No abstract

This paper presents a method for a team of multiple agents with finite communication range to perform an efficient continuous search of an area of interest. The search is coordinated between agents in a decentralized manner through local probability maps which are maintained by each individual agent. The probability map details the odds that a target exists at any given location within the area of interest. The search algorithm endeavors to find pop-up targets, which is accomplished as agent's plan their motion to search high probability regions of the area of interest, while simultaneously striving to maintain a connected communication topology with the other agents. This algorithm is simulated with the premise of unmanned air vehicles (UAVs), which are limited by a finite communication constraint, searching an area for targets. Simulation results are presented for 5, 20, and 30 agents and compared against a lower bound on the global optimum of the problem presented, as well as an average of random walks.

This paper addresses the multi-target, multi-task assignment problem for unmanned vehicles by presenting the decentralized Extended Consensus-Based Bundle Algorithm (ECBBA). The algorithm converges to conflict-free feasible solutions; furthermore, the algorithm allows local "pop-up" targets to be added and re-converges to a solution without resolving the entire optimization. A method to limit the communication network around targets is introduced. Limiting the communication is shown, through simulation, to produce solutions commensurate with full communication (within 2%) with the added benefit of reduced communication across the network.

No abstract