The Identification of Terrains for Mobile Robots Using Eigenspace and Neural Network Methods

Today’s autonomous vehicles operate within an increasingly larger set of environments compared to earlier research in which environments were more controlled. In particular, unmanned ground vehicles (UGV’s) must be able to travel on whatever terrain the mission offers, including sand, mud, or even snow. These terrains can affect the performance and controllability of the vehicle. Like a human driver who feels his vehicle’s response to the terrain and takes appropriate steps to compensate, a UGV that can autonomously perceive its terrain can also make necessary changes to its control strategy. This article focuses on the development of a terrain detection algorithm based on features extracted from terrain induced vehicle vibration. Research is conducted to reduce correlation of traversing terrains at different speeds. Procedures are presented to remove the dependencies of speed through eigendecompositon methods and applying the probabilistic neural network for classification between nonlinear boundaries. Experimental results based on iRobot’s ATRV Jr demonstrate that the algorithm is able to identify multi-differentiated terrains broadly defined as grass, asphalt, and gravel.