High Efficient Recognition and Docking of Mobile Robot for Multiple Trapezoidal Shelves

Recognition and docking technology of mobile robot is becoming increasing important in the production and manufacture factory. It is necessary to identify the position of shelves efficiently and perform docking accurately. In this paper, we design a recognition method for trapezoidal shelf based on geometric information and a multiple model shelf screening method, which can autonomously identify and dock trapezoidal shelves quickly and accurately. During the recognition process, we combine odometry to eliminate abnormal recognition results and smooth the results. The experiment shows that the method proposed can identify the pose of the trapezoidal shelves efficiently, and the docking accuracy can achieve $\pm 1 \mathrm{~cm}$.