arXiv subjects

Peter Stone

Publications and source records attributed to Peter Stone.

At least 55 records · Page 3Linked to original sources

arXiv2025

MEReQ: Max-Ent Residual-Q Inverse RL for Sample-Efficient Alignment from Intervention

Aligning robot behavior with human preferences is crucial for deploying embodied AI agents in human-centered environments. A promising solution is interactive imitation learning from human intervention, where a human expert observes the policy's execution and provides interventions as feedback. However, existing methods often fail to utilize the prior policy efficiently to facilitate learning, thus hindering sample efficiency. In this work, we introduce MEReQ (Maximum-Entropy Residual-Q Inverse Reinforcement Learning), designed for sample-efficient alignment from human intervention. Instead of inferring the complete human behavior characteristics, MEReQ infers a residual reward function that captures the discrepancy between the human expert's and the prior policy's underlying reward functions. It then employs Residual Q-Learning (RQL) to align the policy with human preferences using this residual reward function. Extensive evaluations on simulated and real-world tasks demonstrate that MEReQ achieves sample-efficient policy alignment from human intervention.

cs.RO↗

arXiv2025

Sequence Modeling for N-Agent Ad Hoc Teamwork

N-agent ad hoc teamwork (NAHT) is a newly introduced challenge in multi-agent reinforcement learning, where controlled subteams of varying sizes must dynamically collaborate with varying numbers and types of unknown teammates without pre-coordination. The existing learning algorithm (POAM) considers only independent learning for its flexibility in dealing with a changing number of agents. However, independent learning fails to fully capture the inter-agent dynamics essential for effective collaboration. Based on our observation that transformers deal effectively with sequences with varying lengths and have been shown to be highly effective for a variety of machine learning problems, this work introduces a centralized, transformer-based method for N-agent ad hoc teamwork. Our proposed approach incorporates historical observations and actions of all controlled agents, enabling optimal responses to diverse and unseen teammates in partially observable environments. Empirical evaluation on a StarCraft II task demonstrates that MAT-NAHT outperforms POAM, achieving superior sample efficiency and generalization, without auxiliary agent-modeling objectives.

cs.MA↗

arXiv2025

Adversarial Reinforcement Learning for Large Language Model Agent Safety

Large Language Model (LLM) agents can leverage tools such as Google Search to complete complex tasks. However, this tool usage introduces the risk of indirect prompt injections, where malicious instructions hidden in tool outputs can manipulate the agent, posing security risks like data leakage. Current defense strategies typically rely on fine-tuning LLM agents on datasets of known attacks. However, the generation of these datasets relies on manually crafted attack patterns, which limits their diversity and leaves agents vulnerable to novel prompt injections. To address this limitation, we propose Adversarial Reinforcement Learning for Agent Safety (ARLAS), a novel framework that leverages adversarial reinforcement learning (RL) by formulating the problem as a two-player zero-sum game. ARLAS co-trains two LLMs: an attacker that learns to autonomously generate diverse prompt injections and an agent that learns to defend against them while completing its assigned tasks. To ensure robustness against a wide range of attacks and to prevent cyclic learning, we employ a population-based learning framework that trains the agent to defend against all previous attacker checkpoints. Evaluated on BrowserGym and AgentDojo, agents fine-tuned with ARLAS achieve a significantly lower attack success rate than the original model while also improving their task success rate. Our analysis further confirms that the adversarial process generates a diverse and challenging set of attacks, leading to a more robust agent compared to the base model.

cs.LG↗

arXiv2025

ComposableNav: Instruction-Following Navigation in Dynamic Environments via Composable Diffusion

This paper considers the problem of enabling robots to navigate dynamic environments while following instructions. The challenge lies in the combinatorial nature of instruction specifications: each instruction can include multiple specifications, and the number of possible specification combinations grows exponentially as the robot's skill set expands. For example, "overtake the pedestrian while staying on the right side of the road" consists of two specifications: "overtake the pedestrian" and "walk on the right side of the road." To tackle this challenge, we propose ComposableNav, based on the intuition that following an instruction involves independently satisfying its constituent specifications, each corresponding to a distinct motion primitive. Using diffusion models, ComposableNav learns each primitive separately, then composes them in parallel at deployment time to satisfy novel combinations of specifications unseen in training. Additionally, to avoid the onerous need for demonstrations of individual motion primitives, we propose a two-stage training procedure: (1) supervised pre-training to learn a base diffusion model for dynamic navigation, and (2) reinforcement learning fine-tuning that molds the base model into different motion primitives. Through simulation and real-world experiments, we show that ComposableNav enables robots to follow instructions by generating trajectories that satisfy diverse and unseen combinations of specifications, significantly outperforming both non-compositional VLM-based policies and costmap composing baselines. Videos and additional materials can be found on the project page: https://amrl.cs.utexas.edu/ComposableNav/

cs.RO↗

arXiv2025

SocialNav-SUB: Benchmarking VLMs for Scene Understanding in Social Robot Navigation

Robot navigation in dynamic, human-centered environments requires socially-compliant decisions grounded in robust scene understanding. Recent Vision-Language Models (VLMs) exhibit promising capabilities such as object recognition, common-sense reasoning, and contextual understanding-capabilities that align with the nuanced requirements of social robot navigation. However, it remains unclear whether VLMs can accurately understand complex social navigation scenes (e.g., inferring the spatial-temporal relations among agents and human intentions), which is essential for safe and socially compliant robot navigation. While some recent works have explored the use of VLMs in social robot navigation, no existing work systematically evaluates their ability to meet these necessary conditions. In this paper, we introduce the Social Navigation Scene Understanding Benchmark (SocialNav-SUB), a Visual Question Answering (VQA) dataset and benchmark designed to evaluate VLMs for scene understanding in real-world social robot navigation scenarios. SocialNav-SUB provides a unified framework for evaluating VLMs against human and rule-based baselines across VQA tasks requiring spatial, spatiotemporal, and social reasoning in social robot navigation. Through experiments with state-of-the-art VLMs, we find that while the best-performing VLM achieves an encouraging probability of agreeing with human answers, it still underperforms simpler rule-based approach and human consensus baselines, indicating critical gaps in social scene understanding of current VLMs. Our benchmark sets the stage for further research on foundation models for social robot navigation, offering a framework to explore how VLMs can be tailored to meet real-world social robot navigation needs. An overview of this paper along with the code and data can be found at https://larg.github.io/socialnav-sub .

cs.RO↗

arXiv2025

Vision-based Manipulation from Single Human Video with Open-World Object Graphs

This work presents an object-centric approach to learning vision-based manipulation skills from human videos. We investigate the problem of robot manipulation via imitation in the open-world setting, where a robot learns to manipulate novel objects from a single video demonstration. We introduce ORION, an algorithm that tackles the problem by extracting an object-centric manipulation plan from a single RGB or RGB-D video and deriving a policy that conditions on the extracted plan. Our method enables the robot to learn from videos captured by daily mobile devices and to generalize the policies to deployment environments with varying visual backgrounds, camera angles, spatial layouts, and novel object instances. We systematically evaluate our method on both short-horizon and long-horizon tasks, using RGB-D and RGB-only demonstration videos. Across varied tasks and demonstration types (RGB-D / RGB), we observe an average success rate of 74.4%, demonstrating the efficacy of ORION in learning from a single human video in the open world. Additional materials can be found on our project website: https://ut-austin-rpl.github.io/ORION-release.

cs.RO↗

arXiv2025

Dyna-LfLH: Learning Agile Navigation in Dynamic Environments from Learned Hallucination

This paper introduces Dynamic Learning from Learned Hallucination (Dyna-LfLH), a self-supervised method for training motion planners to navigate environments with dense and dynamic obstacles. Classical planners struggle with dense, unpredictable obstacles due to limited computation, while learning-based planners face challenges in acquiring high-quality demonstrations for imitation learning or dealing with exploration inefficiencies in reinforcement learning. Building on Learning from Hallucination (LfH), which synthesizes training data from past successful navigation experiences in simpler environments, Dyna-LfLH incorporates dynamic obstacles by generating them through a learned latent distribution. This enables efficient and safe motion planner training. We evaluate Dyna-LfLH on a ground robot in both simulated and real environments, achieving up to a 25% improvement in success rate compared to baselines.

cs.RO↗

arXiv2025

GACL: Grounded Adaptive Curriculum Learning with Active Task and Performance Monitoring

Curriculum learning has emerged as a promising approach for training complex robotics tasks, yet current applications predominantly rely on manually designed curricula, which demand significant engineering effort and can suffer from subjective and suboptimal human design choices. While automated curriculum learning has shown success in simple domains like grid worlds and games where task distributions can be easily specified, robotics tasks present unique challenges: they require handling complex task spaces while maintaining relevance to target domain distributions that are only partially known through limited samples. To this end, we propose Grounded Adaptive Curriculum Learning, a framework specifically designed for robotics curriculum learning with three key innovations: (1) a task representation that consistently handles complex robot task design, (2) an active performance tracking mechanism that allows adaptive curriculum generation appropriate for the robot's current capabilities, and (3) a grounding approach that maintains target domain relevance through alternating sampling between reference and synthetic tasks. We validate GACL on wheeled navigation in constrained environments and quadruped locomotion in challenging 3D confined spaces, achieving 6.8% and 6.1% higher success rates, respectively, than state-of-the-art methods in each domain.

cs.RO↗

arXiv2025

L3M+P: Lifelong Planning with Large Language Models

By combining classical planning methods with large language models (LLMs), recent research such as LLM+P has enabled agents to plan for general tasks given in natural language. However, scaling these methods to general-purpose service robots remains challenging: (1) classical planning algorithms generally require a detailed and consistent specification of the environment, which is not always readily available; and (2) existing frameworks mainly focus on isolated planning tasks, whereas robots are often meant to serve in long-term continuous deployments, and therefore must maintain a dynamic memory of the environment which can be updated with multi-modal inputs and extracted as planning knowledge for future tasks. To address these two issues, this paper introduces L3M+P (Lifelong LLM+P), a framework that uses an external knowledge graph as a representation of the world state. The graph can be updated from multiple sources of information, including sensory input and natural language interactions with humans. L3M+P enforces rules for the expected format of the absolute world state graph to maintain consistency between graph updates. At planning time, given a natural language description of a task, L3M+P retrieves context from the knowledge graph and generates a problem definition for classical planners. Evaluated on household robot simulators and on a real-world service robot, L3M+P achieves significant improvement over baseline methods both on accurately registering natural language state changes and on correctly generating plans, thanks to the knowledge graph retrieval and verification.

cs.RO↗

arXiv2025

Benchmarking Massively Parallelized Multi-Task Reinforcement Learning for Robotics Tasks

Multi-task Reinforcement Learning (MTRL) has emerged as a critical training paradigm for applying reinforcement learning (RL) to a set of complex real-world robotic tasks, which demands a generalizable and robust policy. At the same time, \emph{massively parallelized training} has gained popularity, not only for significantly accelerating data collection through GPU-accelerated simulation but also for enabling diverse data collection across multiple tasks by simulating heterogeneous scenes in parallel. However, existing MTRL research has largely been limited to off-policy methods like SAC in the low-parallelization regime. MTRL could capitalize on the higher asymptotic performance of on-policy algorithms, whose batches require data from the current policy, and as a result, take advantage of massive parallelization offered by GPU-accelerated simulation. To bridge this gap, we introduce a massively parallelized $\textbf{M}$ulti-$\textbf{T}$ask $\textbf{Bench}$mark for robotics (MTBench), an open-sourced benchmark featuring a broad distribution of 50 manipulation tasks and 20 locomotion tasks, implemented using the GPU-accelerated simulator IsaacGym. MTBench also includes four base RL algorithms combined with seven state-of-the-art MTRL algorithms and architectures, providing a unified framework for evaluating their performance. Our extensive experiments highlight the superior speed of evaluating MTRL approaches using MTBench, while also uncovering unique challenges that arise from combining massive parallelism with MTRL. Code is available at https://github.com/Viraj-Joshi/MTBench

cs.RO↗

arXiv2025

Dyn-O: Building Structured World Models with Object-Centric Representations

World models aim to capture the dynamics of the environment, enabling agents to predict and plan for future states. In most scenarios of interest, the dynamics are highly centered on interactions among objects within the environment. This motivates the development of world models that operate on object-centric rather than monolithic representations, with the goal of more effectively capturing environment dynamics and enhancing compositional generalization. However, the development of object-centric world models has largely been explored in environments with limited visual complexity (such as basic geometries). It remains underexplored whether such models can generalize to more complex settings with diverse textures and cluttered scenes. In this paper, we fill this gap by introducing Dyn-O, an enhanced structured world model built upon object-centric representations. Compared to prior work in object-centric representations, Dyn-O improves in both learning representations and modeling dynamics. On the challenging Procgen games, we find that our method can learn object-centric world models directly from pixel observations, outperforming DreamerV3 in rollout prediction accuracy. Furthermore, by decoupling object-centric features into dynamics-agnostic and dynamics-aware components, we enable finer-grained manipulation of these features and generate more diverse imagined trajectories.

cs.LG↗

arXiv2025

LaRS: Latent Reasoning Skills for Chain-of-Thought Reasoning

Chain-of-thought (CoT) prompting is a popular in-context learning (ICL) approach for large language models (LLMs), especially when tackling complex reasoning tasks. Traditional ICL approaches construct prompts using examples that contain questions similar to the input question. However, CoT prompting, which includes crucial intermediate reasoning steps (rationales) within its examples, necessitates selecting examples based on these rationales rather than the questions themselves. Existing methods require human experts or pre-trained LLMs to describe the skill, a high-level abstraction of rationales, to guide the selection. These methods, however, are often costly and difficult to scale. Instead, this paper introduces a new approach named Latent Reasoning Skills (LaRS) that employs unsupervised learning to create a latent space representation of rationales, with a latent variable called a reasoning skill. Concurrently, LaRS learns a reasoning policy to determine the required reasoning skill for a given question. Then the ICL examples are selected by aligning the reasoning skills between past examples and the question. This approach is theoretically grounded and compute-efficient, eliminating the need for auxiliary LLM inference or manual prompt design. Empirical results demonstrate that LaRS consistently outperforms SOTA skill-based selection methods, processing example banks four times faster, reducing LLM inferences during the selection stage by half, and showing greater robustness to sub-optimal example banks.

cs.CL↗

arXiv2025

SimBa: Simplicity Bias for Scaling Up Parameters in Deep Reinforcement Learning

Recent advances in CV and NLP have been largely driven by scaling up the number of network parameters, despite traditional theories suggesting that larger networks are prone to overfitting. These large networks avoid overfitting by integrating components that induce a simplicity bias, guiding models toward simple and generalizable solutions. However, in deep RL, designing and scaling up networks have been less explored. Motivated by this opportunity, we present SimBa, an architecture designed to scale up parameters in deep RL by injecting a simplicity bias. SimBa consists of three components: (i) an observation normalization layer that standardizes inputs with running statistics, (ii) a residual feedforward block to provide a linear pathway from the input to output, and (iii) a layer normalization to control feature magnitudes. By scaling up parameters with SimBa, the sample efficiency of various deep RL algorithms-including off-policy, on-policy, and unsupervised methods-is consistently improved. Moreover, solely by integrating SimBa architecture into SAC, it matches or surpasses state-of-the-art deep RL methods with high computational efficiency across DMC, MyoSuite, and HumanoidBench. These results demonstrate SimBa's broad applicability and effectiveness across diverse RL algorithms and environments.

cs.LG↗

arXiv2025

Hyperspherical Normalization for Scalable Deep Reinforcement Learning

Scaling up the model size and computation has brought consistent performance improvements in supervised learning. However, this lesson often fails to apply to reinforcement learning (RL) because training the model on non-stationary data easily leads to overfitting and unstable optimization. In response, we introduce SimbaV2, a novel RL architecture designed to stabilize optimization by (i) constraining the growth of weight and feature norm by hyperspherical normalization; and (ii) using a distributional value estimation with reward scaling to maintain stable gradients under varying reward magnitudes. Using the soft actor-critic as a base algorithm, SimbaV2 scales up effectively with larger models and greater compute, achieving state-of-the-art performance on 57 continuous control tasks across 4 domains. The code is available at https://dojeon-ai.github.io/SimbaV2.

cs.LG↗

arXiv2025

PACER: Preference-conditioned All-terrain Costmap Generation

In autonomous robot navigation, terrain cost assignment is typically performed using a semantics-based paradigm in which terrain is first labeled using a pre-trained semantic classifier and costs are then assigned according to a user-defined mapping between label and cost. While this approach is rapidly adaptable to changing user preferences, only preferences over the types of terrain that are already known by the semantic classifier can be expressed. In this paper, we hypothesize that a machine-learning-based alternative to the semantics-based paradigm above will allow for rapid cost assignment adaptation to preferences expressed over new terrains at deployment time without the need for additional training. To investigate this hypothesis, we introduce and study PACER, a novel approach to costmap generation that accepts as input a single birds-eye view (BEV) image of the surrounding area along with a user-specified preference context and generates a corresponding BEV costmap that aligns with the preference context. Using both real and synthetic data along with a combination of proposed training tasks, we find that PACER is able to adapt quickly to new user preferences while also exhibiting better generalization to novel terrains compared to both semantics-based and representation-learning approaches.

cs.RO↗

arXiv2025

Multi-Agent Inverse Reinforcement Learning in Real World Unstructured Pedestrian Crowds

Social robot navigation in crowded public spaces such as university campuses, restaurants, grocery stores, and hospitals, is an increasingly important area of research. One of the core strategies for achieving this goal is to understand humans' intent--underlying psychological factors that govern their motion--by learning their reward functions, typically via inverse reinforcement learning (IRL). Despite significant progress in IRL, learning reward functions of multiple agents simultaneously in dense unstructured pedestrian crowds has remained intractable due to the nature of the tightly coupled social interactions that occur in these scenarios \textit{e.g.} passing, intersections, swerving, weaving, etc. In this paper, we present a new multi-agent maximum entropy inverse reinforcement learning algorithm for real world unstructured pedestrian crowds. Key to our approach is a simple, but effective, mathematical trick which we name the so-called tractability-rationality trade-off trick that achieves tractability at the cost of a slight reduction in accuracy. We compare our approach to the classical single-agent MaxEnt IRL as well as state-of-the-art trajectory prediction methods on several datasets including the ETH, UCY, SCAND, JRDB, and a new dataset, called Speedway, collected at a busy intersection on a University campus focusing on dense, complex agent interactions. Our key findings show that, on the dense Speedway dataset, our approach ranks 1st among top 7 baselines with >2X improvement over single-agent IRL, and is competitive with state-of-the-art large transformer-based encoder-decoder models on sparser datasets such as ETH/UCY (ranks 3rd among top 7 baselines).

cs.RO↗

arXiv2025

Offline Action-Free Learning of Ex-BMDPs by Comparing Diverse Datasets

While sequential decision-making environments often involve high-dimensional observations, not all features of these observations are relevant for control. In particular, the observation space may capture factors of the environment which are not controllable by the agent, but which add complexity to the observation space. The need to ignore these "noise" features in order to operate in a tractably-small state space poses a challenge for efficient policy learning. Due to the abundance of video data available in many such environments, task-independent representation learning from action-free offline data offers an attractive solution. However, recent work has highlighted theoretical limitations in action-free learning under the Exogenous Block MDP (Ex-BMDP) model, where temporally-correlated noise features are present in the observations. To address these limitations, we identify a realistic setting where representation learning in Ex-BMDPs becomes tractable: when action-free video data from multiple agents with differing policies are available. Concretely, this paper introduces CRAFT (Comparison-based Representations from Action-Free Trajectories), a sample-efficient algorithm leveraging differences in controllable feature dynamics across agents to learn representations. We provide theoretical guarantees for CRAFT's performance and demonstrate its feasibility on a toy example, offering a foundation for practical methods in similar settings.

cs.LG↗

arXiv2025

PRESTO: Fast Motion Planning Using Diffusion Models Based on Key-Configuration Environment Representation

We introduce a learning-guided motion planning framework that generates seed trajectories using a diffusion model for trajectory optimization. Given a workspace, our method approximates the configuration space (C-space) obstacles through an environment representation consisting of a sparse set of task-related key configurations, which is then used as a conditioning input to the diffusion model. The diffusion model integrates regularization terms that encourage smooth, collision-free trajectories during training, and trajectory optimization refines the generated seed trajectories to correct any colliding segments. Our experimental results demonstrate that high-quality trajectory priors, learned through our C-space-grounded diffusion model, enable the efficient generation of collision-free trajectories in narrow-passage environments, outperforming previous learning- and planning-based baselines. Videos and additional materials can be found on the project page: https://kiwi-sherbet.github.io/PRESTO.

cs.RO↗