Top-Down Mechanism Design Study for Multi-UAV Search and Surveillance

摘要

Under this seedling effort, Metron has developed and demonstrated 'function-driven design' technology for complex distributed systems and applied it to an Unmanned Aerial Vehicle (UAV) ground target surveillance scenario. The authors considered three types of UAV interaction mechanisms (auction/bidding methods, swapping-based methods and local optimization methods) and two different target motion models (stationary and mobile, random walkers). There are two primary breakthroughs. The first is a value potential approach to optimizing search paths based on approximating an infinite-horizon search plan. Using this value potential to dictate UAV motion improves the search performance, especially for disjoint, multimodal (patchy) probability distributions on target position. The second innovation introduces dynamic area sectoring, which allows UAVs to partition the search area dynamically and to balance the search workload across UAVs. Sectoring also eliminates the need to deconflict search paths and simplifies collision avoidance because each UAV stays inside its sector. Combining the value potential-based UAV motion and dynamic sectoring reduced the median time to target detection by up to 40 percent in this experimental testing. Based in part on the improved multi- sensor search capability developed under this seedling effort, Metron has been awarded a NAVAIR Phase II SBIR contract to accelerate the transition of this distributed search technology.