Master Thesis A Future AI Agent for Vehicle Architecture Integration

Background

Vehicle integration is one of the most critical phases in truck development process. There
are hundreds of mechanical parts and systems, for example, frame, chassis, power and
energy/power supply, thermal management, and cab, they must work together, meet
performance goals, be easy to build and repair, and be ready for mass production. The
main responsibility for vehicle integration is to combine all the parts and systems
harmoniously. In order to successfully finish this job, it requires all kinds of knowledge
and expert domain knowledge. However, today, knowledge is scattered across drawings,
reports, plans, and the minds of many experts. Choices are made in meetings; late
changes spread slowly; and issues like poor fitness or hard-to-reach maintenance areas
can appear late. An intelligent agent powered by AI that captures needs, defines clear
connections between parts, explores practical layouts, and explains its choices could
assist design, speed development, cut risk, and save knowledge for future models.

Object of the thesis

This thesis targets vehicle architecture integration across mechanical and hardware
domains, including cab, chassis, electrical system, energy storage system, thermal
management, etc., where the real challenge is turning many stakeholders’ requirements
for function, performance, assembly, tolerances, and serviceability into a launch‑ready
product. The objective is to create and evaluate an AI Integration Agent that spans
concept to production readiness: it will translate stakeholder intent into computable
constraints on components, interfaces, assembly sequences, and tolerance chains;
generate and screen manufacturable layouts; check design‑for‑manufacture/assembly
(DFM/DFA), serviceability, and performance envelopes; and explain trade‑offs and
change‑impact in a form that engineers can audit and trust.

You will create the framework of AI Vehicle Integration Agent by defining the vehicle
integration metamodel covering components, interfaces, tolerances, assembly stations,
and service access; map stakeholder briefs and engineering data into explicit constraints
and objectives; implement a closed loop, from generate, simulate/estimate, check to
explain, to evaluate alternatives; validate the agent on case studies; deliver a reusable
prototype, anonymized/synthetic datasets, experimental results with ablations, and a
fully reproducible package.

This is hands‑on, high‑impact project. You will fuse design, engineering, manufacturing
knowledge, and LLM into data‑driven optimization that can support in the earlier product
development process. Expect close interaction with manufacturing, quality, and service
experts; opportunities to publish on design representation, surrogate modeling, and
multi‑objective exploration; and the satisfaction of turning engineering judgment into a
tool that improves launch outcomes.

Candidate profile

Having one or more of core skills:

• Strong in programing (python or similar)
• CAD modeling (Catia)
• Have good knowledge of Neural Networks, e.g., Large Language Model (LLM)
  and Convolutional Neural Network (CNN)
• Bonus points if you have experience developing or deploying AI agents

The thesis project is intended for 2 students.
Location: Göteborg, Sweden

Time Schedule: Jan 2026- Jun 2026

Contact persons:

Manager: Mikael Rydh - Volvo GTT
tel: +46 739028419
mail: mikael.rydh@volvo.com

Supervisor: Wenting Hu - Volvo GTT
tel: +46 765534808
mail: wenting.hu@volvo.com