Risk-Aware On-the-Fly Solving of Physical Vehicle Routing Problems

In missions for a set of autonomous vehicles given a complex environment with obstacles and many waypoints to visit, risk-aware routing plays an important role. In this paper, we consider multi-robot, multi-goal motion planning where unsafe areas should be avoided. We assume a geometric environment for a set of high-dimensional robots, providing a motion model with nonlinear dynamics that, for a given state and a small time step, applies a control action and provides a next state or reports a collision. As there are anonymous goals meaning there is no predefined assignment of goals to the robots, the approach assigns goals to them on-the-fly during the solution process. We study the computational limits and possibilities of our approach, derive a scaling framework system that plans and executes the safe travel for the given fleet of robots, and we conduct experiments for benchmark scenarios.