Study on Terrain Classification Based on Vehicle Suspension Vibration

In order to improve the terrain classification performance using the vibration response induced in suspension of a traversing vehicle on natural terrains, a new feature vector extraction method was proposed by combining time domain features and wavelet packet energy features. The probabilistic neural network(PNN)was utilized to perform classification, comparing the classification effect of the combined feature vector with the other two traditional ones. A road simulator was employed to perform the excitations of the presented six roads. The vibration data was collected by a single axis accelerometer mounted on the suspension arm perpendicular to the ground. The results indicate that the proposed method can result in a satisfactory classification accuracy of 91.3%, which outperforms the other two traditional ones.