CIExplore

Yang, Huanhuan; Shi, Dianxi; Zhao, Chenran; Xie, Guojun; Yang, Shaowu

doi:10.1145/3459637.3482326

Cited by 5 publications

(3 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…• For learning algorithm : ICM method has been implemented with some other algorithms, like COMA (Delos Reyes et al, 2022;Yang et al, 2021), PPO (Zhang et al, 2022). We could implement I-Go-Explore with such different learning algorithms.…”

Section: Concluding Remarks and Future Workmentioning

confidence: 99%

Random curiosity-driven exploration in deep reinforcement learning

Shi

et al. 2020

Neurocomputing

View full text Add to dashboard Cite

Section: Concluding Remarks and Future Workmentioning

confidence: 99%

Random curiosity-driven exploration in deep reinforcement learning

Shi

et al. 2020

Neurocomputing

View full text Add to dashboard Cite

“…Reward sparsity is often a pervasive challenge [18] in multi-agent systems. Consequently, investigating techniques to ensure the stability or enhancement of algorithmic performance in such environments becomes essential for addressing real-world tasks [4,33,30]. The trial-and-error reward model in reinforcement learning typically involves substantial overhead costs during strategy revisions.…”

Section: Introductionmentioning

confidence: 99%

IPERS: Individual Prioritized Experience Replay with Subgoals for Sparse Reward Multi-Agent Reinforcement Learning

Xie,

Zhang,

Qiao

et al. 2023

Frontiers in Artificial Intelligence and Applications

View full text Add to dashboard Cite

Multi-agent reinforcement learning commonly uses a global team reward signal to represent overall collaborative performance. Value decomposition breaks this global reward into estimated individual value functions per agent, enabling efficient training. However, in sparse reward environments, agents struggle to assess if their actions achieve the team goal, slowing convergence. This impedes the algorithm’s convergence rate and overall efficacy. We present IPERS, an Individual Prioritized Experience Replay algorithm with Subgoals for Sparse Reward Multi-Agent Reinforcement Learning. IPERS integrates joint action decomposition and prioritized experience replay, maintaining invariance between global and individual loss gradients. Subgoals serve as intermediate goals that break down complex tasks into simpler steps with dense feedback and provide helpful intrinsic rewards that guide agents. This facilitates learning coordinated policies in challenging collaborative environments with sparse rewards. Experimental evaluations of IPERS in both the SMAC and GRF environments demonstrate rapid adaptation to diverse multi-agent tasks and significant improvements in win rate and convergence performance relative to state-of-the-art algorithms.

show abstract

“…This approach can also be extended to multi-agent exploration problems as a joint curiosity reward shared by all agents [138]. When the ensemble of dynamics models makes consistent predictions of the next states, the agent is less curious, as it has explored the state space with low uncertainty.…”

Section: Surprisingness-and-incongruity Based Curiositymentioning

confidence: 99%

Curiosity-driven learning in artificial intelligence and its applications

Sun

View full text Add to dashboard Cite

I would like to express my heartfelt gratitude to my supervisor, Prof Miao Chunyan, for her invaluable guidance, insightful feedback, and constant encouragement throughout my Ph.D journey. Her vast knowledge, experience, and unwavering support have been instrumental in shaping my research and shaping me into a better researcher. I am also grateful to Dr. Qian Hangwei and Dr. Zeng Zhiwei for their assistance and advice that helped me overcome challenges and obstacles. Furthermore, I am thankful to Alibaba-NTU Singapore Joint Research Institute and the Singapore Economic Development Board (EDB) for their generous support and funding of this research project. Without their assistance, this research would not have been possible. Finally, I would like to express my gratitude to my family and friends for their continuous love, support, and encouragement during this long and challenging journey. Their unwavering belief in me has been a constant source of motivation, and I am grateful to have them in my life.

show abstract

CIExplore

Cited by 5 publications

References 22 publications

Random curiosity-driven exploration in deep reinforcement learning

Random curiosity-driven exploration in deep reinforcement learning

IPERS: Individual Prioritized Experience Replay with Subgoals for Sparse Reward Multi-Agent Reinforcement Learning

Curiosity-driven learning in artificial intelligence and its applications

Contact Info

Product

Resources

About