School of Information Technology
|Email=cristofer.englund@hh.se
|Image=CEnglund800.jpg
|Street Address=https://www.hh.se/information/sok-personal.html?person=688095C0-1FFA-446E-9C94-73EC5E8DA29E
|Office=F505
|Given_Name=Cristofer
|Family_name=Englund
}}
{{AssignApplicationAreas
|ApplicationArea=Intelligent Vehicles
}}
{{ShowPerson}}
{{PublicationsList}}

[[Category:Staff]]

Cristofer Englund

2025-10-24T14:06:43Z

Islab:

{{Person
|Family Name=Englund
|Given Name=Cristofer
|Title=PhD
|Cell Phone=+46 708 560 227
|Position=Adjunct Professor
|Email=cristofer.englund@hh.se
|Image=CEnglund800.jpg
|Street Address=https://www.hh.se/information/sok-personal.html?person=688095C0-1FFA-446E-9C94-73EC5E8DA29E
|Office=F505
|Given_Name=Cristofer
|Family_name=Englund
}}
{{AssignApplicationAreas
|ApplicationArea=Intelligent Vehicles
}}
{{ShowPerson}}
{{PublicationsList}}

[[Category:Staff]]

Knowledge graphs 4 XAI in Healthcare

2025-10-24T10:15:39Z

Islab:

{{StudentProjectTemplate
|Summary=Knowledge graphs in Healthcare
|Keywords=Knowledge graphs, XAI, Healthcare
|TimeFrame=Spring 2026
|References=https://www.sciencedirect.com/science/article/pii/S1566253523001148
https://www.sciencedirect.com/science/article/abs/pii/S1532046425000905
https://www.sciencedirect.com/science/article/pii/S0004370221001788
https://www.sciencedirect.com/science/article/pii/S1532046423001247
https://arxiv.org/abs/2402.12608
|Supervisor=Grzegorz J. Nalepa, Farzaneh Etminani, Amira Soliman
|Level=Master
|Status=Open
}}
Knowledge graphs (KG) are a knowledge representation method that has became increasingly important in the context of the explainable AI approach.
Healthcare (HC) systems are a very special area of application of AI due to data sensitivity andf privacy concerns.
In this thesis we would like to conduct a survey on a recent practical applications of KG i HC for knowledge representation and management.
Furthermore, we would like to explore the role of KG in interoperability and FAIRness of HC data.
Moreover, we would like to consider KG for supporting XAI for HC, specifically what explanation forms are mostly suitable for HC experts and how they can be supported by background knowledge
Finally, we want to explore knowledge augmented retrieval with KG for HC data and knowledge bases where explanations can meet medical data.

It is encouraged that this thesis will result in scientific publications possibly also developed in collaboration with external stakeholders.

Cybersec XAI 4 TS

2025-10-24T09:46:06Z

Islab: Islab moved page Cybersec XAI 4 TS to Timeseries XAI in Cybersecurity and Industry

#REDIRECT [[Timeseries XAI in Cybersecurity and Industry]]

Timeseries XAI in Cybersecurity and Industry

2025-10-24T09:46:06Z

Islab: Islab moved page Cybersec XAI 4 TS to Timeseries XAI in Cybersecurity and Industry

{{StudentProjectTemplate
|Summary=Timeseries data analysis with XAI in Cybersecurity and Industry
|Keywords=Cybersecurity, XAI, timeseries, industry
|TimeFrame=Spring 2026
|References=https://www.sciencedirect.com/science/article/pii/S1566253523001148
https://www.nature.com/articles/s41597-025-04603-x
|Supervisor=Grzegorz J. Nalepa, Prayag Tawari, Aurora Esteban
|Level=Master
|Status=Open
}}
Time series data is ubiquitous — from industrial monitoring systems and energy networks to cybersecurity systems and user activity traces. Understanding temporal patterns is crucial for detecting anomalies, anticipating failures, and supporting human decision-making. Yet, the increasing complexity of time series models makes them difficult to interpret and trust.
Industrial and Cybersecurity systems have clearly became a very important area of AI applications recently. From the engineering perspective they produce a large amount of data that can only be analyzed by AI methods.

Explainable Artificial Intelligence (XAI) aims to make models more transparent by uncovering the why behind their predictions. While explainability methods are well-studied for tabular and image data, time series explanations remain a significant open challenge. Temporal dependencies, non-stationarity, and concept drift make it difficult to represent and communicate model reasoning to domain experts.

This project will explore explainable learning and reasoning for time series data, with several possible research directions depending on the student’s interests and available datasets:
- Characterising domain-specific dynamics: analysing how time series from different domains (e.g., industrial processes vs. cybersecurity traffic) differ in variability, drifts, or anomaly structure.
- Representation learning for interpretability: studying prototypes, motifs, or symbolic rules that capture meaningful temporal patterns.
- Counterfactual explanations: developing or adapting methods (e.g., genetic algorithms, motif transformations, gradient perturbations) to generate realistic “what-if” scenarios for time series.
- Explainable anomaly detection: integrating interpretability into models that identify abnormal or critical events over time.
- Concept drift and model evolution: explaining how and why model behavior changes as time series distributions shift.

The work will be done in connection with the KEEPER project using data from our industrial partners such as Volvo, HMS, Toyota, etc. [[https://www.hh.se/english/research/our-research/research-at-the-school-of-information-technology/technology-area-aware-intelligent-systems/research-projects-within-aware-intelligent-systems/keeper---knowledge-creation-for-efficient-and-predictable-industrial-operations-.html]]. The project may use as well public data such as Numenta Anomaly Benchmark or UCR/UEA archive.

It is encouraged that this thesis will result in scientific publications possibly also developed in collaboration with external stakeholders.

Timeseries XAI in Cybersecurity and Industry

2025-10-24T09:43:10Z

Islab:

AGENTIC-AI TOWARDS INTERPRETATION OF SERVICE CANVAS AND AUTOMATION OF TRUSTWORTHY ML-PIPELINES

2025-10-23T15:55:15Z

Islab:

{{StudentProjectTemplate
|Summary=This project allows the MSc student(s) to study and experiment on Agentic AI methods for the generation of trustworthy ML-pipelines
|Keywords=Agentic AI
|TimeFrame=Start 2025 Autumn
|References=[1] Brohi S, Mastoi Q-u-a, Jhanjhi NZ, Pillai TR. A Research Landscape of Agentic AI and Large Language Models: Applications, Challenges and Future Directions. Algorithms. 2025; 18(8):499. https://doi.org/10.3390/a1808049
[2] Trirat, Patara, Wonyong Jeong, and Sung Ju Hwang. "Automl-agent: A multi-agent llm framework for full-pipeline automl." arXiv preprint arXiv:2410.02958 (2024).
|Supervisor=Jens Lundström, Peyman Mashhadi, Stefan Byttner
|Level=Master
|Status=Open
}}
With the emergence of LLM-based Agentic AI [1], language model technology is being used not just as passive tools for generating text, but as a part of autonomous agents capable of reasoning, decision-making, and interacting with environments or other systems. The generation of machine learning pipelines is no exception in the scope of applications covered by Agentic AI [2]. However, with the adoption of legal frameworks related to AI, e.g. EU AI Act it is of importance to develop algorithmic solutions for generation of ML-pipelines that are trustworthy and aligns with the requirements of associated legally-defined risk-levels. To iteratively align with such requirements agents need to consider trustworthy aspects already in the phase of service design. This MSc-thesis explores knowledge-creation in the realm of Agentic AI systems for trustworthy ML-pipelines starting from service canvas and centers around human-in-the-loop and agent specifications. Currently, it is unclear how the degree of autonomy, architectural complexity and level of reasoning affects the performance and trustworthiness of generated ML-pipelines. This thesis will also focus on agents creating their own synthetic data for reasoning and decision-making and tools for augmented service canvas creation. The thesis work will be carried out as two interdisciplinary projects with students from other disciplines.

AGENTIC-AI TOWARDS INTERPRETATION OF SERVICE CANVAS AND AUTOMATION OF TRUSTWORTHY ML-PIPELINES

2025-10-23T15:11:23Z

Islab:

{{StudentProjectTemplate
|Summary=This project allows the MSc student(s) to study and experiment on Agentic AI methods for the generation of trustworthy ML-pipelines
|Keywords=Agentic AI
|TimeFrame=Start 2025 Autumn
|References=[1] Brohi S, Mastoi Q-u-a, Jhanjhi NZ, Pillai TR. A Research Landscape of Agentic AI and Large Language Models: Applications, Challenges and Future Directions. Algorithms. 2025; 18(8):499. https://doi.org/10.3390/a1808049
[2] Trirat, Patara, Wonyong Jeong, and Sung Ju Hwang. "Automl-agent: A multi-agent llm framework for full-pipeline automl." arXiv preprint arXiv:2410.02958 (2024).
|Supervisor=Jens Lundström, Peyman Mashadi, Stefan Byttner
|Level=Master
|Status=Open
}}
With the emergence of LLM-based Agentic AI [1], language model technology is being used not just as passive tools for generating text, but as a part of autonomous agents capable of reasoning, decision-making, and interacting with environments or other systems. The generation of machine learning pipelines is no exception in the scope of applications covered by Agentic AI [2]. However, with the adoption of legal frameworks related to AI, e.g. EU AI Act it is of importance to develop algorithmic solutions for generation of ML-pipelines that are trustworthy and aligns with the requirements of associated legally-defined risk-levels. To iteratively align with such requirements agents need to consider trustworthy aspects already in the phase of service design. This MSc-thesis explores knowledge-creation in the realm of Agentic AI systems for trustworthy ML-pipelines starting from service canvas and centers around human-in-the-loop and agent specifications. Currently, it is unclear how the degree of autonomy, architectural complexity and level of reasoning affects the performance and trustworthiness of generated ML-pipelines. This thesis will also focus on agents creating their own synthetic data for reasoning and decision-making and tools for augmented service canvas creation. The thesis work will be carried out as two interdisciplinary projects with students from other disciplines.

AGENTIC-AI TOWARDS INTERPRETATION OF SERVICE CANVAS AND AUTOMATION OF TRUSTWORTHY ML-PIPELINES

2025-10-23T15:09:29Z

Islab: Created page with "{{StudentProjectTemplate |Summary=This project allows the study and experimentation on Agentic AI methods for the germination of trustworthy ML-pipelines |Keywords=Agentic AI ..."

{{StudentProjectTemplate
|Summary=This project allows the study and experimentation on Agentic AI methods for the germination of trustworthy ML-pipelines
|Keywords=Agentic AI
|TimeFrame=Start 2025 Autumn
|References=[1] Brohi S, Mastoi Q-u-a, Jhanjhi NZ, Pillai TR. A Research Landscape of Agentic AI and Large Language Models: Applications, Challenges and Future Directions. Algorithms. 2025; 18(8):499. https://doi.org/10.3390/a1808049
[2] Trirat, Patara, Wonyong Jeong, and Sung Ju Hwang. "Automl-agent: A multi-agent llm framework for full-pipeline automl." arXiv preprint arXiv:2410.02958 (2024).
|Supervisor=Jens Lundström, Peyman Mashadi, Stefan Byttner
|Level=Master
|Status=Open
}}
With the emergence of LLM-based Agentic AI [1], language model technology is being used not just as passive tools for generating text, but as a part of autonomous agents capable of reasoning, decision-making, and interacting with environments or other systems. The generation of machine learning pipelines is no exception in the scope of applications covered by Agentic AI [2]. However, with the adoption of legal frameworks related to AI, e.g. EU AI Act it is of importance to develop algorithmic solutions for generation of ML-pipelines that are trustworthy and aligns with the requirements of associated legally-defined risk-levels. To iteratively align with such requirements agents need to consider trustworthy aspects already in the phase of service design. This MSc-thesis explores knowledge-creation in the realm of Agentic AI systems for trustworthy ML-pipelines starting from service canvas and centers around human-in-the-loop and agent specifications. Currently, it is unclear how the degree of autonomy, architectural complexity and level of reasoning affects the performance and trustworthiness of generated ML-pipelines. This thesis will also focus on agents creating their own synthetic data for reasoning and decision-making and tools for augmented service canvas creation. The thesis work will be carried out as two interdisciplinary projects with students from other disciplines.

Timeseries XAI in Cybersecurity and Industry

2025-10-22T22:45:15Z

Islab: Created page with "{{StudentProjectTemplate |Summary=Cybersecurity data analysis with XAI |Keywords=Cybersecurity, XAI, timeseries |TimeFrame=Spring 2026 |References=https://www.sciencedirect.co..."

{{StudentProjectTemplate
|Summary=Cybersecurity data analysis with XAI
|Keywords=Cybersecurity, XAI, timeseries
|TimeFrame=Spring 2026
|References=https://www.sciencedirect.com/science/article/pii/S1566253523001148
https://www.nature.com/articles/s41597-025-04603-x
|Supervisor=Grzegorz J. Nalepa, Prayag Tawari, Aurora Esteban
|Level=Master
|Status=Open
}}
Cybersecurity systems have clearly became a very important area of AI applications recently.
From the engineering perspective they produce a large amount of data that can only be analyzed by AI methods.
Thus in this thesis we would like to consider AI-based timeseries analysis with applications and data from cybersecurity systems.
Some of the general questions are
1) what makes this specific domain distinct wrt to data and its characteristics? e.g. rapid changes, anomalies, drifts, data imbalance?
2) what types of explanations could be most useful for cybersec experts, e.g. counterfactuals, prototypes

It is encouraged that this thesis will result in scientific publications possibly also developed in collaboration with external stakeholders.

Knowledge graphs 4 XAI in Healthcare

2025-10-22T22:42:07Z

Islab: Created page with "{{StudentProjectTemplate |Summary=Knowledge graphs in Healthcare |Keywords=Knowledge graphs, XAI, Healthcare |TimeFrame=Spring 2026 |References=https://www.sciencedirect.com/s..."

{{StudentProjectTemplate
|Summary=Knowledge graphs in Healthcare
|Keywords=Knowledge graphs, XAI, Healthcare
|TimeFrame=Spring 2026
|References=https://www.sciencedirect.com/science/article/pii/S1566253523001148
https://www.sciencedirect.com/science/article/abs/pii/S1532046425000905
https://www.sciencedirect.com/science/article/pii/S0004370221001788
https://www.sciencedirect.com/science/article/pii/S1532046423001247
https://arxiv.org/abs/2402.12608
|Supervisor=Grzegorz J. Nalepa, Farzaneh Etminani
|Level=Master
|Status=Open
}}
Knowledge graphs (KG) are a knowledge representation method that has became increasingly important in the context of the explainable AI approach.
Healthcare (HC) systems are a very special area of application of AI due to data sensitivity andf privacy concerns.
In this thesis we would like to conduct a survey on a recent practical applications of KG i HC for knowledge representation and management.
Furthermore, we would like to explore the role of KG in interoperability and FAIRness of HC data.
Moreover, we would like to consider KG for supporting XAI for HC, specifically what explanation forms are mostly suitable for HC experts and how they can be supported by background knowledge
Finally, we want to explore knowledge augmented retrieval with KG for HC data and knowledge bases where explanations can meet medical data.

It is encouraged that this thesis will result in scientific publications possibly also developed in collaboration with external stakeholders.

Machine Unlearning Faithfulness with XAI

2025-10-22T22:40:13Z

Islab:

{{StudentProjectTemplate
|Summary=Evaluating Machine Unlearning Faithfulness in Deep Neural Networks using Explainable AI
|Keywords=ML, XAI, forget!
|TimeFrame=Spring 2026
|References=Beierle, Christoph, and Ingo J. Timm. "Intentional forgetting: An emerging field in AI and beyond." KI-Künstliche Intelligenz 33.1 (2019): 5-8.
Bourtoule, Lucas, et al. "Machine unlearning." 2021 IEEE symposium on security and privacy (SP). IEEE, 2021.
Cadet, Xavier F., et al. "Deep Unlearn: Benchmarking Machine Unlearning for Image Classification." 2025 IEEE 10th European Symposium on Security and Privacy (EuroS&P). IEEE, 2025.
Golatkar, Aditya, Alessandro Achille, and Stefano Soatto. "Eternal sunshine of the spotless net: Selective forgetting in deep networks." Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020.
Vidal, Àlex Pujol, et al. "Verifying machine unlearning with explainable AI." International Conference on Pattern Recognition. Cham: Springer Nature Switzerland, 2024

|Supervisor=Grzegorz J. Nalepa, Peyman Mashhadi
|Level=Master
|Status=Open
}}
As deep learning models become larger and their training size grows, the concept of machine unlearning (ML) [Beierle et al.] is becoming more relevant. Machine unlearning refers to the concept of forgetting information about some data that is not desired to be in the training data and consequently in the trained model due to many different reasons including privacy, stale knowledge, copyrighted material, toxic/unsafe content, and so on.

While the ultimate goal of an unlearned model is to achieve performance indistinguishable from a model retrained from scratch on the retained data (target model), evaluating this indistinguishability rigorously remains difficult [Golatkar et al.]. Current evaluation methods primarily focus on utility (accuracy retention) and privacy against Membership Inference Attacks (MIA) [Cadet et al.].

This thesis proposes to take an instance beyond traditional accuracy and privacy metrics by leveraging Explainable AI (XAI) tools to quantify the structural and functional similarity (or "faithfulness") between the unlearned model and the target model [Beierle et al., Vidal et al].

The thesis has both conceptual and practical motivations. Due to privacy considerations or legal requirements (e.g., the GDPR) business stakeholders might in fact soon require the providers and developers of ML models to effectively remove their data samples from the training datasets. The developers in turn will be interested in understanding the impact of the data removal on the models, as well as finding ways to compensate, e.g., supplement additional data. Finally, societally “the right to be forgotten” is becoming increasingly relevant for the new generations, digital ethics and the so-called Delete Culture.

It is encouraged that this thesis will result in scientific publications possibly also developed in collaboration with external stakeholders.

Machine Unlearning Faithfulness with XAI

2025-10-22T22:36:19Z

Islab: Created page with "{{StudentProjectTemplate |Summary=Evaluating Machine Unlearning Faithfulness in Deep Neural Networks using Explainable AI |Keywords=ML, XAI, forget! |TimeFrame=Spring 2026 |Re..."

{{StudentProjectTemplate
|Summary=Evaluating Machine Unlearning Faithfulness in Deep Neural Networks using Explainable AI
|Keywords=ML, XAI, forget!
|TimeFrame=Spring 2026
|References=Beierle, Christoph, and Ingo J. Timm. "Intentional forgetting: An emerging field in AI and beyond." KI-Künstliche Intelligenz 33.1 (2019): 5-8.
Bourtoule, Lucas, et al. "Machine unlearning." 2021 IEEE symposium on security and privacy (SP). IEEE, 2021.
Cadet, Xavier F., et al. "Deep Unlearn: Benchmarking Machine Unlearning for Image Classification." 2025 IEEE 10th European Symposium on Security and Privacy (EuroS&P). IEEE, 2025.
Golatkar, Aditya, Alessandro Achille, and Stefano Soatto. "Eternal sunshine of the spotless net: Selective forgetting in deep networks." Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020.
Vidal, Àlex Pujol, et al. "Verifying machine unlearning with explainable AI." International Conference on Pattern Recognition. Cham: Springer Nature Switzerland, 2024

|Supervisor=Grzegorz J. Nalepa and Peyman Mashhadi
|Level=Master
|Status=Open
}}
As deep learning models become larger and their training size grows, the concept of machine unlearning (ML) [Beierle et al.] is becoming more relevant. Machine unlearning refers to the concept of forgetting information about some data that is not desired to be in the training data and consequently in the trained model due to many different reasons including privacy, stale knowledge, copyrighted material, toxic/unsafe content, and so on.

While the ultimate goal of an unlearned model is to achieve performance indistinguishable from a model retrained from scratch on the retained data (target model), evaluating this indistinguishability rigorously remains difficult [Golatkar et al.]. Current evaluation methods primarily focus on utility (accuracy retention) and privacy against Membership Inference Attacks (MIA) [Cadet et al.].

This thesis proposes to take an instance beyond traditional accuracy and privacy metrics by leveraging Explainable AI (XAI) tools to quantify the structural and functional similarity (or "faithfulness") between the unlearned model and the target model [Beierle et al., Vidal et al].

The thesis has both conceptual and practical motivations. Due to privacy considerations or legal requirements (e.g., the GDPR) business stakeholders might in fact soon require the providers and developers of ML models to effectively remove their data samples from the training datasets. The developers in turn will be interested in understanding the impact of the data removal on the models, as well as finding ways to compensate, e.g., supplement additional data. Finally, societally “the right to be forgotten” is becoming increasingly relevant for the new generations, digital ethics and the so-called Delete Culture.

It is encouraged that this thesis will result in scientific publications possibly also developed in collaboration with external stakeholders.

Domain Adaptation for Survival Analysis

2025-10-20T14:14:30Z

Islab: Developing robust domain adaptation methods for survival analysis

{{StudentProjectTemplate
|Summary=Developing robust domain adaptation methods for survival analysis
|Keywords=Domain Adaptation, Survival Analysis
|Supervisor=Abdallah Alabdallah, Zahra Taghiyarrenani
|Level=Master
|Status=Open
}}
This thesis explores domain adaptation techniques to enhance the robustness and generalizability of survival models across diverse clinical datasets. The core objective is to develop novel methodologies that learn domain-invariant feature representations, enabling accurate time-to-event predictions in a target domain distinct from the source. A significant focus will be to investigate how specific characteristics of survival data, particularly the presence and distribution of censored observations, impact transferability and pose unique challenges for adaptation algorithms.

Design and Evaluation of an LLM-Based Travel Planner with Dynamic Event and Accommodation Data

2025-10-09T13:22:17Z

Islab:

{{StudentProjectTemplate
|Summary=Design and Evaluation of an LLM-Based Travel Planner with Dynamic Event and Accommodation Data
|TimeFrame=2025-26
|References=1. TravelAgent: An AI Assistant for Personalized Travel Planning by Aili Chen

2. TP-RAG: Benchmarking Retrieval-Augmented Large Language Model Agents for Spatiotemporal-Aware Travel Planning by Hang Ni
|Supervisor=Adeel Zafar, Nuwan Gunasekara
|Level=Master
|Status=Open
}}
This project proposes to develop an AI-powered travel planning app integrating real-time data sources for events and accommodation with a hybrid system using both a small domain-specific LLM and a general-purpose LLM; the focus is on evaluating the trade-offs between latency and efficacy of these LLMs in generating personalised, rational, and comprehensive itineraries. The evaluation will use metrics including itinerary rationality, factual accuracy, coverage, and user satisfaction from human and simulated user studies, guided by state-of-the-art frameworks from recent AI travel systems research. This work aims to provide insights into optimising LLM architectures and evaluation protocols for dynamic, real-time travel recommendation applications.

Scalable Standardized Information Extraction from Free-Text Electronic Health Records

2025-10-06T12:04:48Z

Islab: Created page with "{{StudentProjectTemplate |Summary=Extracting clinical information from free-text notes and providing standardized mapping of extracted concepts using the SNOMED CT medical ont..."

{{StudentProjectTemplate
|Summary=Extracting clinical information from free-text notes and providing standardized mapping of extracted concepts using the SNOMED CT medical ontology.
|TimeFrame=Fall 2025 and Spring 2026
|References=[1] Richter-Pechanski P, Wiesenbach P, Schwab DM, Kiriakou C, Geis N, Dieterich C, Frank A. Clinical information extraction for lower-resource languages and domains with few-shot learning using pretrained language models and prompting. Natural Language Processing. 2025 Sep;31(5):1210-33.
[2] Gu B, Shao V, Liao Z, Carducci V, Brufau SR, Yang J, Desai RJ. Scalable information extraction from free text electronic health records using large language models. BMC Medical Research Methodology. 2025 Jan 28;25(1):23.
[3] Chang E, Sung S. Use of SNOMED CT in Large Language Models: Scoping Review. JMIR Medical Informatics. 2024 Oct 7;12(1):e62924.
[4] Noori A, Devkota P, Mohanty S, Manda P. Automated SNOMED CT Concept Annotation in Clinical Text Using Bi-GRU Neural Networks. arXiv preprint arXiv:2508.02556. 2025 Aug 4.

|Supervisor=Amira Soliman and Mohammad Mansoori
|Level=Master
|Status=Open
}}
A considerable proportion of clinical information remains embedded within unstructured textual data, encompassing measured vital signs, lifestyle factors such as smoking status, alcohol consumption, and substance use, as well as family histories of comorbidities and other health-related observations. The automated extraction of such information poses multiple challenges: the substantial costs associated with clinical expertise for data annotation and model evaluation; limited computational resources for on-premises processing owing to data confidentiality requirements; and the stringent privacy regulations that govern access to sensitive patient information. Recent advances in natural language processing (NLP) and machine learning (ML) have shown great potential in rendering this unstructured data more accessible and usable for both clinical and research applications.

The research questions can be summarized as:
1. How do fine-tuned domain-specific LLMs compare with hybrid rule-based approaches in extracting clinical concepts from unstructured text?
2. What strategies for prompt engineering or model fine-tuning yield the best trade-off between accuracy and computational efficiency?
3. How can extracted clinical concepts be effectively mapped and standardized to established ontologies such as SNOMED CT to ensure interoperability and consistency across EHR systems?

Design and Evaluation of an LLM-Based Travel Planner with Dynamic Event and Accommodation Data

2025-10-03T11:17:56Z

Islab: Created page with "{{StudentProjectTemplate |Summary=Design and Evaluation of an LLM-Based Travel Planner with Dynamic Event and Accommodation Data |TimeFrame=Fall 2025 |References=1. TravelAgen..."

{{StudentProjectTemplate
|Summary=Design and Evaluation of an LLM-Based Travel Planner with Dynamic Event and Accommodation Data
|TimeFrame=Fall 2025
|References=1. TravelAgent: An AI Assistant for Personalized Travel Planning by Aili Chen

2. TP-RAG: Benchmarking Retrieval-Augmented Large Language Model Agents for Spatiotemporal-Aware Travel Planning by Hang Ni
|Supervisor=Adeel Zafar, Nuwan Gunasekara
}}
This project proposes to develop an AI-powered travel planning app integrating real-time data sources for events and accommodation with a hybrid system using both a small domain-specific LLM and a general-purpose LLM; the focus is on evaluating the trade-offs between latency and efficacy of these LLMs in generating personalised, rational, and comprehensive itineraries. The evaluation will use metrics including itinerary rationality, factual accuracy, coverage, and user satisfaction from human and simulated user studies, guided by state-of-the-art frameworks from recent AI travel systems research. This work aims to provide insights into optimising LLM architectures and evaluation protocols for dynamic, real-time travel recommendation applications.

Evaluating Privacy Leakage Attacks on Fine-Tuned Clinical Models with Synthetic Data

2025-09-24T08:43:16Z

Islab: Created page with "{{StudentProjectTemplate |Summary=Generate synthetic clinical data to fine-tune models and systematically evaluate privacy leakage risks using advanced attack techniques. |Tim..."

{{StudentProjectTemplate
|Summary=Generate synthetic clinical data to fine-tune models and systematically evaluate privacy leakage risks using advanced attack techniques.
|TimeFrame=2025-26
|Supervisor=Adeel Zafar
|Level=Master
|Status=Open
}}
This software-focused project investigates how synthetic clinical data can be used to fine-tune language models while minimizing privacy risks. The central research question is: To what extent do models fine-tuned on synthetic health records leak sensitive information when exposed to advanced attack techniques? The work is organized into four parts: (1) Generate high-fidelity synthetic datasets that mimic real clinical records, (2) Fine-tune transformer models on these datasets, (3) Evaluate privacy leakage using state-of-the-art adversarial attacks such as membership inference, and (4) Analyze results to develop mitigation strategies. Deliverables include open-source synthetic datasets, fine-tuned models, privacy risk benchmark reports, and recommendations for safe model deployment.

Analyzing Gender Bias in Pose Estimation Models

2025-09-23T14:19:47Z

Islab: Created page with "{{StudentProjectTemplate |Summary=We will analyze the gender bias of current pose estimation models when trained with unbalanced gender data |Supervisor=Kevin Hernandez Diaz |..."

{{StudentProjectTemplate
|Summary=We will analyze the gender bias of current pose estimation models when trained with unbalanced gender data
|Supervisor=Kevin Hernandez Diaz
|Level=Bachelor
|Status=Open
}}
The idea is to use out-of-the-box pose estimation models and train them from scratch with different levels of gender representations in the data.
Then we will analyze the deviation from the ground truth based on gender.
Finally, we will analyze some basic statistics on the estimated pseudoesqueleton based on gender.

Blood splatter analysis

2025-09-01T13:43:10Z

Islab: Created page with "{{StudentProjectTemplate |Summary=Estimation of direction and distance to origin from blood splatter images |Keywords=computer vision, deep learning, image analysis |TimeFrame..."

{{StudentProjectTemplate
|Summary=Estimation of direction and distance to origin from blood splatter images
|Keywords=computer vision, deep learning, image analysis
|TimeFrame=Fall 2025-Spring 2026
|Prerequisites=Have done or doing the Image Analysis, Computer Vision in 3D and Deep Learning courses.
|Supervisor=Kevin Hernandez-Diaz, Fernando Alonso-Fernandez
|Level=Master
|Status=Open
}}
The project will use a database of blood splatter to estimate the distance to the source and the direction from which it came from using deep learning. The project will focus on data augmentation and simulation when multiple sources may be present in the pattern, and on recognizing and classifying according to the estimated origin.

example of expected work packages on this project:
- Data handling, combination and simulation of multiple sources of blood splatter in a single super-resolution image.
- Training deep learning models for regression of coordinates of origin and compared with previous proposed methods in the literature.
- Classification/clustering analysis of potential number of sources.

Prototype Aligned Embedding for Time Series Forecasting

2025-02-10T17:41:39Z

Islab:

{{StudentProjectTemplate
|Summary=Generate prototype based explanation for time series forecasting
|TimeFrame=Spring 2025
|References=[1] Wenxiang Li and K. L. Eddie Law. Deep learning models for time series forecasting: A review.
IEEE Access, 12:92306–92327, 2024.

[2] Antonios Mamalakis, Elizabeth A Barnes, and Imme Ebert-Uphoff. Investigating the fidelity
of explainable artificial intelligence methods for applications of convolutional neural networks in
geoscience. Artificial Intelligence for the Earth Systems, 1(4):e220012, 2022.

[3] Orfeas Menis Mastromichalakis, Giorgos Filandrianos, Jason Liartis, Edmund Dervakos, and Giorgos
Stamou. Semantic prototypes: Enhancing transparency without black boxes. In Proceedings
of the 33rd ACM International Conference on Information and Knowledge Management, pages
1680–1688, 2024.

[4] Yao Ming, Panpan Xu, Huamin Qu, and Liu Ren. Interpretable and steerable sequence learning
via prototypes. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge
Discovery & Data Mining, pages 903–913, 2019.

[5] Anushka Narayanan and Karianne J Bergen. Prototype-based methods in explainable ai and
emerging opportunities in the geosciences. arXiv preprint arXiv:2410.19856, 2024.

[6] G Pradeep Reddy and YV Pavan Kumar. Explainable ai (xai): Explained. In 2023 IEEE Open
Conference of Electrical, Electronic and Information Sciences (eStream), pages 1–6. IEEE, 2023.
5

[7] Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. ” why should i trust you?” explaining the
predictions of any classifier. In Proceedings of the 22nd ACM SIGKDD international conference
on knowledge discovery and data mining, pages 1135–1144, 2016.

[8] A Saranya and R Subhashini. A systematic review of explainable artificial intelligence models and
applications: Recent developments and future trends. Decision analytics journal, 7:100230, 2023.

[9] Chenxi Sun, Hongyan Li, Yaliang Li, and Shenda Hong. Test: Text prototype aligned embedding
to activate llm’s ability for time series. arXiv preprint arXiv:2308.08241, 2023.

[10] Yuyi Zhang, Qiushi Sun, Dongfang Qi, Jing Liu, Ruimin Ma, and Ovanes Petrosian. Shaptime:
A general xai approach for explainable time series forecasting. In Proceedings of SAI Intelligent
Systems Conference, pages 659–673. Springer, 2023.

[11] Alexandra Zytek, Sara Pid`o, and Kalyan Veeramachaneni. Llms for xai: Future directions for
explaining explanations. arXiv preprint arXiv:2405.06064, 2024.
|Supervisor=Nuwan Gunasekara, Parisa Jamshidi, Yuantao Fan
|Level=Master
|Status=Internal Draft
}}
This project will investigate, and develop a framework that integrates prototype-based reasoning with LLM-generated explanations for improving time-series forecasting. The prototype explanations generated via the proposed approach will be evaluated for its consistency and validity.

Prototype Aligned Embedding for Time Series Forecasting

2025-02-10T17:37:27Z

Islab: Created page with "{{StudentProjectTemplate |Summary=Generate prototype based explanation for time series forecasting |TimeFrame=Spring 2025 |Supervisor=Nuwan Gunasekara, Parisa Jamshidi, Yuanta..."

{{StudentProjectTemplate
|Summary=Generate prototype based explanation for time series forecasting
|TimeFrame=Spring 2025
|Supervisor=Nuwan Gunasekara, Parisa Jamshidi, Yuantao Fan
|Level=Master
|Status=Internal Draft
}}
This project will investigate, and develop a framework that integrates prototype-based reasoning with LLM-generated explanations for improving time-series forecasting. The prototype explanations generated via the proposed approach will be evaluated for its consistency and validity.