Optimal and receding-horizon path planning algorithms for communications relay vehicles in complex environments

Abstract

This thesis presents new algorithms for path planning in a communications constrained environment for teams of unmanned vehicles. This problem involves a lead vehicle that must gather information from a set of locations and relay it back to its operator. In general, these locations and the lead vehicle's position are beyond line of-sight from the operator and non-stationary, which introduces several difficulties to the problem. The proposed solution is to use several additional unmanned vehicles to create a network linkage between the operator and the lead vehicle that can be used to relay information between the two endpoints. Because the operating environment is cluttered with obstacles that block both line-of-sight and vehicle movement, the paths of the vehicles must be carefully planned to meet all constraints. The core problem of interest that is addressed in this thesis is the path planning for these link vehicles. Two solutions are presented in this thesis. The first is a centralized approach based on a numerical solution of optimal control theory. This thesis presents an optimal control problem formulation that balances the competing objectives of minimizing overall mission time and minimizing energy expenditure. Also presented is a new modification of the Rapidly-Exploring Random Tree algorithm that makes it more efficient at finding paths that are applicable to the communications chaining problem. The second solution takes a distributed, receding-horizon approach, where each vehicle solves for its own path using a local optimization that helps the system as a whole achieve the global objective.

(cont.) This solution is applicable to real-time use onboard a team of vehicles. To offset the loss of optimality from this approach, a new heuristic is developed for the linking vehicles. Finally, both solutions are demonstrated in simulation and in flight tests in MIT's RAVEN testbed. These simulations and flight tests demonstrate the performance of the two solution methods as well as their viability for use in real unmanned vehicle systems.