Robust Autonomous Emergency Braking Algorithm using the Tire-road Friction Estimation and the Sensor Uncertainties

Abstract

Safe and collision-free travel is vital in todays society. It is also an important issue in many industrial processes. Therefore the automakers are trying to help drivers to avoid or mitigate collision with active safety systems instead of passive safety systems. For example, ACC (Adaptive Cruise Control) and AEBS (Advanced Emergency Braking System) warn the driver from rear-end collision risk and intervene by partial braking maneuvers have already been implemented in modern passenger vehicles in recent years. Since the active safety system always work with a human driver co-existing, the longitudinal safety system must be acceptable to the driver. Thus the system operation law need to be set based on the human drivers driving characteristics. In order to be acceptable to drivers, vehicle behavior or driving characteristics of the control in target situation needs to be similar to the human drivers. Therefore, in this thesis, the longitudinal safety control algorithm was not only designed by using the physical collision risk but also by drivers characteristic to achieve safe and acceptable control algorithm. In the case of road information, previous research about friction estimation was used to estimate the tire-road friction information. To make up for unreliability in normal steady driving situation, some assumptions were applied in the case of normal steady-straight driving condition. By using the estimated tire-road friction information, safety indices for the control mode decision were redefined. Generally, measured sensor signal has difference with true value due to the measurement noise or uncertainty. To guarantee the robust control performance, RADAR and vision sensor are used. For robust control mode decision, the simple static theory and Kalman filter considering the measurement noise are used in this research. The expected error range of the longitudinal safety index from the measurement noise can be defined from the covariance matrix of the Kalman filter and simple definition of the deviation of the function. By using the expected error, the threshold of new longitudinal safety index was determined for the safety monitoring of the driving situation. The proposed vehicle longitudinal safety algorithm was evaluated through computer simulations using vehicle simulation software, CARSIM and MATLAB/Simulink in two kinds of scenario: 1) emergency braking in steady following driving, 2) stop preceding vehicle in the rainy day situation. Also, to confirm the robustness of the proposed control algorithm, the simulation was conducted 100 iteration in the same scenario. From simulation results, it can be concluded that the proposed longitudinal safety algorithm could enhanced the longitudinal safety and guarantee the robust capacity from the sensor uncertainty and various road condition.