Robocup Rescue - Virtual Robots Team STEEL (USA) MrCS - The Multirobot Control System Prasanna Velagapudi 1 , Alexander Kleiner 1 , Nathan Brooks 1 , Paul Scerri 1 , Michael Lewis 2 , and Katia Sycara 1 1 Carnegie Mellon University, Pittsburgh PA, USA 2 University of Pittsburgh, Pittsburgh PA, USA Abstract. This paper describes the software system supporting the Carnegie Mellon/Univ. of Pittsburgh team of simulated search and res- cue robots in the Robocup Rescue 2010 Virtual Robots competition. Building on the Machinetta agent software, robot command and con- trol is decomposed into a hierarchy of subtasks managed by independent agents both on the robot and colocated with human operators. By en- capsulating all robot and human operator interactions into interfaces to these agents, the system can perform with a high level of robustness and reusability. As in previous years, the entire code base is portable and platform-independent, running entirely in Java. 1 Introduction In human-robot interaction, how and when the operator intervenes in the robotic system are the two predominant issues [Endsley, 1996]. How a human works with the system is a function of the level of autonomy (LOA), which describes the static function assignments between the human and the robot. The LOA can range from full manual control to full autonomy, with intermediate levels of LOA generally being superior to full autonomy or full manual control. This is because an LOA that is too high leads to degradation in manual or mental skill, loss of situation awareness, decision bias, and decrease in vigilance. The low LOA of full manual control leads to high mental demand, human decision bias, complacency, boredom, and inconsistent control behavior, all of which degrade performance. In systems with adaptive autonomy (AA), the allocation of control between the human and the robot can be dynamically changed and is usually triggered by a critical event, performance measurement, operator’s workload, or the operator model. Carefully calibrated AA maximizes the amount of time the human operator can spend doing tasks which humans perform better than robots, such as victim identification and navigation of robots out of stuck or dan- gerous positions, problems present with the current state-of-the-art for robots. If utilized effectively, it is clearly the case that adding robots to a search and rescue team will improve the speed at which an area can be searched, leading to