Increasing interest in ensuring safety of next-generation Artificial Intelligence (AI) systems calls for novel approaches to embedding morality into autonomous agents. Traditionally, this has been done by imposing explicit top-down rules or hard constraints on systems, for example by filtering system outputs through pre-defined ethical rules. Recently, instead, entirely bottom-up methods for learning implicit preferences from human behavior have become increasingly popular, such as those for training and fine-tuning Large Language Models. In this paper, we provide a systematization of existing approaches to the problem of introducing morality in machines - modeled as a continuum, and argue that the majority of popular techniques lie at the extremes - either being fully hard-coded, or entirely learned, where no explicit statement of any moral principle is required. Given the relative strengths and weaknesses of each type of methodology, we argue that more hybrid solutions are needed to create adaptable and robust, yet more controllable and interpretable agents. In particular, we present three case studies of recent works which use learning from experience (i.e., Reinforcement Learning) to explicitly provide moral principles to learning agents - either as intrinsic rewards, moral logical constraints or textual principles for language models. For example, using intrinsic rewards in Social Dilemma games, we demonstrate how it is possible to represent classical moral frameworks for agents. We also present an overview of the existing work in this area in order to provide empirical evidence for the potential of this hybrid approach. We then discuss strategies for evaluating the effectiveness of moral learning agents. Finally, we present open research questions and implications for the future of AI safety and ethics which are emerging from this framework.

Agents built with large language models (LLMs) have recently achieved great advancements. However, most of the efforts focus on single-agent or cooperative settings, leaving more general multi-agent environments underexplored. We propose a new framework powered by reinforcement learning (RL) to develop strategic language agents, i.e., LLM-based agents with strategic thinking ability, for a popular language game, Werewolf. Werewolf is a social deduction game with hidden roles that involves both cooperation and competition and emphasizes deceptive communication and diverse gameplay. Our agent tackles this game by first using LLMs to reason about potential deceptions and generate a set of strategically diverse actions. Then an RL policy, which selects an action from the candidates, is learned by population-based training to enhance the agents' decision-making ability. By combining LLMs with the RL policy, our agent produces a variety of emergent strategies, achieves the highest win rate against other LLM-based agents, and stays robust against adversarial human players in the Werewolf game.

The field of Multi-Agent Reinforcement Learning (MARL) is currently facing a reproducibility crisis. While solutions for standardized reporting have been proposed to address the issue, we still lack a benchmarking tool that enables standardization and reproducibility, while leveraging cutting-edge Reinforcement Learning (RL) implementations. In this paper, we introduce BenchMARL, the first MARL training library created to enable standardized benchmarking across different algorithms, models, and environments. BenchMARL uses TorchRL as its backend, granting it high performance and maintained state-of-the-art implementations while addressing the broad community of MARL PyTorch users. Its design enables systematic configuration and reporting, thus allowing users to create and run complex benchmarks from simple one-line inputs.

Multi-agent Reinforcement Learning (MARL) has gained wide attention in recent years and has made progress in various fields. Specifically, cooperative MARL focuses on training a team of agents to cooperatively achieve tasks that are difficult for a single agent to handle. It has shown great potential in applications such as path planning, autonomous driving, active voltage control, and dynamic algorithm configuration. One of the research focuses in the field of cooperative MARL is how to improve the coordination efficiency of the system, while research work has mainly been conducted in simple, static, and closed environment settings. To promote the application of artificial intelligence in real-world, some research has begun to explore multi-agent coordination in open environments. These works have made progress in exploring and researching the environments where important factors might change. However, the mainstream work still lacks a comprehensive review of the research direction. In this paper, starting from the concept of reinforcement learning, we subsequently introduce multi-agent systems (MAS), cooperative MARL, typical methods, and test environments. Then, we summarize the research work of cooperative MARL from closed to open environments, extract multiple research directions, and introduce typical works. Finally, we summarize the strengths and weaknesses of the current research, and look forward to the future development direction and research problems in cooperative MARL in open environments.

Route planning for a fleet of vehicles is an important task in applications such as package delivery, surveillance, or transportation. This problem is usually modeled as a Combinatorial Optimization problem named as Team Orienteering Problem. The most popular Team Orienteering Problem solvers are mainly based on either linear programming, which provides accurate solutions by employing a large computation time that grows with the size of the problem, or heuristic methods, which usually find suboptimal solutions in a shorter amount of time. In this paper, a multi-agent route planning system capable of solving the Team Orienteering Problem in a very fast and accurate manner is presented. The proposed system is based on a centralized Transformer neural network that can learn to encode the scenario (modeled as a graph) and the context of the agents to provide fast and accurate solutions. Several experiments have been performed to demonstrate that the presented system can outperform most of the state-of-the-art works in terms of computation speed. In addition, the code is publicly available at http://gti.ssr.upm.es/data.

Multi-agent decision problems are typically solved via distributed iterative algorithms, where the agents only communicate between themselves on a peer-to-peer network. Each agent usually maintains a copy of each decision variable, while agreement among the local copies is enforced via consensus protocols. Yet, each agent is often directly influenced by a small portion of the decision variables only: neglecting this sparsity results in redundancy, poor scalability with the network size, communication and memory overhead. To address these challenges, we develop Estimation Network Design (END), a framework for the design and analysis of distributed algorithms, generalizing several recent approaches. END algorithms can be tuned to exploit problem-specific sparsity structures, by optimally allocating copies of each variable only to a subset of agents, to improve efficiency and minimize redundancy. We illustrate the END's potential by designing new algorithms for generalised Nash equilibrium (GNE) seeking under partial-decision information, that can leverage the sparsity in cost functions, constraints and aggregation values. Finally, we test numerically our methods on a unicast rate allocation problem, revealing greatly reduced communication and memory costs.

"social sciences" for agent community may also be derived. Similarity is established with the human social sciences serving as the baseline. Since there exist inherent differences in the way agents and human act, the "social sciences" for agent society may also be Similar to the common methodology in human social different from that for human society.

Information exchange in multi-agent systems improves the cooperation among agents, especially in partially observable settings. In the real world, communication is often carried out over imperfect channels. This requires agents to handle uncertainty due to potential information loss. In this paper, we consider a cooperative multi-agent system where the agents act and exchange information in a decentralized manner using a limited and unreliable channel. To cope with such channel constraints, we propose a novel communication approach based on independent Q-learning. Our method allows agents to dynamically adapt how much information to share by sending messages of different sizes, depending on their local observations and the channel's properties. In addition to this message size selection, agents learn to encode and decode messages to improve their jointly trained policies. We show that our approach outperforms approaches without adaptive capabilities in a novel cooperative digit-prediction environment and discuss its limitations in the traffic junction environment.

Abstract-- This paper provides a perspective on the literature and current challenges in Multi-Agent Systems for interoperable robot navigation in industry. The focus is on the multiagent decision stack for Autonomous Mobile Robots operating in mixed environments with humans, manually driven vehicles, and legacy Automated Guided Vehicles. We provide typical characteristics of such Multi-Agent Systems observed today and how these are expected to change on the short term due to the new standard VDA5050 and the interoperability framework OpenRMF. Approaches to increase the robustness and performance of multi-robot navigation systems for transportation are discussed, and research opportunities are derived. I. INTRODUCTION Multi-robot navigation encompasses an ever-tighter integration of a vast number of disciplines and research as in most of finalized components to storage locations.

Deep reinforcement learning suffers from catastrophic forgetting and sample inefficiency making it less applicable to the ever-changing real world. However, the ability to use previously learned knowledge is essential for AI agents to quickly adapt to novelties. Often, certain spatial information observed by the agent in the previous interactions can be leveraged to infer task-specific rules. Inferred rules can then help the agent to avoid potentially dangerous situations in the previously unseen states and guide the learning process increasing agent's novelty adaptation speed. In this work, we propose a general framework that is applicable to deep reinforcement learning agents. Our framework provides the agent with an autonomous way to discover the task-specific rules in the novel environments and self-supervise it's learning. We provide a rule-driven deep Q-learning agent (RDQ) as one possible implementation of that framework. We show that RDQ successfully extracts task-specific rules as it interacts with the world and uses them to drastically increase its learning efficiency. In our experiments, we show that the RDQ agent is significantly more resilient to the novelties than the baseline agents, and is able to detect and adapt to novel situations faster.