Comparison of Trajectory Determination Approaches for Small UAVs

摘要

In considering the problem of small unmanned aerial vehicle (SUAV) surveillance mission in a target rich environment, it is desirable to follow a trajectory path that maximizes targets coverage and observation time, while minimizing airframe maneuvering. Motivated by this requirement, this paper investigates the merits of multiple vehicle trajectory path schemes. Genetic Algorithms (GAs) and local optimum techniques are compared to a more conventional defined-path approach. The authors also introduce a polygon boundary reflection algorithm (PBRA) and investigate its merits. Given a scenario containing multiple targets of unknown positions, the GA optimization approach determines the waypoints defining a path that best satisfies three goals: (1) maximize the number of targets seen, (2) maximize the average observation time for each target, and (3) minimize the SUAV acceleration history. Were the target locations known apriori, this problem could decompose into a variant of the much-studied traveling salesman problem (TSP). The complication of not knowing the actual target locations apriori means that the optimization tool must find waypoints that best satisfy the multiple objectives with little actual knowledge at initiation. Given this additional complexity and the fact that there are multiple objectives that must be maximized, a GA approach was investigated because it offers the ability to rigorously search for the optimum waypoint locations while simultaneously examining performance against multiple objectives. The GA software used in the analysis is IMPROVE (Implicit Multi-objective Parameter Optimization via Evolution). Comparison results of the GA based approaches, pareto and non-pareto, were investigated and compared with the simple PBRA and the popular Serpentine path approach. The analysis shows the GA optimization benefits and performance tradeoffs for all the path planning approaches that were studied.