Decision-Making Aid for Human-Machine Teaming in Multiplayer Pursuit-Evasion Games

Abstract

Effective decision-making in adversarial multiplayer games is complicated by the large number of degrees-of-freedom in state-action pairs by the players. Difficulty is exacerbated by complex objectives that often develop in strategic deployment of tactics that coordinate multiple objectives. In this paper, we consider the decision-making problem for human-machine teaming to facilitate strategic command and control. We formulate decision-making problems for multiplayer pursuit-evasion games with multiple strategic objectives of varying complexity, such as capturing an evader, or steering it away or towards parts of the domain. We test the complexity of decision-making tactics through human-subject experiments. Operators are able to command a team of pursuers by selecting formations and motion primitives from a library of tactics to influence an evader's behavior. The designed interface allows the operator to perform practice trials in simulation that enable testing various strategies and examine time-risk before deployment on hardware platforms. Through surveys and metrics, we show that such an interface is found to be useful for decision-making by users. Performance was not impacted in any significant manner by the number of practice trials. This result suggests the impact a decision-making aid could have on improving performance in command and control of human-machine teams. Previous chapter