Cooperative Multi-agent Reinforcement Learning Research Articles

Interaction Pattern Disentangling for Multi-Agent Reinforcement Learning.

Deep cooperative multi-agent reinforcement learning has demonstrated its remarkable success over a wide spectrum of complex control tasks. However, recent advances in multi-agent learning mainly focus on value decomposition while leaving entity interactions still intertwined, which easily leads to over-fitting on noisy interactions between entities. In this work, we introduce a novel interactiOn Pattern disenTangling (OPT) method, to disentangle the entity interactions into interaction prototypes, each of which represents an underlying interaction pattern within a subgroup of the entities. OPT facilitates filtering the noisy interactions between irrelevant entities and thus significantly improves generalizability as well as interpretability. Specifically, OPT introduces a sparse disagreement mechanism to encourage sparsity and diversity among discovered interaction prototypes. Then the model selectively restructures these prototypes into a compact interaction pattern by an aggregator with learnable weights. To alleviate the training instability issue caused by partial observability, we propose to maximize the mutual information between the aggregation weights and the history behaviors of each agent. Experiments on single-task, multi-task and zero-shot benchmarks demonstrate that the proposed method yields results superior to the state-of-the-art counterparts.

Skill matters: Dynamic skill learning for multi-agent cooperative reinforcement learning

With the popularization of intelligence, the necessity of cooperation between intelligent machines makes the research of collaborative multi-agent reinforcement learning (MARL) more extensive. Existing approaches typically address this challenge through task decomposition of the environment or role classification of agents. However, these studies may rely on the sharing of parameters between agents, resulting in the homogeneity of agent behavior, which is not effective for complex tasks. Or training that relies on external rewards is difficult to adapt to scenarios with sparse rewards. Based on the above challenges, in this paper we propose a novel dynamic skill learning (DSL) framework for agents to learn more diverse abilities motivated by internal rewards. Specifically, the DSL has two components: (i) Dynamic skill discovery, which encourages the production of meaningful skills by exploring the environment in an unsupervised manner, using the inner product between a skill vector and a trajectory representation to generate intrinsic rewards. Meanwhile, the Lipschitz constraint of the state representation function is used to ensure the proper trajectory of the learned skills. (ii) Dynamic skill assignment, which utilizes a policy controller to assign skills to each agent based on its different trajectory latent variables. In addition, in order to avoid training instability caused by frequent changes in skill selection, we introduce a regularization term to limit skill switching between adjacent time steps. We thoroughly tested the DSL approach on two challenging benchmarks, StarCraft II and Google Research Football. Experimental results show that compared with strong benchmarks such as QMIX and RODE, DSL effectively improves performance and is more adaptable to difficult collaborative scenarios.

SQIX: QMIX Algorithm Activated by General Softmax Operator for Cooperative Multiagent Reinforcement Learning

SQIX: QMIX Algorithm Activated by General Softmax Operator for Cooperative Multiagent Reinforcement Learning

Cooperative multi-agent reinforcement learning for multi-area integrated scheduling in wafer fabs

The existing scheduling methods of wafer fabs focus on single area, achieving local optimisation while failing to realise global optimisation due to neglecting the coordination of multi-area. Therefore, it is necessary to consider the complex opposing relationships between multi-area caused by constraints such as batch processing, re-entrance, and multiple residency times within and between areas to conduct integrated scheduling and shorten the production cycle time. For this issue, this paper proposes a cooperative multi-agent reinforcement learning for multi-area integrated scheduling. Aiming at the dynamic batching and scheduling considering the dynamic arrival lots in multi-area, a multi-agent reinforcement learning algorithm is presented to learn the optimal dynamic batching and scheduling policy firstly. Subsequently, a cooperative multi-agent framework is raised to achieve the global optimisation and coordination of multi-area. Furthermore, an adaptive exploration strategy is constructed to enhance the global exploration capability of the complex solution space caused by residency time constraints and re-entrant property. Moreover, a policy share enhanced Double DQN is employed to improve the generalisation and adaptability of the multi-agent. Finally, the experiments demonstrate that the proposed integrated scheduling method has better comprehensive performance compared to the previous area-separated scheduling methods.

Regional Multi-Agent Cooperative Reinforcement Learning for City-Level Traffic Grid Signal Control

Regional Multi-Agent Cooperative Reinforcement Learning for City-Level Traffic Grid Signal Control

Multiagent Trust Region Policy Optimization.

We extend trust region policy optimization (TRPO) to cooperative multiagent reinforcement learning (MARL) for partially observable Markov games (POMGs). We show that the policy update rule in TRPO can be equivalently transformed into a distributed consensus optimization for networked agents when the agents' observation is sufficient. By using a local convexification and trust-region method, we propose a fully decentralized MARL algorithm based on a distributed alternating direction method of multipliers (ADMM). During training, agents only share local policy ratios with neighbors via a peer-to-peer communication network. Compared with traditional centralized training methods in MARL, the proposed algorithm does not need a control center to collect global information, such as global state, collective reward, or shared policy and value network parameters. Experiments on two cooperative environments demonstrate the effectiveness of the proposed method.

MuDE: Multi-agent decomposed reward-based exploration

In cooperative multi-agent reinforcement learning, agents jointly optimize a centralized value function based on the rewards shared by all agents and learn decentralized policies through value function decomposition. Although such a learning framework is considered effective, estimating individual contribution from the rewards, which is essential for learning highly cooperative behaviors, is difficult. In addition, it becomes more challenging when reinforcement and punishment, help in increasing or decreasing the specific behaviors of agents, coexist because the processes of maximizing reinforcement and minimizing punishment can often conflict in practice. This study proposes a novel exploration scheme called multi-agent decomposed reward-based exploration (MuDE), which preferably explores the action spaces associated with positive sub-rewards based on a modified reward decomposition scheme, thus effectively exploring action spaces not reachable by existing exploration schemes. We evaluate MuDE with a challenging set of StarCraft II micromanagement and modified predator–prey tasks extended to include reinforcement and punishment. The results show that MuDE accurately estimates sub-rewards and outperforms state-of-the-art approaches in both convergence speed and win rates.

Cooperative Multi-Agent Reinforcement Learning for Data Gathering in Energy-Harvesting Wireless Sensor Networks

This study introduces a novel approach to data gathering in energy-harvesting wireless sensor networks (EH-WSNs) utilizing cooperative multi-agent reinforcement learning (MARL). In addressing the challenges of efficient data collection in resource-constrained WSNs, we propose and examine a decentralized, autonomous communication framework where sensors function as individual agents. These agents employ an extended version of the Q-learning algorithm, tailored for a multi-agent setting, enabling independent learning and adaptation of their data transmission strategies. We introduce therein a specialized ϵ-p-greedy exploration method which is well suited for MAS settings. The key objective of our approach is the maximization of report flow, aligning with specific applicative goals for these networks. Our model operates under varying energy constraints and dynamic environments, with each sensor making decisions based on interactions within the network, devoid of explicit inter-sensor communication. The focus is on optimizing the frequency and efficiency of data report delivery to a central collection point, taking into account the unique attributes of each sensor. Notably, our findings present a surprising result: despite the known challenges of Q-learning in MARL, such as non-stationarity and the lack of guaranteed convergence to optimality due to multi-agent related pathologies, the cooperative nature of the MARL protocol in our study obtains high network performance. We present simulations and analyze key aspects contributing to coordination in various scenarios. A noteworthy feature of our system is its perpetual learning capability, which fosters network adaptiveness in response to changes such as sensor malfunctions or new sensor integrations. This dynamic adaptability ensures sustained and effective resource utilization, even as network conditions evolve. Our research lays grounds for learning-based WSNs and offers vital insights into the application of MARL in real-world EH-WSN scenarios, underscoring its effectiveness in navigating the intricate challenges of large-scale, resource-limited sensor networks.

Optimistic sequential multi-agent reinforcement learning with motivational communication

Centralized Training with Decentralized Execution (CTDE) is a prevalent paradigm in the field of fully cooperative Multi-Agent Reinforcement Learning (MARL). Existing algorithms often encounter two major problems: independent strategies tend to underestimate the potential value of actions, leading to the convergence on sub-optimal Nash Equilibria (NE); some communication paradigms introduce added complexity to the learning process, complicating the focus on the essential elements of the messages. To address these challenges, we propose a novel method called Optimistic Sequential Soft Actor Critic with Motivational Communication (OSSMC). The key idea of OSSMC is to utilize a greedy-driven approach to explore the potential value of individual policies, named optimistic Q-values, which serve as an upper bound for the Q-value of the current policy. We then integrate a sequential update mechanism with optimistic Q-value for agents, aiming to ensure monotonic improvement in the joint policy optimization process. Moreover, we establish motivational communication modules for each agent to disseminate motivational messages to promote cooperative behaviors. Finally, we employ a value regularization strategy from the Soft Actor Critic (SAC) method to maximize entropy and improve exploration capabilities. The performance of OSSMC was rigorously evaluated against a series of challenging benchmark sets. Empirical results demonstrate that OSSMC not only surpasses current baseline algorithms but also exhibits a more rapid convergence rate.

HyperComm: Hypergraph-based communication in multi-agent reinforcement learning

In the realm of fully cooperative multi-agent reinforcement learning (MARL), effective communication can induce implicit cooperation among agents and improve overall performance. In current communication strategies, agents are allowed to exchange local observations or latent embeddings, which can augment individual local policy inputs and mitigate uncertainty in local decision-making processes. Unfortunately, in previous communication schemes, agents may potentially receive irrelevant information, which increases training difficulty and leads to poor performance in complex settings. Furthermore, most existing works lack the consideration of the impact of small coalitions formed by agents in the multi-agent system. To address these challenges, we propose HyperComm, a novel framework that uses the hypergraph to model the multi-agent system, improving the accuracy and specificity of communication among agents. Our approach brings the concept of hypergraph for the first time in multi-agent communication for MARL. Within this framework, each agent can communicate more effectively with other agents within the same hyperedge, leading to better cooperation in environments with multiple agents. Compared to those state-of-the-art communication-based approaches, HyperComm demonstrates remarkable performance in scenarios involving a large number of agents.

Attention-Based Intrinsic Reward Mixing Network for Credit Assignment in Multiagent Reinforcement Learning

Credit assignment is a critical problem in cooperative Multi-Agent Reinforcement Learning (MARL). To address this problem, current studies mainly rely on the intrinsic reward, which is directly summed with the global reward to generate a total reward. However, such kinds of intrinsic reward functions ignore the dependence among agents and inevitably limit the adaptivity and effectiveness of MARL methods. In this paper, we propose a novel method, Attention-based Intrinsic Reward Mixing Network (AIRMN), for credit assignment in MARL. Specifically, we design a new intrinsic reward network on the basis of the attention mechanism, in order to enhance the effectiveness of teamwork. Besides, we devise a new mixing network that combines the intrinsic and extrinsic rewards in a nonlinear and dynamic manner, so as to adapt the total reward to the variation of the environment. Experimental results on the battle games of StarCraft II demonstrate that AIRMN outperforms the state-of-the-art methods in terms of the average test win rate, and also validate that AIRMN can dynamically return the precise intrinsic reward to each agent based on their contributions to the team cooperation, thereby better dealing with the credit assignment problem.

Leveraging Organizational Hierarchy to Simplify Reward Design in Cooperative Multi-agent Reinforcement Learning

The effectiveness of multi-agent reinforcement learning (MARL) hinges largely on the meticulous arrangement of objectives. Yet, conventional MARL methods might not completely harness the inherent structures present in environmental states and agent relationships for goal organization. This study is conducted within the domain of military training simulations, which are typically characterized by complexity, heterogeneity, non-stationary and doctrine-driven environments with a clear organizational hierarchy and a top-down chain of command. This research investigates the approximation and integration of the organizational hierarchy into MARL for cooperative training scenarios, with the goal of streamlining the processes of reward engineering and enhancing team coordination. In the preliminary experiments, we employed two-tiered commander-subordinate feudal hierarchical (CSFH) networks to separate the prioritized team goal and individual goals. The empirical results demonstrate that the proposed framework enhances learning efficiency. It guarantees the learning of a prioritized policy for the commander agent and encourages subordinate agents to explore areas of interest more frequently, guided by appropriate soft constraints imposed by the commander.

A Pilot Study of Observation Poisoning on Selective Reincarnation in Multi-Agent Reinforcement Learning

This research explores the vulnerability of selective reincarnation, a concept in Multi-Agent Reinforcement Learning (MARL), in response to observation poisoning attacks. Observation poisoning is an adversarial strategy that subtly manipulates an agent’s observation space, potentially leading to a misdirection in its learning process. The primary aim of this paper is to systematically evaluate the robustness of selective reincarnation in MARL systems against the subtle yet potentially debilitating effects of observation poisoning attacks. Through assessing how manipulated observation data influences MARL agents, we seek to highlight potential vulnerabilities and inform the development of more resilient MARL systems. Our experimental testbed was the widely used HalfCheetah environment, utilizing the Independent Deep Deterministic Policy Gradient algorithm within a cooperative MARL setting. We introduced a series of triggers, namely Gaussian noise addition, observation reversal, random shuffling, and scaling, into the teacher dataset of the MARL system provided to the reincarnating agents of HalfCheetah. Here, the “teacher dataset” refers to the stored experiences from previous training sessions used to accelerate the learning of reincarnating agents in MARL. This approach enabled the observation of these triggers’ significant impact on reincarnation decisions. Specifically, the reversal technique showed the most pronounced negative effect for maximum returns, with an average decrease of 38.08% in Kendall’s tau values across all the agent combinations. With random shuffling, Kendall’s tau values decreased by 17.66%. On the other hand, noise addition and scaling aligned with the original ranking by only 21.42% and 32.66%, respectively. The results, quantified by Kendall’s tau metric, indicate the fragility of the selective reincarnation process under adversarial observation poisoning. Our findings also reveal that vulnerability to observation poisoning varies significantly among different agent combinations, with some exhibiting markedly higher susceptibility than others. This investigation elucidates our understanding of selective reincarnation’s robustness against observation poisoning attacks, which is crucial for developing more secure MARL systems and also for making informed decisions about agent reincarnation.

Optimization of Energy Efficiency for Uplink mURLLC Over Multiple Cells Using Cooperative Multiagent Reinforcement Learning

Optimization of Energy Efficiency for Uplink mURLLC Over Multiple Cells Using Cooperative Multiagent Reinforcement Learning

GHQ: grouped hybrid Q-learning for cooperative heterogeneous multi-agent reinforcement learning

Previous deep multi-agent reinforcement learning (MARL) algorithms have achieved impressive results, typically in symmetric and homogeneous scenarios. However, asymmetric heterogeneous scenarios are prevalent and usually harder to solve. In this paper, the main discussion is about the cooperative heterogeneous MARL problem in asymmetric heterogeneous maps of the Starcraft Multi-Agent Challenges (SMAC) environment. Recent mainstream approaches use policy-based actor-critic algorithms to solve the heterogeneous MARL problem with various individual agent policies. However, these approaches lack formal definition and further analysis of the heterogeneity problem. Therefore, a formal definition of the Local Transition Heterogeneity (LTH) problem is first given. Then, the LTH problem in SMAC environment can be studied. To comprehensively reveal and study the LTH problem, some new asymmetric heterogeneous maps in SMAC are designed. It has been observed that baseline algorithms fail to perform well in the new maps. Then, the authors propose the Grouped Individual-Global-Max (GIGM) consistency and a novel MARL algorithm, Grouped Hybrid Q-Learning (GHQ). GHQ separates agents into several groups and keeps individual parameters for each group. To enhance cooperation between groups, GHQ maximizes the mutual information between trajectories of different groups. A novel hybrid structure for value factorization in GHQ is also proposed. Finally, experiments on the original and the new maps show the fabulous performance of GHQ compared to other state-of-the-art algorithms.

Cooperative Multiagent Reinforcement Learning Coupled With A* Search for Ship Multicabin Equipment Layout Considering Pipe Route

_ The paper presents a novel approach of cooperative multiagent reinforcement learning (CMARL) combined with A* search to address ship multicabin equipment layout considering pipe route, aiming to minimize pipe cost while considering practical requirements. The formulation is established through equipment simplification and grid marking, and A* search is utilized to value the pipe route. By designing equipment states, the equipment layout in each cabin is solved using a CMARL approach that involves three actions. Subsequently, comparative experiments were conducted on an engine room case by CMARL against genetic algorithm and single multiagent reinforcement learning methods under the condition of coupling with A* search. The parameter values for these methods were sampled using Latin Hypercube. The findings demonstrate that CMARL has superior combination properties. Keywords ship equipment layout; multicabin layout; cooperative multiagent reinforcement learning; A* search; pipe route

Optimistic Value Instructors for Cooperative Multi-Agent Reinforcement Learning

In cooperative multi-agent reinforcement learning, decentralized agents hold the promise of overcoming the combinatorial explosion of joint action space and enabling greater scalability. However, they are susceptible to a game-theoretic pathology called relative overgeneralization that shadows the optimal joint action. Although recent value-decomposition algorithms guide decentralized agents by learning a factored global action value function, the representational limitation and the inaccurate sampling of optimal joint actions during the learning process make this problem still. To address this limitation, this paper proposes a novel algorithm called Optimistic Value Instructors (OVI). The main idea behind OVI is to introduce multiple optimistic instructors into the value-decomposition paradigm, which are capable of suggesting potentially optimal joint actions and rectifying the factored global action value function to recover these optimal actions. Specifically, the instructors maintain optimistic value estimations of per-agent local actions and thus eliminate the negative effects caused by other agents' exploratory or sub-optimal non-cooperation, enabling accurate identification and suggestion of optimal joint actions. Based on the instructors' suggestions, the paper further presents two instructive constraints to rectify the factored global action value function to recover these optimal joint actions, thus overcoming the RO problem. Experimental evaluation of OVI on various cooperative multi-agent tasks demonstrates its superior performance against multiple baselines, highlighting its effectiveness.

STAS: Spatial-Temporal Return Decomposition for Solving Sparse Rewards Problems in Multi-agent Reinforcement Learning

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. 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.

Settling Decentralized Multi-Agent Coordinated Exploration by Novelty Sharing

Exploration in decentralized cooperative multi-agent reinforcement learning faces two challenges. One is that the novelty of global states is unavailable, while the novelty of local observations is biased. The other is how agents can explore in a coordinated way. To address these challenges, we propose MACE, a simple yet effective multi-agent coordinated exploration method. By communicating only local novelty, agents can take into account other agents' local novelty to approximate the global novelty. Further, we newly introduce weighted mutual information to measure the influence of one agent's action on other agents' accumulated novelty. We convert it as an intrinsic reward in hindsight to encourage agents to exert more influence on other agents' exploration and boost coordinated exploration. Empirically, we show that MACE achieves superior performance in three multi-agent environments with sparse rewards.

Learning Efficient and Robust Multi-Agent Communication via Graph Information Bottleneck

Efficient communication learning among agents has been shown crucial for cooperative multi-agent reinforcement learning (MARL), as it can promote the action coordination of agents and ultimately improve performance. Graph neural network (GNN) provide a general paradigm for communication learning, which consider agents and communication channels as nodes and edges in a graph, with the action selection corresponding to node labeling. Under such paradigm, an agent aggregates information from neighbor agents, which can reduce uncertainty in local decision-making and induce implicit action coordination. However, this communication paradigm is vulnerable to adversarial attacks and noise, and how to learn robust and efficient communication under perturbations has largely not been studied. To this end, this paper introduces a novel Multi-Agent communication mechanism via Graph Information bottleneck (MAGI), which can optimally balance the robustness and expressiveness of the message representation learned by agents. This communication mechanism is aim at learning the minimal sufficient message representation for an agent by maximizing the mutual information (MI) between the message representation and the selected action, and simultaneously constraining the MI between the message representation and the agent feature. Empirical results demonstrate that MAGI is more robust and efficient than state-of-the-art GNN-based MARL methods.