Automated guided vehicles (AGVs) are an essential component for automation fulfillment centers, a kind of warehouse. Efficient control of the AGV leads to easier management of inventory in the fulfillment center. To increase the productivity of various warehouses including fulfillment centers, we propose a Multi-Agent Reinforcement Learning (MARL)-based algorithm for cooperative control of AGVs. The proposed algorithm is based on a popular cooperative MARL algorithm, and utilizes an additional technique for path control of AGVs to distinguish the sacrifices of each agent and compensate them accordingly. We evaluate the proposed algorithm in comparison with a basic MARL algorithm on two fulfillment center layouts and provide further insight via the visualization of the results.