Research on AGV Task Rescheduling Based on Improved Greedy and Simulated Annealing

This paper addresses multi-AGV task allocation in intelligent manufacturing by proposing a hybrid rescheduling method combining an improved greedy algorithm and simulated annealing. It overcomes limitations of existing methods, such as poor global optimization and uneven system loads in complex road networks. The workshop environment is modeled using a regionalized node model and a directed path network graph. The hybrid algorithm employs a polling greedy approach for rapid, balanced initial task allocation and simulated annealing for near-global optimal solutions, avoiding local optima. Simulations demonstrate the method reduces the longest AGV movement distance by 42.3% compared to FCFS, 23.8% compared to polling greedy, and 39.5% compared to standalone simulated annealing, enhancing efficiency and system balance.