Leveraging Reinforcement Learning to Allocate Bandwidth in the Agent-based Resource Management System

In this research, we utilize an agent-based resource management technique, namely the real-time Network Communication Resource Manager (NCRM) to manage battlefield network resources. The NCRM addresses the allocation problem by providing a process, an architecture, and an algorithm to allocate network resources to a mission based on its requirements and currently available resources. This approach enables resilient centralized command (i.e., mission analyses and resource allocation), distributed control, and decentralized execution (i.e., resource reservation and reallocation). Our solution focuses on maximizing the effectiveness of available (remaining) network assets to maintain integrity and efficiency of mission critical assets. NCRM can: (A) through intelligent path selection, deliver mission traffic over the optimal performing paths based on its characteristics or requirements and real-time path status; and (B) effectively converge to a new optimum that dynamically adapts itself to the latest network condition (e.g., newly introduced or lost assets/links).