Dynamic Obstacle Avoidance for Nonholonomic Vehicles using Collision Cones: Theory and Experiments *

Obstacle avoidance is a critical part in the control of autonomous vehicles. This paper proposes an obstacle avoidance algorithm based on the computationally advantageous collision cone concept, for unicycles which may be subject to strict input constraints and dynamic, non-compliant obstacles which may even actively pursue a collision. The paper presents both a theoretical analysis together with experiments demonstrating the beneficial features of the algorithm. We analytically derive explicit, and practically intuitive, conditions on the control parameters, under which safety is guaranteed. Experiments illustrate the safe, intuitive, and minimally invasive behaviors generated by the proposed algorithm.