基于GMM – BHO与Hybrid A*的海洋牧场无人船集群投喂任务分配方法

Feeding Task Allocation Method of USV Cluster in Marine Ranching Based on GMM – BHO and Hybrid A*

摘要:

针对海洋牧场规模化网箱自动投喂中时间窗集中、无人船载饵容量受限、补给返航频繁及网箱避障约束强等问题,提出一种融合高斯混合模型(GMM)、赏金猎人优化算法(BHO)与Hybrid A*的USV集群投喂任务分配与路径规划方法。首先,构建任务–障碍一体化模型,将网箱同时表征为投喂服务对象和路径规划膨胀障碍;其次,利用GMM生成空间连续的初始责任区,并将其作为结构化精英个体注入BHO,通过边界任务调整优化任务归属、补给次数和时间窗违约代价;最后,结合容量感知TSP、补给节点插入和Hybrid A*生成满足网箱避障、航向连续和环绕投喂约束的可执行航迹。仿真结果表明,在60个等效直径80 m网箱、6艘USV、航渡速度5 m/s条件下,所提方法在180 min时间窗内实现100%任务完成率和零障碍违规;在120 min紧时间窗下,相较K-means和GMM方法,总航程降低6.4%,总延误时间降低19.8%,补给次数由11次降至9次。结果表明,该方法能够提升复杂海洋牧场场景下USV集群自动投喂的时效性、安全性和补给组织效率。

Abstract:

Aiming at the problems of time window concentration, limited USV bait capacity, frequent replenishment and return, and strong obstacle avoidance constraints in the automatic feeding of large-scale cages in marine ranching, a USV cluster feeding task allocation and path planning method combining Gaussian mixture model ( GMM), bounty hunter optimization algorithm (BHO) and Hybrid A* was proposed. Firstly, the task – obstacle integration model was constructed, and the cage was characterized as feeding service object and path planning expansion obstacle at the same time. Secondly, GMM was used to generate a spatially continuous initial responsibility area, which was injected into BHO as a structured elite individual, and the task attribution, replenishment times and time window violation cost were optimized through boundary task adjustment. Finally, the capacity-aware TSP, replenishment node insertion and Hybrid A* were combined to generate an executable path that satisfied the cage obstacle avoidance, heading continuity and surrounding feeding constraints. The simulation results showed that the proposed method achieved 100% task completion rate and zero obstacle violation in a 180 min time window under the conditions of 60 equivalent diameter 80 m cages, 6 USVs and 5 m/s ferry speed. Under the 120 min tight time window, compared with the K-means and GMM methods, the total voyage was reduced by 6.4%, the total delay time was reduced by 19.8%, and the number of replenishments was reduced from 11 to 9. The results showed that this method can improve the timeliness, safety and supply organization efficiency of USV cluster automatic feeding in complex marine ranching scenarios.