Modified Particle Swarm Optimization for Search Missions

There are many applications that can benefit from a well planned search. Whether the search objective is a lost hiker, a stolen vehicle on the interstate, or enemies on the battlefield, some assumptions must be made about the search objective before the search can begin. These assumptions focus the search in areas that have a relatively high likelihood of finding the targets of interest. A common approach to mission planning is to apply an optimization algorithm and obtain a good solution based on these assumptions. In general, the mission planner uses simulated targets to emulate the expected target behavior in order to evaluate candidate search paths. In practice, a major drawback is that the prior distribution of targets is only used to evaluate the search paths rather than to guide the optimization algorithm in generating the search paths. This paper introduces a method that explicitly exploits the sampled target distribution to create search paths, which naturally improves the results since the search paths directly depend on the time varying target locations. Results from a realistic cooperative path planning scenario show that explicit usage of target distributions can improve the performance of particle swarm optimization.