MSc. presentation by C. Alptürk: Risk Averse Path Planning Using Approximated Wassertein Distributionally Robust Deep Q-learning
Place: Seminar Room KC 3N27
Contact: venkatraman [dot] renganathan [at] control [dot] lth [dot] se
Save event to your calendar
Cem Alptürk is defending his Master's thesis at the Dept. of Automatic Control.
Where: Seminar Room KC:3N27
When: June 10th, 14:00-15:00
Author: Cem Alptürk
Title: Risk Averse Path Planning Using Approximated Wassertein Distributionally Robust Deep Q-learning
Advisors: Venkatraman Renganathan
Examiner: Anders Rantzer
We investigate the problem of risk averse robot path planning using the deep reinforcement learning perspective. Our problem formulation involves modeling the robot as a stochastic linear dynamical system and cast the risk averse motion planning problem as a Markov decision process. Specifically, we propose a continuous reward function design that explicitly takes into account the risk of collision with obstacles while encouraging the robot's motion towards the goal. Assuming that a collection of process noise samples is available, we safely learn the risk-averse robot control actions through Lipschitz approximated Wasserstein distributionally robust Q-learning to hedge against the noise uncertainty. The learned control actions result in a safe and risk averse trajectory from the source to the goal avoiding all the obstacles. Various supporting numerical simulations are presented to demonstrate our proposed approach.