arXiv subjects

Kechun Xu

Publications and source records attributed to Kechun Xu.

18 recordsLinked to original sources

arXiv2026

APT: Action Expert Pretraining Improves Instruction Generalization of Vision-Language-Action Policies

Vision-Language-Action (VLA) models that couple pretrained Vision-Language Models (VLMs) with continuous action experts have achieved strong manipulation performance, yet generalization to out-of-distribution (OOD) language instructions remains poor. A known challenge is the structural imbalance in VLA data, where language is far less diverse than visual and action content, making policies prone to visual shortcuts. While discrete-action methods mitigate this through vision-language co-training, continuous action experts lack such protection: they start from random initialization and learn entirely from imbalanced data, producing noisy gradients that corrupt the VLM and fail to exploit its language capability. We address this from a Bayesian perspective, factorizing the policy into a language-agnostic Vision-Action (VA) prior and a language-conditioned VLA likelihood, and propose APT, a two-stage training method emphasizing Action expert PreTraining. In Stage 1, the action expert is pretrained as a VA prior on vision-action pairs from a frozen VLM, bypassing the language imbalance. In Stage 2, language tokens are injected through a gated fusion mechanism that integrates VLM features while preserving the learned visuomotor prior. APT applies to mainstream VLA architectures, including the $\pi$ and GR00T-style architectures. Comprehensive experiments validate that APT achieves consistent gains on unseen instructions and compositional tasks. Project Page: https://xukechun.github.io/papers/APT/

cs.RO↗

arXiv2026

Direction Matters: Learning Force Direction Enables Sim-to-Real Contact-Rich Manipulation

Sim-to-real transfer for contact-rich manipulation remains challenging due to the inherent discrepancy in contact dynamics. While existing methods often rely on costly real-world data or utilize blind compliance through fixed controllers, we propose a framework that leverages expert-designed controller logic for transfer. Inspired by the success of privileged supervision in kinematic tasks, we employ a human-designed finite state machine based position/force controller in simulation to provide privileged guidance. The resulting policy is trained to predict the end-effector pose, contact state, and crucially the desired contact force direction. Unlike force magnitudes, which are highly sensitive to simulation inaccuracies, force directions encode high-level task geometry and remain robust across the sim-to-real gap. At deployment, these predictions configure a force-aware admittance controller. By combining the policy's directional intent with a constant, low-cost manually tuned force magnitude, the system generates adaptive, task-aligned compliance. This tuning is lightweight, typically requiring only a single scalar per contact state. We provide theoretical analysis for stability and robustness to disturbances. Experiments on four real-world tasks, i.e., microwave opening, peg-in-hole, whiteboard wiping, and door opening, demonstrate that our approach significantly outperforms strong baselines in both success rate and robustness. Videos are available at: https://yifei-y.github.io/project-pages/DirectionMatters/.

cs.RO↗

arXiv2025

Seeing to Act, Prompting to Specify: A Bayesian Factorization of Vision Language Action Policy

The pursuit of out-of-distribution generalization in Vision-Language-Action (VLA) models is often hindered by catastrophic forgetting of the Vision-Language Model (VLM) backbone during fine-tuning. While co-training with external reasoning data helps, it requires experienced tuning and data-related overhead. Beyond such external dependencies, we identify an intrinsic cause within VLA datasets: modality imbalance, where language diversity is much lower than visual and action diversity. This imbalance biases the model toward visual shortcuts and language forgetting. To address this, we introduce BayesVLA, a Bayesian factorization that decomposes the policy into a visual-action prior, supporting seeing-to-act, and a language-conditioned likelihood, enabling prompt-to-specify. This inherently preserves generalization and promotes instruction following. We further incorporate pre- and post-contact phases to better leverage pre-trained foundation models. Information-theoretic analysis formally validates our effectiveness in mitigating shortcut learning. Extensive experiments show superior generalization to unseen instructions, objects, and environments compared to existing methods. Project page is available at: https://xukechun.github.io/papers/BayesVLA.

cs.RO↗

arXiv2025

Toward Embodiment Equivariant Vision-Language-Action Policy

Vision-language-action policies learn manipulation skills across tasks, environments and embodiments through large-scale pre-training. However, their ability to generalize to novel robot configurations remains limited. Most approaches emphasize model size, dataset scale and diversity while paying less attention to the design of action spaces. This leads to the configuration generalization problem, which requires costly adaptation. We address this challenge by formulating cross-embodiment pre-training as designing policies equivariant to embodiment configuration transformations. Building on this principle, we propose a framework that (i) establishes a embodiment equivariance theory for action space and policy design, (ii) introduces an action decoder that enforces configuration equivariance, and (iii) incorporates a geometry-aware network architecture to enhance embodiment-agnostic spatial reasoning. Extensive experiments in both simulation and real-world settings demonstrate that our approach improves pre-training effectiveness and enables efficient fine-tuning on novel robot embodiments. Our code is available at https://github.com/hhcaz/e2vla

cs.RO↗

arXiv2025

ExploreVLM: Closed-Loop Robot Exploration Task Planning with Vision-Language Models

The advancement of embodied intelligence is accelerating the integration of robots into daily life as human assistants. This evolution requires robots to not only interpret high-level instructions and plan tasks but also perceive and adapt within dynamic environments. Vision-Language Models (VLMs) present a promising solution by combining visual understanding and language reasoning. However, existing VLM-based methods struggle with interactive exploration, accurate perception, and real-time plan adaptation. To address these challenges, we propose ExploreVLM, a novel closed-loop task planning framework powered by Vision-Language Models (VLMs). The framework is built around a step-wise feedback mechanism that enables real-time plan adjustment and supports interactive exploration. At its core is a dual-stage task planner with self-reflection, enhanced by an object-centric spatial relation graph that provides structured, language-grounded scene representations to guide perception and planning. An execution validator supports the closed loop by verifying each action and triggering re-planning. Extensive real-world experiments demonstrate that ExploreVLM significantly outperforms state-of-the-art baselines, particularly in exploration-centric tasks. Ablation studies further validate the critical role of the reflective planner and structured perception in achieving robust and efficient task execution.

cs.RO↗

arXiv2025

Grounding 3D Object Affordance with Language Instructions, Visual Observations and Interactions

Grounding 3D object affordance is a task that locates objects in 3D space where they can be manipulated, which links perception and action for embodied intelligence. For example, for an intelligent robot, it is necessary to accurately ground the affordance of an object and grasp it according to human instructions. In this paper, we introduce a novel task that grounds 3D object affordance based on language instructions, visual observations and interactions, which is inspired by cognitive science. We collect an Affordance Grounding dataset with Points, Images and Language instructions (AGPIL) to support the proposed task. In the 3D physical world, due to observation orientation, object rotation, or spatial occlusion, we can only get a partial observation of the object. So this dataset includes affordance estimations of objects from full-view, partial-view, and rotation-view perspectives. To accomplish this task, we propose LMAffordance3D, the first multi-modal, language-guided 3D affordance grounding network, which applies a vision-language model to fuse 2D and 3D spatial features with semantic features. Comprehensive experiments on AGPIL demonstrate the effectiveness and superiority of our method on this task, even in unseen experimental settings. Our project is available at https://sites.google.com/view/lmaffordance3d.

cs.CV↗

arXiv2025

Efficient Alignment of Unconditioned Action Prior for Language-conditioned Pick and Place in Clutter

We study the task of language-conditioned pick and place in clutter, where a robot should grasp a target object in open clutter and move it to a specified place. Some approaches learn end-to-end policies with features from vision foundation models, requiring large datasets. Others combine foundation models in a zero-shot setting, suffering from cascading errors. In addition, they primarily leverage vision and language foundation models, focusing less on action priors. In this paper, we aim to develop an effective policy by integrating foundation priors from vision, language, and action. We propose A$^2$, an action prior alignment method that aligns unconditioned action priors with 3D vision-language priors by learning one attention layer. The alignment formulation enables our policy to train with less data and preserve zero-shot generalization capabilities. We show that a shared policy for both pick and place actions enhances the performance for each task, and introduce a policy adaptation scheme to accommodate the multi-modal nature of actions. Extensive experiments in simulation and the real-world show that our policy achieves higher task success rates with fewer steps for both pick and place tasks in clutter, effectively generalizing to unseen objects and language instructions. Videos and codes are available at https://xukechun.github.io/papers/A2.

cs.RO↗

arXiv2025

UniMM: A Unified Mixture Model Framework for Multi-Agent Simulation

Simulation plays a crucial role in assessing autonomous driving systems, where the generation of realistic multi-agent behaviors is a key aspect. In multi-agent simulation, the primary challenges include behavioral multimodality and closed-loop distributional shifts. In this study, we formulate a unified mixture model (UniMM) framework for generating multimodal agent behaviors, which can cover the mainstream methods including regression-based mixture models and discrete NTP models. Furthermore, we introduce a closed-loop sample generation approach tailored for mixture models to mitigate distributional shifts. Within the UniMM framework, we recognize critical configurations from both the model and data perspectives. We conduct a systematic examination of various model configurations, and comprehensively characterize their effects. Moreover, our investigation into the data configuration highlights the pivotal role of closed-loop samples in achieving realistic simulations. To extend the benefits of closed-loop samples across a broader range of mixture models, we further introduce a temporal disentanglement-and-alignment mechanism to address the shortcut learning and off-policy learning issues. Leveraging insights from our exploration, the distinct variants proposed within the UniMM framework, including discrete, anchor-free, and anchor-based models, all achieve state-of-the-art performance on the WOSAC benchmark.

cs.AI↗

arXiv2024

Grasp, See, and Place: Efficient Unknown Object Rearrangement with Policy Structure Prior

We focus on the task of unknown object rearrangement, where a robot is supposed to re-configure the objects into a desired goal configuration specified by an RGB-D image. Recent works explore unknown object rearrangement systems by incorporating learning-based perception modules. However, they are sensitive to perception error, and pay less attention to task-level performance. In this paper, we aim to develop an effective system for unknown object rearrangement amidst perception noise. We theoretically reveal that the noisy perception impacts grasp and place in a decoupled way, and show such a decoupled structure is valuable to improve task optimality. We propose GSP, a dual-loop system with the decoupled structure as prior. For the inner loop, we learn a see policy for self-confident in-hand object matching. For the outer loop, we learn a grasp policy aware of object matching and grasp capability guided by task-level rewards. We leverage the foundation model CLIP for object matching, policy learning and self-termination. A series of experiments indicate that GSP can conduct unknown object rearrangement with higher completion rates and fewer steps.

cs.RO↗

arXiv2023

A Hyper-network Based End-to-end Visual Servoing with Arbitrary Desired Poses

Recently, several works achieve end-to-end visual servoing (VS) for robotic manipulation by replacing traditional controller with differentiable neural networks, but lose the ability to servo arbitrary desired poses. This letter proposes a differentiable architecture for arbitrary pose servoing: a hyper-network based neural controller (HPN-NC). To achieve this, HPN-NC consists of a hyper net and a low-level controller, where the hyper net learns to generate the parameters of the low-level controller and the controller uses the 2D keypoints error for control like traditional image-based visual servoing (IBVS). HPN-NC can complete 6 degree of freedom visual servoing with large initial offset. Taking advantage of the fully differentiable nature of HPN-NC, we provide a three-stage training procedure to servo real world objects. With self-supervised end-to-end training, the performance of the integrated model can be further improved in unseen scenes and the amount of manual annotations can be significantly reduced.

cs.RO↗

arXiv2023

Object-centric Inference for Language Conditioned Placement: A Foundation Model based Approach

We focus on the task of language-conditioned object placement, in which a robot should generate placements that satisfy all the spatial relational constraints in language instructions. Previous works based on rule-based language parsing or scene-centric visual representation have restrictions on the form of instructions and reference objects or require large amounts of training data. We propose an object-centric framework that leverages foundation models to ground the reference objects and spatial relations for placement, which is more sample efficient and generalizable. Experiments indicate that our model can achieve a 97.75% success rate of placement with only ~0.26M trainable parameters. Besides, our method generalizes better to both unseen objects and instructions. Moreover, with only 25% training data, we still outperform the top competing approach.

cs.RO↗

arXiv2023

Failure-aware Policy Learning for Self-assessable Robotics Tasks

Self-assessment rules play an essential role in safe and effective real-world robotic applications, which verify the feasibility of the selected action before actual execution. But how to utilize the self-assessment results to re-choose actions remains a challenge. Previous methods eliminate the selected action evaluated as failed by the self-assessment rules, and re-choose one with the next-highest affordance~(i.e. process-of-elimination strategy [1]), which ignores the dependency between the self-assessment results and the remaining untried actions. However, this dependency is important since the previous failures might help trim the remaining over-estimated actions. In this paper, we set to investigate this dependency by learning a failure-aware policy. We propose two architectures for the failure-aware policy by representing the self-assessment results of previous failures as the variable state, and leveraging recurrent neural networks to implicitly memorize the previous failures. Experiments conducted on three tasks demonstrate that our method can achieve better performances with higher task success rates by less trials. Moreover, when the actions are correlated, learning a failure-aware policy can achieve better performance than the process-of-elimination strategy.

cs.RO↗

arXiv2023

A Joint Modeling of Vision-Language-Action for Target-oriented Grasping in Clutter

We focus on the task of language-conditioned grasping in clutter, in which a robot is supposed to grasp the target object based on a language instruction. Previous works separately conduct visual grounding to localize the target object, and generate a grasp for that object. However, these works require object labels or visual attributes for grounding, which calls for handcrafted rules in planner and restricts the range of language instructions. In this paper, we propose to jointly model vision, language and action with object-centric representation. Our method is applicable under more flexible language instructions, and not limited by visual grounding error. Besides, by utilizing the powerful priors from the pre-trained multi-modal model and grasp model, sample efficiency is effectively improved and the sim2real problem is relived without additional data for transfer. A series of experiments carried out in simulation and real world indicate that our method can achieve better task success rate by less times of motion under more flexible language instructions. Moreover, our method is capable of generalizing better to scenarios with unseen objects and language instructions. Our code is available at https://github.com/xukechun/Vision-Language-Grasping

cs.RO↗

arXiv2022

E-NeRV: Expedite Neural Video Representation with Disentangled Spatial-Temporal Context

Recently, the image-wise implicit neural representation of videos, NeRV, has gained popularity for its promising results and swift speed compared to regular pixel-wise implicit representations. However, the redundant parameters within the network structure can cause a large model size when scaling up for desirable performance. The key reason of this phenomenon is the coupled formulation of NeRV, which outputs the spatial and temporal information of video frames directly from the frame index input. In this paper, we propose E-NeRV, which dramatically expedites NeRV by decomposing the image-wise implicit neural representation into separate spatial and temporal context. Under the guidance of this new formulation, our model greatly reduces the redundant model parameters, while retaining the representation ability. We experimentally find that our method can improve the performance to a large extent with fewer parameters, resulting in a more than $8\times$ faster speed on convergence. Code is available at https://github.com/kyleleey/E-NeRV.

cs.CV↗

arXiv2022

Learning A Simulation-based Visual Policy for Real-world Peg In Unseen Holes

This paper proposes a learning-based visual peg-in-hole that enables training with several shapes in simulation, and adapting to arbitrary unseen shapes in real world with minimal sim-to-real cost. The core idea is to decouple the generalization of the sensory-motor policy to the design of a fast-adaptable perception module and a simulated generic policy module. The framework consists of a segmentation network (SN), a virtual sensor network (VSN), and a controller network (CN). Concretely, the VSN is trained to measure the pose of the unseen shape from a segmented image. After that, given the shape-agnostic pose measurement, the CN is trained to achieve generic peg-in-hole. Finally, when applying to real unseen holes, we only have to fine-tune the SN required by the simulated VSN+CN. To further minimize the transfer cost, we propose to automatically collect and annotate the data for the SN after one-minute human teaching. Simulated and real-world results are presented under the configurations of eye-to/in-hand. An electric vehicle charging system with the proposed policy inside achieves a 10/10 success rate in 2-3s, using only hundreds of auto-labeled samples for the SN transfer.

cs.RO↗

arXiv2022

Efficient Object Manipulation to an Arbitrary Goal Pose: Learning-based Anytime Prioritized Planning

We focus on the task of object manipulation to an arbitrary goal pose, in which a robot is supposed to pick an assigned object to place at the goal position with a specific orientation. However, limited by the execution space of the manipulator with gripper, one-step picking, moving and releasing might be failed, where a reorientation object pose is required as a transition. In this paper, we propose a learning-driven anytime prioritized search-based solver to find a feasible solution with low path cost in a short time. In our work, the problem is formulated as a hierarchical learning problem, with the high level finding a reorientation object pose, and the low level planning paths between adjacent grasps. We learn an offline-training path cost estimator to predict approximate path planning costs, which serve as pseudo rewards to allow for pre-training the high-level planner without interacting with the simulator. To deal with the problem of distribution mismatch of the cost net and the actual execution cost space, a refined training stage is conducted with simulation interaction. A series of experiments carried out in simulation and real world indicate that our system can achieve better performances in the object manipulation task with less time and less cost.

cs.RO↗

arXiv2021

Efficient learning of goal-oriented push-grasping synergy in clutter

We focus on the task of goal-oriented grasping, in which a robot is supposed to grasp a pre-assigned goal object in clutter and needs some pre-grasp actions such as pushes to enable stable grasps. However, in this task, the robot gets positive rewards from environment only when successfully grasping the goal object. Besides, joint pushing and grasping elongates the action sequence, compounding the problem of reward delay. Thus, sample inefficiency remains a main challenge in this task. In this paper, a goal-conditioned hierarchical reinforcement learning formulation with high sample efficiency is proposed to learn a push-grasping policy for grasping a specific object in clutter. In our work, sample efficiency is improved by two means. First, we use a goal-conditioned mechanism by goal relabeling to enrich the replay buffer. Second, the pushing and grasping policies are respectively regarded as a generator and a discriminator and the pushing policy is trained with supervision of the grasping discriminator, thus densifying pushing rewards. To deal with the problem of distribution mismatch caused by different training settings of two policies, an alternating training stage is added to learn pushing and grasping in turn. A series of experiments carried out in simulation and real world indicate that our method can quickly learn effective pushing and grasping policies and outperforms existing methods in task completion rate and goal grasp success rate by less times of motion. Furthermore, we validate that our system can also adapt to goal-agnostic conditions with better performance. Note that our system can be transferred to the real world without any fine-tuning. Our code is available at https://github.com/xukechun/Efficient_goal-oriented_push-grasping_synergy.

cs.RO↗

arXiv2021

Neural Motion Prediction for In-flight Uneven Object Catching

In-flight objects capture is extremely challenging. The robot is required to complete trajectory prediction, interception position calculation and motion planning in sequence within tens of milliseconds. As in-flight uneven objects are affected by various kinds of forces, motion prediction is difficult for a time-varying acceleration. In order to compensate the system's non-linearity, we introduce the Neural Acceleration Estimator (NAE) that estimates the varying acceleration by observing a small fragment of previous deflected trajectory. Moreover, end-to-end training with Differantiable Filter (NAE-DF) gives a supervision for measurement uncertainty and further improves the prediction accuracy. Experimental results show that motion prediction with NAE and NAE-DF is superior to other methods and has a good generalization performance on unseen objects. We test our methods on a robot, performing velocity control in real world and respectively achieve 83.3% and 86.7% success rate on a ploy urethane banana and a gourd. We also release an object in-flight dataset containing 1,500 trajectorys for uneven objects.

cs.RO↗