A Multi-robot Farmland Operation Path Scheduling Method Integrating Jump Point Search and Multi-stage Collaborative Optimized Salp Swarm Algorithm for Balanced Task Allocation
Abstract
To address high path costs, imbalanced task loads, and unstable convergence in multi-robot path planning and scheduling within unstructured farmland environments, this paper presents a balanced scheduling method combining Enhanced Jump Point Search (EJPS) with a Multi-Stage Collaborative Optimization Salp Swarm Algorithm (MCOSSA). Firstly, a farmland grid environment model is constructed to represent spatial accessibility and task locations. EJPS is used for global path planning, incorporating bidirectional jump expansion and adaptive heuristic adjustments. To further refine the path, a path-local segment node reconstruction (PLSNR) is proposed to compress redundant nodes and enhance path smoothness. Secondly, MCOSSA is employed for multi-robot task scheduling. A dual-factor encoding–decoding mechanism is designed, guided by a balanced fitness function minimizing the maximum path cost. The algorithm integrates dynamic subpopulation clustering, adaptive cooperative evolution, non-uniform Gaussian perturbations, and multi-neighborhood local search to improve convergence. Finally, experiments on benchmark functions, farmland maps, and large-scale task allocation scenarios show that the proposed method significantly outperforms particle swarm optimization, salp swarm algorithm, artificial bee colony, and others in terms of path length, load balancing, robustness, and stability, exhibiting strong potential for practical applications.