Specialty
A shared humanoid policy reported higher success than LHM-Humanoid on seen and unseen object-transport tasks, but success fell as the number of objects increased.
Scientific publication4 min read
A preprint describes a robot that watches for meaningful changes after human-object interactions and uses a vision-language model to plan assistance without a task request. The reported comparison is described as comparable, but the available extraction omits the numerical results and uncertainty estimates needed to judge how close the systems were.
Scientific publication4 min read
CrashSim generated safety-critical scenarios that tracked real-world crash behavior and crash-type statistics more closely than competing methods. In closed-loop tests, the resulting nuCrash set produced much wider performance differences among five autonomous-driving planners.
Scientific publication3 min read
In a laboratory test with one human collaborator and a seven-degree-of-freedom robot, an adaptive controller recorded a 95% average success rate, compared with 83% for a fixed-stiffness version and 69% for a deterministic baseline.
Scientific publication4 min read
A preprint proposes a 41-part vocabulary for construction activities and demonstrates four selected physical actions on robot hardware.
Scientific publication3 min read
LM-X recorded higher average task success than the tested controls while exposing three online signals for progress, next events and local action reliability.
Scientific publication3 min read
TacForcing updates unfinished action blocks with execution-time tactile feedback and reported the highest average success rates in both simulated and real-world evaluations. The study was limited to six tasks in each setting and two robot platforms.
Scientific publication4 min read
A test-time adaptation method for vision-language navigation recorded higher reported scores than the cited baseline on three unseen-environment benchmarks, while its real-world evidence remained qualitative.
Scientific publication4 min read
A multi-signal detector for robot image streams reported close-to-per-frame segmentation accuracy while sharply reducing how often the system adapted.
Scientific publication4 min read
A robotics preprint reports higher success and shorter solver times when a planner branches as other agents' intentions are revealed, while stressing that the evidence comes from simulations.
Scientific publication4 min read
A failure-bank method reported higher success and lower policy-induced cost than base policies in ten-task averages for π0.5 and π0, but its mean policy cost was higher than AEGIS.
Scientific publication4 min read
An evaluation in simulation and on a physical robot found that one shared interface could support three instruction modes, with gesture-based modes often performing better when wording, surfaces or objects changed.
Scientific publication4 min read
An arXiv preprint proposes a 3D-grounded test for robot video models. Cosmos led composite scores, but detailed checks found weaker object localization and trajectory accuracy, underscoring the gap between plausible footage and executable behavior.
Scientific publication5 min read
An arXiv preprint tests a multi-critic training method for robots that push and transport objects and open a dishwasher through contact. It reports 94.1% simulation success and 69.0% success in 58 trials on four unseen objects, while the dishwasher test is qualitative.
Scientific publication4 min read
In a randomized experiment, people working with a non-human-shaped robot disclosed more when its small talk contained fewer personal details, while teamwork and coordination ratings were also higher.
Scientific publication4 min read
A four-level robotic bin-picking system cleared all 30 experimental bins, while the study found that bin clearance and individual grasp success told different stories.
Scientific publication4 min read
A simulated comparison found higher success and shorter routes for an LLM-guided UAV than for a conventional lawn-mower scan, but the evaluation did not test unknown environments.
Scientific publication4 min read
A preprint describes MaCoPlanner, a system that turns equipment manuals into structured knowledge, checks robot plans before action, and rejects unresolved candidates. It reported higher success on the hardest benchmark tasks than the listed baselines, but the physical evaluation was small and limited to a no-load simulator.
Scientific publication5 min read
A robotics preprint reports higher task success for a visual-track system in simulation and physical manipulation, along with better future-video prediction scores.
Scientific publication4 min read
A preprint audit of 12 physical-AI benchmarks found positive relationships across every pair, with especially strong agreement for two pairs of tests. The analysis covered 51 models and suggests that benchmark averages can repeat the same strengths, while a compact suite may preserve useful model separation.
Scientific publication5 min read
A two-specimen laboratory study found that a distributed measure of vegetation stiffness stayed stable across robot push heights, while a lumped stiffness varied.
Scientific publication5 min read
FlashVLA, a streaming method for vision-language-action policies, cut reported inference time while maintaining or improving task success across selected simulated benchmarks and three physical-arm tests.
Scientific publication5 min read
A year of field data from a robot shows how snow, seasonal scene changes and false loop closures can undermine autonomous navigation methods.
Scientific publication4 min read
A one-path benchmark reported the lowest observed latency for ROS2 Connect among the tested systems, alongside higher stability and improved scalability under concurrent load.
Scientific publication3 min read
A methods study reports that SOPO-CD was faster than stated comparison methods in tested placement benchmarks, while its design leaves open questions about search breadth and real-world testing.
Scientific publication3 min read
Aerial tracks from 20 intersections were used to train planners and motion predictors for a new city in offline tests, with real-world driving left untested.
Scientific publication4 min read
A laboratory study of one soft pneumatic actuator reported path tracking, user-guided motion, obstacle clearance and local stability checks for a structured learned controller.
Scientific publication4 min read
A preprint reports a GPU simulator that is extremely fast on a simple rigid-body task, while deformable and surgical tests expose limits in scaling, validation and what the results can show.
Scientific publication4 min read
A tendon-driven prototype responded to preset voice commands and repeated finger movements, but its open-loop grip was sensitive to object surfaces and thickness.
Scientific publication3 min read
A preprint reports that RAEM, a LiDAR-equipped quadruped robot system, completed multi-floor exploration in simulation and reached the fifth floor of a real stairwell. The work also reports lower GPU map-construction times, while real-world repetition counts were not provided.
Scientific publication4 min read
LAC is a whole-body humanoid controller designed to adjust responses to pushes and twists separately. The reported tests covered simulated stiffness sweeps, real-robot contact responses and VR-controlled tasks including object carrying and door opening.
Scientific publication4 min read
In simulation, the reported controller reached a target pose even though it was outside the initial obstacle-free box, while measured iteration times stayed below the sampling period.
Scientific publication3 min read
A preprint presents a heat-transfer model for industrial automated tape laying that incorporates finite radiation geometry and mixed convection. It matched a reported temperature series with a 4.02°C RMSE, but independent validation remains outstanding.
Composites Part A: Applied Science and Manufacturing (2026)3 min read
A constrained multibody model suggests that a welding umbilical can add non-negligible modeled joint torques despite its much lower mass. The result comes from a deterministic planar simulation with prescribed robot motion, while experimental validation remains future work.
20th International Symposium on Advances in Robot Kinematics, Jun 2026, Barcelone, Spain3 min read
A preprint reports higher aggregate scores for confidence-guided episode selection in an EVAC robot world-model benchmark, while the Semantics result remains uncertain.
Scientific publication4 min read
A retrieval-based robot-learning approach reported higher average success than comparison systems on held-out LIBERO and UR5e manipulation tasks, while the paper said absolute out-of-distribution success remained low.
Scientific publication5 min read
A simulated social-navigation system uses explicit memory to retain rare, high-cost failures and respond to unfamiliar social behavior, according to an arXiv preprint.
Scientific publication4 min read
A robot-navigation system that combines learned visual waypoints with geometry-aware local control reported higher success and path-efficiency scores than a matched baseline in simulated and office tests.
Scientific publication5 min read
The planner used explicit rules for grasping, placing, sliding and lifting and recorded higher success and faster planning than tested alternatives in a simulated kitchen-table benchmark.
Scientific publication3 min read
GaussianDream++ reported 98.6% average success on LIBERO, 87.8% on LIBERO-Plus Overall and 52.5% pooled success in a physical evaluation, compared with 29.2% for reproduced π0.5.
Scientific publication4 min read