Robotics & Embodied AI - Weekly Roundup

Bicycle robots can now do front-flips, thanks to a reinforcement learning method that bootstraps from dynamically infeasible reference motions.

Robots can now "see" hidden objects and understand articulation by learning from human egocentric video, even if they can't physically explore those areas themselves.

Train drone operators in realistic battlefield environments without ever leaving the simulator, thanks to Unreal Engine's built-in AI.

Forget hand-crafted rewards: MotionVL uses VLMs and LLMs to automatically generate task-aligned reward functions for humanoid robot RL, leading to more human-like and robust motion.

Robots get a 33% speed boost and become significantly more adaptable when you let LLMs handle the reasoning and RL handle the movements.

Train drone operators in realistic battlefield environments without ever leaving the simulator, thanks to Unreal Engine's built-in AI.

Forget hand-crafted rewards: MotionVL uses VLMs and LLMs to automatically generate task-aligned reward functions for humanoid robot RL, leading to more human-like and robust motion.

Robots get a 33% speed boost and become significantly more adaptable when you let LLMs handle the reasoning and RL handle the movements.

Pythonistas rejoice: aggregate programming, a powerful paradigm for distributed systems, finally gets a first-class, easy-to-use library in your favorite language.

Autonomous vehicles can drive more safely and reliably by grounding LLM reasoning in a "Commonsense World" that quantifies and leverages the trustworthiness of LLM outputs.

Achieve superhuman robot dexterity with 10x fewer demonstrations by decoupling intent and action through latent world modeling.

Automating scientific discovery is now more accessible: Owl-AuraID navigates proprietary GUIs to control diverse precision instruments, freeing researchers from tedious manual operation.

Achieve real-time, privacy-aware action detection on edge devices by intelligently fusing fast skeleton tracking with vision-language models, outperforming either approach alone.

Robots can now generalize to unseen objects and categories for manipulation tasks with only a few training examples, thanks to a novel retrieval-augmented affordance prediction framework.

Emulating human movement with 700 muscles reveals that many different control strategies can produce the same observed motion, challenging the assumption that kinematics uniquely define muscle activation.

Smart industrial systems, while promising increased efficiency, introduce unforeseen interoperability side-effects and heightened vulnerability to cyber threats across heterogeneous IIoT systems.

Robots can now learn to reproduce oil paintings with impressive accuracy through self-play and learned dynamics, even without human demonstrations or high-fidelity simulators.

Physical AI systems struggle not with visual recognition, but with understanding space, physics, and action – and PRISM, a new retail video dataset, dramatically closes this gap.

Assistive robots aren't just vulnerable to data breaches; they can be hacked to physically harm the very people they're supposed to protect.

Open-source SurgNavAR slashes the barrier to entry for AR surgical navigation research, offering a ready-to-use framework adaptable to diverse surgical applications.

Synthetic data, when carefully aligned with real-world characteristics, can boost hand-object interaction detection by over 11% even when real labeled data is scarce.

Vision-language models falter at the fine-grained temporal recognition crucial for surgical video understanding, while SurgRec excels.

Even state-of-the-art VLMs exhibit systematic failures in reasoning about the physical feasibility of actions in 3D environments, despite high semantic confidence.

Forget expensive labels: CoRe-DA leverages contrastive learning and self-training to achieve state-of-the-art surgical skill assessment across diverse surgical environments without requiring target domain annotations.

Surgeons can now pinpoint tumor margins with millimeter precision using augmented reality, potentially reducing positive margins in head and neck cancer resections.

Ditching depth map projections for camera-LiDAR calibration unlocks significant gains in accuracy and robustness, especially when starting from poor initial extrinsic estimates.

Quantifying and integrating map uncertainty—both positional and semantic—into trajectory prediction pipelines significantly boosts forecast accuracy, even when using existing baseline models.

LLMs can generate more accurate motion trajectories by clustering them into geometrically consistent families, even without retraining.

Achieve a 60% reduction in trajectory error for monocular SLAM by tightly integrating multi-task dense prediction with a compact perception-to-mapping interface.

Reconstructing dynamic 3D scenes from video just got a whole lot better: MotionScale achieves state-of-the-art fidelity and temporal stability by scaling Gaussian splatting to long, complex sequences.

Forget tedious optimization – LightHarmony3D generates realistic lighting and shadows for inserted 3D objects in a single pass, making scene augmentation feel truly real.

Turn 2D orthographic views into 3D models automatically using corner detection and geometric reconstruction.

Unlock adaptable human augmentation in everyday environments with reconfigurable robotic limbs, guided by quantitative analysis of workspace extension and human-robot collaboration.

A rotating haptic compass on your wrist dramatically improves robotic teleoperation by providing intuitive directional cues, outperforming traditional vibration-based feedback and even improving imitation learning.

You can halve the polygon count of dynamic 3D meshes in VR without users noticing, but existing quality metrics won't tell you that.

Passive iFIR filters learned from just three minutes of robot data can dramatically outperform optimized PID controllers in velocity tracking tasks, offering a fast and stable alternative for robot control.

Get provably safe and dynamically robust robot motions in human environments without the computational bottleneck of online optimization.

Unlock rapid UAV design iteration with MetaMorpher's modular, nonlinear flight dynamics model that accurately simulates diverse wing configurations and flight modes.

Semantic scene understanding can keep your robot from crashing when running LLMs on edge devices.

A long-reach robot arm can gently clean lunar solar panels, even with limited force feedback, opening the door to autonomous maintenance on the moon.

Guaranteeing safety in multi-agent systems with dynamic networks doesn't have to sacrifice performance: this plug-and-play protocol ensures recoverable safety even when agents join/leave or network topologies shift.

Offline RL can now tackle complex, unseen temporal logic tasks without retraining, by stitching together learned short-horizon behaviors into long-horizon plans.

UUVs can navigate communication blackouts with 91% more accuracy by distilling patterns from their past trajectories.

By optimizing PID gains with MPPI, this method achieves comparable performance to conventional MPPI with significantly fewer samples, offering a more sample-efficient approach to learning-based control.

Humanoids can now nimbly navigate real-world clutter and complex terrain using only raw depth data, ditching hand-engineered geometric representations.

Achieve state-of-the-art robotic manipulation with a model orders of magnitude smaller than VLAs by explicitly aligning kinematic and semantic transitions.

Forget brute-force coverage – this method learns from simulated expert guidance to prioritize semantically relevant areas, dramatically speeding up target search in unseen environments.

Legged robots can now navigate more accurately using only internal sensors, even with imperfect foot contact, thanks to a new probabilistic method that dynamically adapts to different contact scenarios.

Automating disassembly of complex, degraded appliances in recycling plants is now feasible, achieving high accuracy without pre-programmed coordinates.

SuperGrasp achieves robust single-view grasping by cleverly combining superquadric-based similarity matching with an end-to-end refinement network, outperforming existing methods in stability and generalization.

Real-time, uncertainty-aware signed distance functions are now possible without sacrificing accuracy, thanks to a novel kernel regression and GP regression hybrid.

Get kilohertz-level dexterous hand teleoperation *with* formal safety guarantees, thanks to a new convex optimization approach.

Policies trained with GenSplat maintain robust performance under severe spatial perturbations where baseline methods completely fail, thanks to its novel 3D Gaussian Splatting-based augmentation.

VLN agents can now "dream ahead" by learning action-conditioned visual dynamics in a latent space, leading to SOTA results and improved real-world navigation.

Ignoring control packet loss in drone communication can lead to trajectory divergence, but this integrated sensing-communication-control scheme achieves decimeter-level accuracy.

Current vibration-based alert systems often misestimate alert durations due to poor damping estimates, but this new information-theoretic method can accurately capture alert duration.

Achieve targeted motion adaptation in physics-based characters by learning a mask-invariant prior, enabling robust control even with missing observations or text-driven partial goals.

Actor-critic methods can achieve state-of-the-art sample complexity in linear MDPs *without* relying on computationally expensive implicit policies or strong assumptions about exploration.

Physics-informed neural networks can now accurately identify impact events on aerospace composites, even with noisy or incomplete data, opening the door to real-time structural health monitoring.

Overcome the curse of dimensionality in offline MARL by learning which agents' actions to replace, achieving state-of-the-art performance with dramatically reduced computation.

VLA models are brittle: even simple synonym substitutions in instructions cause a 22-52% performance drop in robotic manipulation tasks.

A simple DBSCAN model running on real-time bridge sensor data can outperform other ML models in detecting anomalies, suggesting a practical path to preventing catastrophic failures.

SONAR can "see" road damage and material even when cameras and LiDAR are blinded by rain or fog.

RL agents can learn to control complex fluid dynamics 40% faster by pretraining on Koopman-based surrogate models and iteratively refining them with policy-aware data.

Generating realistic, safety-critical maritime scenarios at scale is now possible by combining generative trajectory modeling with automated encounter pairing, moving beyond limited historical data or handcrafted templates.

Get 80% of your oracle feedback for free: ROVED leverages vision-language embeddings to drastically reduce the need for human preferences in reinforcement learning.

Agentic RL agents can learn faster and perform better by dynamically maintaining a skill bank that combines high-level task guidance with low-level step-by-step decision support.

Learning thermomechanical material properties just got easier: this new framework guarantees thermodynamic consistency without needing entropy data or enforcing complex convexity constraints.

MLLMs can now guide visual generative models to imagine what's hidden behind objects, significantly boosting amodal completion performance.

XR's potential for AI-driven assistance risks eroding human autonomy, but Self++ offers a design blueprint to ensure AI augments, rather than replaces, human judgment.

A new swarm-based optimization algorithm, inspired by dogfighting but built on kinematic equations, achieves state-of-the-art performance across diverse benchmark and real-world engineering problems.

Training data no longer needs to choose between realism and accuracy: SHOW3D delivers both for hand-object interaction.

A 40-point mIoU gap between supervised methods and zero-shot segmentation on Industrial3D reveals that foundation models are nowhere near ready for real-world industrial Scan-to-BIM workflows.

Current robot manipulation benchmarks fail to capture the messy reality of real-world deployment, so this work introduces a new benchmark, ManipArena, to close the sim2real gap.

Real-world 3D scene completion is now possible without synthetic training data, thanks to visibility-guided flow matching that handles incomplete scans.

Event cameras can now accurately measure high-speed 3D deformations of structures under extreme lighting, opening up new possibilities for monitoring the safety of critical infrastructure.

Skipping frames without objects boosts nano-drone object detection throughput by 24% with negligible accuracy loss.

Ghost points, often ignored in LiDAR processing, can be effectively identified and removed using full-waveform LiDAR data, leading to substantial improvements in downstream tasks like SLAM and object detection.

View transformation may be sabotaging your NeRF pre-training: directly learning continuous 3D representations with NeRP3D avoids conflicting priors and boosts performance on nuScenes.

Achieve 49% and 19% better Chamfer distance than state-of-the-art dynamic surface reconstruction methods on Hi4D and CMU Panoptic datasets, respectively, by enforcing temporal consistency in Gaussian Splatting.

Event cameras unlock 6D pose tracking of novel objects at 120+ FPS, even with rapid motion, by fusing sparse event streams with depth in a way that generalizes zero-shot from synthetic training.

Event cameras, fused with traditional frames using an energy-aware approach, can significantly boost the accuracy of autonomous vehicle steering prediction.

Achieve dramatically wider field of view for UAVs without adding sensors or complexity by simply spinning the entire drone.

A foldable soft robot achieves unprecedented agility with nine distinct locomotion modes, opening new possibilities for navigating the human body.

What looks like polite robot navigation from above can feel downright rude when you're the pedestrian dodging it.

Forget painstakingly tuning MPC controllers by hand: this method learns optimal humanoid locomotion policies by aligning MPC cost functions with high-fidelity RL data.

Updating motion planning roadmaps in dynamic environments just got an order of magnitude faster with a GPU-accelerated edge validation scheme.

Feel what the robot feels: a new glove lets human operators experience high-resolution tactile feedback during dexterous teleoperation, dramatically improving performance in contact-rich tasks.

Agricultural robots can now more accurately follow paths and avoid crop damage thanks to a new controller that explicitly models the implements they tow.

Tilting your drone's propellers isn't just for agility – it can be a game-changer for maintaining comms under jamming attacks, boosting link reliability by orders of magnitude.

More sensors, more problems: a simple active stereo camera setup beats out complex multi-sensor rigs for humanoid robot imitation learning when data is scarce.

Neurosurgeons gain a compact, sterilizable RCM joint with near-isotropic stiffness, minimizing unwanted motion during delicate procedures.

Bicycle robots can now do front-flips, thanks to a reinforcement learning method that bootstraps from dynamically infeasible reference motions.

Robots can now "see" hidden objects and understand articulation by learning from human egocentric video, even if they can't physically explore those areas themselves.

Finally, a single, open-source platform lets you train and test coordinated air and ground robots in photorealistic urban environments with synchronized physics and sensors.

Robots can now learn complex manipulation tasks from scratch using only video and language, bypassing the need for hand-engineered reward functions, demonstrations, or even task-specific tuning.

Synergy's architecture lets agents evolve through experience by proactively recalling rewarded trajectories, hinting at a new way to build agents that learn and adapt in open, collaborative environments.

Adversarial attacks can cripple robotic perception systems, demanding specialized defenses beyond standard image classification techniques.

Robot color choices are subtly shaped by racial and occupational stereotypes, even when users offer seemingly rational justifications.

Bipedal robots can now walk more stably on slippery surfaces thanks to a new control method that explicitly models and compensates for foot slippage.

Real-time 3D occupancy mapping for edge devices is now possible under a 6mW power budget thanks to Gleanmer, a novel SoC.

Freeing robots from pre-assigned tasks slashes completion times in multi-agent settings, with a new algorithm improving performance on almost 90% of tested scenarios.

Unlock real-time control of off-road vehicles on challenging terrain by representing complex terramechanics with linear Koopman operators learned from simulation data.

Unlabeled LiDAR data can now drive state-of-the-art traffic simulation, unlocking scalable realism without costly annotations.

Context-aware robots that "see something, say something" boost user trust by 82% simply by communicating more intelligently about hazards.