A coding agent reconstructs tasks from a single calibration episode and generates realistic simulation data
Explore technology by industry and role
Newest additions first. Explore past and current notes by industry and role.
All notes
Newest additions first. Explore past and current notes by industry and role.
A novel guided data collection method that actively selects informative data to enhance policy performance.
A goal-conditioned reinforcement learning policy enables high-precision joint control without manual tuning.
Generalist robot policies can improve across diverse tasks simultaneously through reinforcement learning.
A pipeline generates vision-language-motion data across thousands of procedurally generated homes to train autonomous movement
A shared tokenized representation enables synchronized motion between humans and humanoid robots, generating reactive and anticipatory behaviors.
Unlike prior methods, RadarMP jointly models target detection and motion estimation, enabling consistent point cloud generation and 3D scene flow prediction
RoboDriveBench, the first benchmark specifically designed to evaluate robustness of VLM-based autonomous driving systems, is introduced.
Introduces the first device-centric benchmark evaluating closed-loop autonomous driving under realistic deployment perturbations
First systematic black-box study on vulnerability to realistic input corruptions in reasoning-enabled VLA models
Introduces an explicit unknown-space representation for unstructured mining environments
Risk-guided slice coverage directs audit budgets toward high-risk driving conditions.
A new method, ZAP, uses pre-trained vision-language models to learn robot control through simulation without real-world data.
Action reasoning is grounded in chain-of-dynamics, enhanced through world imagination and reinforcement learning.
Code agents generate realistic robotics simulations, enabling scalable and faithful physical modeling.