Reactive Collision-Aware Motion Planning for 7-DOF Manipulator Using NVIDIA cuRobo and Live Point Cloud Perception · Autonomous Mobile Robotics Laboratory, UT Austin

Developing perception-driven ROS 2 motion planning pipeline for 7-DOF Kinova mobile manipulator in Isaac Sim, integrating cuRobo with spherically abstracted 3D LiDAR point cloud for GPU-accelerated collision avoidance
Designing safety override combining depth camera and visual discriminator to predict workspace collisions 3 seconds in advance
Preparing deployment of autonomous system on cobot hardware, benchmarking performance against control loop frequency, trajectory generation latency, and collision avoidance accuracy KPIs