Road friction estimation for automobiles using digital signal processing methods

This dissertation develops a new friction estimation method that can detect a low-friction road surface during normal braking using only measurements of the wheel rotational speeds and the vehicle translational velocity. Such a method could find application in Automated Cruise Control systems or Automated Highway Systems because it would allow these systems to choose a driving style that best suits the road condition.; Detecting low-friction roads is largely a signal processing problem, and the difficulties it presents inspire the creation of a filtering technique that we call the “Optimal FIR Derivative.” The Optimal FIR Derivative is a finite impulse response, noise-attenuating differentiator that is useful when very little information is known about the signal to be differentiated. The only data needed for its design are a noise variance and a second derivative bound for the signal that needs to be differentiated. The filter that results is optimal in a minimax sense. Although the new filter is inspired by the friction estimation problem, it is a fairly general method that can be used in any field where noisy data needs to be differentiated.; The chapters of the thesis make a loop from application to theory and then back to application: The practical problems that arise from friction estimation motivate theoretical tire modelling and signal processing investigations; the resulting theoretical results are then successfully applied to help solve the practical problems that motivated them.; The dissertation first acquaints the reader with the friction estimation problem through a literature review and a series of experimental results from a test vehicle. These results establish a correlation between tire slip data taken from normal braking and the road's peak friction value. The correlation has appeared in several recent papers, but it has so far gone without complete theoretical explanation. The thesis therefore takes an excursion into tire/road contact theory in order to make connections between existing theoretical tire models and the empirical correlation. The remaining chapters are then devoted to exploiting this correlation to estimate road friction. An important obstacle that appears is the problem of estimating tire force during braking. Numerical differentiation proves to be a good solution to this problem, but the particularly small amount of a priori information that is available about the signals involved motivates the development of the Optimal FIR Derivative. The thesis concludes by combining the experimental, tire theoretical, and signal processing results of the earlier chapters into a low-friction road detection method.