I. Introduction
According to the different walking modes, self-driving pipeline robots can also be divided into wheeled type, crawler type, vehicle type, peristaltic type, foot type and spiral wheel type, etc [1]. Among them, wheeled pipeline robots have become the most common walking mode adopted by pipeline robots at this stage due to their advantages of stable movement, large traction, exquisite and simple structure, and easy implementation [2]. The robot walks in the pipeline, and the unevenness of the road surface will cause the robot to vibrate, thereby affecting the clarity of the image and interfering with the entire system; In the process of avoiding obstacles, sudden changes in speed and direction will also affect the stability of the entire system. Therefore, how to optimize the control of pipeline robots in complex environments is the focus of the current problem.