Depth map construction with stereo vision for humanoid robot navigation

Abstract

This article presents how stereoscopic vision is implemented in a humanoid robot with the purpose of selecting the most proper routine of movements to avoid obstacles encountered. The goals were to identify accurately the three-dimensional environment by using a depth map constructed with 2D images, acquired with a pair of cameras mounted on the robot, for their analysis using a Raspberry Pi 2 Model B to execute an algorithm of navigation. The experimental results demonstrate that the stereoscopic vision system is capable of measuring the relative position between objects and the humanoid robot with accuracy. With this information the humanoid robot would be able to avoid and pass through different obstacles, once the system recognizes a suitable solution for each situation. This article presents how stereoscopic vision is implemented in a humanoid robot with the purpose of selecting the most proper routine of movements to avoid obstacles encountered. The goals were to identify accurately the three-dimensional environment by using a depth map constructed with 2D images, acquired with a pair of cameras mounted on the robot, for their analysis using a Raspberry Pi 2 Model B to execute an algorithm of navigation. The experimental results demonstrate that the stereoscopic vision system is capable of measuring the relative position between objects and the humanoid robot with accuracy. With this information the humanoid robot would be able to avoid and pass through different obstacles, once the system recognizes a suitable solution for each situation.