2024-05-05: To address the challenge of the exponential growth of the joint action space and the resulting exponential increase in search space, this paper extends the AlphaZero and MuZero algorithms to more complex multi-agent Markov decision processes, proposing two algorithms, Multiagent Gumbel AlphaZero and Multiagent Gumbel MuZero. These algorithms are based on resettable environment simulators and multi-agent world models constructed by neural networks, enabling efficient search and decision-making. An efficient without-replacement top-$k$ sampling algorithm is proposed for exponential action space, along with a high-quality policy improvement operator adapted to it, reducing the complexity and computational burden of Monte Carlo tree search while enhancing exploration capabilities. Besides, we propose a Centralized Planning with Decentralized Execution (CPDE) paradigm which uses centralized planning to accelerate policy learning while enabling decentralized execution for deployment. The algorithms achieve the best performance in testing environments such as StarCraft compared to model-free multi-agent reinforcement learning algorithms, with a 10-fold increase in learning sample efficiency while maintaining win rates.

2024-05-05: In this work we introduce a general sequence modeling framework for studying sequential decision making through the lens of Hierarchical RL. At the time of making decisions, a high-level policy first proposes an ideal prompt for the current state, a low-level policy subsequently generates an action conditioned on the given prompt. We show DT emerges as a special case of this framework with certain choices of high-level and low-level policies, and discuss the potential failure of these choices. Inspired by these observations, we study how to jointly optimize the high-level and low-level policies to enable the stitching ability, which further leads to the development of new offline RL algorithms. .

2023-10-21: We study Policy-extended Value Function Approximator (PeVFA) in Reinforcement Learning (RL), which extends conventional value function approximator (VFA) to take as input not only the state (and action) but also an explicit policy representation. Such an extension enables PeVFA to preserve values of multiple policies at the same time and brings an appealing characteristic, i.e., *value generalization among policies*.

2023-09-10: Deep Reinforcement Learning (Deep RL) and Evolutionary Algorithms (EA) are two major paradigms of policy optimization with distinct learning principles, i.e., gradient-based v.s. gradient-free. An appealing research direction is integrating Deep RL and EA to devise new methods by fusing their complementary advantages. However, existing works on combining Deep RL and EA have two common drawbacks: 1) the RL agent and EA agents learn their policies individually, neglecting efficient sharing of useful common knowledge; 2) parameter-level policy optimization guarantees no semantic level of behavior evolution for the EA side. In this paper, we propose Evolutionary Reinforcement Learning with Two-scale State Representation and Policy Representation (ERL-Re^2), a novel solution to the aforementioned two drawbacks. The key idea of ERL-Re^2 is two-scale representation: all EA and RL policies share the same nonlinear state representation while maintaining individual} linear policy representations. The state representation conveys expressive common features of the environment learned by all the agents collectively; the linear policy representation provides a favorable space for efficient policy optimization, where novel behavior-level crossover and mutation operations can be performed. Moreover, the linear policy representation allows convenient generalization of policy fitness with the help of the Policy-extended Value Function Approximator (PeVFA), further improving the sample efficiency of fitness estimation. The experiments on a range of continuous control tasks show that ERL-Re^2 consistently outperforms advanced baselines and achieves the State Of The Art (SOTA).