Intelligent Performance, Architecture Analysis, Functional Safety Metrics of Automated Steering Systems for Autonomous Vehicles

The increasing complexities and functionalities of the electrical and/or electronic (E/E) systems in present day automobiles, make it challenging for original equipment manufacturers (OEMs) and suppliers to ensure a high level of safety in the automotive critical safety systems. The steering systems represent a standard functionality on every vehicle to control the direction of the vehicle literally and provide more stability for the vehicle motion. High automated vehicles require intelligent steering systems in which more Advanced Driver Assistance Systems (ADAS) applications are linked together such as cameras, radars, Lidars, and global positioning system (GPS). These integrated systems and applications are required for environmental perception, communications, data fusion, planning, prediction, decision making, and actuation processes all in real-time. Therefore, hardware (HW) and software (SW) solutions are developed and implemented in compliance with ISO 26262 standard, Road Vehicles – Functional Safety. Due to the lack of the steering systems published information and the crucial role of the steering associated with complex functionalities challenges, this dissertation provides a case study of how the steering systems of different automated driving levels can be complied with ISO 26262 given the emerging challenges imposed by the electric vehicle curb weight, increasing trend for the near future. The analysis focused on the safety lifecycle of the E/E components of the steering systems to ensure high availability of the steering systems and avoid any sudden loss of assistance (SLOA). Various safety mechanisms were evaluated and analyzed to improve the functional safety of the steering systems architecture and logic control paths. Based on the proposed controllability metrics performed in this dissertation, it was found that the hazard or malfunction of the steering systems shifted from the Automotive Safety Integrity Level (ASIL) B to ASIL C, the second most critical safety level. To comply with the ISO 26262 and to mitigate the residual risks of E/E systems failure, several solutions proposed in the concept for compliance with the standard such as redundant HW or SW in the controller path. The controllability classes or categories of the high automated vehicles based on the vehicle global position related to the lane marker lines were investigated and redefined to accommodate for the machine or system in the loop controlling the dynamic driving task (DDT) in autonomous vehicle maneuvering. A new wheel offset marker concept was introduced when the vehicle is approaching the lane marker lines. Also, it was found that the are human factors challenges in SAE level 4 and 5 and the interaction between the driver and the automated control systems of the vehicle that require human machine interface (HMI) modalities. The driver – automated control system engagement in the steering system of the vehicles is one of the crucial control complex scenarios that add uncertainties and potential risks when handing over the steering control between the driver and-or the automated control system with the allotted time. This study highlights the need to define the driver intervention in high-automated vehicle of SAE level 4 and 5 in order to sustain the traffic safety and keep the vehicle in the intended trajectory or path. This can be addressed by deploying HMI and the human factor implementation in ISO 26262 to standardize the driver-machine relation with the DDT in real time and interactive environment. Both manual and automated driving modes demand the functional safety implementation of the steering system to mitigate any system malfunction or failure.An artificial neural network (ANN) model was developed to predict the steering torque commands and steering wheel angle (SWA) based on the steering system dataset and vehicle’s parameters. ANN model was developed using Neural Network Training (nntraintool) toolbox of MATLAB to evaluate the intelligent steering system performance. The trained ANN model delivered a regression value of ~ 98.5 % versus the measured SWA. The results showed that the ANN was effective in predicting the steering wheel angle patterns based on the input dataset, considering the non-linearity and complexity of the steering system control. This finding helps to improve the functional safety of autonomous vehicles and introduce the concept of intelligent steering systems for path and trajectory planning. Therefore, ANN should be implemented as an abstraction layer in the control module and deployed in the control and actuation processes to support sensor data fusion and support the prediction and pattern recognition.