Recognize exhibits
Jetson Orin and OpenCV extract exhibit IDs from camera video.
← Back to projects
Perception & interaction · Systems
A quadruped museum guide
Connecting visual recognition, robot movement, and spoken narration in a perception–decision–action loop.
Technical illustration
Go2Navigation + narration
Camera→OpenCV→DDS
- Period
- Feb 2025 — Apr 2025
- Role
- Research contributor
- Institution & location
- Xi’an Jiaotong-Liverpool University · Suzhou
Advisors · Dr. Fanxin Wang, Dr. Qinglei Bu, Dr. Martijn ten Bhömer & Prof. Jie Sun
Background
In a museum, a guide robot must recognize exhibits and coordinate movement, posture, and narration. This Unitree Go2 project connects perception and task execution across computing platforms.
My contributions
- Streamed onboard camera video to Jetson Orin and used OpenCV for QR-code detection and exhibit identification.
- Programmed the RK3588 controller to receive exhibit IDs and coordinate robot motion, posture, and audio playback through DDS.
- Built a perception–decision–action loop to synchronize autonomous navigation and exhibit narration.
Technical approach
Coordinate tasks
RK3588 coordinates execution modules through DDS.
Move & narrate
Synchronize navigation, posture adjustment, and audio playback.
Outcomes & evaluation
Go2Quadruped platform
DDSCross-module task communication
- Integrated an intelligent museum guide system on a physical Go2 platform.
- Connected visual recognition outputs to robot movement and narration.
Resources
Next project