Ant colony optimization for agile motion planning

Abstract

With the need for greater autonomy in unmanned vehicles growing, design of algorithms for mission-level planning becomes essential. The general field of motion planning for unmanned vehicles falls into this category. Of particular interest is the case of operating in hostile environments with unknown threat locations. When a threat appears, a replan must be quickly formulated and executed. The use of terrain masking to hide from the threat is a vital tactic, which a good algorithm should exploit. In addition, the algorithm should be able to accommodate large search spaces and non-linear objective functions. This thesis investigates the suitability of the Ant Colony Optimization (ACO) heuristic for the agile vehicle motion planning problem. An ACO implementation tailored to the motion planning problem was designed and tested against an existing genetic algorithm solution method for validation. Results show that ACO is indeed a viable option for real-time trajectory generation. ACO' ability to incorporate heuristic information, and its method of solution construction, make it better suited to motion planning problems than existing methods.