Robotics 33
☆ LaST$_{0}$: Latent Spatio-Temporal Chain-of-Thought for Robotic Vision-Language-Action Model
Zhuoyang Liu, Jiaming Liu, Hao Chen, Ziyu Guo, Chengkai Hou, Chenyang Gu, Jiale Yu, Xiangju Mi, Renrui Zhang, Zhengping Che, Jian Tang, Pheng-Ann Heng, Shanghang Zhang
Vision-Language-Action (VLA) models have recently demonstrated strong generalization capabilities in robotic manipulation. Some existing VLA approaches attempt to improve action accuracy by explicitly generating linguistic reasoning traces or future visual observations before action execution. However, explicit reasoning typically incurs non-negligible inference latency, which constrains the temporal resolution required for robotic manipulation. Moreover, such reasoning is confined to the linguistic space, imposing a representational bottleneck that struggles to faithfully capture ineffable physical attributes. To mitigate these limitations, we propose LaST$_0$, a framework that enables efficient reasoning before acting through a Latent Spatio-Temporal Chain-of-Thought (CoT), capturing fine-grained physical and robotic dynamics that are often difficult to verbalize. Specifically, we introduce a token-efficient latent CoT space that models future visual dynamics, 3D structural information, and robot proprioceptive states, and further extends these representations across time to enable temporally consistent implicit reasoning trajectories. Furthermore, LaST$_0$ adopts a dual-system architecture implemented via a Mixture-of-Transformers design, where a reasoning expert conducts low-frequency latent inference and an acting expert generates high-frequency actions conditioned on robotics-oriented latent representations. To facilitate coordination, LaST$_0$ is trained with heterogeneous operation frequencies, enabling adaptive switching between reasoning and action inference rates during deployment. Across ten simulated and six real-world manipulation tasks, LaST$_0$ improves mean success rates by 8% and 13% over prior VLA methods, respectively, while achieving substantially faster inference. Project website: https://sites.google.com/view/last0
☆ Generate, Transfer, Adapt: Learning Functional Dexterous Grasping from a Single Human Demonstration
Functional grasping with dexterous robotic hands is a key capability for enabling tool use and complex manipulation, yet progress has been constrained by two persistent bottlenecks: the scarcity of large-scale datasets and the absence of integrated semantic and geometric reasoning in learned models. In this work, we present CorDex, a framework that robustly learns dexterous functional grasps of novel objects from synthetic data generated from just a single human demonstration. At the core of our approach is a correspondence-based data engine that generates diverse, high-quality training data in simulation. Based on the human demonstration, our data engine generates diverse object instances of the same category, transfers the expert grasp to the generated objects through correspondence estimation, and adapts the grasp through optimization. Building on the generated data, we introduce a multimodal prediction network that integrates visual and geometric information. By devising a local-global fusion module and an importance-aware sampling mechanism, we enable robust and computationally efficient prediction of functional dexterous grasps. Through extensive experiments across various object categories, we demonstrate that CorDex generalizes well to unseen object instances and significantly outperforms state-of-the-art baselines.
comment: Project Page: https://cordex-manipulation.github.io/
☆ RoboVIP: Multi-View Video Generation with Visual Identity Prompting Augments Robot Manipulation
Boyang Wang, Haoran Zhang, Shujie Zhang, Jinkun Hao, Mingda Jia, Qi Lv, Yucheng Mao, Zhaoyang Lyu, Jia Zeng, Xudong Xu, Jiangmiao Pang
The diversity, quantity, and quality of manipulation data are critical for training effective robot policies. However, due to hardware and physical setup constraints, collecting large-scale real-world manipulation data remains difficult to scale across diverse environments. Recent work uses text-prompt conditioned image diffusion models to augment manipulation data by altering the backgrounds and tabletop objects in the visual observations. However, these approaches often overlook the practical need for multi-view and temporally coherent observations required by state-of-the-art policy models. Further, text prompts alone cannot reliably specify the scene setup. To provide the diffusion model with explicit visual guidance, we introduce visual identity prompting, which supplies exemplar images as conditioning inputs to guide the generation of the desired scene setup. To this end, we also build a scalable pipeline to curate a visual identity pool from large robotics datasets. Using our augmented manipulation data to train downstream vision-language-action and visuomotor policy models yields consistent performance gains in both simulation and real-robot settings.
☆ Driving on Registers
Ellington Kirby, Alexandre Boulch, Yihong Xu, Yuan Yin, Gilles Puy, Éloi Zablocki, Andrei Bursuc, Spyros Gidaris, Renaud Marlet, Florent Bartoccioni, Anh-Quan Cao, Nermin Samet, Tuan-Hung VU, Matthieu Cord
We present DrivoR, a simple and efficient transformer-based architecture for end-to-end autonomous driving. Our approach builds on pretrained Vision Transformers (ViTs) and introduces camera-aware register tokens that compress multi-camera features into a compact scene representation, significantly reducing downstream computation without sacrificing accuracy. These tokens drive two lightweight transformer decoders that generate and then score candidate trajectories. The scoring decoder learns to mimic an oracle and predicts interpretable sub-scores representing aspects such as safety, comfort, and efficiency, enabling behavior-conditioned driving at inference. Despite its minimal design, DrivoR outperforms or matches strong contemporary baselines across NAVSIM-v1, NAVSIM-v2, and the photorealistic closed-loop HUGSIM benchmark. Our results show that a pure-transformer architecture, combined with targeted token compression, is sufficient for accurate, efficient, and adaptive end-to-end driving. Code and checkpoints will be made available via the project page.
☆ Compensation Effect Amplification Control (CEAC): A movement-based approach for coordinated position and velocity control of the elbow of upper-limb prostheses
Despite advances in upper-limb (UL) prosthetic design, achieving intuitive control of intermediate joints - such as the wrist and elbow - remains challenging, particularly for continuous and velocity-modulated movements. We introduce a novel movement-based control paradigm entitled Compensation Effect Amplification Control (CEAC) that leverages users' trunk flexion and extension as input for controlling prosthetic elbow velocity. Considering that the trunk can be both a functional and compensatory joint when performing upper-limb actions, CEAC amplifies the natural coupling between trunk and prosthesis while introducing a controlled delay that allows users to modulate both the position and velocity of the prosthetic joint. We evaluated CEAC in a generic drawing task performed by twelve able-bodied participants using a supernumerary prosthesis with an active elbow. Additionally a multiple-target-reaching task was performed by a subset of ten participants. Results demonstrate task performances comparable to those obtained with natural arm movements, even when gesture velocity or drawing size were varied, while maintaining ergonomic trunk postures. Analysis revealed that CEAC effectively restores joint coordinated action, distributes movement effort between trunk and elbow, enabling intuitive trajectory control without requiring extreme compensatory movements. Overall, CEAC offers a promising control strategy for intermediate joints of UL prostheses, particularly in tasks requiring continuous and precise coordination.
☆ DAVOS: An Autonomous Vehicle Operating System in the Vehicle Computing Era
Vehicle computing represents a fundamental shift in how autonomous vehicles are designed and deployed, transforming them from isolated transportation systems into mobile computing platforms that support both safety-critical, real-time driving and data-centric services. In this setting, vehicles simultaneously support real-time driving pipelines and a growing set of data-driven applications, placing increased responsibility on the vehicle operating system to coordinate computation, data movement, storage, and access. These demands highlight recurring system considerations related to predictable execution, data and execution protection, efficient handling of high-rate sensor data, and long-term system evolvability, commonly summarized as Safety, Security, Efficiency, and Extensibility (SSEE). Existing vehicle operating systems and runtimes address these concerns in isolation, resulting in fragmented software stacks that limit coordination between autonomy workloads and vehicle data services. This paper presents DAVOS, the Delaware Autonomous Vehicle Operating System, a unified vehicle operating system architecture designed for the vehicle computing context. DAVOS provides a cohesive operating system foundation that supports both real-time autonomy and extensible vehicle computing within a single system framework.
☆ The RoboSense Challenge: Sense Anything, Navigate Anywhere, Adapt Across Platforms IROS 2025
Lingdong Kong, Shaoyuan Xie, Zeying Gong, Ye Li, Meng Chu, Ao Liang, Yuhao Dong, Tianshuai Hu, Ronghe Qiu, Rong Li, Hanjiang Hu, Dongyue Lu, Wei Yin, Wenhao Ding, Linfeng Li, Hang Song, Wenwei Zhang, Yuexin Ma, Junwei Liang, Zhedong Zheng, Lai Xing Ng, Benoit R. Cottereau, Wei Tsang Ooi, Ziwei Liu, Zhanpeng Zhang, Weichao Qiu, Wei Zhang, Ji Ao, Jiangpeng Zheng, Siyu Wang, Guang Yang, Zihao Zhang, Yu Zhong, Enzhu Gao, Xinhan Zheng, Xueting Wang, Shouming Li, Yunkai Gao, Siming Lan, Mingfei Han, Xing Hu, Dusan Malic, Christian Fruhwirth-Reisinger, Alexander Prutsch, Wei Lin, Samuel Schulter, Horst Possegger, Linfeng Li, Jian Zhao, Zepeng Yang, Yuhang Song, Bojun Lin, Tianle Zhang, Yuchen Yuan, Chi Zhang, Xuelong Li, Youngseok Kim, Sihwan Hwang, Hyeonjun Jeong, Aodi Wu, Xubo Luo, Erjia Xiao, Lingfeng Zhang, Yingbo Tang, Hao Cheng, Renjing Xu, Wenbo Ding, Lei Zhou, Long Chen, Hangjun Ye, Xiaoshuai Hao, Shuangzhi Li, Junlong Shen, Xingyu Li, Hao Ruan, Jinliang Lin, Zhiming Luo, Yu Zang, Cheng Wang, Hanshi Wang, Xijie Gong, Yixiang Yang, Qianli Ma, Zhipeng Zhang, Wenxiang Shi, Jingmeng Zhou, Weijun Zeng, Kexin Xu, Yuchen Zhang, Haoxiang Fu, Ruibin Hu, Yanbiao Ma, Xiyan Feng, Wenbo Zhang, Lu Zhang, Yunzhi Zhuge, Huchuan Lu, You He, Seungjun Yu, Junsung Park, Youngsun Lim, Hyunjung Shim, Faduo Liang, Zihang Wang, Yiming Peng, Guanyu Zong, Xu Li, Binghao Wang, Hao Wei, Yongxin Ma, Yunke Shi, Shuaipeng Liu, Dong Kong, Yongchun Lin, Huitong Yang, Liang Lei, Haoang Li, Xinliang Zhang, Zhiyong Wang, Xiaofeng Wang, Yuxia Fu, Yadan Luo, Djamahl Etchegaray, Yang Li, Congfei Li, Yuxiang Sun, Wenkai Zhu, Wang Xu, Linru Li, Longjie Liao, Jun Yan, Benwu Wang, Xueliang Ren, Xiaoyu Yue, Jixian Zheng, Jinfeng Wu, Shurui Qin, Wei Cong, Yao He
Autonomous systems are increasingly deployed in open and dynamic environments -- from city streets to aerial and indoor spaces -- where perception models must remain reliable under sensor noise, environmental variation, and platform shifts. However, even state-of-the-art methods often degrade under unseen conditions, highlighting the need for robust and generalizable robot sensing. The RoboSense 2025 Challenge is designed to advance robustness and adaptability in robot perception across diverse sensing scenarios. It unifies five complementary research tracks spanning language-grounded decision making, socially compliant navigation, sensor configuration generalization, cross-view and cross-modal correspondence, and cross-platform 3D perception. Together, these tasks form a comprehensive benchmark for evaluating real-world sensing reliability under domain shifts, sensor failures, and platform discrepancies. RoboSense 2025 provides standardized datasets, baseline models, and unified evaluation protocols, enabling large-scale and reproducible comparison of robust perception methods. The challenge attracted 143 teams from 85 institutions across 16 countries, reflecting broad community engagement. By consolidating insights from 23 winning solutions, this report highlights emerging methodological trends, shared design principles, and open challenges across all tracks, marking a step toward building robots that can sense reliably, act robustly, and adapt across platforms in real-world environments.
comment: Official IROS 2025 RoboSense Challenge Report; 51 pages, 37 figures, 5 tables; Competition Website at https://robosense2025.github.io/
☆ When to Act: Calibrated Confidence for Reliable Human Intention Prediction in Assistive Robotics
Assistive devices must determine both what a user intends to do and how reliable that prediction is before providing support. We introduce a safety-critical triggering framework based on calibrated probabilities for multimodal next-action prediction in Activities of Daily Living. Raw model confidence often fails to reflect true correctness, posing a safety risk. Post-hoc calibration aligns predicted confidence with empirical reliability and reduces miscalibration by about an order of magnitude without affecting accuracy. The calibrated confidence drives a simple ACT/HOLD rule that acts only when reliability is high and withholds assistance otherwise. This turns the confidence threshold into a quantitative safety parameter for assisted actions and enables verifiable behavior in an assistive control loop.
☆ SKATER: Synthesized Kinematics for Advanced Traversing Efficiency on a Humanoid Robot via Roller Skate Swizzles
Junchi Gu, Feiyang Yuan, Weize Shi, Tianchen Huang, Haopeng Zhang, Xiaohu Zhang, Yu Wang, Wei Gao, Shiwu Zhang
Although recent years have seen significant progress of humanoid robots in walking and running, the frequent foot strikes with ground during these locomotion gaits inevitably generate high instantaneous impact forces, which leads to exacerbated joint wear and poor energy utilization. Roller skating, as a sport with substantial biomechanical value, can achieve fast and continuous sliding through rational utilization of body inertia, featuring minimal kinetic energy loss. Therefore, this study proposes a novel humanoid robot with each foot equipped with a row of four passive wheels for roller skating. A deep reinforcement learning control framework is also developed for the swizzle gait with the reward function design based on the intrinsic characteristics of roller skating. The learned policy is first analyzed in simulation and then deployed on the physical robot to demonstrate the smoothness and efficiency of the swizzle gait over traditional bipedal walking gait in terms of Impact Intensity and Cost of Transport during locomotion. A reduction of $75.86\%$ and $63.34\%$ of these two metrics indicate roller skating as a superior locomotion mode for enhanced energy efficiency and joint longevity.
☆ Zero Wrench Control via Wrench Disturbance Observer for Learning-free Peg-in-hole Assembly
This paper proposes a Dynamic Wrench Disturbance Observer (DW-DOB) designed to achieve highly sensitive zero-wrench control in contact-rich manipulation. By embedding task-space inertia into the observer nominal model, DW-DOB cleanly separates intrinsic dynamic reactions from true external wrenches. This preserves sensitivity to small forces and moments while ensuring robust regulation of contact wrenches. A passivity-based analysis further demonstrates that DW-DOB guarantees stable interactions under dynamic conditions, addressing the shortcomings of conventional observers that fail to compensate for inertial effects. Peg-in-hole experiments at industrial tolerances (H7/h6) validate the approach, yielding deeper and more compliant insertions with minimal residual wrenches and outperforming a conventional wrench disturbance observer and a PD baseline. These results highlight DW-DOB as a practical learning-free solution for high-precision zero-wrench control in contact-rich tasks.
☆ SeqWalker: Sequential-Horizon Vision-and-Language Navigation with Hierarchical Planning
Sequential-Horizon Vision-and-Language Navigation (SH-VLN) presents a challenging scenario where agents should sequentially execute multi-task navigation guided by complex, long-horizon language instructions. Current vision-and-language navigation models exhibit significant performance degradation with such multi-task instructions, as information overload impairs the agent's ability to attend to observationally relevant details. To address this problem, we propose SeqWalker, a navigation model built on a hierarchical planning framework. Our SeqWalker features: i) A High-Level Planner that dynamically selects global instructions into contextually relevant sub-instructions based on the agent's current visual observations, thus reducing cognitive load; ii) A Low-Level Planner incorporating an Exploration-Verification strategy that leverages the inherent logical structure of instructions for trajectory error correction. To evaluate SH-VLN performance, we also extend the IVLN dataset and establish a new benchmark. Extensive experiments are performed to demonstrate the superiority of the proposed SeqWalker.
☆ Nightmare Dreamer: Dreaming About Unsafe States And Planning Ahead
Reinforcement Learning (RL) has shown remarkable success in real-world applications, particularly in robotics control. However, RL adoption remains limited due to insufficient safety guarantees. We introduce Nightmare Dreamer, a model-based Safe RL algorithm that addresses safety concerns by leveraging a learned world model to predict potential safety violations and plan actions accordingly. Nightmare Dreamer achieves nearly zero safety violations while maximizing rewards. Nightmare Dreamer outperforms model-free baselines on Safety Gymnasium tasks using only image observations, achieving nearly a 20x improvement in efficiency.
comment: RSS'25: Multi-Objective Optimization and Planning in Robotics Workshop: 5 pages, 8 figures
☆ Optimizing Path Planning using Deep Reinforcement Learning for UGVs in Precision Agriculture
This study focuses on optimizing path planning for unmanned ground vehicles (UGVs) in precision agriculture using deep reinforcement learning (DRL) techniques in continuous action spaces. The research begins with a review of traditional grid-based methods, such as A* and Dijkstra's algorithms, and discusses their limitations in dynamic agricultural environments, highlighting the need for adaptive learning strategies. The study then explores DRL approaches, including Deep Q-Networks (DQN), which demonstrate improved adaptability and performance in two-dimensional simulations. Enhancements such as Double Q-Networks and Dueling Networks are evaluated to further improve decision-making. Building on these results, the focus shifts to continuous action space models, specifically Deep Deterministic Policy Gradient (DDPG) and Twin Delayed Deep Deterministic Policy Gradient (TD3), which are tested in increasingly complex environments. Experiments conducted in a three-dimensional environment using ROS and Gazebo demonstrate the effectiveness of continuous DRL algorithms in navigating dynamic agricultural scenarios. Notably, the pretrained TD3 agent achieves a 95 percent success rate in dynamic environments, demonstrating the robustness of the proposed approach in handling moving obstacles while ensuring safety for both crops and the robot.
☆ Model of Spatial Human-Agent Interaction with Consideration for Others
Communication robots often need to initiate conversations with people in public spaces. At the same time, such robots must not disturb pedestrians. To handle these two requirements, an agent needs to estimate the communication desires of others based on their behavior and then adjust its own communication activities accordingly. In this study, we construct a computational spatial interaction model that considers others. Consideration is expressed as a quantitative parameter: the amount of adjustment of one's internal state to the estimated internal state of the other. To validate the model, we experimented with a human and a virtual robot interacting in a VR environment. The results show that when the participant moves to the target, a virtual robot with a low consideration value inhibits the participant's movement, while a robot with a higher consideration value did not inhibit the participant's movement. When the participant approached the robot, the robot also exhibited approaching behavior, regardless of the consideration value, thus decreasing the participant's movement. These results appear to verify the proposed model's ability to clarify interactions with consideration for others.
☆ UniBiDex: A Unified Teleoperation Framework for Robotic Bimanual Dexterous Manipulation
We present UniBiDex a unified teleoperation framework for robotic bimanual dexterous manipulation that supports both VRbased and leaderfollower input modalities UniBiDex enables realtime contactrich dualarm teleoperation by integrating heterogeneous input devices into a shared control stack with consistent kinematic treatment and safety guarantees The framework employs nullspace control to optimize bimanual configurations ensuring smooth collisionfree and singularityaware motion across tasks We validate UniBiDex on a longhorizon kitchentidying task involving five sequential manipulation subtasks demonstrating higher task success rates smoother trajectories and improved robustness compared to strong baselines By releasing all hardware and software components as opensource we aim to lower the barrier to collecting largescale highquality human demonstration datasets and accelerate progress in robot learning.
☆ Discrete Fourier Transform-based Point Cloud Compression for Efficient SLAM in Featureless Terrain
Simultaneous Localization and Mapping (SLAM) is an essential technology for the efficiency and reliability of unmanned robotic exploration missions. While the onboard computational capability and communication bandwidth are critically limited, the point cloud data handled by SLAM is large in size, attracting attention to data compression methods. To address such a problem, in this paper, we propose a new method for compressing point cloud maps by exploiting the Discrete Fourier Transform (DFT). The proposed technique converts the Digital Elevation Model (DEM) to the frequency-domain 2D image and omits its high-frequency components, focusing on the exploration of gradual terrains such as planets and deserts. Unlike terrains with detailed structures such as artificial environments, high-frequency components contribute little to the representation of gradual terrains. Thus, this method is effective in compressing data size without significant degradation of the point cloud. We evaluated the method in terms of compression rate and accuracy using camera sequences of two terrains with different elevation profiles.
comment: Author's version of a manuscript accepted at the 11th International Conference on Automation, Robotics, and Applications (ICARA). (c) IEEE
☆ Data-Driven Terramechanics Approach Towards a Realistic Real-Time Simulator for Lunar Rovers
High-fidelity simulators for the lunar surface provide a digital environment for extensive testing of rover operations and mission planning. However, current simulators focus on either visual realism or physical accuracy, which limits their capability to replicate lunar conditions comprehensively. This work addresses that gap by combining high visual fidelity with realistic terrain interaction for a realistic representation of rovers on the lunar surface. Because direct simulation of wheel-soil interactions is computationally expensive, a data-driven approach was adopted, using regression models for slip and sinkage from data collected in both full-rover and single-wheel experiments and simulations. The resulting regression-based terramechanics model accurately reproduced steady-state and dynamic slip, as well as sinkage behavior, on flat terrain and slopes up to 20 degrees, with validation against field test results. Additionally, improvements were made to enhance the realism of terrain deformation and wheel trace visualization. This method supports real-time applications that require physically plausible terrain response alongside high visual fidelity.
comment: Author's version of a manuscript accepted at the International Conference on Space Robotics 2025 (iSpaRo 2025). (c) IEEE
☆ Design and Development of Modular Limbs for Reconfigurable Robots on the Moon
Gustavo H. Diaz, A. Sejal Jain, Matteo Brugnera, Elian Neppel, Shreya Santra, Kentaro Uno, Kazuya Yoshida
In this paper, we present the development of 4-DOF robot limbs, which we call Moonbots, designed to connect in various configurations with each other and wheel modules, enabling adaptation to different environments and tasks. These modular components are intended primarily for robotic systems in space exploration and construction on the Moon in our Moonshot project. Such modular robots add flexibility and versatility for space missions where resources are constrained. Each module is driven by a common actuator characterized by a high torque-to-speed ratio, supporting both precise control and dynamic motion when required. This unified actuator design simplifies development and maintenance across the different module types. The paper describes the hardware implementation, the mechanical design of the modules, and the overall software architecture used to control and coordinate them. Additionally, we evaluate the control performance of the actuator under various load conditions to characterize its suitability for modular robot applications. To demonstrate the adaptability of the system, we introduce nine functional configurations assembled from the same set of modules: 4DOF-limb, 8DOF-limb, vehicle, dragon, minimal, quadruped, cargo, cargo-minimal, and bike. These configurations reflect different locomotion strategies and task-specific behaviors, offering a practical foundation for further research in reconfigurable robotic systems.
comment: Author's version of a manuscript accepted at the International Conference on Space Robotics 2025 (iSpaRo 2025). (c) IEEE
☆ Multiagent Reinforcement Learning with Neighbor Action Estimation
Multiagent reinforcement learning, as a prominent intelligent paradigm, enables collaborative decision-making within complex systems. However, existing approaches often rely on explicit action exchange between agents to evaluate action value functions, which is frequently impractical in real-world engineering environments due to communication constraints, latency, energy consumption, and reliability requirements. From an artificial intelligence perspective, this paper proposes an enhanced multiagent reinforcement learning framework that employs action estimation neural networks to infer agent behaviors. By integrating a lightweight action estimation module, each agent infers neighboring agents' behaviors using only locally observable information, enabling collaborative policy learning without explicit action sharing. This approach is fully compatible with standard TD3 algorithms and scalable to larger multiagent systems. At the engineering application level, this framework has been implemented and validated in dual-arm robotic manipulation tasks: two robotic arms collaboratively lift objects. Experimental results demonstrate that this approach significantly enhances the robustness and deployment feasibility of real-world robotic systems while reducing dependence on information infrastructure. Overall, this research advances the development of decentralized multiagent artificial intelligence systems while enabling AI to operate effectively in dynamic, information-constrained real-world environments.
☆ Fast Continuum Robot Shape and External Load State Estimation on SE(3) ICRA 2026
Previous on-manifold approaches to continuum robot state estimation have typically adopted simplified Cosserat rod models, which cannot directly account for actuation inputs or external loads. We introduce a general framework that incorporates uncertainty models for actuation (e.g., tendon tensions), applied forces and moments, process noise, boundary conditions, and arbitrary backbone measurements. By adding temporal priors across time steps, our method additionally performs joint estimation in both the spatial (arclength) and temporal domains, enabling full \textit{spacetime} state estimation. Discretizing the arclength domain yields a factor graph representation of the continuum robot model, which can be exploited for fast batch sparse nonlinear optimization in the style of SLAM. The framework is general and applies to a broad class of continuum robots; as illustrative cases, we show (i) tendon-driven robots in simulation, where we demonstrate real-time kinematics with uncertainty, tip force sensing from position feedback, and distributed load estimation from backbone strain, and (ii) a surgical concentric tube robot in experiment, where we validate accurate kinematics and tip force estimation, highlighting potential for surgical palpation.
comment: Public preprint for ICRA 2026
♻ ☆ $π_0$: A Vision-Language-Action Flow Model for General Robot Control
Kevin Black, Noah Brown, Danny Driess, Adnan Esmail, Michael Equi, Chelsea Finn, Niccolo Fusai, Lachy Groom, Karol Hausman, Brian Ichter, Szymon Jakubczak, Tim Jones, Liyiming Ke, Sergey Levine, Adrian Li-Bell, Mohith Mothukuri, Suraj Nair, Karl Pertsch, Lucy Xiaoyang Shi, James Tanner, Quan Vuong, Anna Walling, Haohuan Wang, Ury Zhilinsky
Robot learning holds tremendous promise to unlock the full potential of flexible, general, and dexterous robot systems, as well as to address some of the deepest questions in artificial intelligence. However, bringing robot learning to the level of generality required for effective real-world systems faces major obstacles in terms of data, generalization, and robustness. In this paper, we discuss how generalist robot policies (i.e., robot foundation models) can address these challenges, and how we can design effective generalist robot policies for complex and highly dexterous tasks. We propose a novel flow matching architecture built on top of a pre-trained vision-language model (VLM) to inherit Internet-scale semantic knowledge. We then discuss how this model can be trained on a large and diverse dataset from multiple dexterous robot platforms, including single-arm robots, dual-arm robots, and mobile manipulators. We evaluate our model in terms of its ability to perform tasks in zero shot after pre-training, follow language instructions from people and from a high-level VLM policy, and its ability to acquire new skills via fine-tuning. Our results cover a wide variety of tasks, such as laundry folding, table cleaning, and assembling boxes.
comment: See project website for videos: https://physicalintelligence.company/blog/pi0 Published in RSS 2025
♻ ☆ Uncertainty-Aware Robotic World Model Makes Offline Model-Based Reinforcement Learning Work on Real Robots
Reinforcement Learning (RL) has achieved impressive results in robotics, yet high-performing pipelines remain highly task-specific, with little reuse of prior data. Offline Model-based RL (MBRL) offers greater data efficiency by training policies entirely from existing datasets, but suffers from compounding errors and distribution shift in long-horizon rollouts. Although existing methods have shown success in controlled simulation benchmarks, robustly applying them to the noisy, biased, and partially observed datasets typical of real-world robotics remains challenging. We present a principled pipeline for making offline MBRL effective on physical robots. Our RWM-U extends autoregressive world models with epistemic uncertainty estimation, enabling temporally consistent multi-step rollouts with uncertainty effectively propagated over long horizons. We combine RWM-U with MOPO-PPO, which adapts uncertainty-penalized policy optimization to the stable, on-policy PPO framework for real-world control. We evaluate our approach on diverse manipulation and locomotion tasks in simulation and on real quadruped and humanoid, training policies entirely from offline datasets. The resulting policies consistently outperform model-free and uncertainty-unaware model-based baselines, and fusing real-world data in model learning further yields robust policies that surpass online model-free baselines trained solely in simulation.
♻ ☆ ReinFlow: Fine-tuning Flow Matching Policy with Online Reinforcement Learning
We propose ReinFlow, a simple yet effective online reinforcement learning (RL) framework that fine-tunes a family of flow matching policies for continuous robotic control. Derived from rigorous RL theory, ReinFlow injects learnable noise into a flow policy's deterministic path, converting the flow into a discrete-time Markov Process for exact and straightforward likelihood computation. This conversion facilitates exploration and ensures training stability, enabling ReinFlow to fine-tune diverse flow model variants, including Rectified Flow [35] and Shortcut Models [19], particularly at very few or even one denoising step. We benchmark ReinFlow in representative locomotion and manipulation tasks, including long-horizon planning with visual input and sparse reward. The episode reward of Rectified Flow policies obtained an average net growth of 135.36% after fine-tuning in challenging legged locomotion tasks while saving denoising steps and 82.63% of wall time compared to state-of-the-art diffusion RL fine-tuning method DPPO [43]. The success rate of the Shortcut Model policies in state and visual manipulation tasks achieved an average net increase of 40.34% after fine-tuning with ReinFlow at four or even one denoising step, whose performance is comparable to fine-tuned DDIM policies while saving computation time for an average of 23.20%. Project webpage: https://reinflow.github.io/
comment: 38 pages
♻ ☆ ImagineNav++: Prompting Vision-Language Models as Embodied Navigator through Scene Imagination
Visual navigation is a fundamental capability for autonomous home-assistance robots, enabling long-horizon tasks such as object search. While recent methods have leveraged Large Language Models (LLMs) to incorporate commonsense reasoning and improve exploration efficiency, their planning remains constrained by textual representations, which cannot adequately capture spatial occupancy or scene geometry--critical factors for navigation decisions. We explore whether Vision-Language Models (VLMs) can achieve mapless visual navigation using only onboard RGB/RGB-D streams, unlocking their potential for spatial perception and planning. We achieve this through an imagination-powered navigation framework, ImagineNav++, which imagines future observation images from candidate robot views and translates navigation planning into a simple best-view image selection problem for VLMs. First, a future-view imagination module distills human navigation preferences to generate semantically meaningful viewpoints with high exploration potential. These imagined views then serve as visual prompts for the VLM to identify the most informative viewpoint. To maintain spatial consistency, we develop a selective foveation memory mechanism, which hierarchically integrates keyframe observations via a sparse-to-dense framework, constructing a compact yet comprehensive memory for long-term spatial reasoning. This approach transforms goal-oriented navigation into a series of tractable point-goal navigation tasks. Extensive experiments on open-vocabulary object and instance navigation benchmarks show that ImagineNav++ achieves SOTA performance in mapless settings, even surpassing most map-based methods, highlighting the importance of scene imagination and memory in VLM-based spatial reasoning.
comment: 17 pages, 10 figures. arXiv admin note: text overlap with arXiv:2410.09874
♻ ☆ Cognitive-Hierarchy Guided End-to-End Planning for Autonomous Driving
While end-to-end autonomous driving has advanced significantly, prevailing methods remain fundamentally misaligned with human cognitive principles in both perception and planning. In this paper, we propose CogAD, a novel end-to-end autonomous driving model that emulates the hierarchical cognition mechanisms of human drivers. CogAD implements dual hierarchical mechanisms: global-to-local context processing for human-like perception and intent-conditioned multi-mode trajectory generation for cognitively-inspired planning. The proposed method demonstrates three principal advantages: comprehensive environmental understanding through hierarchical perception, robust planning exploration enabled by multi-level planning, and diverse yet reasonable multi-modal trajectory generation facilitated by dual-level uncertainty modeling. Extensive experiments on nuScenes and Bench2Drive demonstrate that CogAD achieves state-of-the-art performance in end-to-end planning, exhibiting particular superiority in long-tail scenarios and robust generalization to complex real-world driving conditions.
♻ ☆ Solving Robotics Tasks with Prior Demonstration via Exploration-Efficient Deep Reinforcement Learning
This paper proposes an exploration-efficient Deep Reinforcement Learning with Reference policy (DRLR) framework for learning robotics tasks that incorporates demonstrations. The DRLR framework is developed based on an algorithm called Imitation Bootstrapped Reinforcement Learning (IBRL). We propose to improve IBRL by modifying the action selection module. The proposed action selection module provides a calibrated Q-value, which mitigates the bootstrapping error that otherwise leads to inefficient exploration. Furthermore, to prevent the RL policy from converging to a sub-optimal policy, SAC is used as the RL policy instead of TD3. The effectiveness of our method in mitigating bootstrapping error and preventing overfitting is empirically validated by learning two robotics tasks: bucket loading and open drawer, which require extensive interactions with the environment. Simulation results also demonstrate the robustness of the DRLR framework across tasks with both low and high state-action dimensions, and varying demonstration qualities. To evaluate the developed framework on a real-world industrial robotics task, the bucket loading task is deployed on a real wheel loader. The sim2real results validate the successful deployment of the DRLR framework.
comment: This paper has been accepted for Journal publication in Frontiers in Robotics and AI
♻ ☆ Properties of Lyapunov Subcenter Manifolds in Conservative Mechanical Systems
Multi-body mechanical systems have rich internal dynamics, whose solutions can be exploited as efficient control targets. Yet, solutions non-trivially depend on system parameters, obscuring feasible properties for use as target trajectories. For periodic regulation tasks in robotics applications, we investigate properties of nonlinear normal modes (NNMs) collected in Lyapunov subcenter manifolds (LSMs) of conservative mechanical systems. Using a time-symmetry of conservative mechanical systems, we show that mild non-resonance conditions guarantee LSMs to be Eigenmanifolds, in which NNMs are guaranteed to oscillate between two points of zero velocity. We also prove the existence of a unique generator, which is a connected, 1D manifold that collects these points of zero velocity for a given Eigenmanifold. Furthermore, we show that an additional spatial symmetry provides LSMs with yet stronger properties of Rosenberg manifolds. Here all brake trajectories pass through a unique equilibrium configuration, which can be favorable for control applications. These theoretical results are numerically confirmed on two mechanical systems: a double pendulum and a 5-link pendulum.
comment: 18 pages, 27 figures, submitted to Automatica
♻ ☆ SynDroneVision: A Synthetic Dataset for Image-Based Drone Detection
Developing robust drone detection systems is often constrained by the limited availability of large-scale annotated training data and the high costs associated with real-world data collection. However, leveraging synthetic data generated via game engine-based simulations provides a promising and cost-effective solution to overcome this issue. Therefore, we present SynDroneVision, a synthetic dataset specifically designed for RGB-based drone detection in surveillance applications. Featuring diverse backgrounds, lighting conditions, and drone models, SynDroneVision offers a comprehensive training foundation for deep learning algorithms. To evaluate the dataset's effectiveness, we perform a comparative analysis across a selection of recent YOLO detection models. Our findings demonstrate that SynDroneVision is a valuable resource for real-world data enrichment, achieving notable enhancements in model performance and robustness, while significantly reducing the time and costs of real-world data acquisition. SynDroneVision will be publicly released upon paper acceptance.
♻ ☆ RoboReward: General-Purpose Vision-Language Reward Models for Robotics
A well-designed reward is critical for effective reinforcement learning-based policy improvement. In real-world robotics, obtaining such rewards typically requires either labor-intensive human labeling or brittle, handcrafted objectives. Vision-language models (VLMs) have shown promise as automatic reward models, yet their effectiveness on real robot tasks is poorly understood. In this work, we aim to close this gap by introducing (1) RoboReward, a robotics reward dataset and benchmark built on large-scale real-robot corpora from Open X-Embodiment (OXE) and RoboArena, and (2) vision-language reward models trained on this dataset (RoboReward 4B/8B). Because OXE is success-heavy and lacks failure examples, we propose a negative examples data augmentation pipeline that generates calibrated negative and near-misses via counterfactual relabeling of successful episodes and temporal clipping to create partial-progress outcomes from the same videos. Using this framework, we build a large training and evaluation dataset spanning diverse tasks and embodiments to test whether state-of-the-art VLMs can reliably provide rewards for robot learning. Our evaluation of open and proprietary VLMs finds that no model excels across tasks, highlighting substantial room for improvement. We then train general-purpose 4B- and 8B-parameter models that outperform much larger VLMs in assigning rewards for short-horizon robotic tasks. Finally, we deploy the 8B model in real-robot reinforcement learning and find that it improves policy learning over Gemini Robotics-ER 1.5 while narrowing the gap to RL training with human-provided rewards. We release the full dataset, trained reward models, and evaluation suite on our website to advance the development of general-purpose reward models in robotics: https://crfm.stanford.edu/helm/robo-reward-bench (project website).
♻ ☆ Human-Aided Trajectory Planning for Automated Vehicles through Teleoperation and Arbitration Graphs
Teleoperation enables remote human support of automated vehicles in scenarios where the automation is not able to find an appropriate solution. Remote assistance concepts, where operators provide discrete inputs to aid specific automation modules like planning, is gaining interest due to its reduced workload on the human remote operator and improved safety. However, these concepts are challenging to implement and maintain due to their deep integration and interaction with the automated driving system. In this paper, we propose a solution to facilitate the implementation of remote assistance concepts that intervene on planning level and extend the operational design domain of the vehicle at runtime. Using arbitration graphs, a modular decision-making framework, we integrate remote assistance into an existing automated driving system without modifying the original software components. Our simulative implementation demonstrates this approach in two use cases, allowing operators to adjust planner constraints and enable trajectory generation beyond nominal operational design domains.
comment: 7 pages, 8 figures, presented at IEEE Intelligent Vehicles Symposium 2025, video demonstration available at https://www.youtube.com/watch?v=fVSO-YOeGMk
♻ ☆ Bio-Skin: A Cost-Effective Thermostatic Tactile Sensor with Multi-Modal Force and Temperature Detection IROS2025
Tactile sensors can significantly enhance the perception of humanoid robotics systems by providing contact information that facilitates human-like interactions. However, existing commercial tactile sensors focus on improving the resolution and sensitivity of single-modal detection with high-cost components and densely integrated design, incurring complex manufacturing processes and unaffordable prices. In this work, we present Bio-Skin, a cost-effective multi-modal tactile sensor that utilizes single-axis Hall-effect sensors for planar normal force measurement and bar-shape piezo resistors for 2D shear force measurement. A thermistor coupling with a heating wire is integrated into a silicone body to achieve temperature sensation and thermostatic function analogous to human skin. We also present a cross-reference framework to validate the two modalities of the force sensing signal, improving the sensing fidelity in a complex electromagnetic environment. Bio-Skin has a multi-layer design, and each layer is manufactured sequentially and subsequently integrated, thereby offering a fast production pathway. After calibration, Bio-Skin demonstrates performance metrics-including signal-to-range ratio, sampling rate, and measurement range-comparable to current commercial products, with one-tenth of the cost. The sensor's real-world performance is evaluated using an Allegro hand in object grasping tasks, while its temperature regulation functionality was assessed in a material detection task.
comment: This work has been published by IROS2025
♻ ☆ Forging Spatial Intelligence: A Roadmap of Multi-Modal Data Pre-Training for Autonomous Systems
The rapid advancement of autonomous systems, including self-driving vehicles and drones, has intensified the need to forge true Spatial Intelligence from multi-modal onboard sensor data. While foundation models excel in single-modal contexts, integrating their capabilities across diverse sensors like cameras and LiDAR to create a unified understanding remains a formidable challenge. This paper presents a comprehensive framework for multi-modal pre-training, identifying the core set of techniques driving progress toward this goal. We dissect the interplay between foundational sensor characteristics and learning strategies, evaluating the role of platform-specific datasets in enabling these advancements. Our central contribution is the formulation of a unified taxonomy for pre-training paradigms: ranging from single-modality baselines to sophisticated unified frameworks that learn holistic representations for advanced tasks like 3D object detection and semantic occupancy prediction. Furthermore, we investigate the integration of textual inputs and occupancy representations to facilitate open-world perception and planning. Finally, we identify critical bottlenecks, such as computational efficiency and model scalability, and propose a roadmap toward general-purpose multi-modal foundation models capable of achieving robust Spatial Intelligence for real-world deployment.
comment: Survey; 40 pages, 7 figures, 9 tables; GitHub Repo at https://github.com/worldbench/awesome-spatial-intelligence
♻ ☆ FoldNet: Learning Generalizable Closed-Loop Policy for Garment Folding via Keypoint-Driven Asset and Demonstration Synthesis
Due to the deformability of garments, generating a large amount of high-quality data for robotic garment manipulation tasks is highly challenging. In this paper, we present a synthetic garment dataset that can be used for robotic garment folding. We begin by constructing geometric garment templates based on keypoints and applying generative models to generate realistic texture patterns. Leveraging these keypoint annotations, we generate folding demonstrations in simulation and train folding policies via closed-loop imitation learning. To improve robustness, we propose KG-DAgger, which uses a keypoint-based strategy to generate demonstration data for recovering from failures. KG-DAgger significantly improves the model performance, boosting the real-world success rate by 25\%. After training with 15K trajectories (about 2M image-action pairs), the model achieves a 75\% success rate in the real world. Experiments in both simulation and real-world settings validate the effectiveness of our proposed framework.