Resource Search in Spatial Environments

Modern assistance systems and autonomous vehicles require flexible and dynamic control mechanisms to exploit the variety of information provided by Smart Cities. Therefore, we develop novel planing algorithm based on stochastic optimization and reinforcement learning.

Training Robots with Reinforcement Learning

Training robots with neural networks requires large amounts of samples which are usually generated by simulations. However, transferring a policy trained in an virtual environments bears risks when transferring the trained model to physical systems.