Flatness-Based Mixed-Integer Obstacle Avoidance MPC for Collision-Safe Automated Urban Driving

Abstract

Safe and efficient obstacle avoidance is a challenging feature in the urban automated driving context. This paper proposes an optimal flatness-based modular obstacle-avoidance model-predictive control strategy. Exploiting differential flatness, a unified obstacle-avoidance MPC formulation in the linearized flat coordinates is devised, based on the well-known mixed-integer Big-M formulation to resolve the non-convex avoidance constraints. Various choices of flat outputs (Cartesian vs. Frenet path coordinates) can be utilized with appropriate transformations, allowing a unified, efficient, and globally optimal solution of the obstacle-avoidance MPC problem. The resulting control law is tested in a multi-agent traffic simulation and shows excellent performance in a typical complex urban intersection scenario.