Road Condition Classification Using A New Global Alignment Kernel
Takato Goto, Yasushi Hanatsuka, Tomoyuki Higuchi, Tomoko Matsui

Abstract:
The development of the so-called intelligent tire has changed the role of the tire. Here we discuss a real-time road condition classification system that employs monitoring tire acceleration. Because the tire acceleration is non-stationary and is warped non-linearly in the time domain, we applied the time alignment algorithm to it similarly to speech recognition. High accuracy classification systems robust against tire-size range variations have been achieved in previous studies, in which the same pattern tires with prescribed pressure and several vehicles were used. However, there is no discussion on the execution speed of such a system. Therefore, we validated the algorithm and improved it for use in real-time applications. Specially, the calculation time was reduced by limiting the path of the global alignment kernel, which handles a sequence with time warping. Furthermore, we used SVM with L" norm regularization instead of L2 norm regularization. We succeeded in achieving a real-time road condition classification system, which was verified in a vehicle test.