Autonomous Navigation for Hexapod Robots
A real-world perception-to-action stack integrating visual SLAM, path planning, and legged locomotion.
Tongji University
Role: Core Developer, Perception & Planning Lead · Advisor: Dr. Jia Xu
Period: May 2024 – September 2025
I deployed and calibrated a real-time visual SLAM system based on ORB-SLAM3 on a physical hexapod robot. The deployment required handling sensor noise, IMU drift, and unmodeled leg-ground contact dynamics that do not appear in simulation.
I co-developed a hierarchical navigation stack whose local planner connects high-level, SLAM-based global pose estimation with low-level legged locomotion control. This integration enabled closed-loop perception-to-action behavior on real hardware.
I also carried out end-to-end hardware-software integration and systematic debugging across the perception, planning, and control pipeline, supporting collision-free indoor navigation and reliable pose tracking.
Technologies: C++, Python, ORB-SLAM3, OpenCV, ROS, visual-inertial SLAM