arXiv subjects

Sitong Mao

Publications and source records attributed to Sitong Mao.

18 recordsLinked to original sources

arXiv2026

CR-VLA-Force: Learning Control-aware Compliance VLA Model for Robust Contact-rich Robotic Manipulation

Integrating visuomotor policies or Vision-Language-Action (VLA) models with force/torque (F/T) perception has demonstrated significant progress in imitation learning for robotic manipulation. However, existing force-aware VLA models frequently exhibit limited capability in precise force tracking and rapid successive adjustments. This deficiency stems from the limitations of action-chunk execution strategies and the substantial latency between perception and real-time control. Such limitations can lead to task failures and safety risks, particularly when the execution of an action chunk exerts excessive interaction forces without timely adjustment. To overcome this challenge, we propose the Control-aware Compliance VLA (CC-VLA) framework for reactive control. The CC-VLA model employs a multimodal mixture-of-experts (MoE) to encode force signal sequences and vision-language fused feature. Furthermore, it utilizes a multi-stage training strategy to ensure robust perception within the visual-semantic space and effective force perception under sparse sampling conditions. Additionally, a VLA-guided adaptive compliance controller is designed to facilitate precise position tracking during contact-free motion and optimal force-position tracking for contact-rich tasks. To facilitate high-precision F/T data acquisition, we also implement an adversaria shared teleoperation strategy for contact-rich demonstrations that bolsters system safety and interactivity. Extensive real-world experiments demonstrate that CC-VLA significantly improves success rates in challenging force-perception tasks and enhances force-control precision, while providing multi-level safety and robustness under the tested partial-OOD pose-shift settings.

cs.RO↗

arXiv2026

Revisiting Topological Graphs for Macro Action based Closed-loop Reinforcement Learning of Vision Language Navigation in Continuous Environment

Vision-Language Navigation in Continuous Environments (VLN-CE) requires an agent to follow natural language instructions through unseen environments. Existing imitation learning (IL) pipelines struggle in this closed-loop setting: behavior cloning suffers from distribution shift, and DAgger's expert actions become ambiguous upon trajectory deviation. While Reinforcement Learning (RL) offers a natural paradigm to address this, directly applying RL to micro action spaces is sample-inefficient due to reward sparsity. To overcome this bottleneck, we reformulate VLN-CE as a Hierarchical Markov Decision Process (MDP), explicitly decoupling high-level planning from low-level control. By abstracting the environment into a topological graph, our high-level policy operates on a macro action space of frontier nodes, with a training-free low-level controller acting as its state transition, which significantly compresses the decision horizon and makes closed-loop RL tractable. To support RL optimization on the macro MDP, we propose an action-aware value head to effectively evaluate state values under the dynamic frontier action space, powering a graph-based PPO. Extensive experiments demonstrate the effectiveness of our architecture. Finally, our model achieves state-of-the-art performance on the R2R-CE and RxR-CE benchmarks.

cs.RO↗

arXiv2026

EventVLA: Event-Driven Visual Evidence Memory for Long-Horizon Vision-Language-Action Policies

Memory remains a critical bottleneck for long-horizon robotic manipulation, as standard Vision-Language-Action (VLA) policies often fail when task-relevant cues become occluded or unobservable over time. While existing memory-augmented methods utilize historical context, they either suffer from severe information bottlenecks, incur high latency via decoupled dual systems, or rely on unselective buffers that accumulate massive visual redundancies. To address these limitations, we introduce EventVLA, an end-to-end framework founded on the concept of sparse visual evidence memory that comprises two core components: foundational visual anchors to retain initial and short-term contexts, and a dynamic Keyframe Evidence Memory (KEM) module. Specifically, KEM directly predicts future keyframe probabilities from the VLA's latent embeddings to autonomously capture and store sparse, task-critical visual events. This foresight-driven mechanism empowers the policy to dynamically evaluate the future causal utility of current observations, preserving transient visual evidence before it becomes unobservable. Furthermore, we propose RoboTwin-MeM, a diagnostic benchmark specifically designed to evaluate non-Markovian manipulation tasks with interactive visual evidence. Extensive evaluations show that across 17 memory-requiring simulation tasks and 4 real-world bimanual tasks, EventVLA achieves an average success rate improvement of +40% over state-of-the-art memory-augmented VLAs.

cs.CV↗

arXiv2026

HyperSim: A Holistic Sim-To-Real Framework For Robust Robotic Manipulation

Scaling data volume and diversity is critical for generalizing embodied intelligence. While synthetic data generation offers a scalable alternative to expensive physical data acquisition, transferring robotic manipulation policies from simulation to the real world (sim-to-real) remains a formidable challenge due to the domain gap. This paper presents HyperSim, a holistic framework spanning from synthetic data generation to policy training and seamless real-world deployment. To systematically bridge the sim-to-real gap, HyperSim is realized through three core pillars: high-fidelity environment synthesis, adversarial trajectory generation, and sim-and-real co-training. Collectively, these modules address domain discrepancies by enhancing visual fidelity, expanding data coverage, and enforcing domain-invariant representations. We rigorously validate HyperSim through a large-scale empirical study involving 400 real-world task executions across two representative manipulation models. Assessed across three fine-grained metrics, our complete pipeline achieves remarkable sim-to-real success rates of 80% and 95% with ACT and \pi_{0}, respectively. Furthermore, policies trained on our adversarial trajectories exhibit significantly enhanced robustness against dynamic uncertainties, achieving a 35% higher completion rate under physical perturbations.

cs.RO↗

arXiv2025

ETP-R1: Evolving Topological Planning with Reinforcement Fine-tuning for Vision-Language Navigation in Continuous Environments

Vision-Language Navigation in Continuous Environments (VLN-CE) requires an embodied agent to navigate towards target in continuous environments, following natural language instructions. While current graph-based methods offer an efficient, structured approach by abstracting the environment into a topological map and simplifying the action space to waypoint selection, they lag behind methods based on Large Vision-Language Models (LVLMs) in leveraging large-scale data and advanced training paradigms. In this paper, we try to bridge this gap by introducing ETP-R1, a framework that applies the paradigm of scaling up data and Reinforcement Fine-Tuning (RFT) to a graph-based VLN-CE model. To build a strong foundation, we first construct a high-quality, large-scale pretraining dataset using the Gemini API. This dataset consists of diverse, low-hallucination instructions for topological trajectories, providing rich supervision for our graph-based policy to map language to topological paths. This foundation is further strengthened by unifying data from both R2R and RxR tasks for joint pretraining. Building on this, we introduce a three-stage training paradigm, which culminates in the first application of closed-loop, online RFT to a graph-based VLN-CE model, powered by the Group Relative Policy Optimization (GRPO) algorithm. Extensive experiments demonstrate that our approach is highly effective, establishing new state-of-the-art performance across all major metrics on both the R2R-CE and RxR-CE benchmarks. Our code is available at https://github.com/Cepillar/ETP-R1.

cs.RO↗

arXiv2025

Spatial Reasoning with Vision-Language Models in Ego-Centric Multi-View Scenes

Understanding 3D spatial relationships remains a major limitation of current Vision-Language Models (VLMs). Prior work has addressed this issue by creating spatial question-answering (QA) datasets based on single images or indoor videos. However, real-world embodied AI agents such as robots and self-driving cars typically rely on ego-centric, multi-view observations. To this end, we introduce Ego3D-Bench, a new benchmark designed to evaluate the spatial reasoning abilities of VLMs using ego-centric, multi-view outdoor data. Ego3D-Bench comprises over 8,600 QA pairs, created with significant involvement from human annotators to ensure quality and diversity. We benchmark 16 SOTA VLMs, including GPT-4o, Gemini1.5-Pro, InternVL3, and Qwen2.5-VL. Our results reveal a notable performance gap between human level scores and VLM performance, highlighting that current VLMs still fall short of human level spatial understanding. To bridge this gap, we propose Ego3D-VLM, a post-training framework that enhances 3D spatial reasoning of VLMs. Ego3D-VLM generates cognitive map based on estimated global 3D coordinates, resulting in 12% average improvement on multi-choice QA and 56% average improvement on absolute distance estimation. Ego3D-VLM is modular and can be integrated with any existing VLM. Together, Ego3D-Bench and Ego3D-VLM offer valuable tools for advancing toward human level spatial understanding in real-world, multi-view environments.

cs.CV↗

arXiv2025

Discrete Diffusion VLA: Bringing Discrete Diffusion to Action Decoding in Vision-Language-Action Policies

Vision-Language-Action (VLA) models adapt large vision-language backbones to map images and instructions into robot actions. However, prevailing VLAs either generate actions autoregressively in a fixed left-to-right order with poor performance or attach separate diffusion heads outside the backbone that fragments information pathways and hinders unified, scalable architectures. Instead, we present Discrete Diffusion VLA that discretizes action chunks and models them with discrete diffusion pattern retaining progressive refinement inside the unified transformer backbone. Our method achieves an adaptive decoding order that resolves high-confidence action elements before harder ones and employs secondary re-masking to revisit uncertain predictions, enabling robust error correction. This design preserves pretrained vision-language priors, supports parallel decoding, and improves the efficiency. Discrete Diffusion VLA achieves 96.4% avg. success on LIBERO, 71.2% visual matching on SimplerEnv-Fractal, and 54.2% overall on SimplerEnv-Bridge. On out-of-distribution tests of LIBERO-Goal, our method exhibits only 0.8% language degradation versus 8.0% of parallel decoding, and 20.4% vision degradation versus 29.0% for continuous diffusion, demonstrating well retention of pretrained vision-language capabilities. We also conduct two real-robot evaluations on AgileX Cobot Magic platform to show the method's effectiveness.

cs.CV↗

arXiv2025

Dynamic Mixture of Progressive Parameter-Efficient Expert Library for Lifelong Robot Learning

A generalist agent must continuously learn and adapt throughout its lifetime, achieving efficient forward transfer while minimizing catastrophic forgetting. Previous work within the dominant pretrain-then-finetune paradigm has explored parameter-efficient fine-tuning for single-task adaptation, effectively steering a frozen pretrained model with a small number of parameters. However, in the context of lifelong learning, these methods rely on the impractical assumption of a test-time task identifier and restrict knowledge sharing among isolated adapters. To address these limitations, we propose Dynamic Mixture of Progressive Parameter-Efficient Expert Library (DMPEL) for lifelong robot learning. DMPEL progressively builds a low-rank expert library and employs a lightweight router to dynamically combine experts into an end-to-end policy, enabling flexible and efficient lifelong forward transfer. Furthermore, by leveraging the modular structure of the fine-tuned parameters, we introduce expert coefficient replay, which guides the router to accurately retrieve frozen experts for previously encountered tasks. This technique mitigates forgetting while being significantly more storage- and computation-efficient than experience replay over the entire policy. Extensive experiments on the lifelong robot learning benchmark LIBERO demonstrate that our framework outperforms state-of-the-art lifelong learning methods in success rates during continual adaptation, while utilizing minimal trainable parameters and storage.

cs.LG↗

arXiv2025

PanopticSplatting: End-to-End Panoptic Gaussian Splatting

Open-vocabulary panoptic reconstruction is a challenging task for simultaneous scene reconstruction and understanding. Recently, methods have been proposed for 3D scene understanding based on Gaussian splatting. However, these methods are multi-staged, suffering from the accumulated errors and the dependence of hand-designed components. To streamline the pipeline and achieve global optimization, we propose PanopticSplatting, an end-to-end system for open-vocabulary panoptic reconstruction. Our method introduces query-guided Gaussian segmentation with local cross attention, lifting 2D instance masks without cross-frame association in an end-to-end way. The local cross attention within view frustum effectively reduces the training memory, making our model more accessible to large scenes with more Gaussians and objects. In addition, to address the challenge of noisy labels in 2D pseudo masks, we propose label blending to promote consistent 3D segmentation with less noisy floaters, as well as label warping on 2D predictions which enhances multi-view coherence and segmentation accuracy. Our method demonstrates strong performances in 3D scene panoptic reconstruction on the ScanNet-V2 and ScanNet++ datasets, compared with both NeRF-based and Gaussian-based panoptic reconstruction methods. Moreover, PanopticSplatting can be easily generalized to numerous variants of Gaussian splatting, and we demonstrate its robustness on different Gaussian base models.

cs.CV↗

arXiv2024

Scale Disparity of Instances in Interactive Point Cloud Segmentation

Interactive point cloud segmentation has become a pivotal task for understanding 3D scenes, enabling users to guide segmentation models with simple interactions such as clicks, therefore significantly reducing the effort required to tailor models to diverse scenarios and new categories. However, in the realm of interactive segmentation, the meaning of instance diverges from that in instance segmentation, because users might desire to segment instances of both thing and stuff categories that vary greatly in scale. Existing methods have focused on thing categories, neglecting the segmentation of stuff categories and the difficulties arising from scale disparity. To bridge this gap, we propose ClickFormer, an innovative interactive point cloud segmentation model that accurately segments instances of both thing and stuff categories. We propose a query augmentation module to augment click queries by a global query sampling strategy, thus maintaining consistent performance across different instance scales. Additionally, we employ global attention in the query-voxel transformer to mitigate the risk of generating false positives, along with several other network structure improvements to further enhance the model's segmentation performance. Experiments demonstrate that ClickFormer outperforms existing interactive point cloud segmentation methods across both indoor and outdoor datasets, providing more accurate segmentation results with fewer user clicks in an open-world setting.

cs.CV↗

arXiv2024

Let Occ Flow: Self-Supervised 3D Occupancy Flow Prediction

Accurate perception of the dynamic environment is a fundamental task for autonomous driving and robot systems. This paper introduces Let Occ Flow, the first self-supervised work for joint 3D occupancy and occupancy flow prediction using only camera inputs, eliminating the need for 3D annotations. Utilizing TPV for unified scene representation and deformable attention layers for feature aggregation, our approach incorporates a novel attention-based temporal fusion module to capture dynamic object dependencies, followed by a 3D refine module for fine-gained volumetric representation. Besides, our method extends differentiable rendering to 3D volumetric flow fields, leveraging zero-shot 2D segmentation and optical flow cues for dynamic decomposition and motion optimization. Extensive experiments on nuScenes and KITTI datasets demonstrate the competitive performance of our approach over prior state-of-the-art methods. Our project page is available at https://eliliu2233.github.io/letoccflow/

cs.CV↗

arXiv2024

PanopticRecon: Leverage Open-vocabulary Instance Segmentation for Zero-shot Panoptic Reconstruction

Panoptic reconstruction is a challenging task in 3D scene understanding. However, most existing methods heavily rely on pre-trained semantic segmentation models and known 3D object bounding boxes for 3D panoptic segmentation, which is not available for in-the-wild scenes. In this paper, we propose a novel zero-shot panoptic reconstruction method from RGB-D images of scenes. For zero-shot segmentation, we leverage open-vocabulary instance segmentation, but it has to face partial labeling and instance association challenges. We tackle both challenges by propagating partial labels with the aid of dense generalized features and building a 3D instance graph for associating 2D instance IDs. Specifically, we exploit partial labels to learn a classifier for generalized semantic features to provide complete labels for scenes with dense distilled features. Moreover, we formulate instance association as a 3D instance graph segmentation problem, allowing us to fully utilize the scene geometry prior and all 2D instance masks to infer global unique pseudo 3D instance ID. Our method outperforms state-of-the-art methods on the indoor dataset ScanNet V2 and the outdoor dataset KITTI-360, demonstrating the effectiveness of our graph segmentation method and reconstruction network.

cs.CV↗

arXiv2024

SCALE: Self-Correcting Visual Navigation for Mobile Robots via Anti-Novelty Estimation

Although visual navigation has been extensively studied using deep reinforcement learning, online learning for real-world robots remains a challenging task. Recent work directly learned from offline dataset to achieve broader generalization in the real-world tasks, which, however, faces the out-of-distribution (OOD) issue and potential robot localization failures in a given map for unseen observation. This significantly drops the success rates and even induces collision. In this paper, we present a self-correcting visual navigation method, SCALE, that can autonomously prevent the robot from the OOD situations without human intervention. Specifically, we develop an image-goal conditioned offline reinforcement learning method based on implicit Q-learning (IQL). When facing OOD observation, our novel localization recovery method generates the potential future trajectories by learning from the navigation affordance, and estimates the future novelty via random network distillation (RND). A tailored cost function searches for the candidates with the least novelty that can lead the robot to the familiar places. We collect offline data and conduct evaluation experiments in three real-world urban scenarios. Experiment results show that SCALE outperforms the previous state-of-the-art methods for open-world navigation with a unique capability of localization recovery, significantly reducing the need for human intervention. Code is available at https://github.com/KubeEdge4Robotics/ScaleNav.

cs.RO↗

arXiv2023

NF-Atlas: Multi-Volume Neural Feature Fields for Large Scale LiDAR Mapping

LiDAR Mapping has been a long-standing problem in robotics. Recent progress in neural implicit representation has brought new opportunities to robotic mapping. In this paper, we propose the multi-volume neural feature fields, called NF-Atlas, which bridge the neural feature volumes with pose graph optimization. By regarding the neural feature volume as pose graph nodes and the relative pose between volumes as pose graph edges, the entire neural feature field becomes both locally rigid and globally elastic. Locally, the neural feature volume employs a sparse feature Octree and a small MLP to encode the submap SDF with an option of semantics. Learning the map using this structure allows for end-to-end solving of maximum a posteriori (MAP) based probabilistic mapping. Globally, the map is built volume by volume independently, avoiding catastrophic forgetting when mapping incrementally. Furthermore, when a loop closure occurs, with the elastic pose graph based representation, only updating the origin of neural volumes is required without remapping. Finally, these functionalities of NF-Atlas are validated. Thanks to the sparsity and the optimization based formulation, NF-Atlas shows competitive performance in terms of accuracy, efficiency and memory usage on both simulation and real-world datasets.

cs.RO↗

arXiv2020

Against Adversarial Learning: Naturally Distinguish Known and Unknown in Open Set Domain Adaptation

Open set domain adaptation refers to the scenario that the target domain contains categories that do not exist in the source domain. It is a more common situation in the reality compared with the typical closed set domain adaptation where the source domain and the target domain contain the same categories. The main difficulty of open set domain adaptation is that we need to distinguish which target data belongs to the unknown classes when machine learning models only have concepts about what they know. In this paper, we propose an "against adversarial learning" method that can distinguish unknown target data and known data naturally without setting any additional hyper parameters and the target data predicted to the known classes can be classified at the same time. Experimental results show that the proposed method can make significant improvement in performance compared with several state-of-the-art methods.

cs.LG↗

arXiv2020

Mixed Set Domain Adaptation

In the settings of conventional domain adaptation, categories of the source dataset are from the same domain (or domains for multi-source domain adaptation), which is not always true in reality. In this paper, we propose \textbf{\textit{Mixed Set Domain Adaptation} (MSDA)}. Under the settings of MSDA, different categories of the source dataset are not all collected from the same domain(s). For instance, category $1\sim k$ are collected from domain $α$ while category $k+1\sim c$ are collected from domain $β$. Under such situation, domain adaptation performance will be further influenced because of the distribution discrepancy inside the source data. A feature element-wise weighting (FEW) method that can reduce distribution discrepancy between different categories is also proposed for MSDA. Experimental results and quality analysis show the significance of solving MSDA problem and the effectiveness of the proposed method.

cs.LG↗

arXiv2020

Deep Adversarial Domain Adaptation Based on Multi-layer Joint Kernelized Distance

Domain adaptation refers to the learning scenario that a model learned from the source data is applied on the target data which have the same categories but different distribution. While it has been widely applied, the distribution discrepancy between source data and target data can substantially affect the adaptation performance. The problem has been recently addressed by employing adversarial learning and distinctive adaptation performance has been reported. In this paper, a deep adversarial domain adaptation model based on a multi-layer joint kernelized distance metric is proposed. By utilizing the abstract features extracted from deep networks, the multi-layer joint kernelized distance (MJKD) between the $j$th target data predicted as the $m$th category and all the source data of the $m'$th category is computed. Base on MJKD, a class-balanced selection strategy is utilized in each category to select target data that are most likely to be classified correctly and treat them as labeled data using their pseudo labels. Then an adversarial architecture is used to draw the newly generated labeled training data and the remaining target data close to each other. In this way, the target data itself provide valuable information to enhance the domain adaptation. An analysis of the proposed method is also given and the experimental results demonstrate that the proposed method can achieve a better performance than a number of state-of-the-art methods.

cs.LG↗

arXiv2020

Network Together: Node Classification via Cross network Deep Network Embedding

Network embedding is a highly effective method to learn low-dimensional node vector representations with original network structures being well preserved. However, existing network embedding algorithms are mostly developed for a single network, which fail to learn generalized feature representations across different networks. In this paper, we study a cross-network node classification problem, which aims at leveraging the abundant labeled information from a source network to help classify the unlabeled nodes in a target network. To succeed in such a task, transferable features should be learned for nodes across different networks. To this end, a novel cross-network deep network embedding (CDNE) model is proposed to incorporate domain adaptation into deep network embedding so as to learn label-discriminative and network-invariant node vector representations. On one hand, CDNE leverages network structures to capture the proximities between nodes within a network, by mapping more strongly connected nodes to have more similar latent vector representations. On the other hand, node attributes and labels are leveraged to capture the proximities between nodes across different networks by making the same labeled nodes across networks have aligned latent vector representations. Extensive experiments have been conducted, demonstrating that the proposed CDNE model significantly outperforms the state-of-the-art network embedding algorithms in cross-network node classification.

cs.SI↗