Robotics 30
☆ Fast-ThinkAct: Efficient Vision-Language-Action Reasoning via Verbalizable Latent Planning
Vision-Language-Action (VLA) tasks require reasoning over complex visual scenes and executing adaptive actions in dynamic environments. While recent studies on reasoning VLAs show that explicit chain-of-thought (CoT) can improve generalization, they suffer from high inference latency due to lengthy reasoning traces. We propose Fast-ThinkAct, an efficient reasoning framework that achieves compact yet performant planning through verbalizable latent reasoning. Fast-ThinkAct learns to reason efficiently with latent CoTs by distilling from a teacher, driven by a preference-guided objective to align manipulation trajectories that transfers both linguistic and visual planning capabilities for embodied control. This enables reasoning-enhanced policy learning that effectively connects compact reasoning to action execution. Extensive experiments across diverse embodied manipulation and reasoning benchmarks demonstrate that Fast-ThinkAct achieves strong performance with up to 89.3\% reduced inference latency over state-of-the-art reasoning VLAs, while maintaining effective long-horizon planning, few-shot adaptation, and failure recovery.
comment: Project page: https://jasper0314-huang.github.io/fast-thinkact/
☆ Sim2real Image Translation Enables Viewpoint-Robust Policies from Fixed-Camera Datasets
Vision-based policies for robot manipulation have achieved significant recent success, but are still brittle to distribution shifts such as camera viewpoint variations. Robot demonstration data is scarce and often lacks appropriate variation in camera viewpoints. Simulation offers a way to collect robot demonstrations at scale with comprehensive coverage of different viewpoints, but presents a visual sim2real challenge. To bridge this gap, we propose MANGO -- an unpaired image translation method with a novel segmentation-conditioned InfoNCE loss, a highly-regularized discriminator design, and a modified PatchNCE loss. We find that these elements are crucial for maintaining viewpoint consistency during sim2real translation. When training MANGO, we only require a small amount of fixed-camera data from the real world, but show that our method can generate diverse unseen viewpoints by translating simulated observations. In this domain, MANGO outperforms all other image translation methods we tested. Imitation-learning policies trained on data augmented by MANGO are able to achieve success rates as high as 60\% on views that the non-augmented policy fails completely on.
☆ Multimodal Signal Processing For Thermo-Visible-Lidar Fusion In Real-time 3D Semantic Mapping
In complex environments, autonomous robot navigation and environmental perception pose higher requirements for SLAM technology. This paper presents a novel method for semantically enhancing 3D point cloud maps with thermal information. By first performing pixel-level fusion of visible and infrared images, the system projects real-time LiDAR point clouds onto this fused image stream. It then segments heat source features in the thermal channel to instantly identify high temperature targets and applies this temperature information as a semantic layer on the final 3D map. This approach generates maps that not only have accurate geometry but also possess a critical semantic understanding of the environment, making it highly valuable for specific applications like rapid disaster assessment and industrial preventive maintenance.
comment: 5 pages,7 figures. Under review
☆ Learning Whole-Body Human-Humanoid Interaction from Human-Human Demonstrations
Wei-Jin Huang, Yue-Yi Zhang, Yi-Lin Wei, Zhi-Wei Xia, Juantao Tan, Yuan-Ming Li, Zhilin Zhao, Wei-Shi Zheng
Enabling humanoid robots to physically interact with humans is a critical frontier, but progress is hindered by the scarcity of high-quality Human-Humanoid Interaction (HHoI) data. While leveraging abundant Human-Human Interaction (HHI) data presents a scalable alternative, we first demonstrate that standard retargeting fails by breaking the essential contacts. We address this with PAIR (Physics-Aware Interaction Retargeting), a contact-centric, two-stage pipeline that preserves contact semantics across morphology differences to generate physically consistent HHoI data. This high-quality data, however, exposes a second failure: conventional imitation learning policies merely mimic trajectories and lack interactive understanding. We therefore introduce D-STAR (Decoupled Spatio-Temporal Action Reasoner), a hierarchical policy that disentangles when to act from where to act. In D-STAR, Phase Attention (when) and a Multi-Scale Spatial module (where) are fused by the diffusion head to produce synchronized whole-body behaviors beyond mimicry. By decoupling these reasoning streams, our model learns robust temporal phases without being distracted by spatial noise, leading to responsive, synchronized collaboration. We validate our framework through extensive and rigorous simulations, demonstrating significant performance gains over baseline approaches and a complete, effective pipeline for learning complex whole-body interactions from HHI data.
☆ CLARE: Continual Learning for Vision-Language-Action Models via Autonomous Adapter Routing and Expansion
To teach robots complex manipulation tasks, it is now a common practice to fine-tune a pre-trained vision-language-action model (VLA) on task-specific data. However, since this recipe updates existing representations, it is unsuitable for long-term operation in the real world, where robots must continually adapt to new tasks and environments while retaining the knowledge they have already acquired. Existing continual learning methods for robotics commonly require storing previous data (exemplars), struggle with long task sequences, or rely on task identifiers for deployment. To address these limitations, we propose CLARE, a general, parameter-efficient framework for exemplar-free continual learning with VLAs. CLARE introduces lightweight modular adapters into selected feedforward layers and autonomously expands the model only where necessary when learning a new task, guided by layer-wise feature similarity. During deployment, an autoencoder-based routing mechanism dynamically activates the most relevant adapters without requiring task labels. Through extensive experiments on the LIBERO benchmark, we show that CLARE achieves high performance on new tasks without catastrophic forgetting of earlier tasks, significantly outperforming even exemplar-based methods. Code and data are available at https://tum-lsy.github.io/clare.
comment: Project page: https://tum-lsy.github.io/clare. 9 pages, 5 figures
☆ Data Scaling for Navigation in Unknown Environments
Generalization of imitation-learned navigation policies to environments unseen in training remains a major challenge. We address this by conducting the first large-scale study of how data quantity and data diversity affect real-world generalization in end-to-end, map-free visual navigation. Using a curated 4,565-hour crowd-sourced dataset collected across 161 locations in 35 countries, we train policies for point goal navigation and evaluate their closed-loop control performance on sidewalk robots operating in four countries, covering 125 km of autonomous driving.
Our results show that large-scale training data enables zero-shot navigation in unknown environments, approaching the performance of policies trained with environment-specific demonstrations. Critically, we find that data diversity is far more important than data quantity. Doubling the number of geographical locations in a training set decreases navigation errors by ~15%, while performance benefit from adding data from existing locations saturates with very little data. We also observe that, with noisy crowd-sourced data, simple regression-based models outperform generative and sequence-based architectures. We release our policies, evaluation setup and example videos on the project page.
☆ ReflexDiffusion: Reflection-Enhanced Trajectory Planning for High-lateral-acceleration Scenarios in Autonomous Driving AAAI 2026
Generating safe and reliable trajectories for autonomous vehicles in long-tail scenarios remains a significant challenge, particularly for high-lateral-acceleration maneuvers such as sharp turns, which represent critical safety situations. Existing trajectory planners exhibit systematic failures in these scenarios due to data imbalance. This results in insufficient modelling of vehicle dynamics, road geometry, and environmental constraints in high-risk situations, leading to suboptimal or unsafe trajectory prediction when vehicles operate near their physical limits. In this paper, we introduce ReflexDiffusion, a novel inference-stage framework that enhances diffusion-based trajectory planners through reflective adjustment. Our method introduces a gradient-based adjustment mechanism during the iterative denoising process: after each standard trajectory update, we compute the gradient between the conditional and unconditional noise predictions to explicitly amplify critical conditioning signals, including road curvature and lateral vehicle dynamics. This amplification enforces strict adherence to physical constraints, particularly improving stability during high-lateral-acceleration maneuvers where precise vehicle-road interaction is paramount. Evaluated on the nuPlan Test14-hard benchmark, ReflexDiffusion achieves a 14.1% improvement in driving score for high-lateral-acceleration scenarios over the state-of-the-art (SOTA) methods. This demonstrates that inference-time trajectory optimization can effectively compensate for training data sparsity by dynamically reinforcing safety-critical constraints near handling limits. The framework's architecture-agnostic design enables direct deployment to existing diffusion-based planners, offering a practical solution for improving autonomous vehicle safety in challenging driving conditions.
comment: Accepted by AAAI 2026
☆ Feedback-Based Mobile Robot Navigation in 3-D Environments Using Artificial Potential Functions Technical Report
This technical report presents the construction and analysis of polynomial navigation functions for motion planning in 3-D workspaces populated by spherical and cylindrical obstacles. The workspace is modeled as a bounded spherical region, and obstacles are encoded using smooth polynomial implicit functions. We establish conditions under which the proposed navigation functions admit a unique non-degenerate minimum at the target while avoiding local minima, including in the presence of pairwise intersecting obstacles. Gradient and Hessian analyses are provided, and the theoretical results are validated through numerical simulations in obstacle rich 3-D environments.
☆ Online Trajectory Optimization for Arbitrary-Shaped Mobile Robots via Polynomial Separating Hypersurfaces
An emerging class of trajectory optimization methods enforces collision avoidance by jointly optimizing the robot's configuration and a separating hyperplane. However, as linear separators only apply to convex sets, these methods require convex approximations of both the robot and obstacles, which becomes an overly conservative assumption in cluttered and narrow environments. In this work, we unequivocally remove this limitation by introducing nonlinear separating hypersurfaces parameterized by polynomial functions. We first generalize the classical separating hyperplane theorem and prove that any two disjoint bounded closed sets in Euclidean space can be separated by a polynomial hypersurface, serving as the theoretical foundation for nonlinear separation of arbitrary geometries. Building on this result, we formulate a nonlinear programming (NLP) problem that jointly optimizes the robot's trajectory and the coefficients of the separating polynomials, enabling geometry-aware collision avoidance without conservative convex simplifications. The optimization remains efficiently solvable using standard NLP solvers. Simulation and real-world experiments with nonconvex robots demonstrate that our method achieves smooth, collision-free, and agile maneuvers in environments where convex-approximation baselines fail.
☆ Vision-Conditioned Variational Bayesian Last Layer Dynamics Models
Agile control of robotic systems often requires anticipating how the environment affects system behavior. For example, a driver must perceive the road ahead to anticipate available friction and plan actions accordingly. Achieving such proactive adaptation within autonomous frameworks remains a challenge, particularly under rapidly changing conditions. Traditional modeling approaches often struggle to capture abrupt variations in system behavior, while adaptive methods are inherently reactive and may adapt too late to ensure safety. We propose a vision-conditioned variational Bayesian last-layer dynamics model that leverages visual context to anticipate changes in the environment. The model first learns nominal vehicle dynamics and is then fine-tuned with feature-wise affine transformations of latent features, enabling context-aware dynamics prediction. The resulting model is integrated into an optimal controller for vehicle racing. We validate our method on a Lexus LC500 racing through water puddles. With vision-conditioning, the system completed all 12 attempted laps under varying conditions. In contrast, all baselines without visual context consistently lost control, demonstrating the importance of proactive dynamics adaptation in high-performance applications.
comment: 9 pages, 7 figures, currently under review
☆ CEI: A Unified Interface for Cross-Embodiment Visuomotor Policy Learning in 3D Space
Robotic foundation models trained on large-scale manipulation datasets have shown promise in learning generalist policies, but they often overfit to specific viewpoints, robot arms, and especially parallel-jaw grippers due to dataset biases. To address this limitation, we propose Cross-Embodiment Interface (\CEI), a framework for cross-embodiment learning that enables the transfer of demonstrations across different robot arm and end-effector morphologies. \CEI introduces the concept of \textit{functional similarity}, which is quantified using Directional Chamfer Distance. Then it aligns robot trajectories through gradient-based optimization, followed by synthesizing observations and actions for unseen robot arms and end-effectors. In experiments, \CEI transfers data and policies from a Franka Panda robot to \textbf{16} different embodiments across \textbf{3} tasks in simulation, and supports bidirectional transfer between a UR5+AG95 gripper robot and a UR5+Xhand robot across \textbf{6} real-world tasks, achieving an average transfer ratio of 82.4\%. Finally, we demonstrate that \CEI can also be extended with spatial generalization and multimodal motion generation capabilities using our proposed techniques. Project website: https://cross-embodiment-interface.github.io/
☆ Vision Foundation Models for Domain Generalisable Cross-View Localisation in Planetary Ground-Aerial Robotic Teams
Accurate localisation in planetary robotics enables the advanced autonomy required to support the increased scale and scope of future missions. The successes of the Ingenuity helicopter and multiple planetary orbiters lay the groundwork for future missions that use ground-aerial robotic teams. In this paper, we consider rovers using machine learning to localise themselves in a local aerial map using limited field-of-view monocular ground-view RGB images as input. A key consideration for machine learning methods is that real space data with ground-truth position labels suitable for training is scarce. In this work, we propose a novel method of localising rovers in an aerial map using cross-view-localising dual-encoder deep neural networks. We leverage semantic segmentation with vision foundation models and high volume synthetic data to bridge the domain gap to real images. We also contribute a new cross-view dataset of real-world rover trajectories with corresponding ground-truth localisation data captured in a planetary analogue facility, plus a high volume dataset of analogous synthetic image pairs. Using particle filters for state estimation with the cross-view networks allows accurate position estimation over simple and complex trajectories based on sequences of ground-view images.
comment: 7 pages, 10 figures. Presented at the International Conference on Space Robotics (iSpaRo) 2025 in Sendai, Japan. Dataset available: https://doi.org/10.5281/zenodo.17364038
☆ Design Methodology of Hydraulically-driven Soft Robotic Gripper for a Large and Heavy Object
This paper presents a design methodology of a hydraulically-driven soft robotic gripper for grasping a large and heavy object -- approximately 10 - 20 kg with 20 - 30 cm diameter. Most existing soft grippers are pneumatically actuated with several hundred kPa pressure, and cannot generate output force sufficient for such a large and heavy object. Instead of pneumatic actuation, hydraulic actuation has a potential to generate much larger power by several MPa pressure. In this study, we develop a hydraulically-driven soft gripper, in which its basic design parameters are determined based on a mathematical model that represents the relationship among the driving pressure, bending angle, object mass and grasping force. Moreover, we selected materials suitable for grasping a heavier object, based on the finite element analysis result of the detailed design. We report experimental results on a 20 kg object grasping and closed-loop control of the finger bending angle.
♻ ☆ Causality-enhanced Decision-Making for Autonomous Mobile Robots in Dynamic Environments
The growing integration of robots in shared environments - such as warehouses, shopping centres, and hospitals - demands a deep understanding of the underlying dynamics and human behaviours, including how, when, and where individuals engage in various activities and interactions. This knowledge goes beyond simple correlation studies and requires a more comprehensive causal analysis. By leveraging causal inference to model cause-and-effect relationships, we can better anticipate critical environmental factors and enable autonomous robots to plan and execute tasks more effectively. To this end, we propose a novel causality-based decision-making framework that reasons over a learned causal model to assist the robot in deciding when and how to complete a given task. In the examined use case - i.e., a warehouse shared with people - we exploit the causal model to estimate battery usage and human obstructions as factors influencing the robot's task execution. This reasoning framework supports the robot in making informed decisions about task timing and strategy. To achieve this, we developed also PeopleFlow, a new Gazebo-based simulator designed to model context-sensitive human-robot spatial interactions in shared workspaces. PeopleFlow features realistic human and robot trajectories influenced by contextual factors such as time, environment layout, and robot state, and can simulate a large number of agents. While the simulator is general-purpose, in this paper we focus on a warehouse-like environment as a case study, where we conduct an extensive evaluation benchmarking our causal approach against a non-causal baseline. Our findings demonstrate the efficacy of the proposed solutions, highlighting how causal reasoning enables autonomous robots to operate more efficiently and safely in dynamic environments shared with humans.
comment: Causal Discovery and Inference - Robot Autonomy - Human-Robot Spatial Interaction - Decision-Making
♻ ☆ SPARK: Safe Protective and Assistive Robot Kit
Yifan Sun, Rui Chen, Kai S. Yun, Yikuan Fang, Sebin Jung, Feihan Li, Bowei Li, Weiye Zhao, Changliu Liu
This paper introduces the Safe Protective and Assistive Robot Kit (SPARK), a comprehensive benchmark designed to ensure safety in humanoid autonomy and teleoperation. Humanoid robots pose significant safety risks due to their physical capabilities of interacting with complex environments. The physical structures of humanoid robots further add complexity to the design of general safety solutions. To facilitate safe deployment of complex robot systems, SPARK can be used as a toolbox that comes with state-of-the-art safe control algorithms in a modular and composable robot control framework. Users can easily configure safety criteria and sensitivity levels to optimize the balance between safety and performance. To accelerate humanoid safety research and development, SPARK provides simulation benchmarks that compare safety approaches in a variety of environments, tasks, and robot models. Furthermore, SPARK allows quick deployment of synthesized safe controllers on real robots. For hardware deployment, SPARK supports Apple Vision Pro (AVP) or a Motion Capture System as external sensors, while offering interfaces for seamless integration with alternative hardware setups at the same time. This paper demonstrates SPARK's capability with both simulation experiments and case studies with a Unitree G1 humanoid robot. Leveraging these advantages of SPARK, users and researchers can significantly improve the safety of their humanoid systems as well as accelerate relevant research. The open source code is available at: https://github.com/intelligent-control-lab/spark.
comment: Presented at IFAC Symposium on Robotics
♻ ☆ Environment as Policy: Learning to Race in Unseen Tracks ICRA
Reinforcement learning (RL) has achieved outstanding success in complex robot control tasks, such as drone racing, where the RL agents have outperformed human champions in a known racing track. However, these agents fail in unseen track configurations, always requiring complete retraining when presented with new track layouts. This work aims to develop RL agents that generalize effectively to novel track configurations without retraining. The naive solution of training directly on a diverse set of track layouts can overburden the agent, resulting in suboptimal policy learning as the increased complexity of the environment impairs the agent's ability to learn to fly. To enhance the generalizability of the RL agent, we propose an adaptive environment-shaping framework that dynamically adjusts the training environment based on the agent's performance. We achieve this by leveraging a secondary RL policy to design environments that strike a balance between being challenging and achievable, allowing the agent to adapt and improve progressively. Using our adaptive environment shaping, one single racing policy efficiently learns to race in diverse challenging tracks. Experimental results validated in both simulation and the real world show that our method enables drones to successfully fly complex and unseen race tracks, outperforming existing environment-shaping techniques. Project page: http://rpg.ifi.uzh.ch/env_as_policy.
comment: Accepted at IEEE International Conference on Robotics and Automation (ICRA), 2025
♻ ☆ Periodic robust robotic rock chop via virtual model control
Robotic cutting is a challenging contact-rich manipulation task where the robot must simultaneously negotiate unknown object mechanics, large contact forces, and precise motion requirements. We introduce a new active virtual-model control scheme that enables knife rocking motion for robot manipulators, without pre-planned trajectories or precise information of the environment. Motion is generated and controlled through switching virtual coupling with virtual mechanisms, given by virtual springs, dampers, and masses arranged in a suitable way. Through analysis and experiments, we demonstrate that the controlled robot behavior settles into a periodic motion. Experiments with a Franka manipulator demonstrate robust cuts with five different vegetables, and sub-millimeter slice accuracy from 1 mm to 6 mm at nearly one cut per second. The same controller survives changes in knife shape and cutting board height, and adaptation to a different humanoid manipulator, demonstrating robustness and platform independence.
♻ ☆ Shape-Space Graphs: Fast and Collision-Free Path Planning for Soft Robots
Soft robots, inspired by elephant trunks or octopus arms, offer extraordinary flexibility to bend, twist, and elongate in ways that rigid robots cannot. However, their motion planning remains a challenge, especially in cluttered environments with obstacles, due to their highly nonlinear and infinite-dimensional kinematics. Here, we present a graph-based path planning tool for an elephant-trunk-inspired soft robot designed with three artificial muscle fibers that allow for continuous deformation through contraction. Using a biomechanical model that integrates morphoelastic and active filament theories, we precompute a shape library and construct a k-nearest neighbor graph in \emph{shape space}, ensuring that each node corresponds to a valid robot shape. For the graph, we use signed distance functions to prune nodes and edges colliding with obstacles, and define multi-objective edge costs based on geometric distance and actuation effort, enabling energy-aware planning with collision avoidance. We demonstrate that our algorithm reliably avoids obstacles and generates feasible paths within milliseconds from precomputed graphs using Dijkstra's algorithm. We show that including energy costs can drastically reduce the actuation effort compared to geometry-only planning, at the expense of longer tip trajectories. Our results highlight the potential of shape-space graph search for fast and reliable path planning in the field of soft robotics, paving the way for real-time applications in surgical, industrial, and assistive settings.
comment: revised version
♻ ☆ UniConFlow: A Unified Constrained Flow-Matching Framework for Certified Motion Planning
Generative models have become increasingly powerful tools for robot motion generation, enabling flexible and multimodal trajectory generation across various tasks. Yet, most existing approaches remain limited in handling multiple types of constraints, such as collision avoidance, actuation limits, and dynamic consistency, which are typically addressed individually or heuristically. In this work, we propose UniConFlow, a unified constrained flow matching-based framework for trajectory generation that systematically incorporates both equality and inequality constraints. Moreover, UniConFlow introduces a novel prescribed-time zeroing function that shapes a time-varying guidance field during inference, allowing the generation process to adapt to varying system models and task requirements. Furthermore, to further address the computational challenges of long-horizon and high-dimensional trajectory generation, we propose two practical strategies for the terminal constraint enforcement and inference process: a violation-segment extraction protocol that precisely localizes and refines only the constraint-violating portions of trajectories, and a trajectory compression method that accelerates optimization in a reduced-dimensional space while preserving high-fidelity reconstruction after decoding. Empirical validation across three experiments, including a double inverted pendulum, a real-to-sim car racing task, and a sim-to-real manipulation task, demonstrates that UniConFlow outperforms state-of-the-art generative planners and conventional optimization baselines, achieving superior performance on certified motion planning metrics such as safety, kinodynamic consistency, and action feasibility. Project page is available at: https://uniconflow.github.io.
♻ ☆ Learning on the Fly: Rapid Policy Adaptation via Differentiable Simulation
Learning control policies in simulation enables rapid, safe, and cost-effective development of advanced robotic capabilities. However, transferring these policies to the real world remains difficult due to the sim-to-real gap, where unmodeled dynamics and environmental disturbances can degrade policy performance. Existing approaches, such as domain randomization and Real2Sim2Real pipelines, can improve policy robustness, but either struggle under out-of-distribution conditions or require costly offline retraining. In this work, we approach these problems from a different perspective. Instead of relying on diverse training conditions before deployment, we focus on rapidly adapting the learned policy in the real world in an online fashion. To achieve this, we propose a novel online adaptive learning framework that unifies residual dynamics learning with real-time policy adaptation inside a differentiable simulation. Starting from a simple dynamics model, our framework refines the model continuously with real-world data to capture unmodeled effects and disturbances such as payload changes and wind. The refined dynamics model is embedded in a differentiable simulation framework, enabling gradient backpropagation through the dynamics and thus rapid, sample-efficient policy updates beyond the reach of classical RL methods like PPO. All components of our system are designed for rapid adaptation, enabling the policy to adjust to unseen disturbances within 5 seconds of training. We validate the approach on agile quadrotor control under various disturbances in both simulation and the real world. Our framework reduces hovering error by up to 81% compared to L1-MPC and 55% compared to DATT, while also demonstrating robustness in vision-based control without explicit state estimation.
♻ ☆ JuggleRL: Mastering Ball Juggling with a Quadrotor via Deep Reinforcement Learning
Shilong Ji, Yinuo Chen, Chuqi Wang, Jiayu Chen, Ruize Zhang, Feng Gao, Wenhao Tang, Shu'ang Yu, Sirui Xiang, Xinlei Chen, Chao Yu, Yu Wang
Aerial robots interacting with objects must perform precise, contact-rich maneuvers under uncertainty. In this paper, we study the problem of aerial ball juggling using a quadrotor equipped with a racket, a task that demands accurate timing, stable control, and continuous adaptation. We propose JuggleRL, the first reinforcement learning-based system for aerial juggling. It learns closed-loop policies in large-scale simulation using systematic calibration of quadrotor and ball dynamics to reduce the sim-to-real gap. The training incorporates reward shaping to encourage racket-centered hits and sustained juggling, as well as domain randomization over ball position and coefficient of restitution to enhance robustness and transferability. The learned policy outputs mid-level commands executed by a low-level controller and is deployed zero-shot on real hardware, where an enhanced perception module with a lightweight communication protocol reduces delays in high-frequency state estimation and ensures real-time control. Experiments show that JuggleRL achieves an average of $311$ hits over $10$ consecutive trials in the real world, with a maximum of $462$ hits observed, far exceeding a model-based baseline that reaches at most $14$ hits with an average of $3.1$. Moreover, the policy generalizes to unseen conditions, successfully juggling a lighter $5$ g ball with an average of $145.9$ hits. This work demonstrates that reinforcement learning can empower aerial robots with robust and stable control in dynamic interaction tasks.
♻ ☆ SAC Flow: Sample-Efficient Reinforcement Learning of Flow-Based Policies via Velocity-Reparameterized Sequential Modeling
Yixian Zhang, Shu'ang Yu, Tonghe Zhang, Mo Guang, Haojia Hui, Kaiwen Long, Yu Wang, Chao Yu, Wenbo Ding
Training expressive flow-based policies with off-policy reinforcement learning is notoriously unstable due to gradient pathologies in the multi-step action sampling process. We trace this instability to a fundamental connection: the flow rollout is algebraically equivalent to a residual recurrent computation, making it susceptible to the same vanishing and exploding gradients as RNNs. To address this, we reparameterize the velocity network using principles from modern sequential models, introducing two stable architectures: Flow-G, which incorporates a gated velocity, and Flow-T, which utilizes a decoded velocity. We then develop a practical SAC-based algorithm, enabled by a noise-augmented rollout, that facilitates direct end-to-end training of these policies. Our approach supports both from-scratch and offline-to-online learning and achieves state-of-the-art performance on continuous control and robotic manipulation benchmarks, eliminating the need for common workarounds like policy distillation or surrogate objectives.
♻ ☆ Where Did I Leave My Glasses? Open-Vocabulary Semantic Exploration in Real-World Semi-Static Environments
Benjamin Bogenberger, Oliver Harrison, Orrin Dahanaggamaarachchi, Lukas Brunke, Jingxing Qian, Siqi Zhou, Angela P. Schoellig
Robots deployed in real-world environments, such as homes, must not only navigate safely but also understand their surroundings and adapt to changes in the environment. To perform tasks efficiently, they must build and maintain a semantic map that accurately reflects the current state of the environment. Existing research on semantic exploration largely focuses on static scenes without persistent object-level instance tracking. In this work, we propose an open-vocabulary, semantic exploration system for semi-static environments. Our system maintains a consistent map by building a probabilistic model of object instance stationarity, systematically tracking semi-static changes, and actively exploring areas that have not been visited for an extended period. In addition to active map maintenance, our approach leverages the map's semantic richness with large language model (LLM)-based reasoning for open-vocabulary object-goal navigation. This enables the robot to search more efficiently by prioritizing contextually relevant areas. We compare our approach against state-of-the-art baselines using publicly available object navigation and mapping datasets, and we further demonstrate real-world transferability in three real-world environments. Our approach outperforms the compared baselines in both success rate and search efficiency for object-navigation tasks and can more reliably handle changes in mapping semi-static environments. In real-world experiments, our system detects 95% of map changes on average, improving efficiency by more than 29% as compared to random and patrol strategies.
♻ ☆ IKDiffuser: a Diffusion-based Generative Inverse Kinematics Solver for Kinematic Trees
Solving Inverse Kinematics (IK) for arbitrary kinematic trees presents significant challenges due to their high-dimensionality, redundancy, and complex inter-branch constraints. Conventional optimization-based solvers can be sensitive to initialization and suffer from local minima or conflicting gradients. At the same time, existing learning-based approaches are often tied to a predefined number of end-effectors and a fixed training objective, limiting their reusability across various robot morphologies and task requirements. To address these limitations, we introduce IKDiffuser, a scalable IK solver built upon conditional diffusion-based generative models, which learns the distribution of the configuration space conditioned on end-effector poses. We propose a structure-agnostic formulation that represents end-effector poses as a sequence of tokens, leading to a unified framework that handles varying numbers of end-effectors while learning the implicit kinematic structures entirely from data. Beyond standard IK generation, IKDiffuser handles partially specified goals via a masked marginalization mechanism that conditions only on a subset of end-effector constraints. Furthermore, it supports adding task objectives at inference through objective-guided sampling, enabling capabilities such as warm-start initialization and manipulability maximization without retraining. Extensive evaluations across seven diverse robotic platforms demonstrate that IKDiffuser significantly outperforms state-of-the-art baselines in accuracy, solution diversity, and collision avoidance. Moreover, when used to initialize optimization-based solvers, IKDiffuser significantly boosts success rates on challenging redundant systems with high Degrees of Freedom (DoF), such as the 29-DoF Unitree G1 humanoid, from 21.01% to 96.96% while reducing computation time to the millisecond range.
comment: under review
♻ ☆ Tackling the Kidnapped Robot Problem via Sparse Feasible Hypothesis Sampling and Reliable Batched Multi-Stage Inference
This paper addresses the Kidnapped Robot Problem (KRP), a core localization challenge of relocalizing a robot in a known map without prior pose estimate when localization loss or at SLAM initialization. For this purpose, a passive 2-D global relocalization framework is proposed. It estimates the global pose efficiently and reliably from a single LiDAR scan and an occupancy grid map while the robot remains stationary, thereby enhancing the long-term autonomy of mobile robots. The proposed framework casts global relocalization as a non-convex problem and solves it via the multi-hypothesis scheme with batched multi-stage inference and early termination, balancing completeness and efficiency. The Rapidly-exploring Random Tree (RRT), under traversability constraints, asymptotically covers the reachable space to generate sparse, uniformly distributed feasible positional hypotheses, fundamentally reducing the sampling space. The hypotheses are preliminarily ordered by the proposed Scan Mean Absolute Difference (SMAD), a coarse beam-error level metric that facilitates the early termination by prioritizing high-likelihood candidates. The SMAD computation is optimized for limited scan measurements. The Translation-Affinity Scan-to-Map Alignment Metric (TAM) is proposed for reliable orientation selection at hypothesized positions and accurate final global pose evaluation to mitigate degradation in conventional likelihood-field metrics under translational uncertainty induced by sparse hypotheses, as well as non-panoramic LiDAR scan and environmental changes. Real-world experiments on a resource-constrained mobile robot with non-panoramic LiDAR scans show that the proposed framework achieves competitive performance in success rate, robustness under measurement uncertainty, and computational efficiency.
comment: 10 pages, 8 figures. This work has been submitted to the IEEE for possible publication
♻ ☆ Autonomous Robotic Bone Micro-Milling System with Automatic Calibration and 3D Surface Fitting
Automating bone micro-milling using a robotic system presents challenges due to the uncertainties in both the external and internal features of bone tissue. For example, during mouse cranial window creation, a circular path with a radius of 2 to 4 mm needs to be milled on the mouse skull using a microdrill. The uneven surface and non-uniform thickness of the mouse skull make it difficult to fully automate this process, requiring the system to possess advanced perceptual and adaptive capabilities. In this study, we address this challenge by integrating a Microscopic Stereo Camera System (MSCS) into the robotic bone micro-milling system and proposing a novel online pre-measurement pipeline for the target surface. Starting from uncalibrated cameras, the pipeline enables automatic calibration and 3D surface fitting through a convolutional neural network (CNN)-based keypoint detection. Combined with the existing feedback-based system, we develop the world's first autonomous robotic bone micro-milling system capable of rapidly, in real-time perceiving and adapting to surface unevenness and non-uniform thickness, thereby enabling an end-to-end autonomous cranial window creation workflow without human assistance. Validation experiments on euthanized mice demonstrate that the improved system achieves a success rate of 85.7 % and an average milling time of 2.1 minutes, showing not only significant performance improvements over the previous system but also exceptional accuracy, speed, and stability compared to human operators.
comment: 8 pages, 8 figures, accepted by RA-L. Please refer to the DOI to access the accepted version
♻ ☆ Large Multimodal Models for Embodied Intelligent Driving: The Next Frontier in Self-Driving?
The advent of Large Multimodal Models (LMMs) offers a promising technology to tackle the limitations of modular design in autonomous driving, which often falters in open-world scenarios requiring sustained environmental understanding and logical reasoning. Besides, embodied artificial intelligence facilitates policy optimization through closed-loop interactions to achieve the continuous learning capability, thereby advancing autonomous driving toward embodied intelligent (El) driving. However, such capability will be constrained by relying solely on LMMs to enhance EI driving without joint decision-making. This article introduces a novel semantics and policy dual-driven hybrid decision framework to tackle this challenge, ensuring continuous learning and joint decision. The framework merges LMMs for semantic understanding and cognitive representation, and deep reinforcement learning (DRL) for real-time policy optimization. We starts by introducing the foundational principles of EI driving and LMMs. Moreover, we examine the emerging opportunities this framework enables, encompassing potential benefits and representative use cases. A case study is conducted experimentally to validate the performance superiority of our framework in completing lane-change planning task. Finally, several future research directions to empower EI driving are identified to guide subsequent work.
♻ ☆ Do What You Say: Steering Vision-Language-Action Models via Runtime Reasoning-Action Alignment Verification
Reasoning Vision Language Action (VLA) models improve robotic instruction-following by generating step-by-step textual plans before low-level actions, an approach inspired by Chain-of-Thought (CoT) reasoning in language models. Yet even with a correct textual plan, the generated actions can still miss the intended outcomes in the plan, especially in out-of-distribution (OOD) scenarios. We formalize this phenomenon as a lack of embodied CoT faithfulness, and introduce a training-free, runtime policy steering method for reasoning-action alignment. Given a reasoning VLA's intermediate textual plan, our framework samples multiple candidate action sequences from the same model, predicts their outcomes via simulation, and uses a pre-trained Vision-Language Model (VLM) to select the sequence whose outcome best aligns with the VLA's own textual plan. Only executing action sequences that align with the textual reasoning turns our base VLA's natural action diversity from a source of error into a strength, boosting robustness to semantic and visual OOD perturbations and enabling novel behavior composition without costly re-training. We also contribute a reasoning-annotated extension of LIBERO-100, environment variations tailored for OOD evaluation, and demonstrate up to 15% performance gain over prior work on behavior composition tasks and scales with compute and data diversity. Project Website at: https://yilin-wu98.github.io/steering-reasoning-vla/
♻ ☆ Virtual-force Based Visual Servo for Multiple Peg-in-Hole Assembly with Tightly Coupled Multi-Manipulator
Multiple Peg-in-Hole (MPiH) assembly is one of the fundamental tasks in robotic assembly. In the MPiH tasks for large-size parts, it is challenging for a single manipulator to simultaneously align multiple distant pegs and holes, necessitating tightly coupled multi-manipulator systems. For such MPiH tasks using tightly coupled multiple manipulators, we propose a collaborative visual servo control framework that uses only the monocular in-hand cameras of each manipulator to reduce positioning errors. Initially, we train a state classification neural network and a positioning neural network. The former divides the states of the peg and hole in the image into three categories: obscured, separated, and overlapped, while the latter determines the position of the peg and hole in the image. Based on these findings, we propose a method to integrate the visual features of multiple manipulators using virtual forces, which can naturally combine with the cooperative controller of the multi-manipulator system. To generalize our approach to holes of different appearances, we varied the appearance of the holes during the dataset generation process. The results confirm that by considering the appearance of the holes, classification accuracy and positioning precision can be improved. Finally, the results show that our method achieves 100\% success rate in dual-manipulator dual peg-in-hole tasks with a clearance of 0.2 mm, while robust to camera calibration errors.
comment: 8 pages, 11 figures, this paper has been published by IEEE Robotics and Automation Letters
♻ ☆ AURASeg: Attention Guided Upsampling with Residual Boundary-Assistive Refinement for Drivable-Area Segmentation
Free space ground segmentation is essential to navigate autonomous robots, recognize drivable zones, and traverse efficiently. Fine-grained features remain challenging for existing segmentation models, particularly for robots in indoor and structured environments. These difficulties arise from ineffective multi-scale processing, suboptimal boundary refinement, and limited feature representation. To address this, we propose Attention-Guided Upsampling with Residual Boundary-Assistive Refinement (AURASeg), a ground-plane semantic segmentation framework designed to improve border precision while preserving strong region accuracy. Built on a ResNet-50 backbone, AURASeg introduces (i) a Residual Border Refinement Module (RBRM) that enhances edge delineation through boundary-assistive feature refinement, and (ii) Attention Progressive Upsampling Decoder (APUD) blocks that progressively fuse multi-level features during decoding. Additionally, we integrate a (iii) lightweight ASPPLite module to capture multi-scale context with minimal overhead. Extensive experiments on CARL-D, the Ground Mobile Robot Perception (GMRP) dataset, and a custom Gazebo indoor dataset show that AURASeg consistently outperforms strong baselines, with notable gains in boundary metrics. Finally, we demonstrate real-time deployment on a Kobuki TurtleBot, validating practical usability. The code is available at https://github.com/Narendhiranv04/AURASeg
comment: 6 pages, 4 figures, 4 tables