RAPID: An Algorithm for Quick Replanning Under Changed Dynamical Constraints

There are many scenarios where the dynamical constraints of a vehicle may change mid-mission either due to added payload, vehicle damage or changes to the environment. The vehicle may need to come back to the base safely without undergoing collisions with obstacles. We present RAPID (RApid Planning for Interchanged Dynamics), a sampling based motion planning algorithm for quick replanning when the dynamical constraints associated with a vehicle change. RAPID uses the planning information associated with the original dynamics as a heuristic for quick replanning. RAPID is useful when a vehicle must replan to account for changes to its dynamics but does not have enough time or resources to run a whole planning algorithm (like RRT or SST) from scratch. We compare RAPID with brute-force replan using RRT and SST using a combination of hardware and software experiments. The hardware experiments are performed using a modified Parrot Bebop 2 quadrotor fitted with an Intel UP board traversing an obstacle course. Our experimental results demonstrate RAPID provides considerable improvements over planning from scratch (RRT, SST) in solution completion time, path length, and flight time performance metrics.