Agent-Based Distributed Framework for Collaborative Planning

Motivated by the Navy's new emphasis on networked distributed planning capabilities in maritime operations centers (MOC), this paper presents an agent-based distributed planning framework applicable to a range of missions. A mission is modeled as a task graph, where each task is specified in terms of a vector of resource requirements, processing time, geographic location and an opportunity window. Each planning agent has a set of mobile assets, where each asset is specified in terms of a vector of resource capabilities it provides, its velocity, and its current location. Multiple agents need to collaboratively solve the resulting dynamic asset allocation and task scheduling problems to maximize task execution accuracy, measured in terms of how well the task requirements are matched to the asset capabilities, minimize the completion time (timeliness metric) or find a trade-off between accuracy and timeliness. The agent-based distributed planning framework presented here uses a blackboard-based communication paradigm. It has two distinguishable components intra-agent and inter-agent. The intra-agent component employs an optimization-based scheduling and an M-best asset allocation algorithm to match the tasks with its own assets. The inter-agent component coordinates the exchange of information and assets to improve the local schedules and allocations using the blackboard architecture, an asset pricing mechanism, and pair-wise exchange of assets among agents to realize supporting-supported relationships among agents.