Design and implementation of an RL agent to optimize the characterization of electric motors

Context:

With a view to the transformation towards a CO2-neutral society, the electrification of drive systems is of key importance. Electric motors (e-motors) are key components in this context, and their effective and efficient operation requires precise measurement of key parameters, known as characterization. Due to the large number of relevant operating parameters and the high level of detail required for measurement, this currently requires lengthy and cost-intensive work on the test bench.

Reinforcement learning (RL) enables the development of reactive agents that support and optimize the characterization of electric motors. With the help of simulation environments and historical measurement data, the agents are able to learn and abstract complex dependencies of electric motor behavior and integrate them into dynamic measurement strategies. This results in optimized measurement processes in which an automated and efficient refinement of the resulting maps takes place, while adverse system effects, such as temperature effects, are dynamically avoided or compensated for.

The work is being carried out in cooperation with weg//weiser GmbH, a technology start-up that offers advanced testing and analysis services for electric motors and aims to expand these in a targeted manner through the integration of AI. As part of the collaboration, there is the opportunity to optimize and evaluate the developed solution in practice on a real test bench under industry-like conditions.

Objectives:

• Literature research on the current state of the art in the field of RL agents for e-motor control and parameterization strategies
• Preparation of the database and needs-based adaptation of the simulation environment
• Design and implementation of an agent to optimize measurement processes
• Integration and training of an agent in an RL training environment
• Comparison and (possibly practical) evaluation of the developed method

Requirements:

• Independent and solution-oriented way of working
• Programming skills in Python
• Knowledge in the field of machine learning, ideally reinforcement learning
• Good written and spoken German and English skills
• Ideally knowledge of the basic functioning of electric motors