Reinforcement Learning

Global Path Planning in a Digital-Twin

MSc thesis at RSL Lab, supervised by Marco Bjelonic et al. I designed a global-path planning algorithm that can learn navigation costs for a specific robot policy in a digital twin of the environment and that can be transferred to the real world. This allowed robot to plan paths that are 50-100m far reliably.

Learning to calibrate battery models using Deep Reinforcement Learning

Learning to calibrate battery models using lyapunov constrained actor-critic RL. Intial results has been presented at NeurIPS 2020 workshop on ML4Eng

Battery Model Calibration with Deep Reinforcement Learning

__NeurIPS'20 Workshop \& Energies Journal__