Collision avoidance as a robust reachability problem under model uncertainty

The paper presents an approach to the design of an optimal collision avoidance maneuver under model uncertainty. The dynamical model is assumed to be only partially known and the missing components are modeled with a polynomial expansion whose coefficients are recovered from sparse observations. The resulting optimal control problem is then translated into a robust reachability problem in which a controlled object has to avoid the region of possible collisions, in a given time, with a given target. The paper will present a solution for a circular orbit in the case in which the reachable set is given by the level set of an artificial potential function