In recent years, robot technology has undergone rapid development. Compared with a single-robot system, a multi-robot system has incomparable advantages. However, the key to whether a multi-robot system can outperform a single-robot system is whether the robot members in the multi-robot system can work together and efficiently complete global tasks. This article takes the task allocation algorithm of multi-robot systems as the research object, and based on an in-depth analysis of the current research status at home and abroad, a utility evaluation-based task allocation algorithm for multi-robot systems in mine maintenance is proposed to address the characteristics of high difficulty in mine maintenance and the dynamic complexity of the mine environment. This article designs a utility evaluation network structure based on fuzzy neural Q-learning, which improves the evaluation accuracy through real-time learning and parameter adjustment. Design a global reward function to enable each robot to allocate tasks based on their respective task evaluations, achieving global task allocation. It was applied to mine maintenance, achieving the effectiveness evaluation and task selection of maintenance tasks. The feasibility and accuracy of the method were verified through algorithm comparison.