부산대학교 도서관

This paper is an initial investigation of social network analysis approaches to computer testing and evaluation of unmanned robot swarms, specifically unmanned aerial vehicles (UAV). Swarm level Measures of Performance (MoP) and Measures of Effectiveness (MoE) require testing and evaluation methods that extend beyond those sufficient for testing individual UAV function. MoPs and MoEs at the swarm level may be regarded as driven by characteristics of the network as a whole. Fundamental parameters of the network as a whole include average number and speed of UAV to UAV communication linkages. The present study compared differences in MoPs that underlie the effective performance of 30-UAV swarms using simulation experimentation method in a search and rescue mission scenario. Thirty different swarm configurations were composed by varying percentages of entities with one of three different operational behaviors relevant to a search and rescue mission (Social Searcher, Relay, Anti-Social Searcher). Except for the search behaviors, the UAVs were identical in function. Results indicated significant differences in time-to-mission-completion (MoE) among the 30 configurations. Statistical analyses revealed significant differences in network metric MoPs; that is, the best performing swarm configuration had a greater number of communication linkages and faster speed of communication pathways through the swarm network.