Papers

Low-cost inertial navigation systems (INS) are prone to sensor biases and measurement noise, which lead to rapid degradation of navigation accuracy during global positioning system (GPS) outages. To address this challenge and improve positioning continuity in GPS-denied environments, this paper proposes a brain-inspired GPS/INS fusion network (BGFN) based on spiking neural networks (SNNs). The BGFN architecture integrates a spiking Transformer with a spiking encoder to simultaneously extract spatial features from inertial measurement unit (IMU) signals and capture their temporal dynamics. By modeling the relationship between vehicle attitude, specific force, angular rate, and GPS-derived position increments, the network leverages both current and historical IMU data to estimate vehicle motion. The effectiveness of the proposed method is evaluated through real-world field tests and experiments on public datasets. Compared to conventional deep learning approaches, the results demonstrate that BGFN achieves higher accuracy and enhanced reliability in navigation performance, particularly under prolonged GPS outages.

This paper proposes a robust, high-precision positioning methodology to address localization failures arising from complex background interference in large-scale flight navigation and the computational inefficiency inherent in conventional sliding window matching techniques. The proposed methodology employs a three-tiered framework incorporating multi-layer corner screening and adaptive template matching. Firstly, dimensionality is reduced through illumination equalization and structural information extraction. A coarse-to-fine candidate selection strategy minimizes sliding window computational costs, enabling rapid estimation of the marker's position. Finally, adaptive templates are generated for candidate points, achieving subpixel precision through improved template matching with correlation coefficient extremum fitting. Experimental results demonstrate the method's effectiveness in extracting and localizing diagonal markers in complex, large-scale environments, making it ideal for field-of-view measurement in navigation tasks.

This paper presents a gripper capable of grasping and recognizing terrain shapes for mobile robots in extreme environments. Multi-limbed climbing robots with grippers are effective on rough terrains, such as cliffs and cave walls. However, such robots may fall over by misgrasping the surface or getting stuck owing to the loss of graspable points in unknown natural environments. To overcome these issues, we need a gripper capable of adaptive grasping to irregular terrains, not only for grasping but also for measuring the shape of the terrain surface accurately. We developed a gripper that can grasp both convex and concave terrains and simultaneously measure the terrain shape by introducing a pin-array structure. We demonstrated the mechanism of the gripper and evaluated its grasping and terrain recognition performance using a prototype. Moreover, the proposed pin-array design works well for 3D terrain mapping as well as adaptive grasping for irregular terrains.

We present an efficient incremental SLAM back-end that achieves the accuracy of full batch optimization while substantially reducing computational cost. The proposed approach combines two complementary ideas: information-guided gating (IGG) and selective partial optimization (SPO). IGG employs an information-theoretic criterion based on the log-determinant of the information matrix to quantify the contribution of new measurements, triggering global optimization only when a significant information gain is observed. This avoids unnecessary relinearization and factorization when incoming data provide little additional information. SPO executes multi-iteration Gauss-Newton (GN) updates but restricts each iteration to the subset of variables most affected by the new measurements, dynamically refining this active set until convergence. Together, these mechanisms retain all measurements to preserve global consistency while focusing computation on parts of the graph where it yields the greatest benefit. We provide theoretical analysis showing that the proposed approach maintains the convergence guarantees of full GN. Extensive experiments on benchmark SLAM datasets show that our approach consistently matches the estimation accuracy of batch solvers, while achieving significant computational savings compared to conventional incremental approaches. The results indicate that the proposed approach offers a principled balance between accuracy and efficiency, making it a robust and scalable solution for real-time operation in dynamic data-rich environments.

Large Action Models (LAMs) have shown immense potential in autonomous navigation by bridging high-level reasoning with low-level control. However, deploying these multi-billion parameter models on edge devices remains a significant challenge due to memory constraints and latency requirements. In this paper, we propose EdgeNav-QE, a novel framework that integrates Quantized Low-Rank Adaptation (QLoRA) with a dynamic early-exit (DEE) mechanism to optimize LAMs for real-time edge navigation. By quantizing the backbone to 4-bit precision and strategically placing early-exit branches, we enable the model to terminate inference early for simple navigation tasks while retaining full depth for complex decision-making. Experimental results on the Habitat-Sim environment with Matterport3D dataset using OpenVLA-7B backbone, demonstrate that EdgeNav-QE reduces inference latency by 82.7% and memory footprint by 66.7% compared to full-precision baselines, while maintaining 81.8% navigation success rate. Furthermore, it outperforms state-of-the-art static early-exit method by 17.9% in latency, demonstrating the superiority of content-aware adaptive computation for safety-critical applications.

Attitude and Heading Reference Systems (AHRSs) are broadly applied wherever reliable orientation and motion sensing is required. In this paper, we present an improved Cubature Kalman Filter (CKF) with lower computational cost while maintaining estimation accuracy, which is named "Kaisoku Cubature Kalman Filter (KCKF)". The computationally efficient equations of the KCKF are derived by simplifying those of the CKF, while preserving equivalent mathematical relations. The lightweight prediction equations in the KCKF are derived by expanding the summation terms in the CKF and simplifying the result. This paper shows that the KCKF requires fewer floating-point operations (FLOPs) than the CKF. The controlled experimental results show that the KCKF reduces the computation time by approximately 19% compared to the CKF on a high-performance computer, whereas the KCKF reduces the computation time by approximately 15% compared to the CKF on a low-cost single-board computer. In addition, the KCKF maintains the attitude estimation accuracy of the CKF.

Animal behavior reflects interactions between the nervous system, body, and environment. Therefore, biomechanics and environmental context must be considered to understand algorithms for behavioral control. Computational models that embed artificial neural controllers within body models in simulated environments are a powerful tool for this purpose. Here, we review advances in biorealistic neuromechanical models while also highlighting emerging opportunities ahead. We first show how these models enable inference of biophysical variables that are difficult to measure experimentally. Through systematic perturbations, one can generate new experimentally testable hypotheses using these models. We then examine how neuromechanical models facilitate the exchange among neuroscience, robotics, and machine learning, and showcase their applications in healthcare. We envision that coupling experimental studies with active probing of their neuromechanical surrogates will significantly accelerate progress in neuroscience.

The rapid deployment of drones poses significant challenges for airspace management, security, and surveillance. Current detection and classification technologies, including cameras, LiDAR, and conventional radar systems, often struggle to reliably identify and differentiate drones, especially those of similar models, under diverse environmental conditions and at extended ranges. Moreover, low radar cross sections and clutter further complicate accurate drone identification. To address these limitations, we propose a novel drone classification method based on artificial micro-Doppler signatures encoded by resonant electromagnetic stickers attached to drone blades. These tags generate distinctive, configuration-specific radar returns, enabling robust identification. We develop a tailored convolutional neural network (CNN) capable of processing raw radar signals, achieving high classification accuracy. Extensive experiments were conducted both in anechoic chambers with 43 tag configurations and outdoors under realistic flight trajectories and noise conditions. Dimensionality reduction techniques, including Principal Component Analysis (PCA) and Uniform Manifold Approximation and Projection (UMAP), provided insight into code separability and robustness. Our results demonstrate reliable drone classification performance at signal-to-noise ratios as low as 7 dB, indicating the feasibility of long-range detection with advanced surveillance radar systems. Preliminary range estimations indicate potential operational distances of several kilometers, suitable for critical applications such as airport airspace monitoring. The integration of electromagnetic tagging with machine learning enables scalable and efficient drone identification, paving the way for enhanced aerial traffic management and security in increasingly congested airspaces.

We introduce Fiducial Exoskeletons, an image-based reformulation of 3D robot state estimation that replaces cumbersome procedures and motor-centric pipelines with single-image inference. Traditional approaches - especially robot-camera extrinsic estimation - often rely on high-precision actuators and require time-consuming routines such as hand-eye calibration. In contrast, modern learning-based robot control is increasingly trained and deployed from RGB observations on lower-cost hardware. Our key insight is twofold. First, we cast robot state estimation as 6D pose estimation of each link from a single RGB image: the robot-camera base transform is obtained directly as the estimated base-link pose, and the joint state is recovered via a lightweight global optimization that enforces kinematic consistency with the observed link poses (optionally warm-started with encoder readings). Second, we make per-link 6D pose estimation robust and simple - even without learning - by introducing the fiducial exoskeleton: a lightweight 3D-printed mount with a fiducial marker on each link and known marker-link geometry. This design yields robust camera-robot extrinsics, per-link SE(3) poses, and joint-angle state from a single image, enabling robust state estimation even on unplugged robots. Demonstrated on a low-cost robot arm, fiducial exoskeletons substantially simplify setup while improving calibration, state accuracy, and downstream 3D control performance. We release code and printable hardware designs to enable further algorithm-hardware co-design.

We propose a deterministic and time-efficient contact-aware path planner for neurovascular navigation. The algorithm leverages information from pre- and intra-operative images of the vessels to navigate pre-bent passive tools, by intelligently predicting and exploiting interactions with the anatomy. A kinematic model is derived and employed by the sampling-based planner for tree expansion that utilizes simplified motion primitives. This approach enables fast computation of the feasible path, with negligible loss in accuracy, as demonstrated in diverse and representative anatomies of the vessels. In these anatomical demonstrators, the algorithm shows a 100% convergence rate within 22.8s in the worst case, with sub-millimeter tracking errors (less than 0.64 mm), and is found effective on anatomical phantoms representative of around 94% of patients.

Video world models have emerged as promising candidates for high-fidelity world models, offering the potential to synthesize high-quality videos capturing fine-grained interactions between agents and their environments conditioned on multi-modal user inputs. Their impressive capabilities address many of the long-standing challenges faced by physics-based simulators, driving broad adoption in many problem domains, e.g., robotics. For example, video models can generate photorealistic, physically consistent deformable-body simulation without making prohibitive simplifying assumptions, which is a major bottleneck in physics-based simulation. Moreover, video models can serve as foundation world models that capture the dynamics of the world in a fine-grained and expressive way. They thus overcome the limited expressiveness of language-only abstractions in describing intricate physical interactions. However, despite their potential, video models still generate physics-violating future predictions, often manifesting as hallucinations. In this survey, we provide a review of video models and their applications as embodied world models in robotics, including efficient data generation and policy learning, dynamics and rewards modeling in reinforcement learning, policy evaluation, and visual planning. Further, we highlight important challenges hindering the trustworthy integration of video models, such as poor instruction following, hallucinations like violations of physics, unsafe content generation, in addition to significant data and compute overhead. We present potential future directions to address these open research challenges to motivate research and ultimately facilitate broader applications, especially in safety-critical settings. We provide a curated bibliography at https://github.com/irom-princeton/awesome-robotics-video-world-model-papers .

Post-training algorithms based on deep reinforcement learning can push the limits of robotic models for specific objectives, such as generalizability, accuracy, and robustness. However, Intervention-requiring Failures (IR Failures) (e.g., a robot spilling water or breaking fragile glass) during real-world exploration happen inevitably, hindering the practical deployment of such a paradigm. To tackle this, we introduce Failure-Aware Offline-to-Online Reinforcement Learning (FARL), a new paradigm minimizing failures during real-world reinforcement learning. We create FailureBench, a benchmark that incorporates common failure scenarios requiring human intervention, and propose an algorithm that integrates a world-model-based safety critic and a recovery policy trained offline to prevent failures during online exploration. Extensive simulation and real-world experiments demonstrate the effectiveness of FARL in significantly reducing IR Failures while improving performance and generalization during online reinforcement learning post-training. FARL reduces IR Failures by 73.1% while elevating performance by 11.3% on average during real-world RL post-training. Videos and code are available at https://failure-aware-rl.github.io.

This paper presents a data-driven methodology for the control of static hydraulic impact hammers, also known as rock breakers, which are commonly used in the mining industry. The task addressed in this work is that of controlling the rock-breaker so its end-effector reaches arbitrary target poses, which is required in normal operation to place the hammer on top of rocks that need to be fractured. The proposed approach considers several constraints, such as unobserved state variables due to limited sensing and the strict requirement of using a discrete control interface at the joint level. First, the proposed methodology addresses the problem of system identification to obtain an approximate dynamic model of the hydraulic arm. This is done via supervised learning, using only teleoperation data. The learned dynamic model is then exploited to obtain a controller capable of reaching target end-effector poses. For policy synthesis, both reinforcement learning (RL) and model predictive control (MPC) algorithms are utilized and contrasted. As a case study, we consider the automation of a Bobcat E10 mini-excavator arm with a hydraulic impact hammer attached as end-effector. Using this machine, both the system identification and policy synthesis stages are studied in simulation and in the real world. The best RL-based policy consistently reaches target end-effector poses with position errors below 12 cm and pitch angle errors below 0.08 rad in the real world. Considering that the impact hammer has a 4 cm diameter chisel, this level of precision is sufficient for breaking rocks. Notably, this is accomplished by relying only on approximately 68 min of teleoperation data to train and 8 min to evaluate the dynamic model, and without performing any adjustments for a successful policy Sim2Real transfer. A demonstration of policy execution in the real world can be found in https://youtu.be/e-7tDhZ4ZgA.

Structural vibration testing plays a key role in aerospace engineering for evaluating dynamic behaviour, ensuring reliability and verifying structural integrity. These tests rely on accurate and robust data acquisition systems (DAQ) to capture high-quality acceleration data. However, commercial DAQs that provide the required performance and features are often expensive and complex, limiting their accessibility for small-scale research and experimental applications. This work presents the design and experimental validation of an affordable and in-house-developed acceleration DAQ, tested on a small fixed-wing UAV through several Taxi Vibration Test (TVT) runs and ambient vibration measurements. The proposed system integrates several OrangePi 3 LTS single-board computers with multiple LSM6DS3TR-C MEMS inertial measurement units operating simultaneously via an Inter-Integrated Circuit (I2C) communication interface, managed under a Python-based master/slave architecture. Data is acquired at a stable sampling rate of approximately 208 Hz and post-processed using Welch's method to estimate their Power Spectral Density (PSD). Results confirm the system ability to provide consistent multi-sensor acceleration data and repeatable PSD profiles under the same test conditions; thus, demonstrating its reliability. With a total hardware cost below 600 EUR (approximately 690 USD), the developed DAQ offers a compact, scalable and cost-effective alternative for aerospace vibration analysis and structural testing.

The teleoperation of robotic hands is limited by the high costs of depth cameras and sensor gloves, commonly used to estimate hand relative joint positions (XYZ). We present a novel, cost-effective approach using three webcams for triangulation-based tracking to approximate relative joint angles (theta) of human fingers. We also introduce a modified DexHand, a low-cost robotic hand from TheRobotStudio, to demonstrate THETA's real-time application. Data collection involved 40 distinct hand gestures using three 640x480p webcams arranged at 120-degree intervals, generating over 48,000 RGB images. Joint angles were manually determined by measuring midpoints of the MCP, PIP, and DIP finger joints. Captured RGB frames were processed using a DeepLabV3 segmentation model with a ResNet-50 backbone for multi-scale hand segmentation. The segmented images were then HSV-filtered and fed into THETA's architecture, consisting of a MobileNetV2-based CNN classifier optimized for hierarchical spatial feature extraction and a 9-channel input tensor encoding multi-perspective hand representations. The classification model maps segmented hand views into discrete joint angles, achieving 97.18% accuracy, 98.72% recall, F1 Score of 0.9274, and a precision of 0.8906. In real-time inference, THETA captures simultaneous frames, segments hand regions, filters them, and compiles a 9-channel tensor for classification. Joint-angle predictions are relayed via serial to an Arduino, enabling the DexHand to replicate hand movements. Future research will increase dataset diversity, integrate wrist tracking, and apply computer vision techniques such as OpenAI-Vision. THETA potentially ensures cost-effective, user-friendly teleoperation for medical, linguistic, and manufacturing applications.

This work develops a unified nonlinear estimation-guidance-control framework for cooperative simultaneous interception of a stationary target under a heterogeneous sensing topology, where sensing capabilities are non-uniform across interceptors. Specifically, only a subset of agents is instrumented with onboard seekers (informed/seeker-equipped agents), whereas the rest of them (seeker-less agents) acquire the information about the target indirectly via the informed agents and execute a distributed cooperative guidance for simultaneous target interception. To address the resulting partial observability, a predefined-time distributed observer is leveraged, guaranteeing convergence of the target state estimates for seeker-less agents through information exchange with seeker-equipped neighbors over a directed communication graph. Thereafter, an improved time-to-go estimate accounting for wide launch envelopes is utilized to design the distributed cooperative guidance commands. This estimate is coupled with a predefined-time consensus protocol, ensuring consensus in the agents' time-to-go values. The temporal upper bounds within which both observer error and time-to-go consensus error converge to zero can be prescribed as design parameters. Furthermore, the cooperative guidance commands are realized by means of an autopilot, wherein the interceptor is steered by canard actuation. The corresponding fin deflection commands are generated using a predefined-time convergent sliding mode control law. This enables the autopilot to precisely track the commanded lateral acceleration within a design-specified time, while maintaining non-singularity of the overall design. Theoretical guarantees are supported by numerical simulations across diverse engagement geometries, verifying the estimation accuracy, the cooperative interception performance, and the autopilot response using the proposed scheme.

This paper presents a method for carrying fair comparisons of the accuracy of pose estimation using fiducial markers. These comparisons rely on large sets of high-fidelity synthetic images enabling deep exploration of the 6 degrees of freedom. A low-discrepancy sampling of the space allows to check the correlations between each degree of freedom and the pose errors by plotting the 36 pairs of combinations. The images are rendered using a physically based ray tracing code that has been specifically developed to use the standard calibration coefficients of any camera directly. The software reproduces image distortions, defocus and diffraction blur. Furthermore, sub-pixel sampling is applied to sharp edges to enhance the fidelity of the rendered image. After introducing the rendering algorithm and its experimental validation, the paper proposes a method for evaluating the pose accuracy. This method is applied to well-known markers, revealing their strengths and weaknesses for pose estimation. The code is open source and available on GitHub.

Achieving robust humanoid locomotion in complex, unstructured environments requires transitioning from reactive proprioception to proactive perception. However, integrating exteroception remains a significant challenge: existing approaches often fail to achieve high-speed, stable locomotion across diverse, unstructured terrains. In this work, we present Hiking in the Wild, an extensible, end-to-end perceptive legged parkour framework designed for robust humanoid hiking. Our approach maps depth inputs and proprioception directly to joint actions, without relying on external localization and mapping. To ensure safety and training stability, we introduce two key mechanisms: a foothold safety mechanism combining extensible Terrain Edge Detection with Foot Volume Points to reduce catastrophic slippage on edges, and a Flat Patch Sampling strategy that mitigates reward hacking by generating feasible navigation targets. Extensive experiments on the Unitree G1 humanoid robot demonstrate that our policy enables robust traversal of complex terrains at speeds up to 2.5 m/s.

Current approaches to humanoid control generally fall into two paradigms: perceptive locomotion, which handles terrain well but is limited to pedal gaits, and general motion tracking, which reproduces complex skills but ignores environmental capabilities. This work unites these paradigms to achieve perceptive general motion control. We present a framework where exteroceptive sensing is integrated into whole-body motion tracking, permitting a humanoid to perform highly dynamic, non-locomotion tasks on uneven terrain. By training a single policy to perform multiple distinct motions across varied terrestrial features, we demonstrate the non-trivial benefit of integrating perception into the control loop. Our results show that this framework enables robust, highly dynamic multi-contact motions, such as vaulting and dive-rolling, on unstructured terrain, significantly expanding the robot's traversability beyond simple walking or running. https://project-instinct.github.io/deep-whole-body-parkour

Morphological computing, the use of the physical design of a robot to ease the realization of a given task has been proven to be a relevant concept in the context of swarm robotics. Here we demonstrate both experimentally and numerically, that the success of such a strategy may heavily rely on the type of policy adopted by the robots, as well as on the details of the physical design. To do so, we consider a swarm of robots, composed of Kilobots embedded in an exoskeleton, the design of which controls the propensity of the robots to align or anti-align with the direction of the external force they experience. We find experimentally that the contrast that was observed between the two morphologies in the success rate of a simple phototactic task, where the robots were programmed to stop when entering a light region, becomes dramatic, if the robots are not allowed to stop, and can only slow down. Building on a faithful physical model of the self-aligning dynamics of the robots, we perform numerical simulations and demonstrate on one hand that a precise tuning of the self-aligning strength around a sweet spot is required to achieve an efficient phototactic behavior, on the other hand that exploring a range of self-alignment strength allows for a rich expressivity of collective behaviors.

Electric prosthetic hands should be lightweight to decrease the burden on the user, shaped like human hands for cosmetic purposes, and designed with motors enclosed inside to protect them from damage and dirt. Additionally, in-hand manipulation is necessary to perform daily activities such as transitioning between different postures, particularly through rotational movements, such as reorienting a pen into a writing posture after picking it up from a desk. We previously developed PLEXUS hand (Precision-Lateral dEXteroUS manipulation hand), a lightweight (311 g) prosthetic hand driven by four motors. This prosthetic performed reorientation between precision and lateral grasps with various objects. However, its controller required predefined object widths and was limited to handling lightweight objects (of weight up to 34 g). This study addresses these limitations by employing motor current feedback. Combined with the hand's previously optimized single-axis thumb, this approach achieves more stable manipulation by estimating the object's width and adjusting the index finger position to maintain stable object holding during the reorientation. Experimental validation using primitive objects of various widths (5-30 mm) and shapes (cylinders and prisms) resulted in a 100% success rate with lightweight objects and maintained a high success rate (>=80) even with heavy aluminum prisms (of weight up to 289 g). By contrast, the performance without index finger coordination dropped to just 40% on the heaviest 289 g prism. The hand also successfully executed several daily tasks, including closing bottle caps and orienting a pen for writing.

Autonomous 3D scanning of open-world target structures via drones remains challenging despite broad applications. Existing paradigms rely on restrictive assumptions or effortful human priors, limiting practicality, efficiency, and adaptability. Recent foundation models (FMs) offer great potential to bridge this gap. This paper investigates a critical research problem: What system architecture can effectively integrate FM knowledge for this task? We answer it with FlyCo, a principled FM-empowered perception-prediction-planning loop enabling fully autonomous, prompt-driven 3D target scanning in diverse unknown open-world environments. FlyCo directly translates low-effort human prompts (text, visual annotations) into precise adaptive scanning flights via three coordinated stages: (1) perception fuses streaming sensor data with vision-language FMs for robust target grounding and tracking; (2) prediction distills FM knowledge and combines multi-modal cues to infer the partially observed target's complete geometry; (3) planning leverages predictive foresight to generate efficient and safe paths with comprehensive target coverage. Building on this, we further design key components to boost open-world target grounding efficiency and robustness, enhance prediction quality in terms of shape accuracy, zero-shot generalization, and temporal stability, and balance long-horizon flight efficiency with real-time computability and online collision avoidance. Extensive challenging real-world and simulation experiments show FlyCo delivers precise scene understanding, high efficiency, and real-time safety, outperforming existing paradigms with lower human effort and verifying the proposed architecture's practicality. Comprehensive ablations validate each component's contribution. FlyCo also serves as a flexible, extensible blueprint, readily leveraging future FM and robotics advances. Code will be released.

Autonomous nano-drones, powered by vision-based tiny machine learning (TinyML) models, are a novel technology gaining momentum thanks to their broad applicability and pushing scientific advancement on resource-limited embedded systems. Their small form factor, i.e., a few tens of grams, severely limits their onboard computational resources to sub-100mW microcontroller units (MCUs). The Bitcraze Crazyflie nano-drone is the de facto standard, offering a rich set of programmable MCUs for low-level control, multi-core processing, and radio transmission. However, roboticists very often underutilize these onboard precious resources due to the absence of a simple yet efficient software layer capable of time-optimal pipelining of multi-buffer image acquisition, multi-core computation, intra-MCUs data exchange, and Wi-Fi streaming, leading to sub-optimal control performances. Our NanoCockpit framework aims to fill this gap, increasing the throughput and minimizing the system's latency, while simplifying the developer experience through coroutine-based multi-tasking. In-field experiments on three real-world TinyML nanorobotics applications show our framework achieves ideal end-to-end latency, i.e. zero overhead due to serialized tasks, delivering quantifiable improvements in closed-loop control performance (-30% mean position error, mission success rate increased from 40% to 100%).

Reliable humanoid-robot interaction (HRI) in household environments is constrained by two fundamental requirements, namely robustness to unconstrained user positions and preservation of user privacy. Millimeter-wave (mmWave) sensing inherently supports privacy-preserving interaction, making it a promising modality for room-scale HRI. However, existing mmWave-based interaction-sensing systems exhibit poor spatial generalization at unseen distances or viewpoints. To address this challenge, we introduce WaveMan, a spatially adaptive room-scale perception system that restores reliable human interaction sensing across arbitrary user positions. WaveMan integrates viewpoint alignment and spectrogram enhancement for spatial consistency, with dual-channel attention for robust feature extraction. Experiments across five participants show that, under fixed-position evaluation, WaveMan achieves the same cross-position accuracy as the baseline with five times fewer training positions. In random free-position testing, accuracy increases from 33.00% to 94.33%, enabled by the proposed method. These results demonstrate the feasibility of reliable, privacy-preserving interaction for household humanoid robots across unconstrained user positions.

Autonomous flight of micro air vehicles (MAVs) in unknown, cluttered environments remains challenging for time-critical missions due to conservative maneuvering strategies. This article presents an integrated planning and control framework for high-speed, time-optimal autonomous flight of MAVs in cluttered environments. In each replanning cycle (100 Hz), a time-optimal trajectory under polynomial presentation is generated as a reference, with the time-allocation process accelerated by imitation learning. Subsequently, a time-optimal model predictive contouring control (MPCC) incorporates safe flight corridor (SFC) constraints at variable horizon steps to enable aggressive yet safe maneuvering, while fully exploiting the MAV's dynamics. We validate the proposed framework extensively on a custom-built LiDAR-based MAV platform. Simulation results demonstrate superior aggressiveness compared to the state of the art, while real-world experiments achieve a peak speed of 18 m/s in a cluttered environment and succeed in 10 consecutive trials from diverse start points. The video is available at the following link: https://youtu.be/vexXXhv99oQ.

When swimming at low Reynolds numbers, inertial effects are negligible and reciprocal movements cannot induce net motion. Instead, symmetry breaking is necessary to achieve net propulsion. Directed swimming can be supported by magnetic fields, which simultaneously provide a versatile means of remote actuation. Thus, we analyze the motion of a straight microswimmer composed of three magnetizable beads connected by two elastic links. The swimming mechanism is based on oriented external magnetic fields that oscillate in magnitude. Through induced reversible hysteretic collapse of the two segments of the swimmer, the two pairs of beads jump into contact and separate nonreciprocally. Due to higher-order hydrodynamic interactions, net displacement results after each cycle. Different microswimmers can be tuned to different driving amplitudes and frequencies, allowing for simultaneous independent control by just one external magnetic field. The swimmer geometry and magnetic field shape are optimized for maximum swimming speed using an evolutionary optimization strategy. Thanks to the simple working principle, an experimental realization of such a microrobot seems feasible and may open new approaches for microinvasive medical interventions such as targeted drug delivery.

Volcanic gas emissions are key precursors of eruptive activity. Yet, obtaining accurate near-surface measurements remains hazardous and logistically challenging, motivating the need for autonomous solutions. Limited mobility in rough volcanic terrain has prevented wheeled systems from performing reliable in situ gas measurements, reducing their usefulness as sensing platforms. We present a legged robotic system for autonomous volcanic gas analysis, utilizing the quadruped ANYmal, equipped with a quadrupole mass spectrometer system. Our modular autonomy stack integrates a mission planning interface, global planner, localization framework, and terrain-aware local navigation. We evaluated the system on Mount Etna across three autonomous missions in varied terrain, achieving successful gas-source detections with autonomy rates of 93-100%. In addition, we conducted a teleoperated mission in which the robot measured natural fumaroles, detecting sulfur dioxide and carbon dioxide. We discuss lessons learned from the gas-analysis and autonomy perspectives, emphasizing the need for adaptive sensing strategies, tighter integration of global and local planning, and improved hardware design.

6D object pose estimation plays a crucial role in scene understanding for applications such as robotics and augmented reality. To support the needs of ever-changing object sets in such context, modern zero-shot object pose estimators were developed to not require object-specific training but only rely on CAD models. Such models are hard to obtain once deployed, and a continuously changing and growing set of objects makes it harder to reliably identify the instance model of interest. To address this challenge, we introduce an Open-Set CAD Retrieval from a Language Prompt and a Single Image (OSCAR), a novel training-free method that retrieves a matching object model from an unlabeled 3D object database. During onboarding, OSCAR generates multi-view renderings of database models and annotates them with descriptive captions using an image captioning model. At inference, GroundedSAM detects the queried object in the input image, and multi-modal embeddings are computed for both the Region-of-Interest and the database captions. OSCAR employs a two-stage retrieval: text-based filtering using CLIP identifies candidate models, followed by image-based refinement using DINOv2 to select the most visually similar object. In our experiments we demonstrate that OSCAR outperforms all state-of-the-art methods on the cross-domain 3D model retrieval benchmark MI3DOR. Furthermore, we demonstrate OSCAR's direct applicability in automating object model sourcing for 6D object pose estimation. We propose using the most similar object model for pose estimation if the exact instance is not available and show that OSCAR achieves an average precision of 90.48% during object retrieval on the YCB-V object dataset. Moreover, we demonstrate that the most similar object model can be utilized for pose estimation using Megapose achieving better results than a reconstruction-based approach.

Autonomous mobile manipulation in unstructured warehouses requires a balance between efficient large-scale navigation and high-precision object interaction. Traditional end-to-end learning approaches often struggle to handle the conflicting demands of these distinct phases. Navigation relies on robust decision-making over large spaces, while manipulation needs high sensitivity to fine local details. Forcing a single network to learn these different objectives simultaneously often causes optimization interference, where improving one task degrades the other. To address these limitations, we propose a Heterogeneous Multi-Expert Reinforcement Learning (HMER) framework tailored for autonomous forklifts. HMER decomposes long-horizon tasks into specialized sub-policies controlled by a Semantic Task Planner. This structure separates macro-level navigation from micro-level manipulation, allowing each expert to focus on its specific action space without interference. The planner coordinates the sequential execution of these experts, bridging the gap between task planning and continuous control. Furthermore, to solve the problem of sparse exploration, we introduce a Hybrid Imitation-Reinforcement Training Strategy. This method uses expert demonstrations to initialize the policy and Reinforcement Learning for fine-tuning. Experiments in Gazebo simulations show that HMER significantly outperforms sequential and end-to-end baselines. Our method achieves a task success rate of 94.2% (compared to 62.5% for baselines), reduces operation time by 21.4%, and maintains placement error within 1.5 cm, validating its efficacy for precise material handling.

The central challenge in reusing human actions across robots is the need to realize shared intent through incompatible kinematics, dynamics, and control interfaces. We present SHARE, which decomposes action sharing into two connected stages—reference generation and feedback execution—and uses native robot references as their interface to achieve shared structure, native realization. In the reference-generation stage, AdaMorph extracts human-motion intent with a shared temporal backbone and generates embodiment-compatible references through robot-conditioned decoding. In the feedback-execution stage, AdaTrack combines these references with proprioception, action history, and morphology, then uses variable-length dynamic representations, morphology-conditioned attention, expert distillation, and closed-loop reinforcement learning to emit native actions from one shared Actor. On G1, HI, AD, and T1, AdaMorph attains seen/unseen root-speed Pearson correlation coefficients (PCCs) of 0.951/0.894; AdaTrack reaches 94.00%/91.96% closed-loop Success Rate with 21.15% of the total parameters in four specialist Actors, while cross-embodiment pretraining improves seen/unseen adaptation by 9.67/13.62 points with only 1% target data.