Indoor low-speed self-driving vehicle
Developed an edge-AI indoor mobile robot for hotel corridor navigation, combining visual perception, OCR-based door recognition, ultrasonic safety sensing, and task-state machines on the RDK S100 platform.
Computer vision programming lead · Dan Zai Pai DUI (self-driving vehicle) · Software/Electrical · 2026-05-01 – 2026-07-10
Context
Indoor service robots need to navigate hotel-like corridors, avoid dynamic obstacles, and identify target rooms using low-cost onboard sensors.
Approach
Built an RDK S100-based mobile robot integrating YOLO object detection, SORT tracking, OCR door-number recognition, ultrasonic safety sensing, and task-state machines for obstacle avoidance, path selection, and target-room stopping.
Outcome
Achieved stable real-time perception at ~15 FPS with YOLO inference around 11 ms on BPU, demonstrated three autonomous scenarios including visual obstacle avoidance, ultrasonic-based path selection, and OCR-based target door selection.
