Locally optimal decomposition for autonomous obstacle avoidance with the Tunnel-MILP algorithm

The Tunnel-MILP algorithm is a three stage path planning method for 2-D environments that relies on the identification of a sequence of convex polygons to form an obstacle free tunnel through which to plan a dynamically feasible path. This work investigates two aspects of the algorithm. First, a greedy cut method is proposed for improved decomposition of the environment, resulting in fewer regions than existing algorithms. Second, the effect of the decomposition on the resulting solution is investigated, and conditions are presented to demonstrate that the resulting tunnel cannot be improved to yield a better solution. This ensures that the tunnel provided does not adversely affect the resulting dynamically feasible trajectory, and guarantees local optimality of the solution.