Config demonstrates human-present training data for collaborative robot behavior
Evidence notes
- Config demonstrated training data collected in human-present environments, where people enter the scene, give direction or work alongside the robot.
- The company said models trained on this data learned to pause, wait and yield in shared spaces without those behaviors being explicitly programmed.
- Config said it planned to open-source the dataset in the near future.
Company context
Config builds robotics data infrastructure and foundation-model technology for bimanual manipulation. Its platform covers the workflow from task and gripper definition through human and teleoperation data acquisition, data validation, action estimation, model training and evaluation.