Knowledge-based agents in a command and control simulation

Abstract

To facilitate research into multi-agent planning issues, we have developed the Command and Control Testbed Using Simulation (CACTUS) and the Situation-based Autonomous Reasoner in a GBB Environment (SARGE). CACTUS is a simulation of combat operations for platoon-sized units in battalion-level engagements. We model in detail those command and control problems impacting the way planning is done by a formation commander. These include incomplete and uncertain information, imperfect control, differing agent characteristics, and limited inter-unit communications.The testbed originally required the user to input specific movement and firing instructions for each unit simulated. An interface to the simulation permits the integration of automated planners in place of user input. The SARGE commander was built to relieve the user of the need to provide detailed instructions at the simulation system level. Units with SARGE commanders take goal level instructions and SARGE then explictly reasons about plans to achieve those goals and submits the instructions necessary to implement those plans in CACTUS. SARGE uses the blackboard model of reasoning as implemented in the Generic Blackboard (GBB) tool. User access to the knowledge affecting SARGE behavior allows us to assess the impact of different tactical and doctrinal schemes on planning at higher levels of command.The SARGE and CACTUS systems provide the foundations for developing automated planners for multiple unit formations by raising the level of abstraction for inputting instruction to subordinate units. We will use this incremental approach to address command and control issues at progressively higher levels in the chain of command and to develop, refine, and evaluate artificial intelligence techniques to integrate human and machine capabilities. Our ultimate goal is to collect applicable techniques into a system supporting rapid prototyping of tactical decision aids. SARGE and CACTUS are vital steps in this process.