Intelligent Agents as Supervisory Assets for Multiple Uninhabited Systems: RoboLeader

Report No. ARL-TR-5886
Authors: Michael J. Barnes and Jessie Y. C. Chen
Date/Pages: January 2012; 34 pages
Abstract: The report reviews the literature on intelligent agents, documenting the advantages and limitations of various architectural approaches for military applications. The specific application considered is the control of multiple unmanned systems in complex environments. We discuss flexible automation, intelligent agents, swarms, hierarchical systems of agents, cognitive architectures, and agent/human teams. Because of the volatility of combat and its inherent uncertainty, we concluded that only agents that are under the control of human supervisors would be useful in such an environment. As an exemplar, we discussed RoboLeader, an intelligent agent and simulation environment that was developed to evaluate human factors issues related to controlling up to eight robots during urban reconnaissance missions while under the direct control of a human operator. We reviewed three RoboLeader studies that manipulated number of controlled robots, reliability level (false alarms and misses), multitasking difficulty, and target mobility. Robots were required to re-route during the mission as various intelligence updates were received; in the third experiment, they were required to entrap a moving vehicle by using multiple robots to block the vehicle's escape route. The results and their implications are discussed and, in conjunction with the literature review, are used to generate guidelines for developing human/agent architectures.
Distribution: Approved for public release
  Download Report ( 0.843 MBytes )
If you are visually impaired or need a physical copy of this report, please visit and contact DTIC.
 

Last Update / Reviewed: January 1, 2012