Offline Multi-agent Reinforcement Learning (MARL) is valuable in scenarios where online interaction is impractical or risky. While independent learning in MARL offers flexibility and scalability, accurately assigning credit to individual agents in offline settings poses challenges because interactions with an environment are prohibited. In this paper, we propose a new framework, namely Multi-Agent Causal Credit Assignment (MACCA), to address credit assignment in the offline MARL setting. Our approach, MACCA, characterizing the generative process as a Dynamic Bayesian Network, captures relationships between environmental variables, states, actions, and rewards. Estimating this model on offline data, MACCA can learn each agent's contribution by analyzing the causal relationship of their individual rewards, ensuring accurate and interpretable credit assignment. Additionally, the modularity of our approach allows it to seamlessly integrate with various offline MARL methods. Theoretically, we proved that under the setting of the offline dataset, the underlying causal structure and the function for generating the individual rewards of agents are identifiable, which laid the foundation for the correctness of our modeling. In our experiments, we demonstrate that MACCA not only outperforms state-of-the-art methods but also enhances performance when integrated with other backbones.

This paper aims to enable multi-agent systems to effectively utilize past memories to adapt to novel collaborative tasks in a data-efficient fashion. We propose the Multi-Agent Coordination Skill Database, a repository for storing a collection of coordinated behaviors associated with key vectors distinctive to them. Our Transformer-based skill encoder effectively captures spatio-temporal interactions that contribute to coordination and provides a unique skill representation for each coordinated behavior. By leveraging a small number of demonstrations of the target task, the database enables us to train the policy using a dataset augmented with the retrieved demonstrations. Experimental evaluations demonstrate that our method achieves a significantly higher success rate in push manipulation tasks compared with baseline methods like few-shot imitation learning. Furthermore, we validate the effectiveness of our retrieve-and-learn framework in a real environment using a team of wheeled robots.

The necessity for cooperation among intelligent machines has popularised cooperative multi-agent reinforcement learning (MARL) in AI research. However, many research endeavours heavily rely on parameter sharing among agents, which confines them to only homogeneous-agent setting and leads to training instability and lack of convergence guarantees. To achieve effective cooperation in the general heterogeneous-agent setting, we propose Heterogeneous-Agent Reinforcement Learning (HARL) algorithms that resolve the aforementioned issues. Central to our findings are the multi-agent advantage decomposition lemma and the sequential update scheme . Based on these, we develop the provably correct Heterogeneous-Agent Trust Region Learning (HATRL), and derive HATRPO and HAPPO by tractable approximations. Furthermore, we discover a novel framework named Heterogeneous-Agent Mirror Learning (HAML), which strengthens theoretical guarantees for HATRPO and HAPPO and provides a general template for cooperative MARL algorithmic designs. We prove that all algorithms derived from HAML inherently enjoy monotonic improvement of joint return and convergence to Nash Equilibrium. As its natural outcome, HAML validates more novel algorithms in addition to HATRPO and HAPPO, including HAA2C, HADDPG, and HATD3, which generally outperform their existing MAcounterparts. We comprehensively test HARL algorithms on six challenging benchmarks and demonstrate their superior effectiveness and stability for coordinating heterogeneous agents compared to strong baselines such as MAPPO and QMIX.

The vision of AI collaborators is a staple of mythology and science fiction, where artificial agents with special talents assist human partners and teams. In this dream, sophisticated AIs understand nuances of collaboration and human communication. The AI as collaborator dream is different from computer tools that augment human intelligence (IA) or intermediate human collaboration. Such tools have their roots in the 1960s and helped to drive an information technology revolution. They can be useful but they are not intelligent and do not collaborate as effectively as skilled people. With the increase of hybrid and remote work since the COVID pandemic, the benefits and requirements for better coordination, collaboration, and communication are becoming a hot topic in the workplace. Employers and workers face choices and trade-offs as they negotiate the options for working from home versus working at the office. Many factors such as the high costs of homes near employers are impeding a mass return to the office. Government advisory groups and leaders in AI have advocated for years that AIs should be transparent and effective collaborators. Nonetheless, robust AIs that collaborate like talented people remain out of reach. Are AI teammates part of a solution? How artificially intelligent (AI) could and should they be? This position paper reviews the arc of technology and public calls for human-machine teaming. It draws on earlier research in psychology and the social sciences about what human-like collaboration requires. This paper sets a context for a second science-driven paper that advocates a radical shift in technology and methodology for creating resilient, intelligent, and human-compatible AIs (Stefik & Price, 2023). The aspirational goal is that such AIs would learn, share what they learn, and collaborate to achieve high capabilities.

Deep reinforcement learning (DRL) faces significant challenges in addressing the hard-exploration problems in tasks with sparse or deceptive rewards and large state spaces. These challenges severely limit the practical application of DRL. Most previous exploration methods relied on complex architectures to estimate state novelty or introduced sensitive hyperparameters, resulting in instability. To mitigate these issues, we propose an efficient adaptive trajectory-constrained exploration strategy for DRL. The proposed method guides the policy of the agent away from suboptimal solutions by leveraging incomplete offline demonstrations as references. This approach gradually expands the exploration scope of the agent and strives for optimality in a constrained optimization manner. Additionally, we introduce a novel policy-gradient-based optimization algorithm that utilizes adaptively clipped trajectory-distance rewards for both single- and multi-agent reinforcement learning. We provide a theoretical analysis of our method, including a deduction of the worst-case approximation error bounds, highlighting the validity of our approach for enhancing exploration. To evaluate the effectiveness of the proposed method, we conducted experiments on two large 2D grid world mazes and several MuJoCo tasks. The extensive experimental results demonstrate the significant advantages of our method in achieving temporally extended exploration and avoiding myopic and suboptimal behaviors in both single- and multi-agent settings. Notably, the specific metrics and quantifiable results further support these findings. The code used in the study is available at \url{https://github.com/buaawgj/TACE}.

Many applications, e.g., in shared mobility, require coordinating a large number of agents. Mean-field reinforcement learning addresses the resulting scalability challenge by optimizing the policy of a representative agent interacting with the infinite population of identical agents instead of considering individual pairwise interactions. In this paper, we address an important generalization where there exist global constraints on the distribution of agents (e.g., requiring capacity constraints or minimum coverage requirements to be met). We propose Safe-M$^3$-UCRL, the first model-based mean-field reinforcement learning algorithm that attains safe policies even in the case of unknown transitions. As a key ingredient, it uses epistemic uncertainty in the transition model within a log-barrier approach to ensure pessimistic constraints satisfaction with high probability. Beyond the synthetic swarm motion benchmark, we showcase Safe-M$^3$-UCRL on the vehicle repositioning problem faced by many shared mobility operators and evaluate its performance through simulations built on vehicle trajectory data from a service provider in Shenzhen. Our algorithm effectively meets the demand in critical areas while ensuring service accessibility in regions with low demand.

A wide array of modern machine learning applications - from adversarial models to multi-agent reinforcement learning - can be formulated as non-cooperative games whose Nash equilibria represent the system's desired operational states. Despite having a highly non-convex loss landscape, many cases of interest possess a latent convex structure that could potentially be leveraged to yield convergence to an equilibrium. Driven by this observation, our paper proposes a flexible first-order method that successfully exploits such "hidden structures" and achieves convergence under minimal assumptions for the transformation connecting the players' control variables to the game's latent, convex-structured layer. The proposed method - which we call preconditioned hidden gradient descent (PHGD) - hinges on a judiciously chosen gradient preconditioning scheme related to natural gradient methods. Importantly, we make no separability assumptions for the game's hidden structure, and we provide explicit convergence rate guarantees for both deterministic and stochastic environments.

This research paper addresses the challenges of exploration and navigation in unknown environments from an evolutionary swarm robotics perspective. Path formation plays a crucial role in enabling cooperative swarm robots to accomplish these tasks. The paper presents a method called the sub-goal-based path formation, which establishes a path between two different locations by exploiting visually connected sub-goals. Simulation experiments conducted in the Argos simulator demonstrate the successful formation of paths in the majority of trials. Furthermore, the paper tackles the problem of inter-collision (traffic) among a large number of robots engaged in path formation, which negatively impacts the performance of the sub-goal-based method. To mitigate this issue, a task allocation strategy is proposed, leveraging local communication protocols and light signal-based communication. The strategy evaluates the distance between points and determines the required number of robots for the path formation task, reducing unwanted exploration and traffic congestion. The performance of the sub-goal-based path formation and task allocation strategy is evaluated by comparing path length, time, and resource reduction against the A* algorithm. The simulation experiments demonstrate promising results, showcasing the scalability, robustness, and fault tolerance characteristics of the proposed approach.

In an environment with certain locations of higher priority, it is required to patrol these locations as frequently as possible due to their importance. However, the Non-Priority locations are often neglected during the task. It is necessary to balance the patrols on both kinds of sites to avoid breaches in security. We present a distributed online algorithm that assigns the routes to agents that ensures a finite time visit to the Non-Priority locations along with Priority Patrolling. The proposed algorithm generates offline patrol routes (Rabbit Walks) with three segments (Hops) to explore non-priority locations. The generated number of offline walks depends exponentially on a parameter introduced in the proposed algorithm, thereby facilitating the scalable implementation based on the onboard resources available on each patrolling robot. A systematic performance evaluation through simulations and experimental results validates the proportionately balanced visits and suggests the proposed algorithm's versatile applicability in the implementation of deterministic and non-deterministic scenarios.

Stochastic partial observability poses a major challenge for decentralized coordination in multi-agent reinforcement learning but is largely neglected in state-of-the-art research due to a strong focus on state-based centralized training for decentralized execution (CTDE) and benchmarks that lack sufficient stochasticity like StarCraft Multi-Agent Challenge (SMAC). In this paper, we propose Attention-based Embeddings of Recurrence In multi-Agent Learning (AERIAL) to approximate value functions under stochastic partial observability. AERIAL replaces the true state with a learned representation of multi-agent recurrence, considering more accurate information about decentralized agent decisions than state-based CTDE. We then introduce MessySMAC, a modified version of SMAC with stochastic observations and higher variance in initial states, to provide a more general and configurable benchmark regarding stochastic partial observability. We evaluate AERIAL in Dec-Tiger as well as in a variety of SMAC and MessySMAC maps, and compare the results with state-based CTDE. Furthermore, we evaluate the robustness of AERIAL and state-based CTDE against various stochasticity configurations in MessySMAC.