Pommerman is a multi-agent environment that has received considerable attention from researchers in recent years. This environment is an ideal benchmark for multi-agent training, providing a battleground for two teams with communication capabilities among allied agents. Pommerman presents significant challenges for model-free reinforcement learning due to delayed action effects, sparse rewards, and false positives, where opponent players can lose due to their own mistakes. This study introduces a system designed to train multi-agent systems to play Pommerman using a combination of curriculum learning and population-based self-play. We also tackle two challenging problems when deploying the multi-agent training system for competitive games: sparse reward and suitable matchmaking mechanism. Specifically, we propose an adaptive annealing factor based on agents' performance to adjust the dense exploration reward during training dynamically. Additionally, we implement a matchmaking mechanism utilizing the Elo rating system to pair agents effectively. Our experimental results demonstrate that our trained agent can outperform top learning agents without requiring communication among allied agents.

Visual navigation tasks are critical for household service robots. As these tasks become increasingly complex, effective communication and collaboration among multiple robots become imperative to ensure successful completion. In recent years, large language models (LLMs) have exhibited remarkable comprehension and planning abilities in the context of embodied agents. However, their application in household scenarios, specifically in the use of multiple agents collaborating to complete complex navigation tasks through communication, remains unexplored. Therefore, this paper proposes a framework for decentralized multi-agent navigation, leveraging LLM-enabled communication and collaboration. By designing the communication-triggered dynamic leadership organization structure, we achieve faster team consensus with fewer communication instances, leading to better navigation effectiveness and collaborative exploration efficiency. With the proposed novel communication scheme, our framework promises to be conflict-free and robust in multi-object navigation tasks, even when there is a surge in team size.

Understanding user enjoyment is crucial in human-robot interaction (HRI), as it can impact interaction quality and influence user acceptance and long-term engagement with robots, particularly in the context of conversations with social robots. However, current assessment methods rely solely on self-reported questionnaires, failing to capture interaction dynamics. This work introduces the Human-Robot Interaction Conversational User Enjoyment Scale (HRI CUES), a novel scale for assessing user enjoyment from an external perspective during conversations with a robot. Developed through rigorous evaluations and discussions of three annotators with relevant expertise, the scale provides a structured framework for assessing enjoyment in each conversation exchange (turn) alongside overall interaction levels. It aims to complement self-reported enjoyment from users and holds the potential for autonomously identifying user enjoyment in real-time HRI. The scale was validated on 25 older adults' open-domain dialogue with a companion robot that was powered by a large language model for conversations, corresponding to 174 minutes of data, showing moderate to good alignment. The dataset is available online. Additionally, the study offers insights into understanding the nuances and challenges of assessing user enjoyment in robot interactions, and provides guidelines on applying the scale to other domains.

Learning effective negotiation strategies poses two key challenges: the exploration-exploitation dilemma and dealing with large action spaces. However, there is an absence of learning-based approaches that effectively address these challenges in negotiation. This paper introduces a comprehensive formulation to tackle various negotiation problems. Our approach leverages contextual combinatorial multi-armed bandits, with the bandits resolving the exploration-exploitation dilemma, and the combinatorial nature handles large action spaces. Building upon this formulation, we introduce NegUCB, a novel method that also handles common issues such as partial observations and complex reward functions in negotiation. NegUCB is contextual and tailored for full-bandit feedback without constraints on the reward functions. Under mild assumptions, it ensures a sub-linear regret upper bound. Experiments conducted on three negotiation tasks demonstrate the superiority of our approach.

Fleets of autonomous vehicles can mitigate traffic congestion through simple actions, thus improving many socioeconomic factors such as commute time and gas costs. However, these approaches are limited in practice as they assume precise control over autonomous vehicle fleets, incur extensive installation costs for a centralized sensor ecosystem, and also fail to account for uncertainty in driver behavior. To this end, we develop a class of learned residual policies that can be used in cooperative advisory systems and only require the use of a single vehicle with a human driver. Our policies advise drivers to behave in ways that mitigate traffic congestion while accounting for diverse driver behaviors, particularly drivers' reactions to instructions, to provide an improved user experience. To realize such policies, we introduce an improved reward function that explicitly addresses congestion mitigation and driver attitudes to advice. We show that our residual policies can be personalized by conditioning them on an inferred driver trait that is learned in an unsupervised manner with a variational autoencoder. Our policies are trained in simulation with our novel instruction adherence driver model, and evaluated in simulation and through a user study (N=16) to capture the sentiments of human drivers. Our results show that our approaches successfully mitigate congestion while adapting to different driver behaviors, with up to 20% and 40% improvement as measured by a combination metric of speed and deviations in speed across time over baselines in our simulation tests and user study, respectively. Our user study further shows that our policies are human-compatible and personalize to drivers.

We present AVOCADO (AdaptiVe Optimal Collision Avoidance Driven by Opinion), a novel navigation approach to address holonomic robot collision avoidance when the degree of cooperation of the other agents in the environment is unknown. AVOCADO departs from a Velocity Obstacle's formulation akin to the Optimal Reciprocal Collision Avoidance method. However, instead of assuming reciprocity, AVOCADO poses an adaptive control problem that aims at adapting in real-time to the cooperation degree of other robots and agents. Adaptation is achieved through a novel nonlinear opinion dynamics design that relies solely on sensor observations. As a by-product, based on the nonlinear opinion dynamics, we propose a novel method to avoid the deadlocks under geometrical symmetries among robots and agents. Extensive numerical simulations show that AVOCADO surpasses existing geometrical, learning and planning-based approaches in mixed cooperative/non-cooperative navigation environments in terms of success rate, time to goal and computational time. In addition, we conduct multiple real experiments that verify that AVOCADO is able to avoid collisions in environments crowded with other robots and humans.

Consider $N$ players each with a $d$-dimensional action set. Each of the players' utility functions includes their reward function and a linear term for each dimension, with coefficients that are controlled by the manager. We assume that the game is strongly monotone, so if each player runs gradient descent, the dynamics converge to a unique Nash equilibrium (NE). The NE is typically inefficient in terms of global performance. The resulting global performance of the system can be improved by imposing $K$-dimensional linear constraints on the NE. We therefore want the manager to pick the controlled coefficients that impose the desired constraint on the NE. However, this requires knowing the players' reward functions and their action sets. Obtaining this game structure information is infeasible in a large-scale network and violates the users' privacy. To overcome this, we propose a simple algorithm that learns to shift the NE of the game to meet the linear constraints by adjusting the controlled coefficients online. Our algorithm only requires the linear constraints violation as feedback and does not need to know the reward functions or the action sets. We prove that our algorithm, which is based on two time-scale stochastic approximation, guarantees convergence with probability 1 to the set of NE that meet target linear constraints. We then provide a mean square convergence rate of $O(t^{-1/4})$ for our algorithm. This is the first such bound for two time-scale stochastic approximation where the slower time-scale is a fixed point iteration with a non-expansive mapping. We demonstrate how our scheme can be applied to optimizing a global quadratic cost at NE and load balancing in resource allocation games. We provide simulations of our algorithm for these scenarios.

Artificially intelligent agents deployed in the real-world will require the ability to reliably \textit{cooperate} with humans (as well as other, heterogeneous AI agents). To provide formal guarantees of successful cooperation, we must make some assumptions about how partner agents could plausibly behave. Any realistic set of assumptions must account for the fact that other agents may be just as adaptable as our agent is. In this work, we consider the problem of cooperating with a \textit{population} of agents in a finitely-repeated, two player general-sum matrix game with private utilities. Two natural assumptions in such settings are that: 1) all agents in the population are individually rational learners, and 2) when any two members of the population are paired together, with high-probability they will achieve at least the same utility as they would under some Pareto efficient equilibrium strategy. Our results first show that these assumptions alone are insufficient to ensure \textit{zero-shot} cooperation with members of the target population. We therefore consider the problem of \textit{learning} a strategy for cooperating with such a population using prior observations its members interacting with one another. We provide upper and lower bounds on the number of samples needed to learn an effective cooperation strategy. Most importantly, we show that these bounds can be much stronger than those arising from a "naive'' reduction of the problem to one of imitation learning.

Message passing graph neural networks (GNNs) would appear to be powerful tools to learn distributed algorithms via gradient descent, but generate catastrophically incorrect predictions when nodes update asynchronously during inference. This failure under asynchrony effectively excludes these architectures from many potential applications, such as learning local communication policies between resource-constrained agents in, e.g., robotic swarms or sensor networks. In this work we explore why this failure occurs in common GNN architectures, and identify "implicitly-defined" GNNs as a class of architectures which is provably robust to partially asynchronous "hogwild" inference, adapting convergence guarantees from work in asynchronous and distributed optimization, e.g., Bertsekas (1982); Niu et al. (2011). We then propose a novel implicitly-defined GNN architecture, which we call an energy GNN. We show that this architecture outperforms other GNNs from this class on a variety of synthetic tasks inspired by multi-agent systems, and achieves competitive performance on real-world datasets.

Pursuing artificial intelligence for biomedical science, a.k.a. AI Scientist, draws increasing attention, where one common approach is to build a copilot agent driven by Large Language Models (LLMs). However, to evaluate such systems, people either rely on direct Question-Answering (QA) to the LLM itself, or in a biomedical experimental manner. How to precisely benchmark biomedical agents from an AI Scientist perspective remains largely unexplored. To this end, we draw inspiration from one most important abilities of scientists, understanding the literature, and introduce BioKGBench. In contrast to traditional evaluation benchmark that only focuses on factual QA, where the LLMs are known to have hallucination issues, we first disentangle "Understanding Literature" into two atomic abilities, i) "Understanding" the unstructured text from research papers by performing scientific claim verification, and ii) Ability to interact with structured Knowledge-Graph Question-Answering (KGQA) as a form of "Literature" grounding. We then formulate a novel agent task, dubbed KGCheck, using KGQA and domain-based Retrieval-Augmented Generation (RAG) to identify the factual errors of existing large-scale knowledge graph databases. We collect over two thousand data for two atomic tasks and 225 high-quality annotated data for the agent task. Surprisingly, we discover that state-of-the-art agents, both daily scenarios and biomedical ones, have either failed or inferior performance on our benchmark. We then introduce a simple yet effective baseline, dubbed BKGAgent. On the widely used popular knowledge graph, we discover over 90 factual errors which provide scenarios for agents to make discoveries and demonstrate the effectiveness of our approach. The code and data are available at https://github.com/westlake-autolab/BioKGBench.