Real-Time Computer Vision-Based 6DOF Pose Estimation and Target Tracking for a 6-Axis Robotic Arm.

The latest advancement in my neural network-based algorithm for 6DOF object pose estimation enables precise prediction of a stud's 6DOF pose, achieving millimeter-level accuracy without relying on a CAD model. This innovation is a crucial component of our intelligent construction automation robot system. The computer vision node transmits the target's 6DOF pose and reconstructed 3D geometry to ROS/ROS 2, allowing the motion planning algorithm to control the robot arm for real-time target tracking. Even when running on a laptop with limited computational power, the system consistently achieves millimeter-level precision. The frame rate would improve significantly with deployment on a more powerful computer, which I currently lack. My next step is to implement this on a real robot for further testing.