Hierarchical Task Allocation Framework with Adaptive Network Recovery under Node Failures

Multi-agent task allocation in communication constrained environments poses significant challenges when agents fail, leaving critical tasks unfinished. While decentralized methods such as Consensus-Based Bundle Algorithms (CBBA) ensure scalable initial assignments, they lack built-in fault recovery under dynamic connectivity. To address this gap, a hierarchical, three-layer framework is introduced, combining a Fast Local Rescue (FLR) algorithm for rapid, mode-based reassignment, and a real-time MILP optimizer at the base station for global oversight. Within each connected subnet, a centrality-elected arbiter node coordinates local recovery, balancing communication reachability, topological importance, and spare capacity. A Monte Carlo simulation demonstrates that our approach sustains high task completion and swift recovery across diverse failure patterns without imposing prohibitive overhead. The proposed framework offers robust performance and real-time guarantees, paving the way for resilient operations in mission-critical multi-agent systems.