A Novel Mobile Robot Navigation Method Based on Hand-Drawn Paths

In the real world, a human tourist can quickly arrive at a target along the route planned by his/her instinctive cognition of prepared maps. While on the mobile robot, it is required to know every way-points of the static or dynamic planning paths to reach a goal point. The previous navigation modes are mostly from sensor location, speech interaction, grid maps and so on. They rely on sensing performance, the speech instruction of site personnel or complex algorithms for the path searching and tracking. For improving the performance of interaction and tele-operated path tracking, in this paper, a novel interactive navigation based on the recognition of hand-drawn paths is proposed for mobile robots. Pen, paper, and camera are employed as the interface between users and the mobile robot. Initially, the image of a hand drawn path in the paper map is captured by a camera and segmented from the background by the color detection model. It is hypothesized that the captured frame of hand-drawn paths is slant to the original orthogonal map. Therefore, this method should firstly project the smoothed skeleton of slant path into the coordinate space of orthogonal map to guide the robot to perform path tracking. Then the hand-drawn path is rectified and remapped into the registered reference image of the orthogonal map through feature matching and perspective transformation. The combination of SIFT and RANSAC algorithms are employed for improving the accuracy of the projection of slant paths in the orthogonal map. Sequentially, a limited number of way-points are picked out from the corrected path in an orthogonal 2D map space usable for 2D navigation. The oriented sequential motion vectors are estimated by linking all successive way-points and employed for the path tracking along the predefined route step by step. Eventually, random navigation paths are designed to validate the robustness, effectiveness and accuracy of the constructed interactive navigation system.