Experiments in real time path planning for a small unmanned helicopter using mixed integer linear programming

Abstract

We use mathematical programming to perform simulated and actual flight experiments with the MIT autonomous helicopter platform. The experimental platform mechanical hardware, avionics and software architecture are described. Mixed Integer Linear Programming formulations for guidance experiments in obstacle avoidance and threat evasion are presented. Results from an experimental flight and flight simulations are presented and discussed. The performance of the GNU Linear Programming Kit (GLPK) software library, used in the experiments, is examined. The solvers internal branching heuristic, as it pertains to our formulations, is illustrated graphically. Finally, we discuss ways to improve the solvers performance by adding stronger objective bounds to the subproblems it creates while applying the branch and bound algorithm.