New robotics and physical-AI papers, with what each one means for the industry.
We focus on legible robot motion generation in social navigation settings. Legibility in human-robot interaction (HRI) is often described as the property of robot motion that enables an observer to confidently infer the robot's intent.
Voluntary movements have long been hypothesised to be comprised of discrete primitives called submovements, as a descriptive model of human motor behaviour. However, existing methods scale poorly, and no principled method exists to determine whether a decomposition is informative.
Egocentric human data provides a scalable source of experience for robot learning, but varies substantially in human-robot alignment, behavioral coverage, and available supervision. Existing work shows favorable scaling with increasing human data, but it remains unclear which data properties drive downstream robot gains and how to use such data throughout the training pipeline.
Continuum robots are well suited for gentle manipulation because of their inherent compliance and ability to adapt to complex environments. However, their continuously deformable structure makes accurate configuration estimation challenging, particularly when external vision systems are unavailable or obstructed.
Video generation models (VGMs) offer strong spatiotemporal priors for embodied observation--action modeling. However, joint-space action vectors lack explicit image-space structure and vary in dimensionality and semantics across embodiments, making it challenging to directly leverage the rich spatiotemporal priors of VGMs.
This preprint introduces a topology optimization method for designing multi-directional soft pneumatic actuators using Darcy-law pressure modeling and a robust three-field formulation. Optimized chambers are assembled into a simulated three-arm gripper; Abaqus tests at 0.15 MPa show single-arm tip displacements up to about 15.8 mm, with bending directions tracking the vector sum of activated chamber pressures. If experimentally validated, it could replace heuristic soft actuator design with systematic multi-axis bending for soft grippers.
Marine litter is a growing threat to the underwater ecosystem, driving demand for autonomous survey methods that can locate debris efficiently over large areas. Existing survey methods typically follow predefined paths or operate with a single sensing modality, typically a camera (with image quality suffering in poor-visibility conditions) or sonar (usually noisy and low-resolution).
This preprint introduces Magic-W0, a robot manipulation foundation model that couples continuous action generation with explicit latent predictions of current 3D geometry, action-induced 3D motion, and future semantics. On RoboDojo-Sim it posts the highest average score among world-action models at 27.10, and on five real-robot tasks it reaches 94.6% average success after fine-tuning, beating π0.5 at 91.8%. The result matters because it shifts policies from blind imitation toward predicting how candidate actions will change the physical world, with gains in generalization and long-horizon manipulation.
This preprint introduces GroundingPI, a 4B grounding foundation model that predicts object points and boxes as quantized token coordinates rather than using a general-purpose vision-language backbone. Across 34 grounding benchmarks it averages 73.68%, ahead of a larger GPT-6 Astra baseline at 71.54%, and as a visual backbone it improves downstream manipulation (RoboTwin 2.0, RoboCasa-GR1) and autonomous driving (nuScenes L2 0.296 m). The result matters because it argues grounding is a distinct perceptual layer that can make embodied foundation models more precise.
This preprint introduces a framework for scene-scale aerial manipulation on an articulated uncrewed aerial manipulator, combining synthetic whole-body VLA training, measured-progress-aligned trajectory realization, and scene-graph-guided mission composition. In simulation, local skills achieve 39/60 successes under oracle handoff, and full multi-site missions achieve 21/50 (42.0%), while a monolithic whole-task VLA baseline achieves 0/50. Physical trials validate representative tasks without platform demonstrations, reducing risk and data cost for aerial manipulation.
We present PrefPI (Preference-Guided Policy Iteration), an iterative framework for steering pretrained generative robot policies using only relative preferences over self-generated trajectories. Unlike prior preference-learning methods that primarily sharpen modes already represented by the policy, we study steering beyond the initial effective support, where desired behaviors are rarely or never observed under the initial policy.
Vision-Language Navigation (VLN) requires embodied agents to generate actions based on instructions and observations. General-purpose multimodal agents offer a promising basis for this task, but selecting plausible local actions does not ensure that execution remains consistent with the intended route, particularly in long-horizon tasks.
This preprint introduces MVP-SLAM, an online visual-inertial SLAM system for construction sites that uses an as-planned floor plan to correct drift using only two back-to-back fisheye cameras and an IMU. It ranked 2nd of 22 teams in the Localization task of the Hilti–Trimble SLAM Challenge 2026 with 0.29 m mean RMSE, and 5th of 62 teams in SLAM with 0.24 m, the best among systems that integrate floor plans online. The work matters because it enables drift-bounded, plan-aligned localization during active site traversal without depth sensors.
This preprint presents an 8×8 array of 64 three-degree-of-freedom delta robots coupled by a stretchable membrane, creating a continuous surface that can manipulate objects both larger and smaller than the 43.3 mm actuator spacing. On real hardware, a learned policy transferred from simulation and succeeded on 76% of 50 trials across objects from 30 to 90 mm, while local hand-designed primitives routed 15 mm cubes at 7.6 mm/s. The work matters because it removes the centre-to-centre spacing lower bound on object size for distributed manipulation.
Cable-suspended aerial manipulation offers a lightweight architecture for cooperative transportation and physical interaction, yet the passive mechanical response perceived at the load remains insufficiently understood and systematically exploited. This work interprets aerial vehicles as movable compliant anchors and develops a gravity-aware quasi-static theory for predicting and shaping the passive Cartesian stiffness of a suspended load.
This preprint introduces vision-based pose-aware state estimators for neighboring multirotor UAVs, integrating visual tilt measurements to infer thrust direction, including a new linear thrust-constrained Kalman filter. Across two real-world and one simulated dataset, pose-aware methods cut mean velocity and acceleration errors by 40% and 57% versus position-only filters, remove a ~300 ms acceleration delay, and enable simulated follower tracking of >2 g lateral maneuvers where position-only estimation fails. This matters for collision avoidance and coordination in agile multi-UAV operations.
Autonomous exploration demands that robots continuously evaluate candidate viewpoints based on their expected information gain and execution cost. Sampling-based planners estimate this gain by volumetric raycasting and, due to its computational cost, evaluate candidates under an assumption of mutual independence, ignoring the overlap between viewpoints along the same path.
This preprint extends policy-constrained TD-MPC into PL-MPC by changing critic training targets, MPPI terminal values, and actor distillation without altering the world model or planner. On HumanoidBench, it lifts total average return from 98±18 to 387±255 on balance-hard and from 199±13 to 466±200 on hurdle; in zero-shot sim-to-real wrench–nut alignment on a KUKA IIWA14 it achieves 74.2% vs 61.3% success on the training object size. The result matters because it shows targeted interactions in the planning–learning loop can improve hard robot control tasks beyond the usual planner–policy alignment.
Neural trajectory predictors can reach low prediction error while violating dynamics, actuator limits, or state constraints, especially when controls are unobserved and dynamics are partially specified. We introduce the Markovian Dynamics Enforcer (MaDE), a time-invariant post-hoc operator mapping state-transition proposals onto a learned feasible dynamics manifold, trained on feasible states without ground-truth controls.
This preprint presents a quadruped traversability system that predicts five foot-contact outcomes from pre-contact camera images using a frozen DINOv3 backbone and an evidential regressor, then continually updates the model through replay with a historical validation gate. On a sequential real-robot stream across three unseen surface types, the gate cut anchor negative log-likelihood degradation by 23.1% versus replay without the gate while keeping new-terrain adaptation nearly equal. It matters for legged robots that must operate safely on unfamiliar ground without forgetting earlier experience.
Papers come from arXiv robotics feeds. Where our summary is not written yet, you see the opening lines of the abstract.