Deploying teams of unmanned aerial vehicles (UAVs) to harvest data from distributed Internet of Things (IoT) devices requires efficient trajectory planning and coordination algorithms. Multi-agent reinforcement learning (MARL) has emerged as a solution, but requires extensive and costly real-world training data. To tackle this challenge, we propose a novel model-aided federated MARL algorithm to coordinate multiple UAVs on a data harvesting mission with only limited knowledge about the environment. The proposed algorithm alternates between building an environment simulation model from real-world measurements, specifically learning the radio channel characteristics and estimating unknown IoT device positions, and federated QMIX training in the simulated environment. Each UAV agent trains a local QMIX model in its simulated environment and continuously consolidates it through federated learning with other agents, accelerating the learning process. A performance comparison with standard MARL algorithms demonstrates that our proposed model-aided FedQMIX algorithm reduces the need for real-world training experiences by around three magnitudes while attaining similar data collection performance.
Recent approaches have utilized self-supervised auxiliary tasks as representation learning to improve the performance and sample efficiency of vision-based reinforcement learning algorithms in single-agent settings. However, in multi-agent reinforcement learning (MARL), these techniques face challenges because each agent only receives partial observation from an environment influenced by others, resulting in correlated observations in the agent dimension. So it is necessary to consider agent-level information in representation learning for MARL. In this paper, we propose an effective framework called \textbf{M}ulti-\textbf{A}gent \textbf{M}asked \textbf{A}ttentive \textbf{C}ontrastive \textbf{L}earning (MA2CL), which encourages learning representation to be both temporal and agent-level predictive by reconstructing the masked agent observation in latent space. Specifically, we use an attention reconstruction model for recovering and the model is trained via contrastive learning. MA2CL allows better utilization of contextual information at the agent level, facilitating the training of MARL agents for cooperation tasks. Extensive experiments demonstrate that our method significantly improves the performance and sample efficiency of different MARL algorithms and outperforms other methods in various vision-based and state-based scenarios. Our code can be found in \url{https://github.com/ustchlsong/MA2CL}
In fully cooperative multi-agent reinforcement learning (MARL) settings, environments are highly stochastic due to the partial observability of each agent and the continuously changing policies of other agents. To address the above issues, we proposed a unified framework, called DFAC, for integrating distributional RL with value function factorization methods. This framework generalizes expected value function factorization methods to enable the factorization of return distributions. To validate DFAC, we first demonstrate its ability to factorize the value functions of a simple matrix game with stochastic rewards. Then, we perform experiments on all Super Hard maps of the StarCraft Multi-Agent Challenge and six self-designed Ultra Hard maps, showing that DFAC is able to outperform a number of baselines.
Executing actions in a correlated manner is a common strategy for human coordination that often leads to better cooperation, which is also potentially beneficial for cooperative multi-agent reinforcement learning (MARL). However, the recent success of MARL relies heavily on the convenient paradigm of purely decentralized execution, where there is no action correlation among agents for scalability considerations. In this work, we introduce a Bayesian network to inaugurate correlations between agents' action selections in their joint policy. Theoretically, we establish a theoretical justification for why action dependencies are beneficial by deriving the multi-agent policy gradient formula under such a Bayesian network joint policy and proving its global convergence to Nash equilibria under tabular softmax policy parameterization in cooperative Markov games. Further, by equipping existing MARL algorithms with a recent method of differentiable directed acyclic graphs (DAGs), we develop practical algorithms to learn the context-aware Bayesian network policies in scenarios with partial observability and various difficulty. We also dynamically decrease the sparsity of the learned DAG throughout the training process, which leads to weakly or even purely independent policies for decentralized execution. Empirical results on a range of MARL benchmarks show the benefits of our approach.
This paper proposes a step toward approximate Bayesian inference in on-policy actor-critic deep reinforcement learning. It is implemented through three changes to the Asynchronous Advantage Actor-Critic (A3C) algorithm: (1) applying a ReLU function to advantage estimates, (2) spectral normalization of actor-critic weights, and (3) incorporating \emph{dropout as a Bayesian approximation}. We prove under standard assumptions that restricting policy updates to positive advantages optimizes for value by maximizing a lower bound on the value function plus an additive term. We show that the additive term is bounded proportional to the Lipschitz constant of the value function, which offers theoretical grounding for spectral normalization of critic weights. Finally, our application of dropout corresponds to approximate Bayesian inference over both the actor and critic parameters, which enables \textit{adaptive state-aware} exploration around the modes of the actor via Thompson sampling. We demonstrate significant improvements for median and interquartile mean metrics over A3C, PPO, SAC, and TD3 on the MuJoCo continuous control benchmark and improvement over PPO in the challenging ProcGen generalization benchmark.
Centralized Training with Decentralized Execution (CTDE) has recently emerged as a popular framework for cooperative Multi-Agent Reinforcement Learning (MARL), where agents can use additional global state information to guide training in a centralized way and make their own decisions only based on decentralized local policies. Despite the encouraging results achieved, CTDE makes an independence assumption on agent policies, which limits agents to adopt global cooperative information from each other during centralized training. Therefore, we argue that existing CTDE methods cannot fully utilize global information for training, leading to an inefficient joint-policy exploration and even suboptimal results. In this paper, we introduce a novel Centralized Advising and Decentralized Pruning (CADP) framework for multi-agent reinforcement learning, that not only enables an efficacious message exchange among agents during training but also guarantees the independent policies for execution. Firstly, CADP endows agents the explicit communication channel to seek and take advices from different agents for more centralized training. To further ensure the decentralized execution, we propose a smooth model pruning mechanism to progressively constraint the agent communication into a closed one without degradation in agent cooperation capability. Empirical evaluations on StarCraft II micromanagement and Google Research Football benchmarks demonstrate that the proposed framework achieves superior performance compared with the state-of-the-art counterparts. Our code will be made publicly available.
As multi-agent reinforcement learning (MARL) systems are increasingly deployed throughout society, it is imperative yet challenging for users to understand the emergent behaviors of MARL agents in complex environments. This work presents an approach for generating policy-level contrastive explanations for MARL to answer a temporal user query, which specifies a sequence of tasks completed by agents with possible cooperation. The proposed approach encodes the temporal query as a PCTL logic formula and checks if the query is feasible under a given MARL policy via probabilistic model checking. Such explanations can help reconcile discrepancies between the actual and anticipated multi-agent behaviors. The proposed approach also generates correct and complete explanations to pinpoint reasons that make a user query infeasible. We have successfully applied the proposed approach to four benchmark MARL domains (up to 9 agents in one domain). Moreover, the results of a user study show that the generated explanations significantly improve user performance and satisfaction.
Few multi-agent reinforcement learning (MARL) research on Google Research Football (GRF) focus on the 11v11 multi-agent full-game scenario and to the best of our knowledge, no open benchmark on this scenario has been released to the public. In this work, we fill the gap by providing a population-based MARL training pipeline and hyperparameter settings on multi-agent football scenario that outperforms the bot with difficulty 1.0 from scratch within 2 million steps. Our experiments serve as a reference for the expected performance of Independent Proximal Policy Optimization (IPPO), a state-of-the-art multi-agent reinforcement learning algorithm where each agent tries to maximize its own policy independently across various training configurations. Meanwhile, we open-source our training framework Light-MALib which extends the MALib codebase by distributed and asynchronized implementation with additional analytical tools for football games. Finally, we provide guidance for building strong football AI with population-based training and release diverse pretrained policies for benchmarking. The goal is to provide the community with a head start for whoever experiment their works on GRF and a simple-to-use population-based training framework for further improving their agents through self-play. The implementation is available at https://github.com/Shanghai-Digital-Brain-Laboratory/DB-Football.
Cooperative multi-agent reinforcement learning (CMARL) has shown to be promising for many real-world applications. Previous works mainly focus on improving coordination ability via solving MARL-specific challenges (e.g., non-stationarity, credit assignment, scalability), but ignore the policy perturbation issue when testing in a different environment. This issue hasn't been considered in problem formulation or efficient algorithm design. To address this issue, we firstly model the problem as a limited policy adversary Dec-POMDP (LPA-Dec-POMDP), where some coordinators from a team might accidentally and unpredictably encounter a limited number of malicious action attacks, but the regular coordinators still strive for the intended goal. Then, we propose Robust Multi-Agent Coordination via Evolutionary Generation of Auxiliary Adversarial Attackers (ROMANCE), which enables the trained policy to encounter diversified and strong auxiliary adversarial attacks during training, thus achieving high robustness under various policy perturbations. Concretely, to avoid the ego-system overfitting to a specific attacker, we maintain a set of attackers, which is optimized to guarantee the attackers high attacking quality and behavior diversity. The goal of quality is to minimize the ego-system coordination effect, and a novel diversity regularizer based on sparse action is applied to diversify the behaviors among attackers. The ego-system is then paired with a population of attackers selected from the maintained attacker set, and alternately trained against the constantly evolving attackers. Extensive experiments on multiple scenarios from SMAC indicate our ROMANCE provides comparable or better robustness and generalization ability than other baselines.
There is a lack of standard benchmarks for Multi-Agent Reinforcement Learning (MARL) algorithms. The Starcraft Multi-Agent Challenge (SMAC) has been widely used in MARL research, but is built on top of a heavy, closed-source computer game, StarCraft II. Thus, SMAC is computationally expensive and requires knowledge and the use of proprietary tools specific to the game for any meaningful alteration or contribution to the environment. We introduce SMAClite -- a challenge based on SMAC that is both decoupled from Starcraft II and open-source, along with a framework which makes it possible to create new content for SMAClite without any special knowledge. We conduct experiments to show that SMAClite is equivalent to SMAC, by training MARL algorithms on SMAClite and reproducing SMAC results. We then show that SMAClite outperforms SMAC in both runtime speed and memory.
Policy optimization methods with function approximation are widely used in multi-agent reinforcement learning. However, it remains elusive how to design such algorithms with statistical guarantees. Leveraging a multi-agent performance difference lemma that characterizes the landscape of multi-agent policy optimization, we find that the localized action value function serves as an ideal descent direction for each local policy. Motivated by the observation, we present a multi-agent PPO algorithm in which the local policy of each agent is updated similarly to vanilla PPO. We prove that with standard regularity conditions on the Markov game and problem-dependent quantities, our algorithm converges to the globally optimal policy at a sublinear rate. We extend our algorithm to the off-policy setting and introduce pessimism to policy evaluation, which aligns with experiments. To our knowledge, this is the first provably convergent multi-agent PPO algorithm in cooperative Markov games.
Evolutionary science provides evidence that diversity confers resilience in natural systems. Yet, traditional multi-agent reinforcement learning techniques commonly enforce homogeneity to increase training sample efficiency. When a system of learning agents is not constrained to homogeneous policies, individuals may develop diverse behaviors, resulting in emergent complementarity that benefits the system. Despite this, there is a surprising lack of tools that quantify behavioral diversity. Such techniques would pave the way towards understanding the impact of diversity in collective artificial intelligence and enabling its control. In this paper, we introduce System Neural Diversity (SND): a measure of behavioral heterogeneity in multi-agent systems. We discuss and prove its theoretical properties, and compare it with alternate, state-of-the-art behavioral diversity metrics used in the robotics domain. Through simulations of a variety of cooperative multi-robot tasks, we show how our metric constitutes an important tool that enables measurement and control of behavioral heterogeneity. In dynamic tasks, where the problem is affected by repeated disturbances during training, we show that SND allows us to measure latent resilience skills acquired by the agents, while other proxies, such as task performance (reward), fail to. Finally, we show how the metric can be employed to control diversity, allowing us to enforce a desired heterogeneity set-point or range. We demonstrate how this paradigm can be used to bootstrap the exploration phase, finding optimal policies faster, thus enabling novel and more efficient MARL paradigms.
Multi-agent Reinforcement learning (MARL) studies the behaviour of multiple learning agents that coexist in a shared environment. MARL is more challenging than single-agent RL because it involves more complex learning dynamics: the observations and rewards of each agent are functions of all other agents. In the context of MARL, Real-Time Strategy (RTS) games represent very challenging environments where multiple players interact simultaneously and control many units of different natures all at once. In fact, RTS games are so challenging for the current RL methods, that just being able to tackle them with RL is interesting. This project provides the end-to-end experience of applying RL in the Lux AI v2 Kaggle competition, where competitors design agents to control variable-sized fleets of units and tackle a multi-variable optimization, resource gathering, and allocation problem in a 1v1 scenario against other competitors. We use a centralized approach for training the RL agents, and report multiple design decisions along the process. We provide the source code of the project: https://github.com/roger-creus/centralized-control-lux.
Traditional multi-agent reinforcement learning algorithms are difficultly applied in a large-scale multi-agent environment. The introduction of mean field theory has enhanced the scalability of multi-agent reinforcement learning in recent years. This paper considers partially observable multi-agent reinforcement learning (MARL), where each agent can only observe other agents within a fixed range. This partial observability affects the agent's ability to assess the quality of the actions of surrounding agents. This paper focuses on developing a method to capture more effective information from local observations in order to select more effective actions. Previous work in this field employs probability distributions or weighted mean field to update the average actions of neighborhood agents, but it does not fully consider the feature information of surrounding neighbors and leads to a local optimum. In this paper, we propose a novel multi-agent reinforcement learning algorithm, Partially Observable Mean Field Multi-Agent Reinforcement Learning based on Graph-Attention (GAMFQ) to remedy this flaw. GAMFQ uses a graph attention module and a mean field module to describe how an agent is influenced by the actions of other agents at each time step. This graph attention module consists of a graph attention encoder and a differentiable attention mechanism, and this mechanism outputs a dynamic graph to represent the effectiveness of neighborhood agents against central agents. The mean-field module approximates the effect of a neighborhood agent on a central agent as the average effect of effective neighborhood agents. Experiments show that GAMFQ outperforms baselines including the state-of-the-art partially observable mean-field reinforcement learning algorithms. The code for this paper is here \url{https://github.com/yangmin32/GPMF}.
Recent research in multi-agent reinforcement learning (MARL) has shown success in learning social behavior and cooperation. Social dilemmas between agents in mixed-sum settings have been studied extensively, but there is little research into social dilemmas in fullycooperative settings, where agents have no prospect of gaining reward at another agent's expense. While fully-aligned interests are conducive to cooperation between agents, they do not guarantee it. We propose a measure of "stubbornness" between agents that aims to capture the human social behavior from which it takes its name: a disagreement that is gradually escalating and potentially disastrous. We would like to promote research into the tendency of agents to be stubborn, the reactions of counterpart agents, and the resulting social dynamics. In this paper we present Stubborn, an environment for evaluating stubbornness between agents with fully-aligned incentives. In our preliminary results, the agents learn to use their partner's stubbornness as a signal for improving the choices that they make in the environment.
The necessity for cooperation among intelligent machines has popularised cooperative multi-agent reinforcement learning (MARL) in AI research. However, many research endeavours heavily rely on parameter sharing among agents, which confines them to only homogeneous-agent setting and leads to training instability and lack of convergence guarantees. To achieve effective cooperation in the general heterogeneous-agent setting, we propose Heterogeneous-Agent Reinforcement Learning (HARL) algorithms that resolve the aforementioned issues. Central to our findings are the multi-agent advantage decomposition lemma and the sequential update scheme. Based on these, we develop the provably correct Heterogeneous-Agent Trust Region Learning (HATRL), and derive HATRPO and HAPPO by tractable approximations. Furthermore, we discover a novel framework named Heterogeneous-Agent Mirror Learning (HAML), which strengthens theoretical guarantees for HATRPO and HAPPO and provides a general template for cooperative MARL algorithmic designs. We prove that all algorithms derived from HAML inherently enjoy monotonic improvement of joint return and convergence to Nash Equilibrium. As its natural outcome, HAML validates more novel algorithms in addition to HATRPO and HAPPO, including HAA2C, HADDPG, and HATD3, which generally outperform their existing MA-counterparts. We comprehensively test HARL algorithms on six challenging benchmarks and demonstrate their superior effectiveness and stability for coordinating heterogeneous agents compared to strong baselines such as MAPPO and QMIX.
Centralized Training with Decentralized Execution (CTDE) has been proven to be an effective paradigm in cooperative multi-agent reinforcement learning (MARL). One of the major challenges is credit assignment, which aims to credit agents by their contributions. While prior studies have shown great success, their methods typically fail to work in episodic reinforcement learning scenarios where global rewards are revealed only at the end of the episode. They lack the functionality to model complicated relations of the delayed global reward in the temporal dimension and suffer from inefficiencies. To tackle this, we introduce Spatial-Temporal Attention with Shapley (STAS), a novel method that learns credit assignment in both temporal and spatial dimensions. It first decomposes the global return back to each time step, then utilizes the Shapley Value to redistribute the individual payoff from the decomposed global reward. To mitigate the computational complexity of the Shapley Value, we introduce an approximation of marginal contribution and utilize Monte Carlo sampling to estimate it. We evaluate our method on an Alice & Bob example and MPE environments across different scenarios. Our results demonstrate that our method effectively assigns spatial-temporal credit, outperforming all state-of-the-art baselines.
This paper presents a decentralized Multi-Agent Reinforcement Learning (MARL) approach to an incentive-based Demand Response (DR) program, which aims to maintain the capacity limits of the electricity grid and prevent grid congestion by financially incentivizing residential consumers to reduce their energy consumption. The proposed approach addresses the key challenge of coordinating heterogeneous preferences and requirements from multiple participants while preserving their privacy and minimizing financial costs for the aggregator. The participant agents use a novel Disjunctively Constrained Knapsack Problem optimization to curtail or shift the requested household appliances based on the selected demand reduction. Through case studies with electricity data from $25$ households, the proposed approach effectively reduced energy consumption's Peak-to-Average ratio (PAR) by $14.48$% compared to the original PAR while fully preserving participant privacy. This approach has the potential to significantly improve the efficiency and reliability of the electricity grid, making it an important contribution to the management of renewable energy resources and the growing electricity demand.
Rayleigh-B\'enard convection (RBC) is a recurrent phenomenon in several industrial and geoscience flows and a well-studied system from a fundamental fluid-mechanics viewpoint. However, controlling RBC, for example by modulating the spatial distribution of the bottom-plate heating in the canonical RBC configuration, remains a challenging topic for classical control-theory methods. In the present work, we apply deep reinforcement learning (DRL) for controlling RBC. We show that effective RBC control can be obtained by leveraging invariant multi-agent reinforcement learning (MARL), which takes advantage of the locality and translational invariance inherent to RBC flows inside wide channels. The MARL framework applied to RBC allows for an increase in the number of control segments without encountering the curse of dimensionality that would result from a naive increase in the DRL action-size dimension. This is made possible by the MARL ability for re-using the knowledge generated in different parts of the RBC domain. We show in a case study that MARL DRL is able to discover an advanced control strategy that destabilizes the spontaneous RBC double-cell pattern, changes the topology of RBC by coalescing adjacent convection cells, and actively controls the resulting coalesced cell to bring it to a new stable configuration. This modified flow configuration results in reduced convective heat transfer, which is beneficial in several industrial processes. Therefore, our work both shows the potential of MARL DRL for controlling large RBC systems, as well as demonstrates the possibility for DRL to discover strategies that move the RBC configuration between different topological configurations, yielding desirable heat-transfer characteristics. These results are useful for both gaining further understanding of the intrinsic properties of RBC, as well as for developing industrial applications.
Role-based learning is a promising approach to improving the performance of Multi-Agent Reinforcement Learning (MARL). Nevertheless, without manual assistance, current role-based methods cannot guarantee stably discovering a set of roles to effectively decompose a complex task, as they assume either a predefined role structure or practical experience for selecting hyperparameters. In this article, we propose a mathematical Structural Information principles-based Role Discovery method, namely SIRD, and then present a SIRD optimizing MARL framework, namely SR-MARL, for multi-agent collaboration. The SIRD transforms role discovery into a hierarchical action space clustering. Specifically, the SIRD consists of structuralization, sparsification, and optimization modules, where an optimal encoding tree is generated to perform abstracting to discover roles. The SIRD is agnostic to specific MARL algorithms and flexibly integrated with various value function factorization approaches. Empirical evaluations on the StarCraft II micromanagement benchmark demonstrate that, compared with state-of-the-art MARL algorithms, the SR-MARL framework improves the average test win rate by 0.17%, 6.08%, and 3.24%, and reduces the deviation by 16.67%, 30.80%, and 66.30%, under easy, hard, and super hard scenarios.