Multi-Robot Path Planning in Agriculture: Benchmarking State-of-the-Art MAPF Algorithms

The agricultural sector faces challenges such as labour scarcity and the need for increased productivity, driving the adoption of Multi-Agent Path Finding (MAPF) algorithms in agricultural robotics. However, deploying MAPF algorithms in complex agricultural environments requires efficient and scalable path planning solutions capable of handling diverse field conditions and varying operational demands. This study benchmarks state-of-the-art MAPF algorithms using metrics including runtime, scalability (maximum agent capacity), and solution quality (Sum of Costs) under varying agricultural conditions. By conducting rigorous evaluations, the research highlights practical trade-offs between optimality and scalability for real-world field deployments. The main contribution is the identification of MAPF-LNS2 as the most viable algorithm for agriculture, offering rapid, collision-free solutions with good scalability, especially in time-critical, large-scale operations where strict optimality is not critical. This finding supports the use of sub-optimal yet scalable solutions for agricultural robotics where timely task completion is vital. The results provide valuable insights for researchers and practitioners seeking robust, efficient, and practical path planning algorithms suited for dynamic farming environments. Overall, this work contributes to advancing the deployment of collaborative multi-robot systems, enhancing productivity and sustainability in precision agriculture while addressing current limitations in labour and resource management.