Integrating task and motion planning for unmanned aerial vehicles

The process of integrating task and motion planning for unmanned aerial vehicles is examined. Here, an unmanned aerial vehicle is modeled as a Dubins vehicle: a vehicle with a minimum turn radius and the inability to go backward. Given a starting position and orientation for a Dubins vehicle and a set of stationary targets, the main problem is to determine the shortest flyable path that visits each target. This problem is called the Dubins traveling salesman problem, an extension of the well-known traveling salesman problem. A number of algorithms with different approaches, including a hierarchical approach, a generalized traveling salesman problem reformulation approach, and a search with an upper bound Dubins cost approach, are developed and contrasted. Each of the eight algorithms investigated generates either a point to point Dubins trajectory or a point to point relaxed Dubins trajectory. Monte Carlo simulations were performed for a range of vehicle turn radii. Simulations results show that integrating two plausible kinematic satisfying paths as an upper bound to determine the cost-so-far into a search algorithm generally improves performance in terms of the shortest path cost and search complexity.