Xinao Wang and Joseph G. Walters, Nottingham Geospatial Institute, University of Nottingham, UK

View Abstract Sign in for premium content


Being able to recognise terrain and react accordingly is crucial for the successful operation of mobile robots in unstructured outdoor environments. In this article we introduce a terrain recognition system utilising a low-cost LiDAR for a mobile robot. A conventional remote-controlled vehicle was converted into a robot rover and fitted with a single-line LiDAR that can scan for distance information on the terrain immediately in front of the vehicle. LiDAR data from four different types of terrain was collected and used to train a number of popular machine learning algorithms in order to find the model with the best performance for predicting unknown terrains. The results show that the Random Forest algorithm exhibits a reasonable classification performance, with 82% accuracy with 10-fold cross validation, and 77% accuracy when tested with a previously unseen dataset. The result indicates that LiDAR data can indeed be utilised to derive valuable terrain information for outdoor mobile robots. Further work has also been proposed to improve the accuracy of the prediction model.