Workshop Automation together with Volvo Group

Title	Workshop Automation together with Volvo Group
Summary	This project, in collaboration with Volvo Group, investigates how automation can raise workshop throughput, repair quality, and technician experience through data-driven perception and pragmatic use of automation.
Keywords
TimeFrame	Fall 2025
References
Prerequisites	Good knowledge of machine learning & robotics
Author
Supervisor	Sławomir Nowaczyk, Saeed Gholami Shahbandi
Level	Master
Status	Open

Next-generation Orchestrated Workshop Automation (NOWA)

Introduction

Next-generation Orchestrated Workshop Automation (NOWA) is a pre-study where Volvo and the CAISR group at Halmstad University investigate how automation can raise workshop throughput, repair quality and technician experience through data-driven perception and pragmatic use of automation. We will demonstrate five focused showcases that create immediate value while de-risking future robotics: (1) camera-based visual checks for component inspection; (2) an LLM-supported check-in that structures driver symptom descriptions ; (3) engine idle-sound anomaly pre-screening; (4) a mobile parts & tool runner to reduce technician “walking waste”; and (5) robotic support for oil change operation. Each pilot will be co-designed with technicians, instrumented with clear KPIs, and packaged with site-agnostic Standard Operating Procedures—forming a scalable pathway from today’s assisted workflows to tomorrow’s robot-enabled workshop automation.

Since the five showcases are at different levels of maturity, the expected results will also vary, from the practical demonstration of the quantifiable benefits of robot-guided cameras over existing manual procedures for (1) to requirements specification and initial cost-benefit analysis for (5). Overall, however, NOWA directly advances the call’s focus on productivity, sustainability, and human-centred digitalisation in the heavy-vehicle aftermarket.

This pre-study couples rigorous analysis of existing multimodal data (images and idle-sound) with pragmatic automation showcases (LLM-assisted check-in and robotic tool/parts runner) to deliver near-term, measurable value while laying a scalable foundation for future robotic inspection.

Prospective thesis project topics

PLEASE NOTE: project assignments will be finalized based on the number of students/groups, Volvo priorities, and NOWA planning. You are not guaranteed an exact 1-to-1 match with any single showcase; your thesis may combine elements across the four themes. We will aim to align your project with your interests while meeting project constraints.

Visual Checks: From Ad-Hoc Photos to Diagnostic Visual Protocols

Objective: Define site-agnostic image-capture and quality-assessment protocols that turn workshop photos into diagnostically useful evidence for routine visual checks, without increasing technician burden. The protocol is a step toward workshop automation where future robot systems perform consistent capture; the thesis can explore what requirements and constraints this implies for robot-mounted cameras and paths.

Scope: Audit existing image data, specify minimal quality/completeness metrics (e.g., sharpness, exposure, key-region visibility), and draft capture Standard operating procedures (SOP) for both human-held and robot-mounted cameras (manipulator or mobile base), validating via light bench tests or synthetic studies rather than full implementation.

• Deliveries:
  • KPI and metric definitions tied to "diagnostically useful" images.
  • Data audit plan and curated exemplars for target components.
  • Draft Standard operating procedures (SOP) for capture and review checklists; risk & privacy note.
  • Baseline modeling/evaluation plan (ranking/filtering) with simulated ablations to link quality factors to expected diagnostic value.
  • Go/hold decision criteria, Technology Readiness Level (TRL) and Return on Investment (ROI) roadmap outline.

• Research Questions:
  • Which minimal, site-agnostic image-quality and completeness metrics best predict diagnostic usefulness for routine visual checks?
  • Which capture protocol elements (angles, lighting, standoff) most affect those metrics in workshops?

LLM-Assisted Check-In and Handover: Structure, Safety, and UX

Objective: Design and evaluate a lightweight, human-in-the-loop LLM agent that structures symptom capture, supports planning/verification of required resets, and generates technician-ready handovers—improving reception flow while safeguarding inclusivity and transparency.

Scope: Service blueprinting, Wizard-of-Oz trials, prompt/flow design with guardrails, and inclusive language checks; define structured fields and interoperability stubs without building full back-end integrations. Evaluate with time–motion baselines and scenario tests.

• Deliveries:
  • Reception blueprint, field schema, and UX wireframes.
  • Prompt/flow library with safety guardrails and fallback rules.
  • Evaluation plan for completeness and dwell-time effects; risk & data-protection note.
  • Standard operating procedures (SOP) excerpts for reception/hand-over; indicative Technology Readiness Level (TRL) and Return on Investment (ROI) pathway.

• Research Question(s):
  • What minimal set of structured prompts and agent behaviors achieves high symptom-capture completeness while maintaining or reducing driver dwell time at reception?
  • How should explainability and escalation be designed so technicians trust the agent’s outputs in a noisy, multi-lingual environment?

Idle Sound & Vibration Pre-Screening: Signals for Early Triage

Objective: Specify a generic protocol and analysis plan for using idle-engine sound and basic vibration sensing to pre-screen for anomalies, focusing on robust capture, feature extraction, and triage value rather than model optimization.

Scope: Sensor/capture guidelines (microphone/accelerometer placement, duration, environment), privacy and safety assessment, baseline feature bank (time/frequency), and a simulation plan showing how pre-screening could influence triage and Mean Time to Repair (MTTR) under conservative assumptions. Bench tests preferred over field deployments at this stage. If field access is limited, prioritize alternative data sources: controlled bench recordings on campus rigs or donor vehicles, short capture campaigns with university fleet or partner garages, augmentation and synthesis for baseline method validation, and selective use of public datasets where task-aligned.

• Deliveries:
  • Capture Standard operating procedures (SOP); idle conditions, noise controls, metadata.
  • Feature-extraction and labeling plan; small controlled test design.
  • Discrete-event "what-if" simulation spec connecting alerts to workflow outcomes; risk register.
  • Criteria for go/hold and Technology Readiness Level (TRL) and Return on Investment (ROI) progression.

• Research Question(s):
  • Which capture and feature combinations are likely to yield robust anomaly pre-screening across heterogeneous workshop acoustics?
  • What is the expected effect of pre-screening alerts on triage queues and Mean Time to Repair (MTTR) in simulated reception workflows?

Mobile Parts & Tool Runner: Logistics Blueprint and Impact Estimation

Objective: Develop a minimal, technology-agnostic blueprint for a mobile parts/tool runner (human-operated cart today, robot-ready tomorrow) that reduces technician "walking waste" and time-to-first-wrench, including dispatch rules, location strategies, and safety/ergonomics considerations.

Scope: Process maps, layout assumptions, pick/pack/dispatch logic, and discrete-event simulations to estimate effects on utilization and rework; evaluate feasibility and integration constraints without building hardware. Co-design with technicians to ensure acceptance. Primary evaluation will be via discrete-event simulation; optionally validate selected micro-flows on available platforms (e.g., Baxter for handoff and staging tasks, Robotnik mobile base for point-to-point parts delivery) to de-risk assumptions about dispatch latency and human-robot interaction.

• Deliveries:
  • Service blueprint and SOP draft (request -> pick -> deliver -> confirm).
  • Simulation plan with baseline time–motion data and KPI definitions.
  • Preliminary cost–benefit model and Technology Readiness Level (TRL) and Return on Investment (ROI) roadmap; change-management notes.
  • Safety/ergonomics and equality considerations for roles and workflows.

• Research Question(s):
  • Which dispatch and localization strategies (e.g., zone-based, on-demand batching) minimize technician walking time without disrupting safety and flow?
  • Under conservative assumptions, what reduction in time-to-first-wrench is achievable in simulation for common job families?