RoboticsR&D Collaboration
Vision-Guided Robotics Arm
Pick-and-place robotic arm with real-time computer vision.
Cycle time
1.8s
Detection mAP
92.7
Sim hours
12k
Hardware
6-DOF arm
Overview
A pick-and-place robotic arm guided by real-time computer vision. The control stack runs on ROS2, with YOLOv8 detection and a sim-to-real training pipeline that closes the reality gap through domain randomization.
Key features
- ROS2 control stack with deterministic scheduling
- YOLOv8 object detection with on-device inference
- Sim-to-real training in Gazebo digital twin
- Domain randomization for robust transfer
- Safety supervisor with hard-stop guarantees
