P. Flues and O. Stursberg, “Online Control of Affine Systems in Stochastically Modeled Contexts.,” IFAC-PapersOnline, vol. 53(2), pp. 3416–3422, 2020.



This paper proposes an algorithm for online controller synthesis for autonomous systems with LTI dynamics considering obstacle avoidance. The obstacles are assumed to be other systems with affine probabilistic dynamics. The initial state as well as the disturbances of these systems are Gaussian distributed. To guarantee that the probability of a collision is smaller than a predefined threshold, probabilistic reachable sets are used. Due to the Gaussian distribution, the probabilistic reachability procedure can use the principles of the ellipsoidal calculus. For the autonomous system, these time-varying reachable sets of the other systems are avoided by an approach, which is based on model predictive control and successive convexification of the constraints. Due to high computational times required for the computation of probabilistic reachable sets and the convexification, different techniques to reduce the computational time significantly are also proposed.



 AUTHOR={P. Flues and O. Stursberg},
 TITLE={{Online Control of Affine Systems in Stochastically Modeled Contexts}},
 COMMENT={Conf-ID: 3974}}