Stereo visual odometry system design on humanoid robot Nao

摘要

Navigation is a fundamental problem that relates to localization and positioning for a humanoid robot Nao. Odometry is the one of many techniques that able to solve it. Nao humanoid robot actually doesn't have an odometry module by itself, and odometry sensors that currently available on it and can be used is only IMU module. An odometry sensor that quite interesting in recent times, usually for fusioning together with IMU, easier to mount and not so hard to obtain is a visual sensor camera. In this paper we will show how to implement and place stereo cameras in a Nao robot that has no mounting points for external cameras. By using stereo instead of monocular camera approach, we could obtain odomentry measurement scale in actual size. The software used in this paper is OpenCV that can handle all process regarding from image processing to pose estimation. Stereo visual odometry system that we have designed and implemented on a humaoid robot Nao now can be used to measure simple movement and robot rotation. By extending this research achievement, an odometry system now can be designed and implements on a bigger infrastructure, such as Lumen Robot Friend Platform. This also unlock research availability to do 3D scene reconstruction with a Nao robot and all another researches that based on localization.