Cooperation in multi-agent learning (MAL) is a topic at the intersection of numerous disciplines, including game theory, economics, social sciences, and evolutionary biology. Research in this area aims to understand both how agents can coordinate effectively when goals are aligned and how they may cooperate in settings where gains from working together are possible but possibilities for conflict abound. In this paper we provide an overview of the fundamental concepts, problem settings and algorithms of multi-agent learning. This encompasses reinforcement learning, multi-agent sequential decision-making, challenges associated with multi-agent cooperation, and a comprehensive review of recent progress, along with an evaluation of relevant metrics. Finally we discuss open challenges in the field with the aim of inspiring new avenues for research.

Reinforcement learning (RL) makes an agent learn from trial-and-error experiences gathered during the interaction with the environment. Recently, offline RL has become a popular RL paradigm because it saves the interactions with environments. In offline RL, data providers share large pre-collected datasets, and others can train high-quality agents without interacting with the environments. This paradigm has demonstrated effectiveness in critical tasks like robot control, autonomous driving, etc. However, less attention is paid to investigating the security threats to the offline RL system. This paper focuses on backdoor attacks, where some perturbations are added to the data (observations) such that given normal observations, the agent takes high-rewards actions, and low-reward actions on observations injected with triggers. In this paper, we propose Baffle (Backdoor Attack for Offline Reinforcement Learning), an approach that automatically implants backdoors to RL agents by poisoning the offline RL dataset, and evaluate how different offline RL algorithms react to this attack. Our experiments conducted on four tasks and four offline RL algorithms expose a disquieting fact: none of the existing offline RL algorithms is immune to such a backdoor attack. More specifically, Baffle modifies 10\% of the datasets for four tasks (3 robotic controls and 1 autonomous driving). Agents trained on the poisoned datasets perform well in normal settings. However, when triggers are presented, the agents' performance decreases drastically by 63.2\%, 53.9\%, 64.7\%, and 47.4\% in the four tasks on average. The backdoor still persists after fine-tuning poisoned agents on clean datasets. We further show that the inserted backdoor is also hard to be detected by a popular defensive method. This paper calls attention to developing more effective protection for the open-source offline RL dataset.

Cooperative multi-agent reinforcement learning (MARL) aims to coordinate a system of agents towards optimizing global returns (Vinyals et al., 2019), and has witnessed significant prospects in various domains, such as autonomous vehicles (Zhou et al., 2020), smart grid (Chen et al., 2021a), robotics (Yu et al., 2023), and social science (Leibo et al., 2017). Training reliable control policies for coordinating such systems remains a major challenge. The centralized training with decentralized execution (CTDE) (Foerster et al., 2016) hybrids the merits of independent Q-learning (Foerster et al., 2017) and joint action learning (Sukhbaatar et al., 2016), and becomes a compelling paradigm that exploits the centralized training opportunity for training fully decentralized policies (Wang et al., 2023). Subsequently, numerous popular algorithms are proposed, including VDN (Sunehag et al., 2018), QMIX (Rashid et al., 2020), MAAC (Iqbal & Sha, 2019), and MAPPO (Yu et al., 2022). Sharing policy parameters is crucial for scaling these algorithms to massive agents with accelerated cooperation learning (Fu et al., 2022). However, it is widely observed that agents tend to acquire homogeneous behaviors, which might hinder diversified exploration and sophisticated coordination (Christianos et al., 2021). Some methods (Li et al., 2021; Jiang & Lu, 2021; Liu et al., 2023) attempt to promote individualized behaviors by distinguishing each agent from the others, while they often neglect the prospect of effective team composition with implicit task allocation. Real-world multi-agent tasks usually involve dynamic team composition with the emergence of roles (Shao et al., 2022; Hu et al., 2022).

-- Multi-armed bandit algorithms provide solutions for sequential decision-making where learning takes place by interacting with the environment. In this work, we model a distributed optimization problem as a multi-agent kernelized multi-armed bandit problem with a heterogeneous reward setting. In this setup, the agents collabo-ratively aim to maximize a global objective function which is an average of local objective functions. The agents can access only bandit feedback (noisy reward) obtained from the associated unknown local function with a small norm in reproducing kernel Hilbert space (RKHS). We present a fully decentralized algorithm, Multi-agent IGP-UCB (MA-IGP-UCB), which achieves a sub-linear regret bound for popular classes for kernels while preserving privacy. It does not necessitate the agents to share their actions, rewards, or estimates of their local function. In the proposed approach, the agents sample their individual local functions in a way that benefits the whole network by utilizing a running consensus to estimate the upper confidence bound on the global function. Furthermore, we propose an extension, Multi-agent Delayed IGP-UCB (MAD-IGP-UCB) algorithm, which reduces the dependence of the regret bound on the number of agents in the network. It provides improved performance by utilizing a delay in the estimation update step at the cost of more communication. HE problem of distributed optimization deals with the optimization of a function over a network of agents in which the whole function is not completely known to any single agent [1], [2]. In fact, the "global" function can be expressed as an average of "local" functions associated with each agent which are independent of one another. In particular, our interest lies in the case when these local functions are non-convex, unknown, and expensive to compute or record. To form a feasible problem, we assume that these local functions belong to a reproducing kernel Hilbert space (RKHS), which is a very common assumption in the literature [3]- [5]. When dealing with unknown functions, the problem for each agent can be broken down into two segments: sampling and optimization.

We study the problem of modeling a population of agents pursuing unknown goals subject to unknown computational constraints. In standard models of bounded rationality, sub-optimal decision-making is simulated by adding homoscedastic noise to optimal decisions rather than explicitly simulating constrained inference. In this work, we introduce a latent inference budget model (L-IBM) that models agents' computational constraints explicitly, via a latent variable (inferred jointly with a model of agents' goals) that controls the runtime of an iterative inference algorithm. L-IBMs make it possible to learn agent models using data from diverse populations of suboptimal actors. In three modeling tasks--inferring navigation goals from routes, inferring communicative intents from human utterances, and predicting next moves in human chess games--we show that L-IBMs match or outperform Boltzmann models of decision-making under uncertainty. Inferred inference budgets are themselves meaningful, efficient to compute, and correlated with measures of player skill, partner skill and task difficulty. Building effective models for multi-agent decision-making--whether cooperative or adversarial-- requires understanding other agents' goals and plans.

Multi-agent planning and reinforcement learning can be challenging when agents cannot see the state of the world or communicate with each other due to communication costs, latency, or noise. Partially Observable Stochastic Games (POSGs) provide a mathematical framework for modelling such scenarios. This paper aims to improve the efficiency of planning and reinforcement learning algorithms for POSGs by identifying the underlying structure of optimal state-value functions. The approach involves reformulating the original game from the perspective of a trusted third party who plans on behalf of the agents simultaneously. From this viewpoint, the original POSGs can be viewed as Markov games where states are occupancy states, i.e., posterior probability distributions over the hidden states of the world and the stream of actions and observations that agents have experienced so far. This study mainly proves that the optimal state-value function is a convex function of occupancy states expressed on an appropriate basis in all zero-sum, common-payoff, and Stackelberg POSGs.

We study the performance of the gradient play algorithm for stochastic games (SGs), where each agent tries to maximize its own total discounted reward by making decisions independently based on current state information which is shared between agents. Policies are directly parameterized by the probability of choosing a certain action at a given state. We show that Nash equilibria (NEs) and first-order stationary policies are equivalent in this setting, and give a local convergence rate around strict NEs. Further, for a subclass of SGs called Markov potential games (which includes the setting with identical rewards as an important special case), we design a sample-based reinforcement learning algorithm and give a non-asymptotic global convergence rate analysis for both exact gradient play and our sample-based learning algorithm. Our result shows that the number of iterations to reach an $\epsilon$-NE scales linearly, instead of exponentially, with the number of agents. Local geometry and local stability are also considered, where we prove that strict NEs are local maxima of the total potential function and fully-mixed NEs are saddle points.

We investigate the problem of decentralized multi-agent navigation tasks, where multiple agents need to reach initially unassigned targets in a limited time. Classical planning-based methods suffer from expensive computation overhead at each step and offer limited expressiveness for complex cooperation strategies. In contrast, reinforcement learning (RL) has recently become a popular paradigm for addressing this issue. However, RL struggles with low data efficiency and cooperation when directly exploring (nearly) optimal policies in the large search space, especially with an increased agent number (e.g., 10+ agents) or in complex environments (e.g., 3D simulators). In this paper, we propose Multi-Agent Scalable GNN-based P lanner (MASP), a goal-conditioned hierarchical planner for navigation tasks with a substantial number of agents. MASP adopts a hierarchical framework to divide a large search space into multiple smaller spaces, thereby reducing the space complexity and accelerating training convergence. We also leverage graph neural networks (GNN) to model the interaction between agents and goals, improving goal achievement. Besides, to enhance generalization capabilities in scenarios with unseen team sizes, we divide agents into multiple groups, each with a previously trained number of agents. The results demonstrate that MASP outperforms classical planning-based competitors and RL baselines, achieving a nearly 100% success rate with minimal training data in both multi-agent particle environments (MPE) with 50 agents and a quadrotor 3-dimensional environment (OmniDrones) with 20 agents. Furthermore, the learned policy showcases zero-shot generalization across unseen team sizes.

SOTA multiagent reinforcement algorithms distinguish themselves in many ways from their single-agent equivalences. However, most of them still totally inherit the single-agent exploration-exploitation strategy. Naively inheriting this strategy from single-agent algorithms causes potential collaboration failures, in which the agents blindly follow mainstream behaviors and reject taking minority responsibility. We name this problem the Responsibility Diffusion (RD) as it shares similarities with a same-name social psychology effect. In this work, we start by theoretically analyzing the cause of this RD problem, which can be traced back to the exploration-exploitation dilemma of multiagent systems (especially large-scale multiagent systems). We address this RD problem by proposing a Policy Resonance (PR) approach which modifies the collaborative exploration strategy of agents by refactoring the joint agent policy while keeping individual policies approximately invariant. Next, we show that SOTA algorithms can equip this approach to promote the collaborative performance of agents in complex cooperative tasks. Experiments are performed in multiple test benchmark tasks to illustrate the effectiveness of this approach.

This work uses game theory as a mathematical framework to address interaction modeling in multi-agent motion forecasting and control. Despite its interpretability, applying game theory to real-world robotics, like automated driving, faces challenges such as unknown game parameters. To tackle these, we establish a connection between differential games, optimal control, and energy-based models, demonstrating how existing approaches can be unified under our proposed Energy-based Potential Game formulation. Building upon this, we introduce a new end-to-end learning application that combines neural networks for game-parameter inference with a differentiable game-theoretic optimization layer, acting as an inductive bias. The analysis provides empirical evidence that the game-theoretic layer adds interpretability and improves the predictive performance of various neural network backbones using two simulations and two real-world driving datasets.