Specification and Execution of Multiagent Missions

Abstract

Specifying a purely reactive behavioral configuration for use by a multiagent team executing a mission requires both a careful choice of the behavior set and the creation of a temporal chain of behaviors which executes the mission. This difficult task is simplified by applying an object-oriented approach to the design of sequences of behavioral configurations where a methodology called temporal sequencing is used to partition the mission into discrete operating states and enumerate the perceptual triggers which cause transitions between those states. Several smaller independent configurations can then be created with each implementing one state, completing one step in the sequence. When properly constructed, these configurations (assemblages) become high level primitives reusable in subsequent projects, reducing development time. In the multi-vehicle domain being studied for the ARPA Demo II project, assemblages such as travel_to_location and occupy_location consist of groups of basic behaviors associated with coordination mechanisms that allow the group to be treated as a single coherent behavior. For example, travel_to_location consists of move_to_goal, avoid_obstacle, avoid_robot, noise, and stay_in_formation primitive behaviors moderated by a cooperative coordination operator. Upon instantiation, the assemblage is parameterized with a particular formation, goal location, and termination conditions. A mission coordination operator determines which assemblage to activate based upon the mission being executed and the current state of the system. A scenario language has been developed which allows specifying missions as sequences of steps, where each step invokes a particular assemblage. The missions are specified in a structured user-friendly language targeted for groups of cooperating robotic vehicles executing military-style scout missions. Various multiagent missions have been demonstrated in simulation using this system. Deployment on Denning mobile robots demonstrates the utility of this mission execution system, while later deployment on the ARPA Demo II test platforms will ultimately allow comparisons with software developed using other methods.