PyTorchKR Papers

최신 논문

총 1,970편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2505.17434

Dynamic Manipulation of Deformable Objects in 3D: Simulation, Benchmark and Learning Strategy

Goal-conditioned dynamic manipulation is inherently challenging due to complex system dynamics and stringent task constraints, particularly in deformable object scenarios characterized by high degrees of freedom and underactuation. Prior methods often simplify the problem to low-speed or 2D settings, limiting their applicability to real-world 3D tasks. In this work, we explore 3D goal-conditioned rope manipulation as a representative challenge. To mitigate data scarcity, we introduce a novel simulation framework and benchmark grounded in reduced-order dynamics, which enables compact state representation and facilitates efficient policy learning. Building on this, we propose Dynamics Informed Diffusion Policy (DIDP), a framework that integrates imitation pretraining with physics-informed test-time adaptation. First, we design a diffusion policy that learns inverse dynamics within the reduced-order space, enabling imitation learning to move beyond na\"ive data fitting and capture the underlying physical structure. Second, we propose a physics-informed test-time adaptation scheme that imposes kinematic boundary conditions and structured dynamics priors on the diffusion process, ensuring consistency and reliability in manipulation execution. Extensive experiments validate the proposed approach, demonstrating strong performance in terms of accuracy and robustness in the learned policy.

arXiv:2505.18595

MisoDICE: Multi-Agent Imitation from Unlabeled Mixed-Quality Demonstrations

We study offline imitation learning (IL) in cooperative multi-agent settings, where demonstrations have unlabeled mixed quality - containing both expert and suboptimal trajectories. Our proposed solution is structured in two stages: trajectory labeling and multi-agent imitation learning, designed jointly to enable effective learning from heterogeneous, unlabeled data. In the first stage, we combine advances in large language models and preference-based reinforcement learning to construct a progressive labeling pipeline that distinguishes expert-quality trajectories. In the second stage, we introduce MisoDICE, a novel multi-agent IL algorithm that leverages these labels to learn robust policies while addressing the computational complexity of large joint state-action spaces. By extending the popular single-agent DICE framework to multi-agent settings with a new value decomposition and mixing architecture, our method yields a convex policy optimization objective and ensures consistency between global and local policies. We evaluate MisoDICE on multiple standard multi-agent RL benchmarks and demonstrate superior performance, especially when expert data is scarce.

arXiv:2505.17389

Bootstrapping Imitation Learning for Long-horizon Manipulation via Hierarchical Data Collection Space

Imitation learning (IL) with human demonstrations is a promising method for robotic manipulation tasks. While minimal demonstrations enable robotic action execution, achieving high success rates and generalization requires high cost, e.g., continuously adding data or incrementally conducting human-in-loop processes with complex hardware/software systems. In this paper, we rethink the state/action space of the data collection pipeline as well as the underlying factors responsible for the prediction of non-robust actions. To this end, we introduce a Hierarchical Data Collection Space (HD-Space) for robotic imitation learning, a simple data collection scheme, endowing the model to train with proactive and high-quality data. Specifically, We segment the fine manipulation task into multiple key atomic tasks from a high-level perspective and design atomic state/action spaces for human demonstrations, aiming to generate robust IL data. We conduct empirical evaluations across two simulated and five real-world long-horizon manipulation tasks and demonstrate that IL policy training with HD-Space-based data can achieve significantly enhanced policy performance. HD-Space allows the use of a small amount of demonstration data to train a more powerful policy, particularly for long-horizon manipulation tasks. We aim for HD-Space to offer insights into optimizing data quality and guiding data scaling. project page: https://hd-space-robotics.github.io.

arXiv:2505.16856

Efficient Online RL Fine Tuning with Offline Pre-trained Policy Only

Improving the performance of pre-trained policies through online reinforcement learning (RL) is a critical yet challenging topic. Existing online RL fine-tuning methods require continued training with offline pretrained Q-functions for stability and performance. However, these offline pretrained Q-functions commonly underestimate state-action pairs beyond the offline dataset due to the conservatism in most offline RL methods, which hinders further exploration when transitioning from the offline to the online setting. Additionally, this requirement limits their applicability in scenarios where only pre-trained policies are available but pre-trained Q-functions are absent, such as in imitation learning (IL) pre-training. To address these challenges, we propose a method for efficient online RL fine-tuning using solely the offline pre-trained policy, eliminating reliance on pre-trained Q-functions. We introduce PORL (Policy-Only Reinforcement Learning Fine-Tuning), which rapidly initializes the Q-function from scratch during the online phase to avoid detrimental pessimism. Our method not only achieves competitive performance with advanced offline-to-online RL algorithms and online RL approaches that leverage data or policies prior, but also pioneers a new path for directly fine-tuning behavior cloning (BC) policies.

arXiv:2505.16394

Raw2Drive: Reinforcement Learning with Aligned World Models for End-to-End Autonomous Driving (in CARLA v2)

Reinforcement Learning (RL) can mitigate the causal confusion and distribution shift inherent to imitation learning (IL). However, applying RL to end-to-end autonomous driving (E2E-AD) remains an open problem for its training difficulty, and IL is still the mainstream paradigm in both academia and industry. Recently Model-based Reinforcement Learning (MBRL) have demonstrated promising results in neural planning; however, these methods typically require privileged information as input rather than raw sensor data. We fill this gap by designing Raw2Drive, a dual-stream MBRL approach. Initially, we efficiently train an auxiliary privileged world model paired with a neural planner that uses privileged information as input. Subsequently, we introduce a raw sensor world model trained via our proposed Guidance Mechanism, which ensures consistency between the raw sensor world model and the privileged world model during rollouts. Finally, the raw sensor world model combines the prior knowledge embedded in the heads of the privileged world model to effectively guide the training of the raw sensor policy. Raw2Drive is so far the only RL based end-to-end method on CARLA Leaderboard 2.0, and Bench2Drive and it achieves state-of-the-art performance.

arXiv:2505.15725

UAV-Flow Colosseo: A Real-World Benchmark for Flying-on-a-Word UAV Imitation Learning

Unmanned Aerial Vehicles (UAVs) are evolving into language-interactive platforms, enabling more intuitive forms of human-drone interaction. While prior works have primarily focused on high-level planning and long-horizon navigation, we shift attention to language-guided fine-grained trajectory control, where UAVs execute short-range, reactive flight behaviors in response to language instructions. We formalize this problem as the Flying-on-a-Word (Flow) task and introduce UAV imitation learning as an effective approach. In this framework, UAVs learn fine-grained control policies by mimicking expert pilot trajectories paired with atomic language instructions. To support this paradigm, we present UAV-Flow, the first real-world benchmark for language-conditioned, fine-grained UAV control. It includes a task formulation, a large-scale dataset collected in diverse environments, a deployable control framework, and a simulation suite for systematic evaluation. Our design enables UAVs to closely imitate the precise, expert-level flight trajectories of human pilots and supports direct deployment without sim-to-real gap. We conduct extensive experiments on UAV-Flow, benchmarking VLN and VLA paradigms. Results show that VLA models are superior to VLN baselines and highlight the critical role of spatial grounding in the fine-grained Flow setting.

arXiv:2505.14820

Imitation Learning via Focused Satisficing

Imitation learning often assumes that demonstrations are close to optimal according to some fixed, but unknown, cost function. However, according to satisficing theory, humans often choose acceptable behavior based on their personal (and potentially dynamic) levels of aspiration, rather than achieving (near-) optimality. For example, a lunar lander demonstration that successfully lands without crashing might be acceptable to a novice despite being slow or jerky. Using a margin-based objective to guide deep reinforcement learning, our focused satisficing approach to imitation learning seeks a policy that surpasses the demonstrator's aspiration levels -- defined over trajectories or portions of trajectories -- on unseen demonstrations without explicitly learning those aspirations. We show experimentally that this focuses the policy to imitate the highest quality (portions of) demonstrations better than existing imitation learning methods, providing much higher rates of guaranteed acceptability to the demonstrator, and competitive true returns on a range of environments.

arXiv:2505.15659

FLARE: Robot Learning with Implicit World Modeling

We introduce $\textbf{F}$uture $\textbf{LA}$tent $\textbf{RE}$presentation Alignment ($\textbf{FLARE}$), a novel framework that integrates predictive latent world modeling into robot policy learning. By aligning features from a diffusion transformer with latent embeddings of future observations, $\textbf{FLARE}$ enables a diffusion transformer policy to anticipate latent representations of future observations, allowing it to reason about long-term consequences while generating actions. Remarkably lightweight, $\textbf{FLARE}$ requires only minimal architectural modifications -- adding a few tokens to standard vision-language-action (VLA) models -- yet delivers substantial performance gains. Across two challenging multitask simulation imitation learning benchmarks spanning single-arm and humanoid tabletop manipulation, $\textbf{FLARE}$ achieves state-of-the-art performance, outperforming prior policy learning baselines by up to 26%. Moreover, $\textbf{FLARE}$ unlocks the ability to co-train with human egocentric video demonstrations without action labels, significantly boosting policy generalization to a novel object with unseen geometry with as few as a single robot demonstration. Our results establish $\textbf{FLARE}$ as a general and scalable approach for combining implicit world modeling with high-frequency robotic control.

arXiv:2505.15275

Learning-based Autonomous Oversteer Control and Collision Avoidance

Oversteer, wherein a vehicle's rear tires lose traction and induce unintentional excessive yaw, poses critical safety challenges. Failing to control oversteer often leads to severe traffic accidents. Although recent autonomous driving efforts have attempted to handle oversteer through stabilizing maneuvers, the majority rely on expert-defined trajectories or assume obstacle-free environments, limiting real-world applicability. This paper introduces a novel end-to-end (E2E) autonomous driving approach that tackles oversteer control and collision avoidance simultaneously. Existing E2E techniques, including Imitation Learning (IL), Reinforcement Learning (RL), and Hybrid Learning (HL), generally require near-optimal demonstrations or extensive experience. Yet even skilled human drivers struggle to provide perfect demonstrations under oversteer, and high transition variance hinders accumulating sufficient data. Hence, we present Q-Compared Soft Actor-Critic (QC-SAC), a new HL algorithm that effectively learns from suboptimal demonstration data and adapts rapidly to new conditions. To evaluate QC-SAC, we introduce a benchmark inspired by real-world driver training: a vehicle encounters sudden oversteer on a slippery surface and must avoid randomly placed obstacles ahead. Experimental results show QC-SAC attains near-optimal driving policies, significantly surpassing state-of-the-art IL, RL, and HL baselines. Our method demonstrates the world's first safe autonomous oversteer control with obstacle avoidance.

arXiv:2505.15418공식 코드 · Papers with Code

Guided Policy Optimization under Partial Observability

Reinforcement Learning (RL) in partially observable environments poses significant challenges due to the complexity of learning under uncertainty. While additional information, such as that available in simulations, can enhance training, effectively leveraging it remains an open problem. To address this, we introduce Guided Policy Optimization (GPO), a framework that co-trains a guider and a learner. The guider takes advantage of privileged information while ensuring alignment with the learner's policy that is primarily trained via imitation learning. We theoretically demonstrate that this learning scheme achieves optimality comparable to direct RL, thereby overcoming key limitations inherent in existing approaches. Empirical evaluations show strong performance of GPO across various tasks, including continuous control with partial observability and noise, and memory-based challenges, significantly outperforming existing methods.

arXiv:2505.13820

Structured Agent Distillation for Large Language Model

Large language models (LLMs) exhibit strong capabilities as decision-making agents by interleaving reasoning and actions, as seen in ReAct-style frameworks. Yet, their practical deployment is constrained by high inference costs and large model sizes. We propose Structured Agent Distillation, a framework that compresses large LLM-based agents into smaller student models while preserving both reasoning fidelity and action consistency. Unlike standard token-level distillation, our method segments trajectories into {[REASON]} and {[ACT]} spans, applying segment-specific losses to align each component with the teacher's behavior. This structure-aware supervision enables compact agents to better replicate the teacher's decision process. Experiments on ALFWorld, HotPotQA-ReAct, and WebShop show that our approach consistently outperforms token-level and imitation learning baselines, achieving significant compression with minimal performance drop. Scaling and ablation results further highlight the importance of span-level alignment for efficient and deployable agents.

arXiv:2505.11719

Zero-Shot Visual Generalization in Robot Manipulation

Training vision-based manipulation policies that are robust across diverse visual environments remains an important and unresolved challenge in robot learning. Current approaches often sidestep the problem by relying on invariant representations such as point clouds and depth, or by brute-forcing generalization through visual domain randomization and/or large, visually diverse datasets. Disentangled representation learning - especially when combined with principles of associative memory - has recently shown promise in enabling vision-based reinforcement learning policies to be robust to visual distribution shifts. However, these techniques have largely been constrained to simpler benchmarks and toy environments. In this work, we scale disentangled representation learning and associative memory to more visually and dynamically complex manipulation tasks and demonstrate zero-shot adaptability to visual perturbations in both simulation and on real hardware. We further extend this approach to imitation learning, specifically Diffusion Policy, and empirically show significant gains in visual generalization compared to state-of-the-art imitation learning methods. Finally, we introduce a novel technique adapted from the model equivariance literature that transforms any trained neural network policy into one invariant to 2D planar rotations, making our policy not only visually robust but also resilient to certain camera perturbations. We believe that this work marks a significant step towards manipulation policies that are not only adaptable out of the box, but also robust to the complexities and dynamical nature of real-world deployment. Supplementary videos are available at https://sites.google.com/view/vis-gen-robotics/home.

arXiv:2505.11709

EgoDex: Learning Dexterous Manipulation from Large-Scale Egocentric Video

Imitation learning for manipulation has a well-known data scarcity problem. Unlike natural language and 2D computer vision, there is no Internet-scale corpus of data for dexterous manipulation. One appealing option is egocentric human video, a passively scalable data source. However, existing large-scale datasets such as Ego4D do not have native hand pose annotations and do not focus on object manipulation. To this end, we use Apple Vision Pro to collect EgoDex: the largest and most diverse dataset of dexterous human manipulation to date. EgoDex has 829 hours of egocentric video with paired 3D hand and finger tracking data collected at the time of recording, where multiple calibrated cameras and on-device SLAM can be used to precisely track the pose of every joint of each hand. The dataset covers a wide range of diverse manipulation behaviors with everyday household objects in 194 different tabletop tasks ranging from tying shoelaces to folding laundry. Furthermore, we train and systematically evaluate imitation learning policies for hand trajectory prediction on the dataset, introducing metrics and benchmarks for measuring progress in this increasingly important area. By releasing this large-scale dataset, we hope to push the frontier of robotics, computer vision, and foundation models.

arXiv:2505.13436

KinTwin: Imitation Learning with Torque and Muscle Driven Biomechanical Models Enables Precise Replication of Able-Bodied and Impaired Movement from Markerless Motion Capture

Broader access to high-quality movement analysis could greatly benefit movement science and rehabilitation, such as allowing more detailed characterization of movement impairments and responses to interventions, or even enabling early detection of new neurological conditions or fall risk. While emerging technologies are making it easier to capture kinematics with biomechanical models, or how joint angles change over time, inferring the underlying physics that give rise to these movements, including ground reaction forces, joint torques, or even muscle activations, is still challenging. Here we explore whether imitation learning applied to a biomechanical model from a large dataset of movements from able-bodied and impaired individuals can learn to compute these inverse dynamics. Although imitation learning in human pose estimation has seen great interest in recent years, our work differences in several ways: we focus on using an accurate biomechanical model instead of models adopted for computer vision, we test it on a dataset that contains participants with impaired movements, we reported detailed tracking metrics relevant for the clinical measurement of movement including joint angles and ground contact events, and finally we apply imitation learning to a muscle-driven neuromusculoskeletal model. We show that our imitation learning policy, KinTwin, can accurately replicate the kinematics of a wide range of movements, including those with assistive devices or therapist assistance, and that it can infer clinically meaningful differences in joint torques and muscle activations. Our work demonstrates the potential for using imitation learning to enable high-quality movement analysis in clinical practice.

arXiv:2505.09723

EnerVerse-AC: Envisioning Embodied Environments with Action Condition

Robotic imitation learning has advanced from solving static tasks to addressing dynamic interaction scenarios, but testing and evaluation remain costly and challenging due to the need for real-time interaction with dynamic environments. We propose EnerVerse-AC (EVAC), an action-conditional world model that generates future visual observations based on an agent's predicted actions, enabling realistic and controllable robotic inference. Building on prior architectures, EVAC introduces a multi-level action-conditioning mechanism and ray map encoding for dynamic multi-view image generation while expanding training data with diverse failure trajectories to improve generalization. As both a data engine and evaluator, EVAC augments human-collected trajectories into diverse datasets and generates realistic, action-conditioned video observations for policy testing, eliminating the need for physical robots or complex simulations. This approach significantly reduces costs while maintaining high fidelity in robotic manipulation evaluation. Extensive experiments validate the effectiveness of our method. Code, checkpoints, and datasets can be found at <https://annaj2178.github.io/EnerverseAC.github.io>.

arXiv:2505.10442

IN-RIL: Interleaved Reinforcement and Imitation Learning for Policy Fine-Tuning

Imitation learning (IL) and reinforcement learning (RL) each offer distinct advantages for robotics policy learning: IL provides stable learning from demonstrations, and RL promotes generalization through exploration. While existing robot learning approaches using IL-based pre-training followed by RL-based fine-tuning are promising, this two-step learning paradigm often suffers from instability and poor sample efficiency during the RL fine-tuning phase. In this work, we introduce IN-RIL, INterleaved Reinforcement learning and Imitation Learning, for policy fine-tuning, which periodically injects IL updates after multiple RL updates and hence can benefit from the stability of IL and the guidance of expert data for more efficient exploration throughout the entire fine-tuning process. Since IL and RL involve different optimization objectives, we develop gradient separation mechanisms to prevent destructive interference during \ABBR fine-tuning, by separating possibly conflicting gradient updates in orthogonal subspaces. Furthermore, we conduct rigorous analysis, and our findings shed light on why interleaving IL with RL stabilizes learning and improves sample-efficiency. Extensive experiments on 14 robot manipulation and locomotion tasks across 3 benchmarks, including FurnitureBench, OpenAI Gym, and Robomimic, demonstrate that \ABBR can significantly improve sample efficiency and mitigate performance collapse during online finetuning in both long- and short-horizon tasks with either sparse or dense rewards. IN-RIL, as a general plug-in compatible with various state-of-the-art RL algorithms, can significantly improve RL fine-tuning, e.g., from 12\% to 88\% with 6.3x improvement in the success rate on Robomimic Transport. Project page: https://github.com/ucd-dare/IN-RIL.

arXiv:2505.09603

DataMIL: Selecting Data for Robot Imitation Learning with Datamodels

Recently, the robotics community has amassed ever larger and more diverse datasets to train generalist robot policies. However, while these policies achieve strong mean performance across a variety of tasks, they often underperform on individual, specialized tasks and require further tuning on newly acquired task-specific data. Combining task-specific data with carefully curated subsets of large prior datasets via co-training can produce better specialized policies, but selecting data naively may actually harm downstream performance. To address this, we introduce DataMIL, a policy-driven data selection framework built on the datamodels paradigm that reasons about data selection in an end-to-end manner, using the policy itself to identify which data points will most improve performance. Unlike standard practices that filter data using human notions of quality (e.g., based on semantic or visual similarity), DataMIL directly optimizes data selection for task success, allowing us to select data that enhance the policy while dropping data that degrade it. To avoid performing expensive rollouts in the environment during selection, we use a novel surrogate loss function on task-specific data, allowing us to use DataMIL in the real world without degrading performance. We validate our approach on a suite of more than 60 simulation and real-world manipulation tasks - most notably showing successful data selection from the Open X-Embodiment datasets-demonstrating consistent gains in success rates and superior performance over multiple baselines. Our results underscore the importance of end-to-end, performance-aware data selection for unlocking the potential of large prior datasets in robotics. More information at https://robin-lab.cs.utexas.edu/datamodels4imitation/

arXiv:2505.09546

Distilling Realizable Students from Unrealizable Teachers

We study policy distillation under privileged information, where a student policy with only partial observations must learn from a teacher with full-state access. A key challenge is information asymmetry: the student cannot directly access the teacher's state space, leading to distributional shifts and policy degradation. Existing approaches either modify the teacher to produce realizable but sub-optimal demonstrations or rely on the student to explore missing information independently, both of which are inefficient. Our key insight is that the student should strategically interact with the teacher --querying only when necessary and resetting from recovery states --to stay on a recoverable path within its own observation space. We introduce two methods: (i) an imitation learning approach that adaptively determines when the student should query the teacher for corrections, and (ii) a reinforcement learning approach that selects where to initialize training for efficient exploration. We validate our methods in both simulated and real-world robotic tasks, demonstrating significant improvements over standard teacher-student baselines in training efficiency and final performance. The project website is available at : https://portal-cornell.github.io/CritiQ_ReTRy/

arXiv:2505.09308

Neural Multivariate Regression: Qualitative Insights from the Unconstrained Feature Model

The Unconstrained Feature Model (UFM) is a mathematical framework that enables closed-form approximations for minimal training loss and related performance measures in deep neural networks (DNNs). This paper leverages the UFM to provide qualitative insights into neural multivariate regression, a critical task in imitation learning, robotics, and reinforcement learning. Specifically, we address two key questions: (1) How do multi-task models compare to multiple single-task models in terms of training performance? (2) Can whitening and normalizing regression targets improve training performance? The UFM theory predicts that multi-task models achieve strictly smaller training MSE than multiple single-task models when the same or stronger regularization is applied to the latter, and our empirical results confirm these findings. Regarding whitening and normalizing regression targets, the UFM theory predicts that they reduce training MSE when the average variance across the target dimensions is less than one, and our empirical results once again confirm these findings. These findings highlight the UFM as a powerful framework for deriving actionable insights into DNN design and data pre-processing strategies.

arXiv:2505.09099

Imitation Learning for Adaptive Control of a Virtual Soft Exoglove

The use of wearable robots has been widely adopted in rehabilitation training for patients with hand motor impairments. However, the uniqueness of patients' muscle loss is often overlooked. Leveraging reinforcement learning and a biologically accurate musculoskeletal model in simulation, we propose a customized wearable robotic controller that is able to address specific muscle deficits and to provide compensation for hand-object manipulation tasks. Video data of a same subject performing human grasping tasks is used to train a manipulation model through learning from demonstration. This manipulation model is subsequently fine-tuned to perform object-specific interaction tasks. The muscle forces in the musculoskeletal manipulation model are then weakened to simulate neurological motor impairments, which are later compensated by the actuation of a virtual wearable robotics glove. Results shows that integrating the virtual wearable robotic glove provides shared assistance to support the hand manipulator with weakened muscle forces. The learned exoglove controller achieved an average of 90.5\% of the original manipulation proficiency.