Vehicle Speed Estimation During Excessive Tyre Slip Conditions

Abstract

Vehicle speed is one of the most important states that needs to be estimated in a vehicle. This quantity is safety critical as it drives a majority of the on-board safety and driver-assistance systems. A normal modern car is equipped with wheel speed sensors, Inertial Measurement Unit (IMU), steering angle sensor and powertrain and brake torque sensors which makes the foundations for the speed estimate. The wheel speed sensors provide a relatively good estimate of the vehicle speed in normal conditions with limited wheel slip. However, in excessive wheel slip conditions the wheels speeds significantly diverge from the true speed of the vehicle. In these situations, no reliable direct measurement of the speed is available and the speed estimate needs to be complemented with e.g. dead reckoning based on accelerometer input. In this thesis, a kinematics based extended Kalman filter (EKF) for longitudinal vehicle speed estimation in excessive all-wheel slip conditions is presented. The filter uses combined vehicle orientation and speed estimation, only considering longitudinal dynamics. The proposed filter utilizes a slip-detection system that detects wheel slip and filters out wheel speed measurements from these slipping wheels. It also has a separate logic for speed estimation in braking on slippery surfaces. Two slip detection approaches are presented. One approach is to assume slip between detection of certain events related to the powertrain torque, wheel acceleration and braking. The other approach makes a decision about slip at every time step. The filter and slip detection systems are tested on real-world driving data recorded from two different all-wheel drive vehicles in excessive slip conditions. The results show that the proposed method provides a better estimate than the reference brake supplier estimate, keeping the estimate within ±4% of the ground truth speed for many cases. Though, none of the slip-detection systems provide flawless slip detection resulting in the filter to sometimes rely on non-representative wheel speed measurements degrading the estimate. It becomes clear that the speed estimation is limited both by the approach of detecting slip and by the limited sensor setup providing no absolute measurement of the speed in excessive all-wheel slip.