Similar Association Set Based Attribute Selection Algorithm for Road Profile Detection

Abstract

In this study, our goal is to develop an algorithm for selecting most appropriate attributes of CAN (Controller Area Network) bus data in association with Uphill/Downhill road detection problem. In turn, drivers are made aware of Uphill/Downhill situations on steering wheel. In a sense, drivers are enabled to have optimal driving performance for roads at different slopes. Initially, the system collects vehicle dynamic data. Then the proposed algorithm is applied to find optimal attributes before estimating road profile on which vehicle travels. In order to estimate Uphill/Downhill case, we initially consider all the vehicle parameters of the system. Therefore, we drive a car at different road slopes. In these experiments, we collect the relevant data to evaluate them in following stages. A new algorithm is developed and applied to eliminate attributes with least significance. Hence, at the first stage correlated attributes are sought as key features. That is, we search for attributes with similar correlation characteristics for both Uphill and Downhill roads. Similar association sets are established to search for redundant attributes that will be eliminated by our feature selection algorithm. In turn, an optimal set of attributes is determined for the goal driven system development. Proposed algorithm can be employed by any system that includes redundant attribute set.