Gyrodometry: a new method for combining data from gyros and odometry in mobile robots

This paper presents a very simple, yet very effective method for combining measurements from a gyro with measurements from wheel encoders (odometry). Sensor-fusion of this kind has been done before, usually by means of a statistical model that describes the behavior of the gyro and the behavior of the odometry component. However, because these systems are based on models, they cannot anticipate the unpredictable and potentially "catastrophic" effect of larger bumps or objects occasionally encountered on the floor. By contrast, our method, called Gyrodometry, has been developed based on a careful study of the physical interaction between the ground and the vehicle. We present experimental evidence that non-systematic odometry error sources (such as bumps) impact the vehicle only during very short periods; typically a fraction of a second for each encounter. During these short instances the readings from the gyro and from odometry differ significantly, while in the absence of large non-systematic errors the readings are very similar. Gyrodometry makes use of this observation by using odometry data only C most of the time, while substituting gyro data only during those brief instances during which gyro and odometry data differ substantially. This way the ill-effects of gyro drift are almost completely eliminated, and our method can thus make use of inexpensive gyros with large drift rates. Experimental data is presented that demonstrates the effectiveness of this approach.