Terrain classification for terrain parameter estimation based on a dynamic testing system

This paper aims to gain the real-time terrain parameters of the battlefield for the evaluation of military vehicle trafficability. In military missions, improvements in vehicle mobility have the potential to greatly increase the military operational capacity, in which vehicle trafficability plays a significant role.

In this framework, an online terrain parameter estimation method based on the Gauss-Newton algorithm is proposed to estimate the primary terrain mechanical parameters. Good estimation results are indicated, unless the initial values involved are properly selected. Correspondingly, a method of terrain classification is then presented to contribute to the selection of the initial values. This method uses the wavelet packet transform technique for feature extraction and adopts the support vector machine algorithm for terrain classification. Once the terrain type is identified, advices can be given on the initial value selection referring to the empirical terrain parameters.

On the basis of a dynamic testing system suitable for real military vehicles, the proposed algorithms are validated. High estimation accuracy of the terrain parameters is indicated on sandy loam, and good classification performance is demonstrated on four tested terrains.

The presented algorithm outperforms the existing methods, which not only realizes the online terrain parameter estimation but also develops the estimation accuracy. Moreover, its effectiveness is confirmed by real vehicle tests in practice.