Autonomous vision based docking of roombots

Abstract

This semester project was focusing on Roombot autonomous docking. The goal was to integrate a camera and a mini computer inside the Roombot to enable pose estimation and autonomous docking. First using a Raspberry Pi Zero was planned because of its small size and good enough performance but the work done on it proved to be useless because of hardware incompatibility between the camera (MU9PC-MH from Ximea require ARMv7) and the Pi Zero’s processor (ARMv6). Because of this I used a laptop for the rest of the project.

The next part of the project was to do extrinsic camera calibration to calculate relative position between predefined markers and the camera, with our configuration. A small application was developed using OpenCV libraries and 3 different marker type with 3 different marker detection processes. The estimated poses were then compared with a ground truth given by a Mocap system.