Robot Manipulation Multimodal Dataset
Object recognition, grasp poses, end-effector trajectories and outcome feedback.

Object recognition, grasp poses, end-effector trajectories and outcome feedback.
Localization, path planning, dynamic obstacle avoidance and map building for mobile robots.
Natural-language commands, gestures, environment feedback and task execution results.
Dual-arm collaboration, handover, grip control and industrial assembly tasks.
Fingertip tactile data, joint states, hand-eye coordination and small-object manipulation.
Real task-view videos, operation intent, action segments and multi-stage state labels.
Synthetic data generated by scene, object and workflow for hybrid training with real data.
Quality sampling, anomaly detection, desensitization and delivery consistency checks.
Vehicles, pedestrians, traffic signs and complex road environments.
Multilingual wake words, short commands, long speech and multi-speaker scenarios.
Task planning, behavior evaluation, preference ranking and model alignment training.
Work orders, nameplates, instrument panels, abnormal characters and complex lighting OCR.