In comparison to single-agent reinforcement learning (RL), MARL poses many challenges, chief of which is scalability [49]. Even if each agent's local state/action spaces are small, the size of the global state/action space can be large, potentially exponentially large in the number of agents, which renders many RL algorithms such as -learning not applicable. A promising approach for addressing the scalability challenge that has received attention in recent years is to exploit application-specific structures, e.g., [16, 32, 35]. A particularly important example of such a structure is a networked structure, e.g., applications in multi-agent networked systems such as social networks [6, 24], communication networks [44, 52], queueing networks [31], and smart transportation networks [51]. In these networked systems, it is often possible to exploit static, local dependency structures [1, 14, 15, 29], e.g., the fact that agents only interact with a fixed set of neighboring agents throughout the game. This sort of dependency structure often leads to scalable, distributed algorithms for optimization and control [1, 14, 29], and has proven effective for designing scalable and distributed MARL algorithms, e.g.

Many complex manipulation tasks can be decomposed into smaller sub-tasks and distributed amongst multiple robotic arms working in parallel in a shared workspace. However, efficiently motion planning for such multi-arm systems remains a challenge due to its high degrees-of-freedom (DoF) and tightly coupled workspaces. While traditional centralized motion planner [1, 2, 3, 4, 5, 6] benefit from having access to all the information a motion planner module might need, these approaches fail to scale efficiently with the number of arms in the system (the team size) because their centralized components can become the bottleneck of the system. This scalability issue has limited multi-arm applications requiring large numbers of robotic arms operating in a tight workspace or in dynamic environments with moving targets. Less explored alternatives are decentralized motion planners, which treat the multi-arm system as a multi-agent system. Here, each arm is controlled by an agent that receives as input a partial observation of the system's state and computes a motion plan for only itself. Naturally, decentralized motion planners scale efficiently, but designing such a controller for a task as complex as generic multi-arm motion planning remains a challenge. Through observations of other arms' states alone, decentralized motion planners must efficiently coordinate to avoid collisions and cooperate to collectively reach their target end-effector poses, all the while having control over only its arm (Figure 1). An ideal candidate for a multi-arm motion planner should have the following characteristics: - Scalability: The runtime should scale efficiently with the number of arms in the system.

We describe a framework for research and evaluation in Embodied AI. Our proposal is based on a canonical task: Rearrangement. A standard task can focus the development of new techniques and serve as a source of trained models that can be transferred to other settings. In the rearrangement task, the goal is to bring a given physical environment into a specified state. The goal state can be specified by object poses, by images, by a description in language, or by letting the agent experience the environment in the goal state. We characterize rearrangement scenarios along different axes and describe metrics for benchmarking rearrangement performance. To facilitate research and exploration, we present experimental testbeds of rearrangement scenarios in four different simulation environments. We anticipate that other datasets will be released and new simulation platforms will be built to support training of rearrangement agents and their deployment on physical systems.

The Persistent Monitoring (PM) problem seeks to find a set of trajectories (or controllers) for robots to persistently monitor a changing environment. Each robot has a limited field-of-view and may need to coordinate with others to ensure no point in the environment is left unmonitored for long periods of time. We model the problem such that there is a penalty that accrues every time step if a point is left unmonitored. However, the dynamics of the penalty are unknown to us. We present a Multi-Agent Reinforcement Learning (MARL) algorithm for the persistent monitoring problem. Specifically, we present a Multi-Agent Graph Attention Proximal Policy Optimization (MA-G-PPO) algorithm that takes as input the local observations of all agents combined with a low resolution global map to learn a policy for each agent. The graph attention allows agents to share their information with others leading to an effective joint policy. Our main focus is to understand how effective MARL is for the PM problem. We investigate five research questions with this broader goal. We find that MA-G-PPO is able to learn a better policy than the non-RL baseline in most cases, the effectiveness depends on agents sharing information with each other, and the policy learnt shows emergent behavior for the agents.

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. MARL 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.

A fundamental challenge in multiagent reinforcement learning is to learn beneficial behaviors in a shared environment with other agents that are also simultaneously learning. In particular, each agent perceives the environment as effectively non-stationary due to the changing policies of other agents. Moreover, each agent is itself constantly learning, leading to natural nonstationarity in the distribution of experiences encountered. In this paper, we propose a novel meta-multiagent policy gradient theorem that directly accommodates for the non-stationary policy dynamics inherent to these multiagent settings. This is achieved by modeling our gradient updates to directly consider both an agent's own non-stationary policy dynamics and the non-stationary policy dynamics of other agents interacting with it in the environment. We find that our theoretically grounded approach provides a general solution to the multiagent learning problem, which inherently combines key aspects of previous state of the art approaches on this topic. We test our method on several multiagent benchmarks and demonstrate a more efficient ability to adapt to new agents as they learn than previous related approaches across the spectrum of mixed incentive, competitive, and cooperative environments.

Multi-agent interaction is a fundamental aspect of autonomous driving in the real world. Despite more than a decade of research and development, the problem of how to competently interact with diverse road users in diverse scenarios remains largely unsolved. Learning methods have much to offer towards solving this problem. But they require a realistic multi-agent simulator that generates diverse and competent driving interactions. To meet this need, we develop a dedicated simulation platform called SMARTS (Scalable Multi-Agent RL Training School). SMARTS supports the training, accumulation, and use of diverse behavior models of road users. These are in turn used to create increasingly more realistic and diverse interactions that enable deeper and broader research on multi-agent interaction. In this paper, we describe the design goals of SMARTS, explain its basic architecture and its key features, and illustrate its use through concrete multi-agent experiments on interactive scenarios. We open-source the SMARTS platform and the associated benchmark tasks and evaluation metrics to encourage and empower research on multi-agent learning for autonomous driving. Our code is available at https://github.com/huawei-noah/SMARTS.

The purpose of this tutorial is to help individuals use the \underline{FireCommander} game environment for research applications. The FireCommander is an interactive, probabilistic joint perception-action reconnaissance environment in which a composite team of agents (e.g., robots) cooperate to fight dynamic, propagating firespots (e.g., targets). In FireCommander game, a team of agents must be tasked to optimally deal with a wildfire situation in an environment with propagating fire areas and some facilities such as houses, hospitals, power stations, etc. The team of agents can accomplish their mission by first sensing (e.g., estimating fire states), communicating the sensed fire-information among each other and then taking action to put the firespots out based on the sensed information (e.g., dropping water on estimated fire locations). The FireCommander environment can be useful for research topics spanning a wide range of applications from Reinforcement Learning (RL) and Learning from Demonstration (LfD), to Coordination, Psychology, Human-Robot Interaction (HRI) and Teaming. There are four important facets of the FireCommander environment that overall, create a non-trivial game: (1) Complex Objectives: Multi-objective Stochastic Environment, (2)Probabilistic Environment: Agents' actions result in probabilistic performance, (3) Hidden Targets: Partially Observable Environment and, (4) Uni-task Robots: Perception-only and Action-only agents. The FireCommander environment is first-of-its-kind in terms of including Perception-only and Action-only agents for coordination. It is a general multi-purpose game that can be useful in a variety of combinatorial optimization problems and stochastic games, such as applications of Reinforcement Learning (RL), Learning from Demonstration (LfD) and Inverse RL (iRL).

We study a decentralized cooperative multi-agent multi-armed bandit problem with $K$ arms and $N$ agents connected over a network. In our model, each arm's reward distribution is same for all agents, and rewards are drawn independently across agents and over time steps. In each round, agents choose an arm to play and subsequently send a message to their neighbors. The goal is to minimize cumulative regret averaged over the entire network. We propose a decentralized Bayesian multi-armed bandit framework that extends single-agent Bayesian bandit algorithms to the decentralized setting. Specifically, we study an information assimilation algorithm that can be combined with existing Bayesian algorithms, and using this, we propose a decentralized Thompson Sampling algorithm and decentralized Bayes-UCB algorithm. We analyze the decentralized Thompson Sampling algorithm under Bernoulli rewards and establish a problem-dependent upper bound on the cumulative regret. We show that regret incurred scales logarithmically over the time horizon with constants that match those of an optimal centralized agent with access to all observations across the network. Our analysis also characterizes the cumulative regret in terms of the network structure. Through extensive numerical studies, we show that our extensions of Thompson Sampling and Bayes-UCB incur lesser cumulative regret than the state-of-art algorithms inspired by the Upper Confidence Bound algorithm. We implement our proposed decentralized Thompson Sampling under gossip protocol, and over time-varying networks, where each communication link has a fixed probability of failure.

Recent studies have shown that introducing communication between agents can significantly improve overall performance in cooperative Multi-agent reinforcement learning (MARL). However, existing communication schemes often require agents to exchange an excessive number of messages at run-time under a reliable communication channel, which hinders its practicality in many real-world situations. In this paper, we present \textit{Temporal Message Control} (TMC), a simple yet effective approach for achieving succinct and robust communication in MARL. TMC applies a temporal smoothing technique to drastically reduce the amount of information exchanged between agents. Experiments show that TMC can significantly reduce inter-agent communication overhead without impacting accuracy. Furthermore, TMC demonstrates much better robustness against transmission loss than existing approaches in lossy networking environments.