Inertial sensor-based outdoor terrain classification for wheeled mobile robots

Nowadays, an increasing usage of autonomous mobile robots in outdoor applications can be noticed. Identification of the terrain type is very important for efficient navigation. In this paper, a novel method is proposed for terrain classification in the case of wheeled mobile robots. The classification algorithm uses frequency domain features, which are extracted in fixed-size windows, and Multi-Layer Perceptron (MLP) neural networks as classifiers. Data from inertial sensors were collected for different outdoor terrain types using a prototype measurement system. The data of the accelerometer and the gyroscope were tested separately and together, and different processing window sizes were also applied. The achieved results show that above 99% classification efficiency can be achieved using the collected data.