SIMULATION ANALYSIS FOR UAV SEARCH ALGORITHM DESIGN USING APPROXIMATE DYNAMIC PROGRAMMING

Many decision-and-control algorithms have been proposed for autonomous unmanned aerial vehicles (UAVs). The nature of this problem, with large decision spaces and the desire for optimal performance criteria, indicates that closed-form analysis of any approach is nearly impossible, suggesting a simulationbased performance evaluation of relevant scenarios. However, while effective simulation practices have been developed in the operations-research community, awareness of them in the decision-and-control community may be lower than desired for routine application. If applied correctly, these simulation techniques could have a major impact on the quality and effectiveness of UAV algorithms. This paper provides a concrete example that demonstrates that the marriage of UAV decision-andcontrol algorithms with proper simulation techniques can be done effectively and with great benefits for the UAV algorithm researcher, both in terms of the validity and quality of simulation results. The example used in this case is a study conducted for a stochastic UAV algorithm design. In particular, the study is to find the “optimal” sensitivity of a “future-gain” factor that attempts to balance future and present gains in a stochastic-approximate dynamic-programming solution to the problem faced by a team of UAVs searching in an uncertain environment for targets. This work can serve as a template for similar simulation experiments in this area. Experimental motivation and demonstration of the results are given.