Anti-lock braking control design using a Nonlinear Model Predictive approach and wheel information

Abstract

Over the past century, cars have become a fundamental part of our society. With the increasingly larger number of vehicles on the road, safety is, now more than ever, a topic of paramount importance. It is estimated that every year 26,000 people are killed on European roads, corresponding to a social cost of 100 billion euros. In the last two decades, awareness about this issue has however increased substantially, and great effort is now put into improving these numbers.
The introduction of driver assistance systems in recent years has led to a significant decrease in fatal accidents. One of the first implemented ones was Anti-lock Braking System (ABS), introduced on the market in 1978. ABS is an active safety system nowadays mandated on every new car sold in Europe. By preventing wheel lock during emergency braking, the system allows for shorter braking distances and steering ability retention. Despite the fact that forty years went by since ABS introduction, the control strategy behind it has changed very little, still retaining a rule-based approach. Extensive review of literature of the subject, highlighted the possibility that significant improvements could be achieved if the control strategy was to be redesigned in a way that takes advantage of the many technological improvements achieved in the last decade.
The aim in this thesis is to verify this statement and quantify any potential improvement in Anti-lock Braking System. In order to achieve the research objective, a novel ABS algorithm was designed. The controller, based on state-of-the-art hardware, uses a model predictive control approach and potentially available wheel force information as the pillars of its design.
A supervisory activation logic is also developed to replicate full ABS functionalities and keep the driver in the loop. The controller is then tested on Toyota’s high-end vehicle simulator and benchmarked against the current industrial state of the art.
The vehicle simulator has its core in a multibody model augmented with kinematics and compliance measurements, and it is largely validated against experiments. A complex multi-physics model of the brake system, together with the short wavelength rigid ring tire model, complete the setup.
A comprehensive set of manoeuvres, including friction jumps and rough road braking scenarios, is developed to assess performance and robustness of the proposed controller. Additionally, a list of aspect-specific performance metrics is drafted and deployed to accurately investigate the simulations results and identify relative gains with respect to the benchmark in different instances of the braking manoeuvre. The analysis showed substantial reduction of the braking distance and improved steering ability in each of the simulated scenarios. Furthermore, robustness of the controller against external factors, such as high frequency noise generated by road irregularities or friction transitions, is demonstrated to be comparable to that of industrial state of the art controllers. Based on the now verified potential of the proposed controller, and a list of the future engineering steps mapping the way until an eventual controller deployment on the market is identified.