An adaptive multi-objective task splitting algorithm for agricultural multi-robot task allocation
针对果园采摘中多机器人任务分配问题，提出自适应多目标任务拆分算法，优化完工时间和能耗，通过混合编码和自适应搜索策略提升效率，实验验证了有效性。
In smart farming, the development of multi-robot systems is essential for improving harvesting efficiency and addressing labor shortages, yet achieving optimal coordination presents significant scheduling complexities. This paper investigates a critical multi-robot task allocation problem in the context of orchard harvesting, focusing on multi-objective optimization (makespan and energy consumption) while considering task divisibility, robot re-utilization for multiple routes, and load-dependent energy consumption. Balancing these conflicting objectives across transport and operations, while managing the immensely expanded decision space due to task splitting, poses severe challenges to algorithm efficiency. To deal with these issues, we propose an adaptive multi-objective task splitting algorithm (AMTSA), which employs a hybrid encoding method to clearly demonstrate robot-route assignments and task-splitting information. Regarding the search modes, AMTSA adopts an adaptive strategy: the early phase emphasizes diverse exploration through route structural optimization, while the later phase shifts computational resources towards in-depth task splitting optimization (TSO) to accelerate convergence. Furthermore, the TSO process integrates two coordinated splitting mechanisms to collaboratively optimize the spatial layout and temporal balance of the task allocation by targeting the bottleneck robot, thereby addressing multi-objective conflicts. The effectiveness and robustness of AMTSA have been validated through comparative experiments on one real-world case study and 15 newly constructed test instances. Generally, this research advances the theoretical comprehension of multi-robot coordination and provides practical insights relevant to agricultural automation.