Aiming at the problems of unreasonable task matching and frequent path conflicts in the collaborative scheduling of multiple AGVs in the packaging workshop, this paper proposes an optimization method based on genetic algorithm and collaborative path selection. Firstly, the genetic algorithm is used to model the task allocation of multiple AGVs, with the goal of minimizing the total travel distance, to achieve the optimal matching of AGVs and packaging machine tasks. Secondly, for AGVs that retrieve pallets and transport materials, a differentiated path selection strategy is designed to reduce the risk of conflicts through path separation. Simulation experiments show that in the scenario where 6 AGVs are selected from 8 AGVs to serve 3 packaging machines, this algorithm can achieve rapid task matching and effectively reduce path overlap. This study provides an integrated solution for task allocation and path coordination optimization in workshop logistics automation, which has practical significance for improving the operational efficiency of AGV groups.