This paper presents a navigation method for an indoor robot to hit a predetermined target by combining two technologies consist of depth map mining and self-created 2D mapping. Experimental robots are used the tactic named Always Move Straight to the Destination (AMDS). First, the system extracts the ground plane from the depth map provided by the RGB-D camera to find the robot direction. If there is not any obstacle, it always tries to move straight to the destination. If the robot meets obstacles, it will switch to the improved avoidance mode to overcome them. Then, the robotic system returns to AMDS mode. The navigation system has established an optimal obstacle avoiding strategy with a success rate of 98.7% which is better than some recent comparison methods based on Artificial Neural Network (ANN) classifiers or method of combination of two algorithm of Dynamic Window Approach (DWA) and Anytime Repairing A* (ARA*). The quantity of navigation direction angles of the proposed method is more than the best compared method twenty times with navigation angle step of 1 degree. The result of this work is a robust proof for integrating the proposed navigation algorithm into low cost robots.
Keyword
depth map, ground plane, navigation, path direction
