Variable Speed Limit (VSL) control acts as a promising highway traffic management strategy with worldwide deployment, which can enhance traffic safety by dynamically adjusting speed limits according to real-time traffic conditions. Most of the deployed VSL control algorithms so far are rule-based, lacking generalizability under varying and complex traffic scenarios. In this work, we propose MARVEL (Multi-Agent Reinforcement-learning for large-scale Variable spEed Limits), a novel framework for large-scale VSL control on highway corridors with real-world deployment settings. MARVEL utilizes only sensing information observable in the real world as state input and learns through a reward structure that incorporates adaptability to traffic conditions, safety, and mobility, thereby enabling multi-agent coordination. With parameter sharing among all VSL agents, the proposed framework scales to cover corridors with many agents. The policies are trained in a microscopic traffic simulation environment, focusing on a short freeway stretch with 8 VSL agents spanning 7 miles. For testing, these policies are applied to a more extensive network with 34 VSL agents spanning 17 miles of I-24 near Nashville, TN, USA. MARVEL-based method improves traffic safety by 63.4% compared to the no control scenario and enhances traffic mobility by 58.6% compared to a state-of-the-practice algorithm that has been deployed on I-24. Besides, we conduct an explainability analysis to examine the decision-making process of the agents and explore the learned policy under different traffic conditions. Finally, we test the response of the policy learned from the simulation-based experiments with real-world data collected from I-24 and illustrate its deployment capability.
In multi-agent systems, agents need to interact and collaborate with other agents in environments. Agent modeling is crucial to facilitate agent interactions and make adaptive cooperation strategies. However, it is challenging for agents to model the beliefs, behaviors, and intentions of other agents in non-stationary environment where all agent policies are learned simultaneously. In addition, the existing methods realize agent modeling through behavior cloning which assume that the local information of other agents can be accessed during execution or training. However, this assumption is infeasible in unknown scenarios characterized by unknown agents, such as competition teams, unreliable communication and federated learning due to privacy concerns. To eliminate this assumption and achieve agent modeling in unknown scenarios, Fact-based Agent modeling (FAM) method is proposed in which fact-based belief inference (FBI) network models other agents in partially observable environment only based on its local information. The reward and observation obtained by agents after taking actions are called facts, and FAM uses facts as reconstruction target to learn the policy representation of other agents through a variational autoencoder. We evaluate FAM on various Multiagent Particle Environment (MPE) and compare the results with several state-of-the-art MARL algorithms. Experimental results show that compared with baseline methods, FAM can effectively improve the efficiency of agent policy learning by making adaptive cooperation strategies in multi-agent reinforcement learning tasks, while achieving higher returns in complex competitive-cooperative mixed scenarios.
This study introduces a resource allocation framework for integrated satellite-terrestrial networks to address these challenges. The framework leverages local cache pool deployments and non-orthogonal multiple access (NOMA) to reduce time delays and improve energy efficiency. Our proposed approach utilizes a multi-agent enabled deep deterministic policy gradient algorithm (MADDPG) to optimize user association, cache design, and transmission power control, resulting in enhanced energy efficiency. The approach comprises two phases: User Association and Power Control, where users are treated as agents, and Cache Optimization, where the satellite (Bs) is considered the agent. Through extensive simulations, we demonstrate that our approach surpasses conventional single-agent deep reinforcement learning algorithms in addressing cache design and resource allocation challenges in integrated terrestrial-satellite networks. Specifically, our proposed approach achieves significantly higher energy efficiency and reduced time delays compared to existing methods.
Over the years, reinforcement learning has emerged as a popular approach to develop signal control and vehicle platooning strategies either independently or in a hierarchical way. However, jointly controlling both in real-time to alleviate traffic congestion presents new challenges, such as the inherent physical and behavioral heterogeneity between signal control and platooning, as well as coordination between them. This paper proposes an innovative solution to tackle these challenges based on heterogeneous graph multi-agent reinforcement learning and traffic theories. Our approach involves: 1) designing platoon and signal control as distinct reinforcement learning agents with their own set of observations, actions, and reward functions to optimize traffic flow; 2) designing coordination by incorporating graph neural networks within multi-agent reinforcement learning to facilitate seamless information exchange among agents on a regional scale; 3) applying alternating optimization for training, allowing agents to update their own policies and adapt to other agents' policies. We evaluate our approach through SUMO simulations, which show convergent results in terms of both travel time and fuel consumption, and superior performance compared to other adaptive signal control methods.
While Large Language Models (LLMs) have demonstrated impressive accomplishments in both reasoning and planning, their abilities in multi-agent collaborations remains largely unexplored. This study evaluates LLM-based agents in a multi-agent cooperative text game with Theory of Mind (ToM) inference tasks, comparing their performance with Multi-Agent Reinforcement Learning (MARL) and planning-based baselines. We observed evidence of emergent collaborative behaviors and high-order Theory of Mind capabilities among LLM-based agents. Our results reveal limitations in LLM-based agents' planning optimization due to systematic failures in managing long-horizon contexts and hallucination about the task state. We explore the use of explicit belief state representations to mitigate these issues, finding that it enhances task performance and the accuracy of ToM inferences for LLM-based agents.
In multi-agent reinforcement learning (MARL), ensuring robustness against unpredictable or worst-case actions by allies is crucial for real-world deployment. Existing robust MARL methods either approximate or enumerate all possible threat scenarios against worst-case adversaries, leading to computational intensity and reduced robustness. In contrast, human learning efficiently acquires robust behaviors in daily life without preparing for every possible threat. Inspired by this, we frame robust MARL as an inference problem, with worst-case robustness implicitly optimized under all threat scenarios via off-policy evaluation. Within this framework, we demonstrate that Mutual Information Regularization as Robust Regularization (MIR3) during routine training is guaranteed to maximize a lower bound on robustness, without the need for adversaries. Further insights show that MIR3 acts as an information bottleneck, preventing agents from over-reacting to others and aligning policies with robust action priors. In the presence of worst-case adversaries, our MIR3 significantly surpasses baseline methods in robustness and training efficiency while maintaining cooperative performance in StarCraft II and robot swarm control. When deploying the robot swarm control algorithm in the real world, our method also outperforms the best baseline by 14.29%.
Multi-agent reinforcement learning (MARL) plays a pivotal role in tackling real-world challenges. However, the seamless transition of trained policies from simulations to real-world requires it to be robust to various environmental uncertainties. Existing works focus on finding Nash Equilibrium or the optimal policy under uncertainty in one environment variable (i.e. action, state or reward). This is because a multi-agent system itself is highly complex and unstationary. However, in real-world situation uncertainty can occur in multiple environment variables simultaneously. This work is the first to formulate the generalised problem of robustness to multi-modal environment uncertainty in MARL. To this end, we propose a general robust training approach for multi-modal uncertainty based on curriculum learning techniques. We handle two distinct environmental uncertainty simultaneously and present extensive results across both cooperative and competitive MARL environments, demonstrating that our approach achieves state-of-the-art levels of robustness.
Recently, Multi-Agent Reinforcement Learning (MARL) has been applied to a large number of scenarios and has shown promising performance. However, existing MARL algorithms still suffer from the severe exploration problem. In this paper, we propose EXploiting POlicy Discrepancy for efficient Exploration (EXPODE), a new multi-agent exploration framework that leverages discrepancy between two different policies to enable the agents to explore the environment more efficiently. In addition, to tackle the mutual influence issue caused by the concurrent exploration of the agents, we propose three different mechanisms to coordinate the agents’ exploration by taking the information of other agents’ states and policies into account when measuring the agent-wise policy discrepancies. Experimental results on three challenging tasks, i.e., Predator Prey, StarCraft II micromanagement tasks, and Google Research Football, demonstrate that EXPODE achieves the state-of-the-art performance.
Generalization poses a significant challenge in Multi-agent Reinforcement Learning (MARL). The extent to which an agent is influenced by unseen co-players depends on the agent's policy and the specific scenario. A quantitative examination of this relationship sheds light on effectively training agents for diverse scenarios. In this study, we present the Level of Influence (LoI), a metric quantifying the interaction intensity among agents within a given scenario and environment. We observe that, generally, a more diverse set of co-play agents during training enhances the generalization performance of the ego agent; however, this improvement varies across distinct scenarios and environments. LoI proves effective in predicting these improvement disparities within specific scenarios. Furthermore, we introduce a LoI-guided resource allocation method tailored to train a set of policies for diverse scenarios under a constrained budget. Our results demonstrate that strategic resource allocation based on LoI can achieve higher performance than uniform allocation under the same computation budget.
Recent advancements in deep learning techniques have opened new possibilities for designing solutions for autonomous cyber defence. Teams of intelligent agents in computer network defence roles may reveal promising avenues to safeguard cyber and kinetic assets. In a simulated game environment, agents are evaluated on their ability to jointly mitigate attacker activity in host-based defence scenarios. Defender systems are evaluated against heuristic attackers with the goals of compromising network confidentiality, integrity, and availability. Value-based Independent Learning and Centralized Training Decentralized Execution (CTDE) cooperative Multi-Agent Reinforcement Learning (MARL) methods are compared revealing that both approaches outperform a simple multi-agent heuristic defender. This work demonstrates the ability of cooperative MARL to learn effective cyber defence tactics against varied threats.
We study multi-agent reinforcement learning (MARL) for the general-sum Markov Games (MGs) under the general function approximation. In order to find the minimum assumption for sample-efficient learning, we introduce a novel complexity measure called the Multi-Agent Decoupling Coefficient (MADC) for general-sum MGs. Using this measure, we propose the first unified algorithmic framework that ensures sample efficiency in learning Nash Equilibrium, Coarse Correlated Equilibrium, and Correlated Equilibrium for both model-based and model-free MARL problems with low MADC. We also show that our algorithm provides comparable sublinear regret to the existing works. Moreover, our algorithm combines an equilibrium-solving oracle with a single objective optimization subprocedure that solves for the regularized payoff of each deterministic joint policy, which avoids solving constrained optimization problems within data-dependent constraints (Jin et al. 2020; Wang et al. 2023) or executing sampling procedures with complex multi-objective optimization problems (Foster et al. 2023), thus being more amenable to empirical implementation.
Offline reinforcement learning (RL) suffers from the distribution shift between the offline dataset and the online environment. In multi-agent RL (MARL), this distribution shift may arise from the nonstationary opponents in the online testing who display distinct behaviors from those recorded in the offline dataset. Hence, the key to the broader deployment of offline MARL is the online adaptation to nonstationary opponents. Recent advances in foundation models, e.g., large language models, have demonstrated the generalization ability of the transformer, an emerging neural network architecture, in sequence modeling, of which offline RL is a special case. One naturally wonders \textit{whether offline-trained transformer-based RL policies adapt to nonstationary opponents online}. We propose a novel auto-regressive training to equip transformer agents with online adaptability based on the idea of self-augmented pre-conditioning. The transformer agent first learns offline to predict the opponent's action based on past observations. When deployed online, such a fictitious opponent play, referred to as the belief, is fed back to the transformer, together with other environmental feedback, to generate future actions conditional on the belief. Motivated by self-confirming equilibrium in game theory, the training loss consists of belief consistency loss, requiring the beliefs to match the opponent's actual actions and best response loss, mandating the agent to behave optimally under the belief. We evaluate the online adaptability of the proposed self-confirming transformer (SCT) in a structured environment, iterated prisoner's dilemma games, to demonstrate SCT's belief consistency and equilibrium behaviors as well as more involved multi-particle environments to showcase its superior performance against nonstationary opponents over prior transformers and offline MARL baselines.
Learning Nash equilibrium (NE) in complex zero-sum games with multi-agent reinforcement learning (MARL) can be extremely computationally expensive. Curriculum learning is an effective way to accelerate learning, but an under-explored dimension for generating a curriculum is the difficulty-to-learn of the subgames -- games induced by starting from a specific state. In this work, we present a novel subgame curriculum learning framework for zero-sum games. It adopts an adaptive initial state distribution by resetting agents to some previously visited states where they can quickly learn to improve performance. Building upon this framework, we derive a subgame selection metric that approximates the squared distance to NE values and further adopt a particle-based state sampler for subgame generation. Integrating these techniques leads to our new algorithm, Subgame Automatic Curriculum Learning (SACL), which is a realization of the subgame curriculum learning framework. SACL can be combined with any MARL algorithm such as MAPPO. Experiments in the particle-world environment and Google Research Football environment show SACL produces much stronger policies than baselines. In the challenging hide-and-seek quadrant environment, SACL produces all four emergent stages and uses only half the samples of MAPPO with self-play. The project website is at https://sites.google.com/view/sacl-rl.
Cooperation is challenging in biological systems, human societies, and multi-agent systems in general. While a group can benefit when everyone cooperates, it is tempting for each agent to act selfishly instead. Prior human studies show that people can overcome such social dilemmas while choosing interaction partners, i.e., strategic network rewiring. However, little is known about how agents, including humans, can learn about cooperation from strategic rewiring and vice versa. Here, we perform multi-agent reinforcement learning simulations in which two agents play the Prisoner's Dilemma game iteratively. Each agent has two policies: one controls whether to cooperate or defect; the other controls whether to rewire connections with another agent. This setting enables us to disentangle complex causal dynamics between cooperation and network rewiring. We find that network rewiring facilitates mutual cooperation even when one agent always offers cooperation, which is vulnerable to free-riding. We then confirm that the network-rewiring effect is exerted through agents' learning of ostracism, that is, connecting to cooperators and disconnecting from defectors. However, we also find that ostracism alone is not sufficient to make cooperation emerge. Instead, ostracism emerges from the learning of cooperation, and existing cooperation is subsequently reinforced due to the presence of ostracism. Our findings provide insights into the conditions and mechanisms necessary for the emergence of cooperation with network rewiring.
Existing multi-agent PPO algorithms lack compatibility with different types of parameter sharing when extending the theoretical guarantee of PPO to cooperative multi-agent reinforcement learning (MARL). In this paper, we propose a novel and versatile multi-agent PPO algorithm for cooperative MARL to overcome this limitation. Our approach is achieved upon the proposed full-pipeline paradigm, which establishes multiple parallel optimization pipelines by employing various equivalent decompositions of the advantage function. This procedure successfully formulates the interconnections among agents in a more general manner, i.e., the interconnections among pipelines, making it compatible with diverse types of parameter sharing. We provide a solid theoretical foundation for policy improvement and subsequently develop a practical algorithm called Full-Pipeline PPO (FP3O) by several approximations. Empirical evaluations on Multi-Agent MuJoCo and StarCraftII tasks demonstrate that FP3O outperforms other strong baselines and exhibits remarkable versatility across various parameter-sharing configurations.
Reinforcement learning generates policies based on reward functions and hyperparameters. Slight changes in these can significantly affect results. The lack of documentation and reproducibility in Reinforcement learning research makes it difficult to replicate once-deduced strategies. While previous research has identified strategies using grounded maneuvers, there is limited work in more complex environments. The agents in this study are simulated similarly to Open Al's hider and seek agents, in addition to a flying mechanism, enhancing their mobility, and expanding their range of possible actions and strategies. This added functionality improves the Hider agents to develop a chasing strategy from approximately 2 million steps to 1.6 million steps and hiders
Zero-shot coordination (ZSC) is a new cooperative multi-agent reinforcement learning (MARL) challenge that aims to train an ego agent to work with diverse, unseen partners during deployment. The significant difference between the deployment-time partners' distribution and the training partners' distribution determined by the training algorithm makes ZSC a unique out-of-distribution (OOD) generalization challenge. The potential distribution gap between evaluation and deployment-time partners leads to inadequate evaluation, which is exacerbated by the lack of appropriate evaluation metrics. In this paper, we present ZSC-Eval, the first evaluation toolkit and benchmark for ZSC algorithms. ZSC-Eval consists of: 1) Generation of evaluation partner candidates through behavior-preferring rewards to approximate deployment-time partners' distribution; 2) Selection of evaluation partners by Best-Response Diversity (BR-Div); 3) Measurement of generalization performance with various evaluation partners via the Best-Response Proximity (BR-Prox) metric. We use ZSC-Eval to benchmark ZSC algorithms in Overcooked and Google Research Football environments and get novel empirical findings. We also conduct a human experiment of current ZSC algorithms to verify the ZSC-Eval's consistency with human evaluation. ZSC-Eval is now available at https://github.com/sjtu-marl/ZSC-Eval.
The performance of existing supervised neuron segmentation methods is highly dependent on the number of accurate annotations, especially when applied to large scale electron microscopy (EM) data. By extracting semantic information from unlabeled data, self-supervised methods can improve the performance of downstream tasks, among which the mask image model (MIM) has been widely used due to its simplicity and effectiveness in recovering original information from masked images. However, due to the high degree of structural locality in EM images, as well as the existence of considerable noise, many voxels contain little discriminative information, making MIM pretraining inefficient on the neuron segmentation task. To overcome this challenge, we propose a decision-based MIM that utilizes reinforcement learning (RL) to automatically search for optimal image masking ratio and masking strategy. Due to the vast exploration space, using single-agent RL for voxel prediction is impractical. Therefore, we treat each input patch as an agent with a shared behavior policy, allowing for multi-agent collaboration. Furthermore, this multi-agent model can capture dependencies between voxels, which is beneficial for the downstream segmentation task. Experiments conducted on representative EM datasets demonstrate that our approach has a significant advantage over alternative self-supervised methods on the task of neuron segmentation. Code is available at \url{https://github.com/ydchen0806/dbMiM}.
Self-play (SP) is a popular multi-agent reinforcement learning (MARL) framework for solving competitive games, where each agent optimizes policy by treating others as part of the environment. Despite the empirical successes, the theoretical properties of SP-based methods are limited to two-player zero-sum games. However, for mixed cooperative-competitive games where agents on the same team need to cooperate with each other, we can show a simple counter-example where SP-based methods cannot converge to a global Nash equilibrium (NE) with high probability. Alternatively, Policy-Space Response Oracles (PSRO) is an iterative framework for learning NE, where the best responses w.r.t. previous policies are learned in each iteration. PSRO can be directly extended to mixed cooperative-competitive settings by jointly learning team best responses with all convergence properties unchanged. However, PSRO requires repeatedly training joint policies from scratch till convergence, which makes it hard to scale to complex games. In this work, we develop a novel algorithm, Fictitious Cross-Play (FXP), which inherits the benefits from both frameworks. FXP simultaneously trains an SP-based main policy and a counter population of best response policies. The main policy is trained by fictitious self-play and cross-play against the counter population, while the counter policies are trained as the best responses to the main policy's past versions. We validate our method in matrix games and show that FXP converges to global NEs while SP methods fail. We also conduct experiments in a gridworld domain, where FXP achieves higher Elo ratings and lower exploitabilities than baselines, and a more challenging football game, where FXP defeats SOTA models with over 94% win rate.
Traffic signal control (TSC) is a challenging problem within intelligent transportation systems and has been tackled using multi-agent reinforcement learning (MARL). While centralized approaches are often infeasible for large-scale TSC problems, decentralized approaches provide scalability but introduce new challenges, such as partial observability. Communication plays a critical role in decentralized MARL, as agents must learn to exchange information using messages to better understand the system and achieve effective coordination. Deep MARL has been used to enable inter-agent communication by learning communication protocols in a differentiable manner. However, many deep MARL communication frameworks proposed for TSC allow agents to communicate with all other agents at all times, which can add to the existing noise in the system and degrade overall performance. In this study, we propose a communication-based MARL framework for large-scale TSC. Our framework allows each agent to learn a communication policy that dictates "which" part of the message is sent "to whom". In essence, our framework enables agents to selectively choose the recipients of their messages and exchange variable length messages with them. This results in a decentralized and flexible communication mechanism in which agents can effectively use the communication channel only when necessary. We designed two networks, a synthetic $4 \times 4$ grid network and a real-world network based on the Pasubio neighborhood in Bologna. Our framework achieved the lowest network congestion compared to related methods, with agents utilizing $\sim 47-65 \%$ of the communication channel. Ablation studies further demonstrated the effectiveness of the communication policies learned within our framework.