Multi-agent reinforcement learning (MARL) has exploded in popularity in recent years. Many approaches have been developed but they can be divided into three main types: centralized training and execution (CTE), centralized training for decentralized execution (CTDE), and Decentralized training and execution (DTE). CTDE methods are the most common as they can use centralized information during training but execute in a decentralized manner -- using only information available to that agent during execution. CTDE is the only paradigm that requires a separate training phase where any available information (e.g., other agent policies, underlying states) can be used. As a result, they can be more scalable than CTE methods, do not require communication during execution, and can often perform well. CTDE fits most naturally with the cooperative case, but can be potentially applied in competitive or mixed settings depending on what information is assumed to be observed. This text is an introduction to CTDE in cooperative MARL. It is meant to explain the setting, basic concepts, and common methods. It does not cover all work in CTDE MARL as the subarea is quite extensive. I have included work that I believe is important for understanding the main concepts in the subarea and apologize to those that I have omitted.
The field of emergent language represents a novel area of research within the domain of artificial intelligence, particularly within the context of multi-agent reinforcement learning. Although the concept of studying language emergence is not new, early approaches were primarily concerned with explaining human language formation, with little consideration given to its potential utility for artificial agents. In contrast, studies based on reinforcement learning aim to develop communicative capabilities in agents that are comparable to or even superior to human language. Thus, they extend beyond the learned statistical representations that are common in natural language processing research. This gives rise to a number of fundamental questions, from the prerequisites for language emergence to the criteria for measuring its success. This paper addresses these questions by providing a comprehensive review of 181 scientific publications on emergent language in artificial intelligence. Its objective is to serve as a reference for researchers interested in or proficient in the field. Consequently, the main contributions are the definition and overview of the prevailing terminology, the analysis of existing evaluation methods and metrics, and the description of the identified research gaps.
Police patrol units need to split their time between performing preventive patrol and being dispatched to serve emergency incidents. In the existing literature, patrol and dispatch decisions are often studied separately. We consider joint optimization of these two decisions to improve police operations efficiency and reduce response time to emergency calls. Methodology/results: We propose a novel method for jointly optimizing multi-agent patrol and dispatch to learn policies yielding rapid response times. Our method treats each patroller as an independent Q-learner (agent) with a shared deep Q-network that represents the state-action values. The dispatching decisions are chosen using mixed-integer programming and value function approximation from combinatorial action spaces. We demonstrate that this heterogeneous multi-agent reinforcement learning approach is capable of learning joint policies that outperform those optimized for patrol or dispatch alone. Managerial Implications: Policies jointly optimized for patrol and dispatch can lead to more effective service while targeting demonstrably flexible objectives, such as those encouraging efficiency and equity in response.
In this paper, we study the shortest path problem (SPP) with multiple source-destination pairs (MSD), namely MSD-SPP, to minimize average travel time of all shortest paths. The inherent traffic capacity limits within a road network contributes to the competition among vehicles. Multi-agent reinforcement learning (MARL) model cannot offer effective and efficient path planning cooperation due to the asynchronous decision making setting in MSD-SPP, where vehicles (a.k.a agents) cannot simultaneously complete routing actions in the previous time step. To tackle the efficiency issue, we propose to divide an entire road network into multiple sub-graphs and subsequently execute a two-stage process of inter-region and intra-region route planning. To address the asynchronous issue, in the proposed asyn-MARL framework, we first design a global state, which exploits a low-dimensional vector to implicitly represent the joint observations and actions of multi-agents. Then we develop a novel trajectory collection mechanism to decrease the redundancy in training trajectories. Additionally, we design a novel actor network to facilitate the cooperation among vehicles towards the same or close destinations and a reachability graph aimed at preventing infinite loops in routing paths. On both synthetic and real road networks, our evaluation result demonstrates that our approach outperforms state-of-the-art planning approaches.
We initiate the study of Preference-Based Multi-Agent Reinforcement Learning (PbMARL), exploring both theoretical foundations and empirical validations. We define the task as identifying the Nash equilibrium from a preference-only offline dataset in general-sum games, a problem marked by the challenge of sparse feedback signals. Our theory establishes the upper complexity bounds for Nash Equilibrium in effective PbMARL, demonstrating that single-policy coverage is inadequate and highlighting the importance of unilateral dataset coverage. These theoretical insights are verified through comprehensive experiments. To enhance the practical performance, we further introduce two algorithmic techniques. (1) We propose a Mean Squared Error (MSE) regularization along the time axis to achieve a more uniform reward distribution and improve reward learning outcomes. (2) We propose an additional penalty based on the distribution of the dataset to incorporate pessimism, improving stability and effectiveness during training. Our findings underscore the multifaceted approach required for PbMARL, paving the way for effective preference-based multi-agent systems.
Robotics can help address the growing worker shortage challenge of the manufacturing industry. As such, machine tending is a task collaborative robots can tackle that can also highly boost productivity. Nevertheless, existing robotics systems deployed in that sector rely on a fixed single-arm setup, whereas mobile robots can provide more flexibility and scalability. In this work, we introduce a multi-agent multi-machine tending learning framework by mobile robots based on Multi-agent Reinforcement Learning (MARL) techniques with the design of a suitable observation and reward. Moreover, an attention-based encoding mechanism is developed and integrated into Multi-agent Proximal Policy Optimization (MAPPO) algorithm to boost its performance for machine tending scenarios. Our model (AB-MAPPO) outperformed MAPPO in this new challenging scenario in terms of task success, safety, and resources utilization. Furthermore, we provided an extensive ablation study to support our various design decisions.
Value factorization is a popular paradigm for designing scalable multi-agent reinforcement learning algorithms. However, current factorization methods make choices without full justification that may limit their performance. For example, the theory in prior work uses stateless (i.e., history) functions, while the practical implementations use state information -- making the motivating theory a mismatch for the implementation. Also, methods have built off of previous approaches, inheriting their architectures without exploring other, potentially better ones. To address these concerns, we formally analyze the theory of using the state instead of the history in current methods -- reconnecting theory and practice. We then introduce DuelMIX, a factorization algorithm that learns distinct per-agent utility estimators to improve performance and achieve full expressiveness. Experiments on StarCraft II micromanagement and Box Pushing tasks demonstrate the benefits of our intuitions.
Mean-field games (MFG) have become significant tools for solving large-scale multi-agent reinforcement learning problems under symmetry. However, the assumption of exact symmetry limits the applicability of MFGs, as real-world scenarios often feature inherent heterogeneity. Furthermore, most works on MFG assume access to a known MFG model, which might not be readily available for real-world finite-agent games. In this work, we broaden the applicability of MFGs by providing a methodology to extend any finite-player, possibly asymmetric, game to an "induced MFG". First, we prove that $N$-player dynamic games can be symmetrized and smoothly extended to the infinite-player continuum via explicit Kirszbraun extensions. Next, we propose the notion of $\alpha,\beta$-symmetric games, a new class of dynamic population games that incorporate approximate permutation invariance. For $\alpha,\beta$-symmetric games, we establish explicit approximation bounds, demonstrating that a Nash policy of the induced MFG is an approximate Nash of the $N$-player dynamic game. We show that TD learning converges up to a small bias using trajectories of the $N$-player game with finite-sample guarantees, permitting symmetrized learning without building an explicit MFG model. Finally, for certain games satisfying monotonicity, we prove a sample complexity of $\widetilde{\mathcal{O}}(\varepsilon^{-6})$ for the $N$-agent game to learn an $\varepsilon$-Nash up to symmetrization bias. Our theory is supported by evaluations on MARL benchmarks with thousands of agents.
Centralized Training for Decentralized Execution where agents are trained offline in a centralized fashion and execute online in a decentralized manner, has become a popular approach in Multi-Agent Reinforcement Learning (MARL). In particular, it has become popular to develop actor-critic methods that train decentralized actors with a centralized critic where the centralized critic is allowed access global information of the entire system, including the true system state. Such centralized critics are possible given offline information and are not used for online execution. While these methods perform well in a number of domains and have become a de facto standard in MARL, using a centralized critic in this context has yet to be sufficiently analyzed theoretically or empirically. In this paper, we therefore formally analyze centralized and decentralized critic approaches, and analyze the effect of using state-based critics in partially observable environments. We derive theories contrary to the common intuition: critic centralization is not strictly beneficial, and using state values can be harmful. We further prove that, in particular, state-based critics can introduce unexpected bias and variance compared to history-based critics. Finally, we demonstrate how the theory applies in practice by comparing different forms of critics on a wide range of common multi-agent benchmarks. The experiments show practical issues such as the difficulty of representation learning with partial observability, which highlights why the theoretical problems are often overlooked in the literature.
In multi-agent reinforcement learning (MARL), achieving multi-task generalization to diverse agents and objectives presents significant challenges. Existing online MARL algorithms primarily focus on single-task performance, but their lack of multi-task generalization capabilities typically results in substantial computational waste and limited real-life applicability. Meanwhile, existing offline multi-task MARL approaches are heavily dependent on data quality, often resulting in poor performance on unseen tasks. In this paper, we introduce HyGen, a novel hybrid MARL framework, Hybrid Training for Enhanced Multi-Task Generalization, which integrates online and offline learning to ensure both multi-task generalization and training efficiency. Specifically, our framework extracts potential general skills from offline multi-task datasets. We then train policies to select the optimal skills under the centralized training and decentralized execution paradigm (CTDE). During this stage, we utilize a replay buffer that integrates both offline data and online interactions. We empirically demonstrate that our framework effectively extracts and refines general skills, yielding impressive generalization to unseen tasks. Comparative analyses on the StarCraft multi-agent challenge show that HyGen outperforms a wide range of existing solely online and offline methods.
Offline multi-agent reinforcement learning (MARL) is increasingly recognized as crucial for effectively deploying RL algorithms in environments where real-time interaction is impractical, risky, or costly. In the offline setting, learning from a static dataset of past interactions allows for the development of robust and safe policies without the need for live data collection, which can be fraught with challenges. Building on this foundational importance, we present EAQ, Episodes Augmentation guided by Q-total loss, a novel approach for offline MARL framework utilizing diffusion models. EAQ integrates the Q-total function directly into the diffusion model as a guidance to maximize the global returns in an episode, eliminating the need for separate training. Our focus primarily lies on cooperative scenarios, where agents are required to act collectively towards achieving a shared goal-essentially, maximizing global returns. Consequently, we demonstrate that our episodes augmentation in a collaborative manner significantly boosts offline MARL algorithm compared to the original dataset, improving the normalized return by +17.3% and +12.9% for medium and poor behavioral policies in SMAC simulator, respectively.
We consider a dense small cell (DSC) network where multi-antenna small cell base stations (SBSs) transmit data to single-antenna users over a shared frequency band. To enhance capacity, a state-of-the-art technique known as noncoherent joint transmission (JT) is applied, enabling users to receive data from multiple coordinated SBSs. However, the sum rate maximization problem with noncoherent JT is inherently nonconvex and NP-hard. While existing optimization-based noncoherent JT algorithms can provide near-optimal performance, they require global channel state information (CSI) and multiple iterations, which makes them difficult to be implemeted in DSC networks.To overcome these challenges, we first prove that the optimal beamforming structure is the same for both the power minimization problem and the sum rate maximization problem, and then mathematically derive the optimal beamforming structure for both problems by solving the power minimization problem.The optimal beamforming structure can effectively reduces the variable dimensions.By exploiting the optimal beamforming structure, we propose a deep deterministic policy gradient-based distributed noncoherent JT scheme to maximize the system sum rate.In the proposed scheme, each SBS utilizes global information for training and uses local CSI to determine beamforming vectors. Simulation results demonstrate that the proposed scheme achieves comparable performance with considerably lower computational complexity and information overhead compared to centralized iterative optimization-based techniques, making it more attractive for practical deployment.
The advancement of Offline Reinforcement Learning (RL) and Offline Multi-Agent Reinforcement Learning (MARL) critically depends on the availability of high-quality, pre-collected offline datasets that represent real-world complexities and practical applications. However, existing datasets often fall short in their simplicity and lack of realism. To address this gap, we propose Hokoff, a comprehensive set of pre-collected datasets that covers both offline RL and offline MARL, accompanied by a robust framework, to facilitate further research. This data is derived from Honor of Kings, a recognized Multiplayer Online Battle Arena (MOBA) game known for its intricate nature, closely resembling real-life situations. Utilizing this framework, we benchmark a variety of offline RL and offline MARL algorithms. We also introduce a novel baseline algorithm tailored for the inherent hierarchical action space of the game. We reveal the incompetency of current offline RL approaches in handling task complexity, generalization and multi-task learning.
Reinforcement Learning (RL) is a potent tool for sequential decision-making and has achieved performance surpassing human capabilities across many challenging real-world tasks. As the extension of RL in the multi-agent system domain, multi-agent RL (MARL) not only need to learn the control policy but also requires consideration regarding interactions with all other agents in the environment, mutual influences among different system components, and the distribution of computational resources. This augments the complexity of algorithmic design and poses higher requirements on computational resources. Simultaneously, simulators are crucial to obtain realistic data, which is the fundamentals of RL. In this paper, we first propose a series of metrics of simulators and summarize the features of existing benchmarks. Second, to ease comprehension, we recall the foundational knowledge and then synthesize the recently advanced studies of MARL-related autonomous driving and intelligent transportation systems. Specifically, we examine their environmental modeling, state representation, perception units, and algorithm design. Conclusively, we discuss open challenges as well as prospects and opportunities. We hope this paper can help the researchers integrate MARL technologies and trigger more insightful ideas toward the intelligent and autonomous driving.
In partially observable multi-agent systems, agents typically only have access to local observations. This severely hinders their ability to make precise decisions, particularly during decentralized execution. To alleviate this problem and inspired by image outpainting, we propose State Inference with Diffusion Models (SIDIFF), which uses diffusion models to reconstruct the original global state based solely on local observations. SIDIFF consists of a state generator and a state extractor, which allow agents to choose suitable actions by considering both the reconstructed global state and local observations. In addition, SIDIFF can be effortlessly incorporated into current multi-agent reinforcement learning algorithms to improve their performance. Finally, we evaluated SIDIFF on different experimental platforms, including Multi-Agent Battle City (MABC), a novel and flexible multi-agent reinforcement learning environment we developed. SIDIFF achieved desirable results and outperformed other popular algorithms.
This paper proposes a Semi-Centralized Multi-Agent Reinforcement Learning (SCMARL) approach for irrigation scheduling in spatially variable agricultural fields, where management zones address spatial variability. The SCMARL framework is hierarchical in nature, with a centralized coordinator agent at the top level and decentralized local agents at the second level. The coordinator agent makes daily binary irrigation decisions based on field-wide conditions, which are communicated to the local agents. Local agents determine appropriate irrigation amounts for specific management zones using local conditions. The framework employs state augmentation approach to handle non-stationarity in the local agents' environments. An extensive evaluation on a large-scale field in Lethbridge, Canada, compares the SCMARL approach with a learning-based multi-agent model predictive control scheduling approach, highlighting its enhanced performance, resulting in water conservation and improved Irrigation Water Use Efficiency (IWUE). Notably, the proposed approach achieved a 4.0% savings in irrigation water while enhancing the IWUE by 6.3%.
Markov Potential Games (MPGs) form an important sub-class of Markov games, which are a common framework to model multi-agent reinforcement learning problems. In particular, MPGs include as a special case the identical-interest setting where all the agents share the same reward function. Scaling the performance of Nash equilibrium learning algorithms to a large number of agents is crucial for multi-agent systems. To address this important challenge, we focus on the independent learning setting where agents can only have access to their local information to update their own policy. In prior work on MPGs, the iteration complexity for obtaining $\epsilon$-Nash regret scales linearly with the number of agents $N$. In this work, we investigate the iteration complexity of an independent policy mirror descent (PMD) algorithm for MPGs. We show that PMD with KL regularization, also known as natural policy gradient, enjoys a better $\sqrt{N}$ dependence on the number of agents, improving over PMD with Euclidean regularization and prior work. Furthermore, the iteration complexity is also independent of the sizes of the agents' action spaces.
Machine learning has driven an exponential increase in computational demand, leading to massive data centers that consume significant amounts of energy and contribute to climate change. This makes sustainable data center control a priority. In this paper, we introduce SustainDC, a set of Python environments for benchmarking multi-agent reinforcement learning (MARL) algorithms for data centers (DC). SustainDC supports custom DC configurations and tasks such as workload scheduling, cooling optimization, and auxiliary battery management, with multiple agents managing these operations while accounting for the effects of each other. We evaluate various MARL algorithms on SustainDC, showing their performance across diverse DC designs, locations, weather conditions, grid carbon intensity, and workload requirements. Our results highlight significant opportunities for improvement of data center operations using MARL algorithms. Given the increasing use of DC due to AI, SustainDC provides a crucial platform for the development and benchmarking of advanced algorithms essential for achieving sustainable computing and addressing other heterogeneous real-world challenges.
This paper introduces an open-source, decentralized framework named SigmaRL, designed to enhance both sample efficiency and generalization of multi-agent Reinforcement Learning (RL) for motion planning of connected and automated vehicles. Most RL agents exhibit a limited capacity to generalize, often focusing narrowly on specific scenarios, and are usually evaluated in similar or even the same scenarios seen during training. Various methods have been proposed to address these challenges, including experience replay and regularization. However, how observation design in RL affects sample efficiency and generalization remains an under-explored area. We address this gap by proposing five strategies to design information-dense observations, focusing on general features that are applicable to most traffic scenarios. We train our RL agents using these strategies on an intersection and evaluate their generalization through numerical experiments across completely unseen traffic scenarios, including a new intersection, an on-ramp, and a roundabout. Incorporating these information-dense observations reduces training times to under one hour on a single CPU, and the evaluation results reveal that our RL agents can effectively zero-shot generalize. Code: github.com/bassamlab/SigmaRL
In multi-agent cooperative tasks, the presence of heterogeneous agents is familiar. Compared to cooperation among homogeneous agents, collaboration requires considering the best-suited sub-tasks for each agent. However, the operation of multi-agent systems often involves a large amount of complex interaction information, making it more challenging to learn heterogeneous strategies. Related multi-agent reinforcement learning methods sometimes use grouping mechanisms to form smaller cooperative groups or leverage prior domain knowledge to learn strategies for different roles. In contrast, agents should learn deeper role features without relying on additional information. Therefore, we propose QTypeMix, which divides the value decomposition process into homogeneous and heterogeneous stages. QTypeMix learns to extract type features from local historical observations through the TE loss. In addition, we introduce advanced network structures containing attention mechanisms and hypernets to enhance the representation capability and achieve the value decomposition process. The results of testing the proposed method on 14 maps from SMAC and SMACv2 show that QTypeMix achieves state-of-the-art performance in tasks of varying difficulty.