An Inertial Sensing Based System for Lane Curvature Estimation

In this paper, an inertial sensing based scheme without visual detection devices for lane curvature estimation is proposed. Our purpose is to obtain the lane curvature continuously when the visual signals are failed to use abruptly. Our proposed algorithm of lane curvature estimation is separated into three steps. The first step is to calculate the lane curvature through Ackermann steering principle after obtaining the speed and orientation of vehicle. In second step, the lane curvature function, which is formulated into quadratic form, is taken into account. After augmenting the vehicle position and lane curvature into a new augmented system, the state of lane curvature can be solved via some iterative mathematical technique. In this paper, an extended Kalman filter (EKF) method is employed to calculate the estimated lane curvature. However, both of the measured and estimated lane curvatures do not match the true curvature exactly. Therefore, a descriptive statistic technique is applied to obtain the evaluated lane curvature in the third step. The standard error of the mean of both measured and estimated lane curvature is ensured to be distributed in a range of Gaussian distribution, so that the evaluated lane curvature is approximated to the true curvature more exactly. Finally, the experimental results are demonstrated to verify the efficacy of our proposed method of lane curvature estimation.