Development of a Fault-Tolerant Vehicle Control for Crash-avoidance in Critical Scenarios for Over-actuated Vehicles

The concept of autonomous vehicles is mainly centered on reducing the number of accidents and fatalities by substituting the driver with intelligent control systems. The intelligent controller is responsible for maneuvering the vehicle continuously in any driving situation. In recent years, new vehicle concepts such as over-actuated are being investigated due to increased comfort, maneuverability, and stability. Therefore, as the complexity of the autonomous vehicle concept grows with the implementation of sensors and multiple actuators, it is necessary to ensure that the autonomous vehicle can still guarantee safe driving even when facing system faults. Therefore, developing and implementing fault vehicle dynamic controls dedicated to critical scenarios are crucial for increasing vehicle occupants' and road users' safety, particularly in evasive maneuvering. The intelligent controller must handle the vehicle up to handling limits and deal with faults to avoid crashes. Present control strategies consider handling limits, fault strategies, critical scenarios, crash avoidance, and driver discomfort. Nevertheless, they do not address the combination of these parameters for an over-actuated autonomous vehicle that faces a short time window for a crash. This study aims to develop and evaluate a new fault-tolerant control strategy for a maximum complexity over-actuated vehicle with 4-wheel independent steering, 4-wheel independent electric drive, and 4-wheel independent brake. The purpose is to adapt and optimize the vehicle actuators for crash avoidance in critical scenarios with up to two concurrent actuator faults. Critical scenarios that result in deaths and injuries are identified via literature research. The design of the intelligent vehicle controller starts by considering all the actuators fully functional when the vehicle faces an emergency maneuver. The vehicle actuators that might fail are classified according to their probability of occurrence and effects on vehicle behavior and crash avoidance. Later, the vehicle controller is further extended using machine learning techniques to consider up to two concurrent actuator faults for velocities up to 130 km/h, which is the maximum intended velocity for autonomous vehicles. The goal is to make the vehicle follow the desired trajectory and avoid a crash with fully functional actuators and in the presence of actuator faults, with safety prioritized over comfort. From the vehicle dynamics perspective, the goal is to keep the tires' forces within or close to saturation limits. State-of-the-art environmental simulation models are used to develop the intelligent vehicle controller. Testing, controller updates, and proof of concept occur in simulation and a scaled vehicle.<p></p>