Robotics 52
☆ Stop Wandering: Efficient Vision-Language Navigation via Metacognitive Reasoning
Training-free Vision-Language Navigation (VLN) agents powered by foundation models can follow instructions and explore 3D environments. However, existing approaches rely on greedy frontier selection and passive spatial memory, leading to inefficient behaviors such as local oscillation and redundant revisiting. We argue that this stems from a lack of metacognitive capabilities: the agent cannot monitor its exploration progress, diagnose strategy failures, or adapt accordingly. To address this, we propose MetaNav, a metacognitive navigation agent integrating spatial memory, history-aware planning, and reflective correction. Spatial memory builds a persistent 3D semantic map. History-aware planning penalizes revisiting to improve efficiency. Reflective correction detects stagnation and uses an LLM to generate corrective rules that guide future frontier selection. Experiments on GOAT-Bench, HM3D-OVON, and A-EQA show that MetaNav achieves state-of-the-art performance while reducing VLM queries by 20.7%, demonstrating that metacognitive reasoning significantly improves robustness and efficiency.
comment: 10 pages, 6 figures
☆ Deep Neural Network Based Roadwork Detection for Autonomous Driving
Road construction sites create major challenges for both autonomous vehicles and human drivers due to their highly dynamic and heterogeneous nature. This paper presents a real-time system that detects and localizes roadworks by combining a YOLO neural network with LiDAR data. The system identifies individual roadwork objects while driving, merges them into coherent construction sites and records their outlines in world coordinates. The model training was based on an adapted US dataset and a new dataset collected from test drives with a prototype vehicle in Berlin, Germany. Evaluations on real-world road construction sites showed a localization accuracy below 0.5 m. The system can support traffic authorities with up-to-date roadwork data and could enable autonomous vehicles to navigate construction sites more safely in the future.
comment: 7 pages, 10 figures
☆ Model-Based Reinforcement Learning for Control under Time-Varying Dynamics
Learning-based control methods typically assume stationary system dynamics, an assumption often violated in real-world systems due to drift, wear, or changing operating conditions. We study reinforcement learning for control under time-varying dynamics. We consider a continual model-based reinforcement learning setting in which an agent repeatedly learns and controls a dynamical system whose transition dynamics evolve across episodes. We analyze the problem using Gaussian process dynamics models under frequentist variation-budget assumptions. Our analysis shows that persistent non-stationarity requires explicitly limiting the influence of outdated data to maintain calibrated uncertainty and meaningful dynamic regret guarantees. Motivated by these insights, we propose a practical optimistic model-based reinforcement learning algorithm with adaptive data buffer mechanisms and demonstrate improved performance on continuous control benchmarks with non-stationary dynamics.
comment: 15 pages, 5 figues, 2 tables. This work has been submitted to the IEEE for possible publication
☆ A virtual-variable-length method for robust inverse kinematics of multi-segment continuum robots
This paper proposes a new, robust method to solve the inverse kinematics (IK) of multi-segment continuum manipulators. Conventional Jacobian-based solvers, especially when initialized from neutral/rest configurations, often exhibit slow convergence and, in certain conditions, may fail to converge (deadlock). The Virtual-Variable-Length (VVL) method proposed here introduces fictitious variations of segments' length during the solution iteration, conferring virtual axial degrees of freedom that alleviate adverse behaviors and constraints, thus enabling or accelerating convergence. Comprehensive numerical experiments were conducted to compare the VVL method against benchmark Jacobian-based and Damped Least Square IK solvers. Across more than $1.8\times 10^6$ randomized trials covering manipulators with two to seven segments, the proposed approach achieved up to a 20$\%$ increase in convergence success rate over the benchmark and a 40-80$\%$ reduction in average iteration count under equivalent accuracy thresholds ($10^{-4}-10^{-8}$). While deadlocks are not restricted to workspace boundaries and may occur at arbitrary poses, our empirical study identifies boundary-proximal configurations as a frequent cause of failed convergence and the VVL method mitigates such occurrences over a statistical sample of test cases.
comment: 8 pages, 6 figures, accepted for presentation in IEEE RoboSoft 2026, Kanazawa, Japan
☆ UAV-Track VLA: Embodied Aerial Tracking via Vision-Language-Action Models
Qiyao Zhang, Shuhua Zheng, Jianli Sun, Chengxiang Li, Xianke Wu, Zihan Song, Zhiyong Cui, Yisheng Lv, Yonglin Tian
Embodied visual tracking is crucial for Unmanned Aerial Vehicles (UAVs) executing complex real-world tasks. In dynamic urban scenarios with complex semantic requirements, Vision-Language-Action (VLA) models show great promise due to their cross-modal fusion and continuous action generation capabilities. To benchmark multimodal tracking in such environments, we construct a dedicated evaluation benchmark and a large-scale dataset encompassing over 890K frames, 176 tasks, and 85 diverse objects. Furthermore, to address temporal feature redundancy and the lack of spatial geometric priors in existing VLA models, we propose an improved VLA tracking model, UAV-Track VLA. Built upon the $π_{0.5}$ architecture, our model introduces a temporal compression net to efficiently capture inter-frame dynamics. Additionally, a parallel dual-branch decoder comprising a spatial-aware auxiliary grounding head and a flow matching action expert is designed to decouple cross-modal features and generate fine-grained continuous actions. Systematic experiments in the CARLA simulator validate the superior end-to-end performance of our method. Notably, in challenging long-distance pedestrian tracking tasks, UAV-Track VLA achieves a 61.76\% success rate and 269.65 average tracking frames, significantly outperforming existing baselines. Furthermore, it demonstrates robust zero-shot generalization in unseen environments and reduces single-step inference latency by 33.4\% (to 0.0571s) compared to the original $π_{0.5}$, enabling highly efficient, real-time UAV control. Data samples and demonstration videos are available at: https://github.com/Hub-Tian/UAV-Track\_VLA.
☆ UniDriveVLA: Unifying Understanding, Perception, and Action Planning for Autonomous Driving
Yongkang Li, Lijun Zhou, Sixu Yan, Bencheng Liao, Tianyi Yan, Kaixin Xiong, Long Chen, Hongwei Xie, Bing Wang, Guang Chen, Hangjun Ye, Wenyu Liu, Haiyang Sun, Xinggang Wang
Vision-Language-Action (VLA) models have recently emerged in autonomous driving, with the promise of leveraging rich world knowledge to improve the cognitive capabilities of driving systems. However, adapting such models for driving tasks currently faces a critical dilemma between spatial perception and semantic reasoning. Consequently, existing VLA systems are forced into suboptimal compromises: directly adopting 2D Vision-Language Models yields limited spatial perception, whereas enhancing them with 3D spatial representations often impairs the native reasoning capacity of VLMs. We argue that this dilemma largely stems from the coupled optimization of spatial perception and semantic reasoning within shared model parameters. To overcome this, we propose UniDriveVLA, a Unified Driving Vision-Language-Action model based on Mixture-of-Transformers that addresses the perception-reasoning conflict via expert decoupling. Specifically, it comprises three experts for driving understanding, scene perception, and action planning, which are coordinated through masked joint attention. In addition, we combine a sparse perception paradigm with a three-stage progressive training strategy to improve spatial perception while maintaining semantic reasoning capability. Extensive experiments show that UniDriveVLA achieves state-of-the-art performance in open-loop evaluation on nuScenes and closed-loop evaluation on Bench2Drive. Moreover, it demonstrates strong performance across a broad range of perception, prediction, and understanding tasks, including 3D detection, online mapping, motion forecasting, and driving-oriented VQA, highlighting its broad applicability as a unified model for autonomous driving. Code and model have been released at https://github.com/xiaomi-research/unidrivevla
comment: code has been released at https://github.com/xiaomi-research/unidrivevla
☆ PRO-SPECT: Probabilistically Safe Scalable Planning for Energy-Aware Coordinated UAV-UGV Teams in Stochastic Environments
We consider energy-aware planning for an unmanned aerial vehicle (UAV) and unmanned ground vehicle (UGV) team operating in a stochastic environment. The UAV must visit a set of air points in minimum time while respecting energy constraints, relying on the UGV as a mobile charging station. Unlike prior work that assumed deterministic travel times or used fixed robustness margins, we model travel times as random variables and bound the probability of failure (energy depletion) across the entire mission to a user-specified risk level. We formulate the problem as a Mixed-Integer Program and propose PRO-SPECT, a polynomial-time algorithm that generates risk-bounded plans. The algorithm supports both offline planning and online re-planning, enabling the team to adapt to disturbances while preserving the risk bound. We provide theoretical results on solution feasibility and time complexity. We also demonstrate the performance of our method via numerical comparisons and simulations.
☆ ROS 2-Based LiDAR Perception Framework for Mobile Robots in Dynamic Production Environments, Utilizing Synthetic Data Generation, Transformation-Equivariant 3D Detection and Multi-Object Tracking
Lukas Bergs, Tan Chung, Marmik Thakkar, Alexander Moriz, Amon Göppert, Chinnawut Nantabut, Robert Schmitt
Adaptive robots in dynamic production environments require robust perception capabilities, including 6D pose estimation and multi-object tracking. To address limitations in real-world data dependency, noise robustness, and spatiotemporal consistency, a LiDAR framework based on the Robot Operating System integrating a synthetic-data-trained Transformation-Equivariant 3D Detection with multi-object-tracking leveraging center poses is proposed. Validated across 72 scenarios with motion capture technology, overall results yield an Intersection over Union of 62.6% for standalone pose estimation, rising to 83.12% with multi-object-tracking integration. Our LiDAR-based framework achieves 91.12% of Higher Order Tracking Accuracy, advancing robustness and versatility of LiDAR-based perception systems for industrial mobile manipulators.
comment: Accepted for publication at CIRP ICME 2025; will appear in Procedia CIRP
☆ Cross-Modal Visuo-Tactile Object Perception
Anirvan Dutta, Simone Tasciotti, Claudia Cusseddu, Ang Li, Panayiota Poirazi, Julijana Gjorgjieva, Etienne Burdet, Patrick van der Smagt, Mohsen Kaboli
Estimating physical properties is critical for safe and efficient autonomous robotic manipulation, particularly during contact-rich interactions. In such settings, vision and tactile sensing provide complementary information about object geometry, pose, inertia, stiffness, and contact dynamics, such as stick-slip behavior. However, these properties are only indirectly observable and cannot always be modeled precisely (e.g., deformation in non-rigid objects coupled with nonlinear contact friction), making the estimation problem inherently complex and requiring sustained exploitation of visuo-tactile sensory information during action. Existing visuo-tactile perception frameworks have primarily emphasized forceful sensor fusion or static cross-modal alignment, with limited consideration of how uncertainty and beliefs about object properties evolve over time. Inspired by human multi-sensory perception and active inference, we propose the Cross-Modal Latent Filter (CMLF) to learn a structured, causal latent state-space of physical object properties. CMLF supports bidirectional transfer of cross-modal priors between vision and touch and integrates sensory evidence through a Bayesian inference process that evolves over time. Real-world robotic experiments demonstrate that CMLF improves the efficiency and robustness of latent physical properties estimation under uncertainty compared to baseline approaches. Beyond performance gains, the model exhibits perceptual coupling phenomena analogous to those observed in humans, including susceptibility to cross-modal illusions and similar trajectories in learning cross-sensory associations. Together, these results constitutes a significant step toward generalizable, robust and physically consistent cross-modal integration for robotic multi-sensory perception.
comment: 23 pages, 8 figures, 1 table. Submitted for review to journal
☆ HyVGGT-VO: Tightly Coupled Hybrid Dense Visual Odometry with Feed-Forward Models
Dense visual odometry (VO), which provides pose estimation and dense 3D reconstruction, serves as the cornerstone for applications ranging from robotics to augmented reality. Recently, feed-forward models have demonstrated remarkable capabilities in dense mapping. However, when these models are used in dense visual SLAM systems, their heavy computational burden restricts them to yielding sparse pose outputs at keyframes while still failing to achieve real-time pose estimation. In contrast, traditional sparse methods provide high computational efficiency and high-frequency pose outputs, but lack the capability for dense reconstruction. To address these limitations, we propose HyVGGT-VO, a novel framework that combines the computational efficiency of sparse VO with the dense reconstruction capabilities of feed-forward models. To the best of our knowledge, this is the first work to tightly couple a traditional VO framework with VGGT, a state-of-the-art feed-forward model. Specifically, we design an adaptive hybrid tracking frontend that dynamically switches between traditional optical flow and the VGGT tracking head to ensure robustness. Furthermore, we introduce a hierarchical optimization framework that jointly refines VO poses and the scale of VGGT predictions to ensure global scale consistency. Our approach achieves an approximately 5x processing speedup compared to existing VGGT-based methods, while reducing the average trajectory error by 85% on the indoor EuRoC dataset and 12% on the outdoor KITTI benchmark. Our code will be publicly available upon acceptance. Project page: https://geneta2580.github.io/HyVGGT-VO.io.
☆ CompassAD: Intent-Driven 3D Affordance Grounding in Functionally Competing Objects
Jingliang Li, Jindou Jia, Tuo An, Chuhao Zhou, Xiangyu Chen, Shilin Shan, Boyu Ma, Bofan Lyu, Gen Li, Jianfei Yang
When told to "cut the apple," a robot must choose the knife over nearby scissors, despite both objects affording the same cutting function. In real-world scenes, multiple objects may share identical affordances, yet only one is appropriate under the given task context. We call such cases confusing pairs. However, existing 3D affordance methods largely sidestep this challenge by evaluating isolated single objects, often with explicit category names provided in the query. We formalize Multi-Object Affordance Grounding under Intent-Driven Instructions, a new 3D affordance setting that requires predicting a per-point affordance mask on the correct object within a cluttered multi-object point cloud, conditioned on implicit natural language intent. To study this problem, we construct CompassAD, the first benchmark centered on implicit intent in confusable multi-object scenes. It comprises 30 confusing object pairs spanning 16 affordance types, 6,422 scenes, and 88K+ query-answer pairs. Furthermore, we propose CompassNet, a framework that incorporates two dedicated modules tailored to this task. Instance-bounded Cross Injection (ICI) constrains language-geometry alignment within object boundaries to prevent cross-object semantic leakage. Bi-level Contrastive Refinement (BCR) enforces discrimination at both geometric-group and point levels, sharpening distinctions between target and confusable surfaces. Extensive experiments demonstrate state-of-the-art results on both seen and unseen queries, and deployment on a robotic manipulator confirms effective transfer to real-world grasping in confusing multi-object scenes.
comment: Code available at: github.com/Lorenzo-0-0/CompassAD
☆ O-ConNet: Geometry-Aware End-to-End Inference of Over-Constrained Spatial Mechanisms
Deep learning has shown strong potential for scientific discovery, but its ability to model macroscopic rigid-body kinematic constraints remains underexplored. We study this problem on spatial over-constrained mechanisms and propose O-ConNet, an end-to-end framework that infers mechanism structural parameters from only three sparse reachable points while reconstructing the full motion trajectory, without explicitly solving constraint equations during inference. On a self-constructed Bennett 4R dataset of 42,860 valid samples, O-ConNet achieves Param-MAE 0.276 +/- 0.077 and Traj-MAE 0.145 +/- 0.018 (mean +/- std over 10 runs), outperforming the strongest sequence baseline (LSTM-Seq2Seq) by 65.1 percent and 88.2 percent, respectively. These results suggest that end-to-end learning can capture closed-loop geometric structure and provide a practical route for inverse design of spatial over-constrained mechanisms under extremely sparse observations.
comment: 8 pages, 5 figures
☆ Bridging Discrete Planning and Continuous Execution for Redundant Robot
Voxel-grid reinforcement learning is widely adopted for path planning in redundant manipulators due to its simplicity and reproducibility. However, direct execution through point-wise numerical inverse kinematics on 7-DoF arms often yields step-size jitter, abrupt joint transitions, and instability near singular configurations. This work proposes a bridging framework between discrete planning and continuous execution without modifying the discrete planner itself. On the planning side, step-normalized 26-neighbor Cartesian actions and a geometric tie-breaking mechanism are introduced to suppress unnecessary turns and eliminate step-size oscillations. On the execution side, a task-priority damped least-squares (TP-DLS) inverse kinematics layer is implemented. This layer treats end-effector position as a primary task, while posture and joint centering are handled as subordinate tasks projected into the null space, combined with trust-region clipping and joint velocity constraints. On a 7-DoF manipulator in random sparse, medium, and dense environments, this bridge raises planning success in dense scenes from about 0.58 to 1.00, shortens representative path length from roughly 1.53 m to 1.10 m, and while keeping end-effector error below 1 mm, reduces peak joint accelerations by over an order of magnitude, substantially improving the continuous execution quality of voxel-based RL paths on redundant manipulators.
comment: 8 pages, 3 figures. Submitted to IFAC World Congress 2026
☆ Integrated Identification of Collaborative Robots for Robot Assisted 3D Printing Processes
In recent years, the integration of additive manufacturing (AM) and industrial robotics has opened new perspectives for the production of complex components, particularly in the automotive sector. Robot-assisted additive manufacturing processes overcome the dimensional and kinematic limitations of traditional Cartesian systems, enabling non-planar deposition and greater geometric flexibility. However, the increasing dynamic complexity of robotic manipulators introduces challenges related to precision, control, and error prediction. This work proposes a model-based approach equipped with an integrated identification procedure of the system's parameters, including the robot, the actuators and the controllers. We show that the integrated modeling procedure allows to obtain a reliable dynamic model even in the presence of sensory and programming limitations typical of collaborative robots. The manipulator's dynamic model is identified through an integrated five step methodology: starting with geometric and inertial analysis, followed by friction and controller parameters identification, all the way to the remaining parameters identification. The proposed procedure intrinsically ensures the physical consistency of the identified parameters. The identification approach is validated on a real world case study involving a 6-Degrees-Of-Freedom (DoFs) collaborative robot used in a thermoplastic extrusion process. The very good matching between the experimental results given by actual robot and those given by the identified model shows the potential enhancement of precision, control, and error prediction in Robot Assisted 3D Printing Processes.
★ World Action Verifier: Self-Improving World Models via Forward-Inverse Asymmetry
Yuejiang Liu, Fan Feng, Lingjing Kong, Weifeng Lu, Jinzhou Tang, Kun Zhang, Kevin Murphy, Chelsea Finn, Yilun Du
General-purpose world models promise scalable policy evaluation, optimization, and planning, yet achieving the required level of robustness remains challenging. Unlike policy learning, which primarily focuses on optimal actions, a world model must be reliable over a much broader range of suboptimal actions, which are often insufficiently covered by action-labeled interaction data. To address this challenge, we propose World Action Verifier (WAV), a framework that enables world models to identify their own prediction errors and self-improve. The key idea is to decompose action-conditioned state prediction into two factors -- state plausibility and action reachability -- and verify each separately. We show that these verification problems can be substantially easier than predicting future states due to two underlying asymmetries: the broader availability of action-free data and the lower dimensionality of action-relevant features. Leveraging these asymmetries, we augment a world model with (i) a diverse subgoal generator obtained from video corpora and (ii) a sparse inverse model that infers actions from a subset of state features. By enforcing cycle consistency among generated subgoals, inferred actions, and forward rollouts, WAV provides an effective verification mechanism in under-explored regimes, where existing methods typically fail. Across nine tasks spanning MiniGrid, RoboMimic, and ManiSkill, our method achieves 2x higher sample efficiency while improving downstream policy performance by 18%.
comment: Project Website: https://world-action-verifier.github.io
☆ Ego-Grounding for Personalized Question-Answering in Egocentric Videos CVPR'26
We present the first systematic analysis of multimodal large language models (MLLMs) in personalized question-answering requiring ego-grounding - the ability to understand the camera-wearer in egocentric videos. To this end, we introduce MyEgo, the first egocentric VideoQA dataset designed to evaluate MLLMs' ability to understand, remember, and reason about the camera wearer. MyEgo comprises 541 long videos and 5K personalized questions asking about "my things", "my activities", and "my past". Benchmarking reveals that competitive MLLMs across variants, including open-source vs. proprietary, thinking vs. non-thinking, small vs. large scales all struggle on MyEgo. Top closed- and open-source models (e.g., GPT-5 and Qwen3-VL) achieve only~46% and 36% accuracy, trailing human performance by near 40% and 50% respectively. Surprisingly, neither explicit reasoning nor model scaling yield consistent improvements. Models improve when relevant evidence is explicitly provided, but gains drop over time, indicating limitations in tracking and remembering "me" and "my past". These findings collectively highlight the crucial role of ego-grounding and long-range memory in enabling personalized QA in egocentric videos. We hope MyEgo and our analyses catalyze further progress in these areas for egocentric personalized assistance. Data and code are available at https://github.com/Ryougetsu3606/MyEgo
comment: To appear at CVPR'26
☆ Learning Spatial Structure from Pre-Beamforming Per-Antenna Range-Doppler Radar Data via Visibility-Aware Cross-Modal Supervision
Automotive radar perception pipelines commonly construct angle-domain representations via beamforming before applying learning-based models. This work instead investigates a representational question: can meaningful spatial structure be learned directly from pre-beamforming per-antenna range-Doppler (RD) measurements? Experiments are conducted on a 6-TX x 8-RX (48 virtual antennas) commodity automotive radar employing an A/B chirp-sequence frequency-modulated continuous-wave (CS-FMCW) transmit scheme, in which the effective transmit aperture varies between chirps (single-TX vs. multi-TX), enabling controlled analysis of chirp-dependent transmit configurations. We operate on pre-beamforming per-antenna RD tensors using a dual-chirp shared-weight encoder trained in an end-to-end, fully data-driven manner, and evaluate spatial recoverability using bird's-eye-view (BEV) occupancy as a geometric probe rather than a performance-driven objective. Supervision is visibility-aware and cross-modal, derived from LiDAR with explicit modeling of the radar field-of-view and occlusion-aware LiDAR observability via ray-based visibility. Through chirp ablations (A-only, B-only, A+B), range-band analysis, and physics-aligned baselines, we assess how transmit configurations affect geometric recoverability. The results indicate that spatial structure can be learned directly from pre-beamforming per-antenna RD tensors without explicit angle-domain construction or hand-crafted signal-processing stages.
☆ Global Geometry of Orthogonal Foliations in the Control Allocation of Signed-Quadratic Systems
This work formalizes the differential topology of redundancy resolution for systems governed by signed-quadratic actuation maps. By analyzing the minimally redundant case, the global topology of the continuous fiber bundle defining the nonlinear actuation null-space is established. The distribution orthogonal to these fibers is proven to be globally integrable and governed by an exact logarithmic potential field. This field foliates the actuator space, inducing a structural stratification of all orthants into transverse layers whose combinatorial sizes follow a strictly binomial progression. Within these layers, adjacent orthants are continuously connected via lower-dimensional strata termed reciprocal hinges, while the layers themselves are separated by boundary hyperplanes, or portals, that act as global sections of the fibers. This partition formally distinguishes extremal and transitional layers, which exhibit fundamentally distinct fiber topologies and foliation properties. Through this geometric framework, classical pseudo-linear static allocation strategies are shown to inevitably intersect singular boundary hyperplanes, triggering infinite-derivative kinetic singularities and fragmenting the task space into an exponential number of singularity-separated sectors. In contrast, allocators derived from the orthogonal manifolds yield continuously differentiable global sections with only a linear number of sectors for transversal layers, or can even form a single global diffeomorphism to the task space in the case of the two extremal layers, thus completely avoiding geometric rank-loss and boundary-crossing singularities. These theoretical results directly apply to the control allocation of propeller-driven architectures, including multirotor UAVs, marine, and underwater vehicles.
comment: Multimedia material attached
☆ Posterior Optimization with Clipped Objective for Bridging Efficiency and Stability in Generative Policy Learning
Expressive generative models have advanced robotic manipulation by capturing complex, multi-modal action distributions over temporally extended trajectories. However, fine-tuning these policies via RL remains challenging due to instability and sample inefficiency. We introduce Posterior Optimization with Clipped Objective (POCO), a principled RL framework that formulates policy improvement as a posterior inference problem tailored for temporal action chunks. Through an Expectation-Maximization procedure, POCO distills a reward-weighted implicit posterior into the policy without likelihood estimation. Furthermore, POCO adopts an offline-to-online paradigm that anchors online exploration to pre-trained priors, and its model-agnostic design scales to fine-tune large VLA models without architectural modifications. Evaluations across 7 simulation benchmarks and 4 contact-rich real-world tasks demonstrate that POCO prevents catastrophic policy collapse, outperforms SOTA baselines, and achieves a 96.7% success rate on real-world tasks. Videos are available at our project website https://cccedric.github.io/poco/.
☆ Preferential Bayesian Optimization with Crash Feedback
Bayesian optimization is a popular black-box optimization method for parameter learning in control and robotics. It typically requires an objective function that reflects the user's optimization goal. However, in practical applications, this objective function is often inaccessible due to complex or unmeasurable performance metrics. Preferential Bayesian optimization (PBO) overcomes this limitation by leveraging human feedback through pairwise comparisons, eliminating the need for explicit performance quantification. When applying PBO to hardware systems, such as in quadcopter control, crashes can cause time-consuming experimental resets, wear and tear, or otherwise undesired outcomes. Standard PBO methods cannot incorporate feedback from such crashed experiments, resulting in the exploration of parameters that frequently lead to experimental crashes. We thus introduce CrashPBO, a user-friendly mechanism that enables users to both express preferences and report crashes during the optimization process. Benchmarking on synthetic functions shows that this mechanism reduces crashes by 63% and increases data efficiency. Through experiments on three robotics platforms, we demonstrate the wide applicability and transferability of CrashPBO, highlighting that it provides a flexible, user-friendly framework for parameter learning with human feedback on preferences and crashes.
☆ DriveDreamer-Policy: A Geometry-Grounded World-Action Model for Unified Generation and Planning
Yang Zhou, Xiaofeng Wang, Hao Shao, Letian Wang, Guosheng Zhao, Jiangnan Shao, Jiagang Zhu, Tingdong Yu, Zheng Zhu, Guan Huang, Steven L. Waslander
Recently, world-action models (WAM) have emerged to bridge vision-language-action (VLA) models and world models, unifying their reasoning and instruction-following capabilities and spatio-temporal world modeling. However, existing WAM approaches often focus on modeling 2D appearance or latent representations, with limited geometric grounding-an essential element for embodied systems operating in the physical world. We present DriveDreamer-Policy, a unified driving world-action model that integrates depth generation, future video generation, and motion planning within a single modular architecture. The model employs a large language model to process language instructions, multi-view images, and actions, followed by three lightweight generators that produce depth, future video, and actions. By learning a geometry-aware world representation and using it to guide both future prediction and planning within a unified framework, the proposed model produces more coherent imagined futures and more informed driving actions, while maintaining modularity and controllable latency. Experiments on the Navsim v1 and v2 benchmarks demonstrate that DriveDreamer-Policy achieves strong performance on both closed-loop planning and world generation tasks. In particular, our model reaches 89.2 PDMS on Navsim v1 and 88.7 EPDMS on Navsim v2, outperforming existing world-model-based approaches while producing higher-quality future video and depth predictions. Ablation studies further show that explicit depth learning provides complementary benefits to video imagination and improves planning robustness.
comment: 11 pages, 4 figures; Project Website: https://drivedreamer-policy.github.io/
☆ Realistic Lip Motion Generation Based on 3D Dynamic Viseme and Coarticulation Modeling for Human-Robot Interaction
Realistic lip synchronization is essential for the natural human-robot non-verbal interaction of humanoid robots. Motivated by this need, this paper presents a lip motion generation framework based on 3D dynamic viseme and coarticulation modeling. By analyzing Chinese pronunciation theory, a 3D dynamic viseme library is constructed based on the ARKit standard, which offers coherent prior trajectories of lips. To resolve motion conflicts within continuous speech streams, a coarticulation mechanism is developed by incorporating initial-final (Shengmu-Yunmu) decoupling and energy modulation. After developing a strategy to retarget high-dimensional spatial lip motion to a 14-DOF lip actuation system of a humanoid head platform, the efficiency and accuracy of the proposed architecture is experimentally validated and demonstrated with quantitative ablation experiments using the metrics of the Pearson Correlation Coefficient (PCC) and the Mean Absolute Jerk (MAJ). This research offers a lightweight, efficient, and highly practical paradigm for the speech-driven lip motion generation of humanoid robots. The 3D dynamic viseme library and real-world deployment videos are available at {https://github.com/yuesheng21/Phoneme-to-Lip-14DOF}
comment: 8 pages,7 figures
☆ Analysis of Efficient Transmission Methods of Grid Maps for Intelligent Vehicles
Grid mapping is a fundamental approach to modeling the environment of intelligent vehicles or robots. Compared with object-based environment modeling, grid maps offer the distinct advantage of representing the environment without requiring any assumptions about objects, such as type or shape. For grid-map-based approaches, the environment is divided into cells, each containing information about its respective area, such as occupancy. This representation of the entire environment is crucial for achieving higher levels of autonomy. However, it has the drawback that modeling the scene at the cell level results in inherently large data sizes. Patched grid maps tackle this issue to a certain extent by adapting cell sizes in specific areas. Nevertheless, the data sizes of patched grid maps are still too large for novel distributed processing setups or vehicle-to-everything (V2X) applications. Our work builds on a patch-based grid-map approach and investigates the size problem from a communication perspective. To address this, we propose a patch-based communication pipeline that leverages existing compression algorithms to transmit grid-map data efficiently. We provide a comprehensive analysis of this pipeline for both intra-vehicle and V2X-based communication. The analysis is verified for these use cases with two real-world experiment setups. Finally, we summarize recommended guidelines for the efficient transmission of grid-map data in intelligent transportation systems.
comment: Accepted for 2026 IEEE Intelligent Vehicles Symposium (IV) - DOI will be added after publication
☆ Causal Scene Narration with Runtime Safety Supervision for Vision-Language-Action Driving
Vision-Language-Action (VLA) models for autonomous driving must integrate diverse textual inputs, including navigation commands, hazard warnings, and traffic state descriptions, yet current systems often present these as disconnected fragments, forcing the model to discover on its own which environmental constraints are relevant to the current maneuver. We introduce Causal Scene Narration (CSN), which restructures VLA text inputs through intent-constraint alignment, quantitative grounding, and structured separation, at inference time with zero GPU cost. We complement CSN with Simplex-based runtime safety supervision and training-time alignment via Plackett-Luce DPO with negative log-likelihood (NLL) regularization. A multi-town closed-loop CARLA evaluation shows that CSN improves Driving Score by +31.1% on original LMDrive and +24.5% on the preference-aligned variant. A controlled ablation reveals that causal structure accounts for 39.1% of this gain, with the remainder attributable to information content alone. A perception noise ablation confirms that CSN's benefit is robust to realistic sensing errors. Semantic safety supervision improves Infraction Score, while reactive Time-To-Collision monitoring degrades performance, demonstrating that intent-aware monitoring is needed for VLA systems.
comment: 18 pages, 6 figures, 4 tables
☆ Hi-LOAM: Hierarchical Implicit Neural Fields for LiDAR Odometry and Mapping
LiDAR Odometry and Mapping (LOAM) is a pivotal technique for embodied-AI applications such as autonomous driving and robot navigation. Most existing LOAM frameworks are either contingent on the supervision signal, or lack of the reconstruction fidelity, which are deficient in depicting details of large-scale complex scenes. To overcome these limitations, we propose a multi-scale implicit neural localization and mapping framework using LiDAR sensor, called Hi-LOAM. Hi-LOAM receives LiDAR point cloud as the input data modality, learns and stores hierarchical latent features in multiple levels of hash tables based on an octree structure, then these multi-scale latent features are decoded into signed distance value through shallow Multilayer Perceptrons (MLPs) in the mapping procedure. For pose estimation procedure, we rely on a correspondence-free, scan-to-implicit matching paradigm to estimate optimal pose and register current scan into the submap. The entire training process is conducted in a self-supervised manner, which waives the model pre-training and manifests its generalizability when applied to diverse environments. Extensive experiments on multiple real-world and synthetic datasets demonstrate the superior performance, in terms of the effectiveness and generalization capabilities, of our Hi-LOAM compared to existing state-of-the-art methods.
comment: This manuscript is the accepted version of IEEE Transactions on Multimedia
☆ OpenGo: An OpenClaw-Based Robotic Dog with Real-Time Skill Switching
Adaptation to complex tasks and multiple scenarios remains a significant challenge for a single robot agent. The ability to acquire organize, and switch between a wide range of skills in real time, particularly in dynamic environments, has become a fundamental requirement for embodied intelligence. We introduce OpenGo, an OpenClaw-powered embodied robotic dog capable of switching skills in real time according to the scene and task instructions. Specifically, the agent is equipped with (1) a customizable skill library with easy skill import and autonomous skill validation, (2) a dispatcher that selects and invokes different skills according to task prompts or language instructions, and (3) a self-learning framework that fine-tunes skills based on task completion and human feedback. We deploy the agent in Unitree's Go2 robotic dog and validate its capabilities in self-checking and switching of skills autonomously. In addition, by integrating Feishu-platform communication, we enable natural-language guidance and human feedback, allowing inexperienced users to control the robotic dog through simple instructions.
comment: 11 pages, 6 figures
☆ 3-D Relative Localization for Multi-Robot Systems with Angle and Self-Displacement Measurements
Realizing relative localization by leveraging inter-robot local measurements is a challenging problem, especially in the presence of measurement noise. Motivated by this challenge, in this paper we propose a novel and systematic 3-D relative localization framework based on inter-robot interior angle and self-displacement measurements. Initially, we propose a linear relative localization theory comprising a distributed linear relative localization algorithm and sufficient conditions for localizability. According to this theory, robots can determine their neighbors' relative positions and orientations in a purely linear manner. Subsequently, in order to deal with measurement noise, we present an advanced Maximum a Posterior (MAP) estimator by addressing three primary challenges existing in the MAP estimator. Firstly, it is common to formulate the MAP problem as an optimization problem, whose inherent non-convexity can result in local optima. To address this issue, we reformulate the linear computation process of the linear relative localization algorithm as a Weighted Total Least Squares (WTLS) optimization problem on manifolds. The optimal solution of the WTLS problem is more accurate, which can then be used as initial values when solving the optimization problem associated with the MAP problem, thereby reducing the risk of falling into local optima. The second challenge is the lack of knowledge of the prior probability density of the robots' relative positions and orientations at the initial time, which is required as an input for the MAP estimator. To deal with it, we combine the WTLS with a Neural Density Estimator (NDE). Thirdly, to prevent the increasing size of the relative positions and orientations to be estimated as the robots continuously move when solving the MAP problem, a marginalization mechanism is designed, which ensures that the computational cost remains constant.
comment: 29 pages, 28 figures
☆ A Graph Neural Network Approach for Solving the Ranked Assignment Problem in Multi-Object Tracking
Associating measurements with tracks is a crucial step in Multi-Object Tracking (MOT) to guarantee the safety of autonomous vehicles. To manage the exponentially growing number of track hypotheses, truncation becomes necessary. In the $δ$-Generalized Labeled Multi-Bernoulli ($δ$-GLMB) filter application, this truncation typically involves the ranked assignment problem, solved by Murty's algorithm or the Gibbs sampling approach, both with limitations in terms of complexity or accuracy, respectively. With the motivation to improve these limitations, this paper addresses the ranked assignment problem arising from data association tasks with an approach that employs Graph Neural Networks (GNNs). The proposed Ranked Assignment Prediction Graph Neural Network (RAPNet) uses bipartite graphs to model the problem, harnessing the computational capabilities of deep learning. The conclusive evaluation compares the RAPNet with Murty's algorithm and the Gibbs sampler, showing accuracy improvements compared to the Gibbs sampler.
comment: 2024 IEEE Intelligent Vehicles Symposium (IV)
☆ Bridging Large-Model Reasoning and Real-Time Control via Agentic Fast-Slow Planning
Large foundation models enable powerful reasoning for autonomous systems, but mapping semantic intent to reliable real-time control remains challenging. Existing approaches either (i) let Large Language Models (LLMs) generate trajectories directly - brittle, hard to verify, and latency-prone - or (ii) adjust Model Predictive Control (MPC) objectives online - mixing slow deliberation with fast control and blurring interfaces. We propose Agentic Fast-Slow Planning, a hierarchical framework that decouples perception, reasoning, planning, and control across natural timescales. The framework contains two bridges. Perception2Decision compresses scenes into ego-centric topologies using an on-vehicle Vision-Language Model (VLM) detector, then maps them to symbolic driving directives in the cloud with an LLM decision maker - reducing bandwidth and delay while preserving interpretability. Decision2Trajectory converts directives into executable paths: Semantic-Guided A* embeds language-derived soft costs into classical search to bias solutions toward feasible trajectories, while an Agentic Refinement Module adapts planner hyperparameters using feedback and memory. Finally, MPC tracks the trajectories in real time, with optional cloud-guided references for difficult cases. Experiments in CARLA show that Agentic Fast-Slow Planning improves robustness under perturbations, reducing lateral deviation by up to 45% and completion time by over 12% compared to pure MPC and an A*-guided MPC baseline. Code is available at https://github.com/cjychenjiayi/icra2026_AFSP.
comment: 8 pages, 12figures
☆ AURA: Multimodal Shared Autonomy for Real-World Urban Navigation
Long-horizon navigation in complex urban environments relies heavily on continuous human operation, which leads to fatigue, reduced efficiency, and safety concerns. Shared autonomy, where a Vision-Language AI agent and a human operator collaborate on maneuvering the mobile machine, presents a promising solution to address these issues. However, existing shared autonomy methods often require humans and AI to operate within the same action space, leading to high cognitive overhead. We present Assistive Urban Robot Autonomy (AURA), a new multi-modal framework that decomposes urban navigation into high-level human instruction and low-level AI control. AURA incorporates a Spatial-Aware Instruction Encoder to align various human instructions with visual and spatial context. To facilitate training, we construct MM-CoS, a large-scale dataset comprising teleoperation and vision-language descriptions. Experiments in simulation and the real world demonstrate that AURA effectively follows human instructions, reduces manual operation effort, and improves navigation stability, while enabling online adaptation. Moreover, under similar takeover conditions, our shared autonomy framework reduces the frequency of takeovers by more than 44%. Demo video and more detail are provided in the project page.
comment: 17 pages, 18 figures, 4 tables, conference
☆ Smooth Feedback Motion Planning with Reduced Curvature
Feedback motion planning over cell decompositions provides a robust method for generating collision-free robot motion with formal guarantees. However, existing algorithms often produce paths with unnecessary bending, leading to slower motion and higher control effort. This paper presents a computationally efficient method to mitigate this issue for a given simplicial decomposition. A heuristic is introduced that systematically aligns and assigns local vector fields to produce more direct trajectories, complemented by a novel geometric algorithm that constructs a maximal star-shaped chain of simplexes around the goal. This creates a large ``funnel'' in which an optimal, direct-to-goal control law can be safely applied. Simulations demonstrate that our method generates measurably more direct paths, reducing total bending by an average of 91.40\% and LQR control effort by an average of 45.47\%. Furthermore, comparative analysis against sampling-based and optimization-based planners confirms the time efficacy and robustness of our approach. While the proposed algorithms work over any finite-dimensional simplicial complex embedded in the collision-free subset of the configuration space, the practical application focuses on low-dimensional ($d\le3$) configuration spaces, where simplicial decomposition is computationally tractable.
comment: Accepted for publication in IEEE Robotics and Automation Letters
☆ F3DGS: Federated 3D Gaussian Splatting for Decentralized Multi-Agent World Modeling CVPR 2026
We present F3DGS, a federated 3D Gaussian Splatting framework for decentralized multi-agent 3D reconstruction. Existing 3DGS pipelines assume centralized access to all observations, which limits their applicability in distributed robotic settings where agents operate independently, and centralized data aggregation may be restricted. Directly extending centralized training to multi-agent systems introduces communication overhead and geometric inconsistency. F3DGS first constructs a shared geometric scaffold by registering locally merged LiDAR point clouds from multiple clients to initialize a global 3DGS model. During federated optimization, Gaussian positions are fixed to preserve geometric alignment, while each client updates only appearance-related attributes, including covariance, opacity, and spherical harmonic coefficients. The server aggregates these updates using visibility-aware aggregation, weighting each client's contribution by how frequently it observed each Gaussian, resolving the partial-observability challenge inherent to multi-agent exploration. To evaluate decentralized reconstruction, we collect a multi-sequence indoor dataset with synchronized LiDAR, RGB, and IMU measurements. Experiments show that F3DGS achieves reconstruction quality comparable to centralized training while enabling distributed optimization across agents. The dataset, development kit, and source code will be publicly released.
comment: Accepted to the CVPR 2026 SPAR-3D Workshop
☆ Boosting Vision-Language-Action Finetuning with Feasible Action Neighborhood Prior CVPR 2026
In real-world robotic manipulation, states typically admit a neighborhood of near-equivalent actions. That is, for each state, there exist a feasible action neighborhood (FAN) rather than a single correct action, within which motions yield indistinguishable progress. However, prevalent VLA training methodologies are directly inherited from linguistic settings and do not exploit the FAN property, thus leading to poor generalization and low sample efficiency. To address this limitation, we introduce a FAN-guided regularizer that shapes the model's output distribution to align with the geometry of FAN. Concretely, we introduce a Gaussian prior that promotes locally smooth and unimodal predictions around the preferred direction and magnitude. In extensive experiments across both reinforced finetuning (RFT) and supervised finetuning (SFT), our method achieves significant improvement in sample efficiency, and success rate in both in-distribution and out-of-distribution (OOD) scenarios. By aligning with the intrinsic action tolerance of physical manipulation, FAN-guided regularization provides a principled and practical method for sample-efficient, and generalizable VLA adaptation.
comment: Accepted by CVPR 2026
☆ AnchorVLA: Anchored Diffusion for Efficient End-to-End Mobile Manipulation
A central challenge in mobile manipulation is preserving multiple plausible action models while remaining reactive during execution. A bottle in a cluttered scene can often be approached and grasped in multiple valid ways. Robust behavior depends on preserving this action diversity while remaining reactive as the scene evolves. Diffusion policies are appealing because they model multimodal action distributions rather than collapsing to one solution. But in practice, full iterative denoising is costly at control time. Action chunking helps amortize inference, yet it also creates partially open-loop behavior, allowing small mismatches to accumulate into drift. We present AnchorVLA, a diffusion-based VLA policy for mobile manipulation built on the core insight that when sampling begins near a plausible solution manifold, extensive denoising is unnecessary to recover multimodal, valid actions. AnchorVLA combines a lightweight VLA adaptation backbone with an anchored diffusion action head, which denoises locally around anchor trajectories using a truncated diffusion schedule. This retains multimodal action generation while reducing inference cost for closed-loop control. Crucially, to mitigate chunking-induced drift, we introduce a test-time self-correction mechanism via a lightweight residual correction module that makes high-frequency, per-step adjustments during rollout. Across diverse mobile manipulation tasks, AnchorVLA improves success and stability under disturbances and distribution shifts while maintaining low-latency inference. The source code is made available at https://github.com/jason-lim26/AnchorVLA.
☆ Robust Autonomous Control of a Magnetic Millirobot in In Vitro Cardiac Flow
Untethered magnetic millirobots offer significant potential for minimally invasive cardiac therapies; however, achieving reliable autonomous control in pulsatile cardiac flow remains challenging. This work presents a vision-guided control framework enabling precise autonomous navigation of a magnetic millirobot in an in vitro heart phantom under physiologically relevant flow conditions. The system integrates UNet-based localization, A* path planning, and a sliding mode controller with a disturbance observer (SMC-DOB) designed for multi-coil electromagnetic actuation. Although drag forces are estimated using steady-state CFD simulations, the controller compensates for transient pulsatile disturbances during closed-loop operation. In static fluid, the SMC-DOB achieved sub-millimeter accuracy (root-mean-square error, RMSE = 0.49 mm), outperforming PID and MPC baselines. Under moderate pulsatile flow (7 cm/s peak, 20 cP), it reduced RMSE by 37% and peak error by 2.4$\times$ compared to PID. It further maintained RMSE below 2 mm (0.27 body lengths) under elevated pulsatile flow (10 cm/s peak, 20 cP) and under low-viscosity conditions (4.3 cP, 7 cm/s peak), where baseline controllers exhibited unstable or failed tracking. These results demonstrate robust closed-loop magnetic control under time-varying cardiac flow disturbances and support the feasibility of autonomous millirobot navigation for targeted drug delivery.
☆ MorphoGuard: A Morphology-Based Whole-Body Interactive Motion Controller
Whole-body control (WBC) has demonstrated significant advantages in complex interactive movements of high-dimensional robotic systems. However, when a robot is required to handle dynamic multi-contact combinations along a single kinematic chain-such as pushing open a door with its elbow while grasping an object-it faces major obstacles in terms of complex contact representation and joint configuration coupling. To address this, we propose a new control approach that explicitly manages arbitrary contact combinations, aiming to endow robots with whole-body interactive capabilities. We develop a morphology-constrained WBC network (MorphoGuard)-which is trained on a self-constructed dual-arm physical and simulation platform. A series of model recommendation experiments are designed to systematically investigate the impact of backbone architecture, fusion strategy, and model scale on network performance. To evaluate the control performance, we adopt a multi-object interaction task as the benchmark, requiring the model to simultaneously manipulate multiple target objects to specified positions. Experimental results show that the proposed method achieves a contact point management error of approximately 1 cm, demonstrating its effectiveness in whole-body interactive control.
♻ ☆ TaCarla: A comprehensive benchmarking dataset for end-to-end autonomous driving
Tugrul Gorgulu, Atakan Dag, M. Esat Kalfaoglu, Halil Ibrahim Kuru, Baris Can Cam, Halil Ibrahim Ozturk, Ozsel Kilinc
Collecting a high-quality dataset is a critical task that demands meticulous attention to detail, as overlooking certain aspects can render the entire dataset unusable. Autonomous driving challenges remain a prominent area of research, requiring further exploration to enhance the perception and planning performance of vehicles. However, existing datasets are often incomplete. For instance, datasets that include perception information generally lack planning data, while planning datasets typically consist of extensive driving sequences where the ego vehicle predominantly drives forward, offering limited behavioral diversity. In addition, many real datasets struggle to evaluate their models, especially for planning tasks, since they lack a proper closed-loop evaluation setup. The CARLA Leaderboard 2.0 challenge, which provides a diverse set of scenarios to address the long-tail problem in autonomous driving, has emerged as a valuable alternative platform for developing perception and planning models in both open-loop and closed-loop evaluation setups. Nevertheless, existing datasets collected on this platform present certain limitations. Some datasets appear to be tailored primarily for limited sensor configuration, with particular sensor configurations. To support end-to-end autonomous driving research, we have collected a new dataset comprising over 2.85 million frames using the CARLA simulation environment for the diverse Leaderboard 2.0 challenge scenarios. Our dataset is designed not only for planning tasks but also supports dynamic object detection, lane divider detection, centerline detection, traffic light recognition, prediction tasks and visual language action models . Furthermore, we demonstrate its versatility by training various models using our dataset. Moreover, we also provide numerical rarity scores to understand how rarely the current state occurs in the dataset.
♻ ★ Olaf: Bringing an Animated Character to Life in the Physical World
David Müller, Espen Knoop, Dario Mylonopoulos, Agon Serifi, Michael A. Hopkins, Ruben Grandia, Moritz Bächer
Animated characters often move in non-physical ways and have proportions that are far from a typical walking robot. This provides an ideal platform for innovation in both mechanical design and stylized motion control. In this paper, we bring Olaf to life in the physical world, relying on reinforcement learning guided by animation references for control. To create the illusion of Olaf's feet moving along his body, we hide two asymmetric legs under a soft foam skirt. To fit actuators inside the character, we use spherical and planar linkages in the arms, mouth, and eyes. Because the walk cycle results in harsh contact sounds, we introduce additional rewards that noticeably reduce impact noise. The large head, driven by small actuators in the character's slim neck, creates a risk of overheating, amplified by the costume. To keep actuators from overheating, we feed temperature values as additional inputs to policies, introducing new rewards to keep them within bounds. We validate the efficacy of our modeling in simulation and on hardware, demonstrating an unmatched level of believability for a costumed robotic character.
♻ ☆ Allometric Scaling Laws for Bipedal Robots
Scaling the design of robots up or down remains a fundamental challenge. While biological systems follow well-established isometric and allometric scaling laws relating mass, stride frequency, velocity, and torque, it is unclear how these relationships translate to robotic systems. In this paper, we generate similar allometric scaling laws for bipedal robots across three orders of magnitude in leg length. First, we conduct a review of legged robots from the literature and extract empirical relationships between leg length (L), body length, mass, and speed. These data show that robot mass scales more closely to L^2, in contrast to the L^3 scaling predicted by isometric scaling. We then perform controlled simulation studies in Drake using three variants of real quasi-passive, hip-actuated walkers with different foot geometries and control strategies. We evaluate the performance of each design scaled with leg length, L. Across all robots, walking velocity follows the expected L^(1/2) trend from dynamic similarity. Minimum required torque scales more closely with m*L than the isometric model of m*L^2. Foot geometry scaled proportionally with L^1. These results provide new insight into how robot designs allometrically scale to different sizes, and how that scaling is different from isometric or biological scaling laws.
♻ ☆ ThinkGrasp: A Vision-Language System for Strategic Part Grasping in Clutter
Robotic grasping in cluttered environments remains a significant challenge due to occlusions and complex object arrangements. We have developed ThinkGrasp, a plug-and-play vision-language grasping system that makes use of GPT-4o's advanced contextual reasoning for heavy clutter environment grasping strategies. ThinkGrasp can effectively identify and generate grasp poses for target objects, even when they are heavily obstructed or nearly invisible, by using goal-oriented language to guide the removal of obstructing objects. This approach progressively uncovers the target object and ultimately grasps it with a few steps and a high success rate. In both simulated and real experiments, ThinkGrasp achieved a high success rate and significantly outperformed state-of-the-art methods in heavily cluttered environments or with diverse unseen objects, demonstrating strong generalization capabilities.
comment: Accepted at CoRL 2024. Project Website:(https://h-freax.github.io/thinkgrasp_page/)
♻ ☆ MaskAdapt: Learning Flexible Motion Adaptation via Mask-Invariant Prior for Physics-Based Characters CVPR 2026
We present MaskAdapt, a framework for flexible motion adaptation in physics-based humanoid control. The framework follows a two-stage residual learning paradigm. In the first stage, we train a mask-invariant base policy using stochastic body-part masking and a regularization term that enforces consistent action distributions across masking conditions. This yields a robust motion prior that remains stable under missing observations, anticipating later adaptation in those regions. In the second stage, a residual policy is trained atop the frozen base controller to modify only the targeted body parts while preserving the original behaviors elsewhere. We demonstrate the versatility of this design through two applications: (i) motion composition, where varying masks enable multi-part adaptation within a single sequence, and (ii) text-driven partial goal tracking, where designated body parts follow kinematic targets provided by a pre-trained text-conditioned autoregressive motion generator. Through experiments, MaskAdapt demonstrates strong robustness and adaptability, producing diverse behaviors under masked observations and delivering superior targeted motion adaptation compared to prior work.
comment: CVPR 2026
♻ ☆ Robot Collapse: Supply Chain Backdoor Attacks Against VLM-based Robotic Manipulation
Xianlong Wang, Hewen Pan, Hangtao Zhang, Minghui Li, Shengshan Hu, Ziqi Zhou, Lulu Xue, Peijin Guo, Aishan Liu, Leo Yu Zhang, Xiaohua Jia
Robotic manipulation policies are increasingly empowered by \textit{large language models} (LLMs) and \textit{vision-language models} (VLMs), leveraging their understanding and perception capabilities. Recently, inference-time attacks against robotic manipulation have been extensively studied, yet backdoor attacks targeting model supply chain security in robotic policies remain largely unexplored. To fill this gap, we propose \texttt{TrojanRobot}, a backdoor injection framework for model supply chain attack scenarios, which embeds a malicious module into modular robotic policies via backdoor relationships to manipulate the LLM-to-VLM pathway and compromise the system. Our vanilla design instantiates this module as a backdoor-finetuned VLM. To further enhance attack performance, we propose a prime scheme by introducing the concept of \textit{LVLM-as-a-backdoor}, which leverages \textit{in-context instruction learning} (ICIL) to steer \textit{large vision-language model} (LVLM) behavior through backdoored system prompts. Moreover, we develop three types of prime attacks, \textit{permutation}, \textit{stagnation}, and \textit{intentional}, achieving flexible backdoor attack effects. Extensive physical-world and simulator experiments on 18 real-world manipulation tasks and 4 VLMs verify the superiority of proposed \texttt{TrojanRobot}
♻ ☆ Constraint-Aware Reinforcement Learning via Adaptive Action Scaling
Safe reinforcement learning (RL) seeks to mitigate unsafe behaviors that arise from exploration during training by reducing constraint violations while maintaining task performance. Existing approaches typically rely on a single policy to jointly optimize reward and safety, which can cause instability due to conflicting objectives, or they use external safety filters that override actions and require prior system knowledge. In this paper, we propose a modular cost-aware regulator that scales the agent's actions based on predicted constraint violations, preserving exploration through smooth action modulation rather than overriding the policy. The regulator is trained to minimize constraint violations while avoiding degenerate suppression of actions. Our approach integrates seamlessly with off-policy RL methods such as SAC and TD3, and achieves state-of-the-art return-to-cost ratios on Safety Gym locomotion tasks with sparse costs, reducing constraint violations by up to 126 times while increasing returns by over an order of magnitude compared to prior methods.
comment: Accepted in 8th Annual Learning for Dynamics & Control Conference (L4DC)
♻ ☆ How Leg Stiffness Affects Energy Economy in Hopping
In the fields of robotics and biomechanics, the integration of elastic elements such as springs and tendons in legged systems has long been recognized for enabling energy-efficient locomotion. Yet, a significant challenge persists: designing a robotic leg that perform consistently across diverse operating conditions, especially varying average forward speeds. It remains unclear whether, for such a range of operating conditions, the stiffness of the elastic elements needs to be varied or if a similar performance can be obtained by changing the motion and actuation while keeping the stiffness fixed. This work explores the influence of the leg stiffness on the energy efficiency of a monopedal robot through an extensive parametric study of its periodic hopping motion. To this end, we formulate an optimal control problem parameterized by average forward speed and leg stiffness, solving it numerically using direct collocation. Our findings indicate that, compared to the use of a fixed stiffness, employing variable stiffness in legged systems improves energy efficiency by 20 % maximally and by 6.8 % on average across a range of speeds.
♻ ☆ Physical Human-Robot Interaction: A Critical Review of Safety Constraints
This paper aims to provide a clear and rigorous understanding of commonly recognized safety constraints in physical human-robot interaction, particularly regarding ISO/TS 15066. We investigate the derivation of these constraints, critically examine the underlying assumptions, and evaluate their practical implications for system-level safety and performance in industrially relevant scenarios. Key design parameters within safety-critical control architectures are identified, and numerical examples are provided to quantify performance degradation arising from typical approximations and design decisions in manufacturing environments. Within this analysis, the fundamental role of energy in safety assessment is emphasized, providing focused insights into energy-based safety methodologies for collaborative industrial robot systems.
♻ ☆ V-OCBF: Learning Safety Filters from Offline Data via Value-Guided Offline Control Barrier Functions
Ensuring safety in autonomous systems requires controllers that aim to satisfy state-wise constraints without relying on online interaction.While existing Safe Offline RL methods typically enforce soft expected-cost constraints, they struggle to ensure strict state-wise safety. Conversely, Control Barrier Functions (CBFs) offer a principled mechanism to enforce forward invariance, but often rely on expert-designed barrier functions or knowledge of the system dynamics. We introduce Value-Guided Offline Control Barrier Functions (V-OCBF), a framework that learns a neural CBF entirely from offline demonstrations. Unlike prior approaches, V-OCBF does not assume access to the dynamics model; instead, it derives a recursive finite-difference barrier update, enabling model-free learning of a barrier that propagates safety information over time. Moreover, V-OCBF incorporates an expectile-based objective that avoids querying the barrier on out-of-distribution actions and restricts updates to the dataset-supported action set. The learned barrier is then used with a Quadratic Program (QP) formulation to synthesize real-time safe control. Across multiple case studies, V-OCBF yields substantially fewer safety violations than baseline methods while maintaining strong task performance, highlighting its scalability for offline synthesis of safety-critical controllers without online interaction or hand-engineered barriers.
comment: 28 pages, 9 figures, 11 tables. Paper accepted at TMLR
♻ ☆ GPA-VGGT:Adapting VGGT to Large Scale Localization by Self-Supervised Learning with Geometry and Physics Aware Loss
Transformer-based general visual geometry frameworks have shown promising performance in camera pose estimation and 3D scene understanding. Recent advancements in Visual Geometry Grounded Transformer (VGGT) models have shown great promise in camera pose estimation and 3D reconstruction. However, these models typically rely on ground truth labels for training, posing challenges when adapting to unlabeled and unseen scenes. In this paper, we propose a self-supervised framework to train VGGT with unlabeled data, thereby enhancing its localization capability in large-scale environments. To achieve this, we extend conventional pair-wise relations to sequence-wise geometric constraints for self-supervised learning. Specifically, in each sequence, we sample multiple source frames and geometrically project them onto different target frames, which improves temporal feature consistency. We formulate physical photometric consistency and geometric constraints as a joint optimization loss to circumvent the requirement for hard labels. By training the model with this proposed method, not only the local and global cross-view attention layers but also the camera and depth heads can effectively capture the underlying multi-view geometry. Experiments demonstrate that the model converges within hundreds of iterations and achieves significant improvements in large-scale localization. Our code will be released at https://github.com/X-yangfan/GPA-VGGT.
♻ ☆ Multi-Staged Framework for Safety Analysis of Offloaded Services in Distributed Intelligent Transportation Systems
The integration of service-oriented architectures (SOA) with function offloading for distributed, intelligent transportation systems (ITS) offers the opportunity for connected autonomous vehicles (CAVs) to extend their locally available services. One major goal of offloading a subset of functions in the processing chain of a CAV to remote devices is to reduce the overall computational complexity on the CAV. The extension of using remote services, however, requires careful safety analysis, since the remotely created data are corrupted more easily, e.g., through an attacker on the remote device or by intercepting the wireless transmission. To tackle this problem, we first analyze the concept of SOA for distributed environments. From this, we derive a safety framework that validates the reliability of remote services and the data received locally. Since it is possible for the autonomous driving task to offload multiple different services, we propose a specific multi-staged framework for safety analysis dependent on the service composition of local and remote services. For efficiency reasons, we directly include the multi-staged framework for safety analysis in our service-oriented function offloading framework (SOFOF) that we have proposed in earlier work. The evaluation compares the performance of the extended framework considering computational complexity, with energy savings being a major motivation for function offloading, and its capability to detect data from corrupted remote services.
comment: 2025 IEEE International Conference on Intelligent Transportation Systems (ITSC)
♻ ☆ Vi-TacMan: Articulated Object Manipulation via Vision and Touch ICRA 2026
Autonomous manipulation of articulated objects remains a fundamental challenge for robots in human environments. Vision-based methods can infer hidden kinematics but can yield imprecise estimates on unfamiliar objects. Tactile approaches achieve robust control through contact feedback but require accurate initialization. This suggests a natural synergy: vision for global guidance, touch for local precision. Yet no framework systematically exploits this complementarity for generalized articulated manipulation. Here we present Vi-TacMan, which uses vision to propose grasps and coarse directions that seed a tactile controller for precise execution. By incorporating surface normals as geometric priors and modeling directions via von Mises-Fisher distributions, our approach achieves significant gains over baselines (all p<0.0001). Critically, manipulation succeeds without explicit kinematic models -- the tactile controller refines coarse visual estimates through real-time contact regulation. Tests on more than 50,000 simulated and diverse real-world objects confirm robust cross-category generalization. This work establishes that coarse visual cues suffice for reliable manipulation when coupled with tactile feedback, offering a scalable paradigm for autonomous systems in unstructured environments.
comment: ICRA 2026
♻ ☆ DualReg: Dual-Space Filtering and Reinforcement for Rigid Registration CVPR 2026
Noisy, partially overlapping data and the need for real-time processing pose major challenges for rigid registration. Considering that feature-based matching can handle large transformation differences but suffers from limited accuracy, while local geometry-based matching can achieve fine-grained local alignment but relies heavily on a good initial transformation, we propose a novel dual-space paradigm to fully leverage the strengths of both approaches. First, we introduce an efficient filtering mechanism consisting of a computationally lightweight one-point RANSAC algorithm and a subsequent refinement module to eliminate unreliable feature-based correspondences. Subsequently, we treat the filtered correspondences as anchor points, extract geometric proxies, and formulate an effective objective function with a tailored solver to estimate the transformation. Experiments verify our method's effectiveness, as demonstrated by a 32x CPU-time speedup over MAC on KITTI with comparable accuracy. Project page: https://ustc3dv.github.io/DualReg/.
comment: Accepted to CVPR 2026, Project page: https://ustc3dv.github.io/DualReg/
♻ ☆ What Capable Agents Must Know: Selection Theorems for Robust Decision-Making under Uncertainty
As artificial agents become increasingly capable, what internal structure is *necessary* for an agent to act competently under uncertainty? Classical results show that optimal control can be *implemented* using belief states or world models, but not that such representations are required. We prove quantitative "selection theorems" showing that strong task performance (low *average-case regret*) forces world models, belief-like memory and -- under task mixtures -- persistent variables resembling core primitives associated with emotion, along with informational modularity under block-structured tasks. Our results cover stochastic policies, partial observability, and evaluation under task distributions, without assuming optimality, determinism, or access to an explicit model. Technically, we reduce predictive modeling to binary "betting" decisions and show that regret bounds limit probability mass on suboptimal bets, enforcing the predictive distinctions needed to separate high-margin outcomes. In fully observed settings, this yields approximate recovery of the interventional transition kernel; under partial observability, it implies necessity of predictive state and belief-like memory, addressing an open question in prior world-model recovery work.
comment: 23 pages; added PSR recovery (Theorems 3 & 4), and updated related work
♻ ☆ Emergent Dexterity via Diverse Resets and Large-Scale Reinforcement Learning
Patrick Yin, Tyler Westenbroek, Zhengyu Zhang, Joshua Tran, Ignacio Dagnino, Eeshani Shilamkar, Numfor Mbiziwo-Tiapo, Simran Bagaria, Xinlei Liu, Galen Mullins, Andrey Kolobov, Abhishek Gupta
Reinforcement learning in massively parallel physics simulations has driven major progress in sim-to-real robot learning. However, current approaches remain brittle and task-specific, relying on extensive per-task engineering to design rewards, curricula, and demonstrations. Even with this engineering, they often fail on long-horizon, contact-rich manipulation tasks and do not meaningfully scale with compute, as performance quickly saturates when training revisits the same narrow regions of state space. We introduce OmniReset, a simple and scalable framework that enables on-policy reinforcement learning to robustly solve a broad class of dexterous manipulation tasks using a single reward function, fixed algorithm hyperparameters, no curricula, and no human demonstrations. Our key insight is that long-horizon exploration can be dramatically simplified by using simulator resets to systematically expose the RL algorithm to the diverse set of robot-object interactions which underlie dexterous manipulation. OmniReset programmatically generates such resets with minimal human input, converting additional compute directly into broader behavioral coverage and continued performance gains. We show that OmniReset gracefully scales to long-horizon dexterous manipulation tasks beyond the capabilities of existing approaches and is able to learn robust policies over significantly wider ranges of initial conditions than baselines. Finally, we distill OmniReset into visuomotor policies which display robust retrying behavior and substantially higher success rates than baselines when transferred to the real world zero-shot. Project webpage: https://weirdlabuw.github.io/omnireset/