Distributed Multi-Robot Dynamic Task Allocation Based on the Improved Artificial Bee Colony Algorithm

Abstract

The dynamic Multi-Robot Task Allocation (MRTA) problem presents a critical bottleneck in distributed control due to high communication overhead, single-point failure risks, and dynamic constraints. This paper proposes a fully decentralized Improved Artificial Bee Colony (I-ABC) algorithm integrated with Finite State Machines (FSMs). The continuous optimization process is mapped into an eight-state Markovian evolution flow, eliminating reliance on a central controller and global clock synchronization. Standard roulette selection is replaced with a Field-of-View (FoV)-based probabilistic recruitment mechanism. Furthermore, an adaptive motivation decay function balances task persistence against dynamic responsiveness—preventing both task inertia and premature thrashing—while quorum consensus sensing suppresses spatial congestion and deadlocks. Extensive Monte Carlo simulations under dynamic Poisson task arrivals and unexpected robot failures demonstrate that I-ABC achieves superior trade-offs in makespan, energy efficiency, and communication overhead compared to standard ABC and state-of-the-art DG-MRTA strategies. The proposed framework offers a highly robust and scalable solution for decentralized swarm deployments.