IMU-Based Smartphone-to-Vehicle Positioning

In this paper, we address the problem of using inertial measurements to position a smartphone with respect to a vehicle-fixed accelerometer. Using rigid body kinematics, this is cast as a nonlinear filtering problem. Unlike previous publications, we consider the complete three-dimensional kinematics, and do not approximate the angular acceleration to be zero. The accuracy of an estimator based on the unscented Kalman filter is compared with the Cramer-Rao bound. As is illustrated, the estimates can be expected to be better in the horizontal plane than in the vertical direction of the vehicle frame. Moreover, implementation issues are discussed and the system model is motivated by observability arguments. The efficiency of the method is demonstrated in a field study which shows that the horizontal RMSE is in the order of 0.5 [m]. Last, the proposed estimator is benchmarked against the state-of-the-art in left/right classification. The framework can be expected to find use in both insurance telematics and distracted driving solutions.