Following the remarkable success of the AlphaGO series, 2019 was a booming year that witnessed significant advances in multi-agent reinforcement learning (MARL) techniques. MARL corresponds to the learning problem in a multi-agent system in which multiple agents learn simultaneously. It is an interdisciplinary domain with a long history that includes game theory, machine learning, stochastic control, psychology, and optimisation. Although MARL has achieved considerable empirical success in solving real-world games, there is a lack of a self-contained overview in the literature that elaborates the game theoretical foundations of modern MARL methods and summarises the recent advances. In fact, the majority of existing surveys are outdated and do not fully cover the recent developments since 2010. In this work, we provide a monograph on MARL that covers both the fundamentals and the latest developments in the research frontier. The goal of our monograph is to provide a self-contained assessment of the current state-of-the-art MARL techniques from a game theoretical perspective. We expect this work to serve as a stepping stone for both new researchers who are about to enter this fast-growing domain and existing domain experts who want to obtain a panoramic view and identify new directions based on recent advances.

An Overview of Multi-Agent Reinforcement Learning from Game Theoretical Perspective

Proximal Policy Optimization (PPO) is a popular on-policy reinforcement learning algorithm but is significantly less utilized than off-policy learning algorithms in multi-agent settings. This is often due the belief that on-policy methods are significantly less sample efficient than their off-policy counterparts in multi-agent problems. In this work, we investigate Multi-Agent PPO (MAPPO), a variant of PPO which is specialized for multi-agent settings. Using a 1-GPU desktop, we show that MAPPO achieves surprisingly strong performance in three popular multi-agent testbeds: the particle-world environments, the Starcraft multi-agent challenge, and the Hanabi challenge, with minimal hyperparameter tuning and without any domain-specific algorithmic modifications or architectures. In the majority of environments, we find that compared to off-policy baselines, MAPPO achieves strong results while exhibiting comparable sample efficiency. Finally, through ablation studies, we present the implementation and algorithmic factors which are most influential to MAPPO's practical performance.

The Surprising Effectiveness of MAPPO in Cooperative, Multi-Agent Games

In intelligent unmanned warehouse goods-to-man systems, the allocation of tasks has an important influence on the efficiency because of the dynamic performance of AGV robots and orders. The paper presents a hierarchical Soft Actor-Critic algorithm to solve the dynamic scheduling problem of orders picking. The method proposed is based on the classic Soft Actor-Critic and hierarchical reinforcement learning algorithm. In this paper, the model is trained at different time scales by introducing sub-goals, with the top-level learning a policy and the bottom level learning a policy to achieve the sub-goals. The actor of the controller aims to maximize expected intrinsic reward while also maximizing entropy. That is, to succeed at the sub-goals while moving as randomly as possible. Finally, experimental results for simulation experiments in different scenes show that the method can make multi-logistics AGV robots work together and improves the reward in sparse environments about 2.61 times compared to the SAC algorithm.

/pdf/a-novel-hierarchical-soft-actor-critic-algorithm-for-multi-4omu1idbf1.pdf

A Novel Hierarchical Soft Actor-Critic Algorithm for Multi-Logistics Robots Task Allocation

Ad hoc teamwork is the challenging problem of designing an autonomous agent which can adapt quickly to collaborate with previously unknown teammates. Prior work in this area has focused on closed teams in which the number of agents is fixed. In this work, we consider open teams by allowing agents of varying types to enter and leave the team without prior notification. Our proposed solution builds on graph neural networks to learn scalable agent models and value decompositions under varying team sizes, which can be jointly trained with a reinforcement learning agent using discounted returns objectives. We demonstrate empirically that our approach results in agent policies which can robustly adapt to dynamic team composition, and is able to effectively generalize to larger teams than were seen during training.

Open Ad Hoc Teamwork using Graph-based Policy Learning

Multi-agent deep reinforcement learning (MARL) suffers from a lack of commonly-used evaluation tasks and criteria, making comparisons between approaches difficult. In this work, we evaluate and compare three different classes of MARL algorithms (independent learners, centralised training with decentralised execution, and value decomposition) in a diverse range of multi-agent learning tasks. Our results show that (1) algorithm performance depends strongly on environment properties and no algorithm learns efficiently across all learning tasks; (2) independent learners often achieve equal or better performance than more complex algorithms; (3) tested algorithms struggle to solve multi-agent tasks with sparse rewards. We report detailed empirical data, including a reliability analysis, and provide insights into the limitations of the tested algorithms.

Comparative Evaluation of Multi-Agent Deep Reinforcement Learning Algorithms.

Exploration in multi-agent reinforcement learning is a challenging problem, especially in environments with sparse rewards. We propose a general method for efficient exploration by sharing experience amongst agents. Our proposed algorithm, called Shared Experience Actor-Critic (SEAC), applies experience sharing in an actor-critic framework. We evaluate SEAC in a collection of sparse-reward multi-agent environments and find that it consistently outperforms two baselines and two state-of-the-art algorithms by learning in fewer steps and converging to higher returns. In some harder environments, experience sharing makes the difference between learning to solve the task and not learning at all.

Shared Experience Actor-Critic for Multi-Agent Reinforcement Learning

/pdf/shared-experience-actor-critic-for-multi-agent-reinforcement-5zp969gevl.pdf

Successful deployment of multi-agent reinforcement learning often requires agents to adapt their behaviour. In this work, we discuss the problem of teamwork adaptation in which a team of agents needs to adapt their policies to solve novel tasks with limited fine-tuning. Motivated by the intuition that agents need to be able to identify and distinguish tasks in order to adapt their behaviour to the current task, we propose to learn multi-agent task embeddings (MATE). These task embeddings are trained using an encoder-decoder architecture optimised for reconstruction of the transition and reward functions which uniquely identify tasks. We show that a team of agents is able to adapt to novel tasks when provided with task embeddings. We propose three MATE training paradigms: independent MATE, centralised MATE, and mixed MATE which vary in the information used for the task encoding. We show that the embeddings learned by MATE identify tasks and provide useful information which agents leverage during adaptation to novel tasks.

Learning Task Embeddings for Teamwork Adaptation in Multi-Agent Reinforcement Learning

—Ensuring robust constraint satisfaction for an inﬁnite time horizon is a challenging, yet crucial task when deploying safety-critical systems. We address this issue by synthesizing robust control invariant sets for perturbed nonlinear sampled-data systems. This task can be encoded as a nonconvex program for which we propose a tailored, computationally efﬁcient successive convexiﬁcation algo- rithm. Based on the zonotopic representation of invariant sets, we obtain an updated candidate for the invariant set and the safety-preserving controller by solving a single convex program. To obtain a possibly large region of safe operation, our algorithm is designed so that the sequence of candidate invariant sets has monotonically increasing volume. We demonstrate the efﬁcacy and scalability of our approach by applying it to a broad range of nonlinear con- trol systems from the literature with up to 20 dimensions.

Lukas Schäfer

Papers

Shared Experience Actor-Critic for Multi-Agent Reinforcement Learning

Comparative Evaluation of Multi-Agent Deep Reinforcement Learning Algorithms.

Shared Experience Actor-Critic for Multi-Agent Reinforcement Learning

Learning Task Embeddings for Teamwork Adaptation in Multi-Agent Reinforcement Learning

Scalable Computation of Robust Control Invariant Sets of Nonlinear Systems