Efficient timing in ride-matching is crucial for improving the performance of ride-hailing and ride-pooling services, as it determines the number of drivers and passengers considered in each matching process. Traditional batched matching methods often use fixed time intervals to accumulate ride requests before assigning matches. While this approach increases the number of available drivers and passengers for matching, it fails to adapt to real-time supply-demand fluctuations, often leading to longer passenger wait times and driver idle periods. To address this limitation, we propose an adaptive ride-matching strategy using deep reinforcement learning (RL) to dynamically determine when to perform matches based on real-time system conditions. Unlike fixed-interval approaches, our method continuously evaluates system states and executes matching at moments that minimize total passenger wait time. Additionally, we incorporate a potential-based reward shaping (PBRS) mechanism to mitigate sparse rewards, accelerating RL training and improving decision quality. Extensive empirical evaluations using a realistic simulator trained on real-world data demonstrate that our approach outperforms fixed-interval matching strategies, significantly reducing passenger waiting times and detour delays, thereby enhancing the overall efficiency of ride-hailing and ride-pooling systems.

The ability of artificial intelligence agents to make optimal decisions and generalise them to different domains and tasks is compromised in complex scenarios. One way to address this issue has focused on learning efficient representations of the world and on how the actions of agents affect them, such as disentangled representations that exploit symmetries. Whereas such representations are procedurally efficient, they are based on the compression of low-level state-action transitions, which lack structural richness. To address this problem, we propose to enrich the agent's ontology and extend the traditional conceptualisation of trajectories to provide a more nuanced view of task execution. Structurally Enriched Trajectories (SETs) extend the encoding of sequences of states and their transitions by incorporating hierarchical relations between objects, interactions and affordances. SETs are built as multi-level graphs, providing a detailed representation of the agent dynamics and a transferable functional abstraction of the task. SETs are integrated into an architecture, Structurally Enriched Trajectory Learning and Encoding (SETLE), that employs a heterogeneous graph-based memory structure of multi-level relational dependencies essential for generalisation. Using reinforcement learning as a data generation tool, we demonstrate that SETLE can support downstream tasks, enabling agents to recognise task-relevant structural patterns across diverse environments.

Learning to manipulate objects efficiently, particularly those involving sustained contact (e.g., pushing, sliding) and articulated parts (e.g., drawers, doors), presents significant challenges. Traditional methods, such as robot-centric reinforcement learning (RL), imitation learning, and hybrid techniques, require massive training and often struggle to generalize across different objects and robot platforms. We propose a novel framework for learning object-centric manipulation policies in force space, decoupling the robot from the object. By directly applying forces to selected regions of the object, our method simplifies the action space, reduces unnecessary exploration, and decreases simulation overhead. This approach, trained in simulation on a small set of representative objects, captures object dynamics -- such as joint configurations -- allowing policies to generalize effectively to new, unseen objects. Decoupling these policies from robot-specific dynamics enables direct transfer to different robotic platforms (e.g., Kinova, Panda, UR5) without retraining. Our evaluations demonstrate that the method significantly outperforms baselines, achieving over an order of magnitude improvement in training efficiency compared to other state-of-the-art methods. Additionally, operating in force space enhances policy transferability across diverse robot platforms and object types. We further showcase the applicability of our method in a real-world robotic setting. For supplementary materials and videos, please visit: https://tufts-ai-robotics-group.github.io/FLEX/

Multi-agent reinforcement learning has shown promise in learning cooperative behaviors in team-based environments. However, such methods often demand extensive training time. For instance, the state-of-the-art method TiZero takes 40 days to train high-quality policies for a football environment. In this paper, we hypothesize that better exploration mechanisms can improve the sample efficiency of multi-agent methods. We propose two different approaches for better exploration in TiZero: a self-supervised intrinsic reward and a random network distillation bonus. Additionally, we introduce architectural modifications to the original algorithm to enhance TiZero's computational efficiency. We evaluate the sample efficiency of these approaches through extensive experiments. Our results show that random network distillation improves training sample efficiency by 18.8% compared to the original TiZero. Furthermore, we evaluate the qualitative behavior of the models produced by both variants against a heuristic AI, with the self-supervised reward encouraging possession and random network distillation leading to a more offensive performance. Our results highlights the applicability of our random network distillation variant in practical settings. Lastly, due to the nature of the proposed method, we acknowledge its use beyond football simulation, especially in environments with strong multi-agent and strategic aspects.

Designing reward functions for continuous-control robotics often leads to subtle misalignments or reward hacking, especially in complex tasks. Preference-based RL mitigates some of these pitfalls by learning rewards from comparative feedback rather than hand-crafted signals, yet scaling human annotations remains challenging. Recent work uses Vision-Language Models (VLMs) to automate preference labeling, but a single final-state image generally fails to capture the agent's full motion. In this paper, we present a two-part solution that both improves feedback accuracy and better aligns reward learning with the agent's policy. First, we overlay trajectory sketches on final observations to reveal the path taken, allowing VLMs to provide more reliable preferences-improving preference accuracy by approximately 15-20% in metaworld tasks. Second, we regularize reward learning by incorporating the agent's performance, ensuring that the reward model is optimized based on data generated by the current policy; this addition boosts episode returns by 20-30% in locomotion tasks. Empirical studies on metaworld demonstrate that our method achieves, for instance, around 70-80% success rate in all tasks, compared to below 50% for standard approaches. These results underscore the efficacy of combining richer visual representations with agent-aware reward regularization.

Action-constrained reinforcement learning (ACRL) is a generic framework for learning control policies with zero action constraint violation, which is required by various safety-critical and resource-constrained applications. The existing ACRL methods can typically achieve favorable constraint satisfaction but at the cost of either high computational burden incurred by the quadratic programs (QP) or increased architectural complexity due to the use of sophisticated generative models. In this paper, we propose a generic and computationally efficient framework that can adapt a standard unconstrained RL method to ACRL through two modifications: (i) To enforce the action constraints, we leverage the classic acceptance-rejection method, where we treat the unconstrained policy as the proposal distribution and derive a modified policy with feasible actions. (ii) To improve the acceptance rate of the proposal distribution, we construct an augmented two-objective Markov decision process (MDP), which include additional self-loop state transitions and a penalty signal for the rejected actions. This augmented MDP incentives the learned policy to stay close to the feasible action sets. Through extensive experiments in both robot control and resource allocation domains, we demonstrate that the proposed framework enjoys faster training progress, better constraint satisfaction, and a lower action inference time simultaneously than the state-of-the-art ACRL methods. We have made the source code publicly available to encourage further research in this direction.

In recent times, reinforcement learning (RL) with physical robots has attracted the attention of a wide range of researchers. However, state-of-the-art RL algorithms do not consider that physical environments do not wait for the RL agent to make decisions or updates. RL agents learn by periodically conducting computationally expensive gradient updates. When decision-making and gradient update tasks are carried out sequentially by the RL agent in a physical robot, it significantly increases the agent's response time. In a rapidly changing environment, this increased response time may be detrimental to the performance of the learning agent. Asynchronous RL methods, which separate the computation of decision-making and gradient updates, are a potential solution to this problem. However, only a few comparisons between asynchronous and synchronous RL have been made with physical robots. For this reason, the exact performance benefits of using asynchronous RL methods over synchronous RL methods are still unclear. In this study, we provide a performance comparison between asynchronous and synchronous RL using a physical robotic arm called Franka Emika Panda. Our experiments show that the agents learn faster and attain significantly more returns using asynchronous RL. Our experiments also demonstrate that the learning agent with a faster response time performs better than the agent with a slower response time, even if the agent with a slower response time performs a higher number of gradient updates.

Transfer Learning (TL) is a powerful tool that enables robots to transfer learned policies across different environments, tasks, or embodiments. To further facilitate this process, efforts have been made to combine it with Learning from Demonstrations (LfD) for more flexible and efficient policy transfer. However, these approaches are almost exclusively limited to offline demonstrations collected before policy transfer starts, which may suffer from the intrinsic issue of covariance shift brought by LfD and harm the performance of policy transfer. Meanwhile, extensive work in the learning-from-scratch setting has shown that online demonstrations can effectively alleviate covariance shift and lead to better policy performance with improved sample efficiency. This work combines these insights to introduce online demonstrations into a policy transfer setting. We present Policy Transfer with Online Demonstrations, an active LfD algorithm for policy transfer that can optimize the timing and content of queries for online episodic expert demonstrations under a limited demonstration budget. We evaluate our method in eight robotic scenarios, involving policy transfer across diverse environment characteristics, task objectives, and robotic embodiments, with the aim to transfer a trained policy from a source task to a related but different target task. The results show that our method significantly outperforms all baselines in terms of average success rate and sample efficiency, compared to two canonical LfD methods with offline demonstrations and one active LfD method with online demonstrations. Additionally, we conduct preliminary sim-to-real tests of the transferred policy on three transfer scenarios in the real-world environment, demonstrating the policy effectiveness on a real robot manipulator.

Lifelong Reinforcement Learning (LRL) holds significant potential for addressing sequential tasks, but it still faces considerable challenges. A key difficulty lies in effectively preventing catastrophic forgetting and facilitating knowledge transfer while maintaining reliable decision-making performance across subsequent tasks in dynamic environments. To tackle this, we propose a novel framework, SDW (Similarity-Driven Weighting Framework), which leverages large-language-model-generated dynamic functions to precisely control the training process. The core of SDW lies in two functions pre-generated by large models: the task similarity function and the weight computation function. The task similarity function extracts multidimensional features from task descriptions to quantify the similarities and differences between tasks in terms of states, actions, and rewards. The weight computation function dynamically generates critical training parameters based on the similarity information, including the proportion of old task data stored in the Replay Buffer and the strategy consistency weight in the loss function, enabling an adaptive balance between learning new tasks and transferring knowledge from previous tasks. By generating function code offline prior to training, rather than relying on large-model inference during the training process, the SDW framework reduces computational overhead while maintaining efficiency in sequential task scenarios. Experimental results on Atari and MiniHack sequential tasks demonstrate that SDW significantly outperforms existing lifelong reinforcement learning methods.

Traditional multi-agent reinforcement learning (MARL) systems can develop cooperative strategies through repeated interactions. However, these systems are unable to perform well on any other setting than the one they have been trained on, and struggle to successfully cooperate with unfamiliar collaborators. This is particularly visible in the Hanabi benchmark, a popular 2-to-5 player cooperative card-game which requires complex reasoning and precise assistance to other agents. Current MARL agents for Hanabi can only learn one specific game-setting (e.g., 2-player games), and play with the same algorithmic agents. This is in stark contrast to humans, who can quickly adjust their strategies to work with unfamiliar partners or situations. In this paper, we introduce Recurrent Replay Relevance Distributed DQN (R3D2), a generalist agent for Hanabi, designed to overcome these limitations. We reformulate the task using text, as language has been shown to improve transfer. We then propose a distributed MARL algorithm that copes with the resulting dynamic observation-and action-space. In doing so, our agent is the first that can play all game settings concurrently, and extend strategies learned from one setting to other ones. As a consequence, our agent also demonstrates the ability to collaborate with different algorithmic agents -- agents that are themselves unable to do so. Humans were able to thrive as a society through their ability to cooperate. Interactions among multiple people or agents are essential components of various aspects of our lives, ranging from everyday activities like commuting to work, to the functioning of fundamental institutions like governments and economic markets. Through repeated interactions, humans can understand their partners, and learn to reason from their perspective. Crucially, humans can generalize their reasonings towards novel partners, in different situations. Artificial agents should be able to do the same for the successful collaboration of artificial and hybrid systems (Dafoe et al., 2020). This is why defining the problem of multi-agent cooperation nicely fits the multi-agent reinforcement learning (MARL) paradigm, as artificial agents learn to collaborate together through repeated interactions, in the same principled manner humans would. In MARL, the game of Hanabi has emerged as a popular benchmark to assess the cooperative abilities of learning agents (Bard et al., 2020).