Publication Details

SELECT * FROM publications WHERE Record_Number=11245
Reference TypeConference Proceedings
Author(s)Celik, O.; Abdulsamad, H.; Peters, J.
TitleChance-Constrained Trajectory Optimization for Nonlinear Systems with Unknown Stochastic Dynamics
Journal/Conference/Book TitleProceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)
AbstractIterative trajectory optimization techniques for non-linear dynamical systems are among the most powerful and sample-efficient methods of model-based reinforcement learning and approximate optimal control. By leveraging time-variant local linear-quadratic approximations of system dynamics and rewards, such methods are able to find both a target-optimal trajectory and time-variant optimal feedback controllers. How- ever, the local linear-quadratic approximations are a major source of optimization bias that leads to catastrophic greedy updates, raising the issue of proper regularization. Moreover, the approximate models’ disregard for any physical state-action limits of the system, causes further aggravation of the problem, as the optimization moves towards unreachable areas of the state-action space. In this paper, we address these drawbacks in the scenario of online-fitted stochastic dynamics. We propose modeling state and action physical limits as probabilistic chance constraints and introduce a new trajectory optimization technique that integrates such probabilistic constraints by opti- mizing a relaxed quadratic program. Our empirical evaluations show a significant improvement in the robustness of the learning process, which enables our approach to perform more effective updates, and avoid premature convergence observed in other state-of-the-art techniques.
Link to PDF


zum Seitenanfang