Integrating on-board diagnostics speed data with sparse GPS measurements for vehicle trajectory estimation

We evaluate the integration of on-board diagnostics (OBD) speed measurements with sparse and/or missing GPS data for vehicle trajectory estimation using an extended Kalman filter. The suggested framework is relevant to the deployment of low-power, low-cost GPS receivers in the context of mobile and pervasive sensing. Our algorithm is capable of handling sensors at different sampling rates and comprises an accelerometer error model that significantly improves trajectory estimation. Based on field data, we evaluate its performance by simulating deliberately reduced GPS sampling rates, random GPS outages, and varying base sampling rates of the Kalman filter estimation loop. We achieve robust performance in estimating the driven distance and a vehicle trajectory with GPS data downsampled from 10 Hz to 0.02 Hz. We also demonstrate that for random GPS omissions of up to two thirds of the samples, the root mean squared error of position estimates is less than 4 m with GPS data downsampled by a factor of 10.