Greedy based Value Representation for Optimal Coordination in Multi-agent Reinforcement Learning

Lipeng Wan, Zeyang Liu, Xingyu Chen, Xuguang Lan, Nanning Zheng
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:22512-22535, 2022.

Abstract

Due to the representation limitation of the joint Q value function, multi-agent reinforcement learning methods with linear value decomposition (LVD) or monotonic value decomposition (MVD) suffer from relative overgeneralization. As a result, they can not ensure optimal consistency (i.e., the correspondence between individual greedy actions and the best team performance). In this paper, we derive the expression of the joint Q value function of LVD and MVD. According to the expression, we draw a transition diagram, where each self-transition node (STN) is a possible convergence. To ensure the optimal consistency, the optimal node is required to be the unique STN. Therefore, we propose the greedy-based value representation (GVR), which turns the optimal node into an STN via inferior target shaping and eliminates the non-optimal STNs via superior experience replay. Theoretical proofs and empirical results demonstrate that given the true Q values, GVR ensures the optimal consistency under sufficient exploration. Besides, in tasks where the true Q values are unavailable, GVR achieves an adaptive trade-off between optimality and stability. Our method outperforms state-of-the-art baselines in experiments on various benchmarks.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-wan22c, title = {Greedy based Value Representation for Optimal Coordination in Multi-agent Reinforcement Learning}, author = {Wan, Lipeng and Liu, Zeyang and Chen, Xingyu and Lan, Xuguang and Zheng, Nanning}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {22512--22535}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/wan22c/wan22c.pdf}, url = {https://proceedings.mlr.press/v162/wan22c.html}, abstract = {Due to the representation limitation of the joint Q value function, multi-agent reinforcement learning methods with linear value decomposition (LVD) or monotonic value decomposition (MVD) suffer from relative overgeneralization. As a result, they can not ensure optimal consistency (i.e., the correspondence between individual greedy actions and the best team performance). In this paper, we derive the expression of the joint Q value function of LVD and MVD. According to the expression, we draw a transition diagram, where each self-transition node (STN) is a possible convergence. To ensure the optimal consistency, the optimal node is required to be the unique STN. Therefore, we propose the greedy-based value representation (GVR), which turns the optimal node into an STN via inferior target shaping and eliminates the non-optimal STNs via superior experience replay. Theoretical proofs and empirical results demonstrate that given the true Q values, GVR ensures the optimal consistency under sufficient exploration. Besides, in tasks where the true Q values are unavailable, GVR achieves an adaptive trade-off between optimality and stability. Our method outperforms state-of-the-art baselines in experiments on various benchmarks.} }
Endnote
%0 Conference Paper %T Greedy based Value Representation for Optimal Coordination in Multi-agent Reinforcement Learning %A Lipeng Wan %A Zeyang Liu %A Xingyu Chen %A Xuguang Lan %A Nanning Zheng %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-wan22c %I PMLR %P 22512--22535 %U https://proceedings.mlr.press/v162/wan22c.html %V 162 %X Due to the representation limitation of the joint Q value function, multi-agent reinforcement learning methods with linear value decomposition (LVD) or monotonic value decomposition (MVD) suffer from relative overgeneralization. As a result, they can not ensure optimal consistency (i.e., the correspondence between individual greedy actions and the best team performance). In this paper, we derive the expression of the joint Q value function of LVD and MVD. According to the expression, we draw a transition diagram, where each self-transition node (STN) is a possible convergence. To ensure the optimal consistency, the optimal node is required to be the unique STN. Therefore, we propose the greedy-based value representation (GVR), which turns the optimal node into an STN via inferior target shaping and eliminates the non-optimal STNs via superior experience replay. Theoretical proofs and empirical results demonstrate that given the true Q values, GVR ensures the optimal consistency under sufficient exploration. Besides, in tasks where the true Q values are unavailable, GVR achieves an adaptive trade-off between optimality and stability. Our method outperforms state-of-the-art baselines in experiments on various benchmarks.
APA
Wan, L., Liu, Z., Chen, X., Lan, X. & Zheng, N.. (2022). Greedy based Value Representation for Optimal Coordination in Multi-agent Reinforcement Learning. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:22512-22535 Available from https://proceedings.mlr.press/v162/wan22c.html.

Related Material