Multi-agent reinforcement learning (MARL) addresses sequential decision-making problems with multiple agents, where each agent optimizes its own objective. In many real-world instances, the agents may not only want to optimize their objectives, but also ensure safe behavior. For example, in traffic routing, each car (agent) aims to reach its destination quickly (objective) while avoiding collisions (safety). Constrained Markov Games (CMGs) are a natural formalism for safe MARL problems, though generally intractable. In this work, we introduce and study Constrained Markov Potential Games (CMPGs), an important class of CMGs. We first show that a Nash policy for CMPGs can be found via constrained optimization. One tempting approach is to solve it by Lagrangian-based primal-dual methods. As we show, in contrast to the single-agent setting, however, CMPGs do not satisfy strong duality, rendering such approaches inapplicable and potentially unsafe. To solve the CMPG problem, we propose our algorithm Coordinate-Ascent for CMPGs (CA-CMPG), which provably converges to a Nash policy in tabular, finite-horizon CMPGs. Furthermore, we provide the first sample complexity bounds for learning Nash policies in unknown CMPGs, and, which under additional assumptions, guarantee safe exploration.

Policy-based algorithms are among the most widely adopted techniques in model-free RL, thanks to their strong theoretical groundings and good properties in continuous action spaces. Unfortunately, these methods require precise and problem-specific hyperparameter tuning to achieve good performance, and tend to struggle when asked to accomplish a series of heterogeneous tasks. In particular, the selection of the step size has a crucial impact on their ability to learn a highly performing policy, affecting the speed and the stability of the training process, and often being the main culprit for poor results. In this paper, we tackle these issues with a Meta Reinforcement Learning approach, by introducing a new formulation, known as meta-MDP, that can be used to solve any hyperparameter selection problem in RL with contextual processes. After providing a theoretical Lipschitz bound to the difference of performance in different tasks, we adopt the proposed framework to train a batch RL algorithm to dynamically recommend the most adequate step size for different policies and tasks. In conclusion, we present an experimental campaign to show the advantages of selecting an adaptive learning rate in heterogeneous environments.

We present Galactic, a large-scale simulation and reinforcement-learning (RL) framework for robotic mobile manipulation in indoor environments. Specifically, a Fetch robot (equipped with a mobile base, 7DoF arm, RGBD camera, egomotion, and onboard sensing) is spawned in a home environment and asked to rearrange objects - by navigating to an object, picking it up, navigating to a target location, and then placing the object at the target location. Galactic is fast. In terms of simulation speed (rendering + physics), Galactic achieves over 421,000 steps-per-second (SPS) on an 8-GPU node, which is 54x faster than Habitat 2.0 (7699 SPS). More importantly, Galactic was designed to optimize the entire rendering + physics + RL interplay since any bottleneck in the interplay slows down training. In terms of simulation+RL speed (rendering + physics + inference + learning), Galactic achieves over 108,000 SPS, which 88x faster than Habitat 2.0 (1243 SPS). These massive speed-ups not only drastically cut the wall-clock training time of existing experiments, but also unlock an unprecedented scale of new experiments. First, Galactic can train a mobile pick skill to >80% accuracy in under 16 minutes, a 100x speedup compared to the over 24 hours it takes to train the same skill in Habitat 2.0. Second, we use Galactic to perform the largest-scale experiment to date for rearrangement using 5B steps of experience in 46 hours, which is equivalent to 20 years of robot experience. This scaling results in a single neural network composed of task-agnostic components achieving 85% success in GeometricGoal rearrangement, compared to 0% success reported in Habitat 2.0 for the same approach. The code is available at github.com/facebookresearch/galactic.

Multi-agent reinforcement learning (MARL) models multiple agents that interact and learn within a shared environment. This paradigm is applicable to various industrial scenarios such as autonomous driving, quantitative trading, and inventory management. However, applying MARL to these real-world scenarios is impeded by many challenges such as scaling up, complex agent interactions, and non-stationary dynamics. To incentivize the research of MARL on these challenges, we develop MABIM (Multi-Agent Benchmark for Inventory Management) which is a multi-echelon, multi-commodity inventory management simulator that can generate versatile tasks with these different challenging properties. Based on MABIM, we evaluate the performance of classic operations research (OR) methods and popular MARL algorithms on these challenging tasks to highlight their weaknesses and potential.

In this work we identify the dormant neuron phenomenon in deep reinforcement learning, where an agent's network suffers from an increasing number of inactive neurons, thereby affecting network expressivity. We demonstrate the presence of this phenomenon across a variety of algorithms and environments, and highlight its effect on learning. To address this issue, we propose a simple and effective method (ReDo) that Recycles Dormant neurons throughout training. Our experiments demonstrate that ReDo maintains the expressive power of networks by reducing the number of dormant neurons and results in improved performance.

There is increasing interest in data-driven approaches for recommending optimal treatment strategies in many chronic disease management and critical care applications. Reinforcement learning methods are well-suited to this sequential decision-making problem, but must be trained and evaluated exclusively on retrospective medical record datasets as direct online exploration is unsafe and infeasible. Despite this requirement, the vast majority of treatment optimization studies use off-policy RL methods (e.g., Double Deep Q Networks (DDQN) or its variants) that are known to perform poorly in purely offline settings. Recent advances in offline RL, such as Conservative Q-Learning (CQL), offer a suitable alternative. But there remain challenges in adapting these approaches to real-world applications where suboptimal examples dominate the retrospective dataset and strict safety constraints need to be satisfied. In this work, we introduce a practical and theoretically grounded transition sampling approach to address action imbalance during offline RL training. We perform extensive experiments on two real-world tasks for diabetes and sepsis treatment optimization to compare performance of the proposed approach against prominent off-policy and offline RL baselines (DDQN and CQL). Across a range of principled and clinically relevant metrics, we show that our proposed approach enables substantial improvements in expected health outcomes and in accordance with relevant practice and safety guidelines.

A major goal of artificial intelligence (AI) is to create an agent capable of acquiring a general understanding of the world. Such an agent would require the ability to continually accumulate and build upon its knowledge as it encounters new experiences. Lifelong or continual learning addresses this setting, whereby an agent faces a continual stream of problems and must strive to capture the knowledge necessary for solving each new task it encounters. If the agent is capable of accumulating knowledge in some form of compositional representation, it could then selectively reuse and combine relevant pieces of knowledge to construct novel solutions. Despite the intuitive appeal of this simple idea, the literatures on lifelong learning and compositional learning have proceeded largely separately. In an effort to promote developments that bridge between the two fields, this article surveys their respective research landscapes and discusses existing and future connections between them.

Domain adaptation in reinforcement learning (RL) mainly deals with the changes of observation when transferring the policy to a new environment. Many traditional approaches of domain adaptation in RL manage to learn a mapping function between the source and target domain in explicit or implicit ways. However, they typically require access to abundant data from the target domain. Besides, they often rely on visual clues to learn the mapping function and may fail when the source domain looks quite different from the target domain. To address these problems, we propose a novel framework Online Prototype Alignment (OPA) to learn the mapping function based on the functional similarity of elements and is able to achieve the few-shot policy transfer within only several episodes. The key insight of OPA is to introduce an exploration mechanism that can interact with the unseen elements of the target domain in an efficient and purposeful manner, and then connect them with the seen elements in the source domain according to their functionalities (instead of visual clues). Experimental results show that when the target domain looks visually different from the source domain, OPA can achieve better transfer performance even with much fewer samples from the target domain, outperforming prior methods.

Green Light Optimal Speed Advisory (GLOSA) system suggests speeds to vehicles to assist them in passing through intersections during green intervals, thus reducing traffic congestion and fuel consumption by minimizing the number of stops and idle times at intersections. However, previous research has focused on optimizing the GLOSA algorithm, neglecting the frequency of speed advisory by the GLOSA system. Specifically, some studies provide speed advisory profile at each decision step, resulting in redundant advisory, while others calculate the optimal speed for the vehicle only once, which cannot adapt to dynamic traffic. In this paper, we propose an Adaptive Frequency GLOSA (AF-GLOSA) model based on Hybrid Proximal Policy Optimization (H-PPO) method, which employs an actor-critic architecture with a hybrid actor network. The hybrid actor network consists of a discrete actor that outputs control gap and a continuous actor that outputs acceleration profiles. Additionally, we design a novel reward function that considers both travel efficiency and fuel consumption. The AF-GLOSA model is evaluated in comparison to traditional GLOSA and learning-based GLOSA methods in a three-lane intersection with a traffic signal in SUMO. The results demonstrate that the AF-GLOSA model performs best in reducing average stop times, fuel consumption and CO2 emissions.

Recent research in pedestrian simulation often aims to develop realistic behaviors in various situations, but it is challenging for existing algorithms to generate behaviors that identify weaknesses in automated vehicles' performance in extreme and unlikely scenarios and edge cases. To address this, specialized pedestrian behavior algorithms are needed. Current research focuses on realistic trajectories using social force models and reinforcement learning based models. However, we propose a reinforcement learning algorithm that specifically targets collisions and better uncovers unique failure modes of automated vehicle controllers. Our algorithm is efficient and generates more severe collisions, allowing for the identification and correction of weaknesses in autonomous driving algorithms in complex and varied scenarios.