Improving 2D Reactive Navigators with Kinect

Most successful mobile robots rely on 2D radial laser scanners for perceiving the environment. The use of these sensors for reactive navigation has a serious limitation: the robot can only detect obstacles in the plane scanned by the sensor, with the consequent risk of collision with objects out of this plane. The recent commercialization of RGB-D cameras, like Kinect, opens new possibilities in this respect. In this paper we address the matter of adding the 3D information provided by these cameras to a reactive navigator designed to work with radial laser scanners. We experimentally analyze the suitability of Kinect to detect small objects and propose a simple but effective method to combine readings from both type of sensors as well as to overcome some of the drawbacks that Kinect presents. Experiments with a real robot and a particular reactive algorithm have been conducted, proving a significant upgrade in performance.