This project requires training detection models, designing recognition algorithms, and controlling the robot dog for perception, obstacle avoidance, and line patrolling.

Fig. 1: Robot dog patrol animation.
  • Trained object detection models under inspection scenarios (e.g., fire point, substation, helmet) and optimized them using various strategies to get good results in the preliminary rounds of the competition.
  • Learned the use and operation of Unitree’s Go1 robot dog. Designed automatic route recognition and motion control algorithms while deploying the detection model so that the robot dog can simulate a human to perform inspections under various work scenarios.
Fig. 2: Trajectory recognition animation.

This absolutely was a wonderful experience, as I was fortunate enough to make it through the preliminary rounds of the competition and was able to get free access to a robot dog that you only see on TV, to develop and control it.