Acceleration Based Collision Detection with a Mobile Robot

Abstract

Data from acceleration sensors can be used for collision detection in autonomous mobile robots. A mobile robot was equipped with a three axis accelerometer. Three different methods for collision detection are evaluated: a simple threshold on the acceleration signal along the driving direction, a running median filter and a frequency based averaging filter. The three methods were first evaluated in an artificial setting and then tested while the robot was driving around in an office environment. The frequency based averaging filter performed best.