Towards Fault Tolerance in Multi-Agent Reinforcement Learning
This work addresses fault tolerance for multi-agent reinforcement learning systems, which is an incremental improvement in a domain-specific area.
The paper tackled the problem of agent faults degrading performance in multi-agent reinforcement learning by enhancing fault tolerance through an attention mechanism for fault detection and a prioritization mechanism for balanced training, achieving effective handling of various fault types, agents, and timings in experiments.
Agent faults pose a significant threat to the performance of multi-agent reinforcement learning (MARL) algorithms, introducing two key challenges. First, agents often struggle to extract critical information from the chaotic state space created by unexpected faults. Second, transitions recorded before and after faults in the replay buffer affect training unevenly, leading to a sample imbalance problem. To overcome these challenges, this paper enhances the fault tolerance of MARL by combining optimized model architecture with a tailored training data sampling strategy. Specifically, an attention mechanism is incorporated into the actor and critic networks to automatically detect faults and dynamically regulate the attention given to faulty agents. Additionally, a prioritization mechanism is introduced to selectively sample transitions critical to current training needs. To further support research in this area, we design and open-source a highly decoupled code platform for fault-tolerant MARL, aimed at improving the efficiency of studying related problems. Experimental results demonstrate the effectiveness of our method in handling various types of faults, faults occurring in any agent, and faults arising at random times.