Go2Fetch

This project presents an autonomous indoor navigation system for a Unitree Go2 quadruped robot equipped with LiDAR and an onboard camera. The robot fuses LiDAR measurements with onboard odometry to construct a real-time two-dimensional occupancy grid of its environment.

A graph-based path planner then computes a collision-free route through the generated map using Dijkstra search and cost maps. The resulting path is converted into high-level motion and yaw commands that guide the robot through the environment while continuously updating its navigation state.

Target objects are detected using a YOLO26 vision model. LiDAR points are projected into the calibrated camera frame to estimate each detected object's position relative to the robot, allowing the system to dynamically update its navigation goal and autonomously approach the target.

Together, these components integrate real-time mapping, perception, sensor fusion, path planning, and high-level control into a complete autonomous fetch pipeline for indoor environments.