Master Thesis - Imitation Learning for Industrial Robotics

30 credits Master Thesis – Smart Factory Lab Imitation Learning for Industrial Robots - Development and Evaluation of a Robotic Test Setup Background Applying the latest AI methods to robotics enables the automation of tasks that are currently difficult or impossible to automate. One promising approach is imitation learning, where robots learn complex tasks from human demonstrations instead of being programmed entirely by hand. Scania has already developed an initial proof-of-concept setup together with an external partner. The infrastructure, including an industrial robot, gripper, cameras, computers, and motion-tracking systems, will be provided. The focus of this thesis is on understanding how to effectively train and evaluate robots using imitation learning to support further automation in Scania’s production. The Smart Factory Lab at Scania is an experimental test environment that explores, assesses, and pilots new technologies before they are adopted in Scania’s production processes. We are looking for motivated and creative students to join our Smart Factory Lab for their Master thesis. You will have the opportunity to work in a dynamic and collaborative environment, where you can test and validate new technologies, methods, and solutions in close collaboration with our internal and external stakeholders. To learn more about the Smart Factory, visit Link. Assignment The objective is to investigate how different training and data-collection factors influence the performance, robustness, and generalisation of imitation-learned robot policies across relevant industrial use cases. Using the existing robotic test setup, the students will select one or more relevant robot tasks and conduct structured experiments. The work may include: Collecting and preparing demonstration data. Investigating the required amount and variation of demonstrations. Evaluating policy performance on different physical setups and environmental conditions. Documenting the results and developing recommendations for future applications. Background and Time Plan Experience with Python, ROS, robot programming, CAD modelling, computer vision, or machine-learning frameworks, is beneficial but not necessarily required. Students in Robotics or Engineering, who bring some background in AI training. Computer Science students who are eager to build physical prototypes. Number of students: 1-2 Start date: January 2027 (20 weeks) Location: Södertälje, Sweden (at least 60% on-site)