Online Terrain Classification For Mobile Robots

Todays unmanned ground vehicles (UGV) must operate in and increasingly general set of circumstances. These UGV’s must be able to provide stable controlability on whatever terrain they may encounter. A UGV might begin a assignment on asphalt but quickly be required to negotiate sand, mud, or even snow. Traversing these terrains can affect the performance and controllability of the vehicle. Like a human driver who can feel his vehicles response to the terrain and make changes to compensate, a UGV that could autonomously determine the current terrain type could also make necessary changes to its control strategy. This article focuses on the development and application of a terrain detection technique based on terrain induced vehicle vibration. Using IRobots ATRV Jr mobile robot equipped with various sensors, we extract the frequencies of the vehicles terrain induced vertical acceleration. These frequencies are input to a fast online terrain classification algorithm based on the Probabilistic Neural Network. We present terrain classification results for the robot traveling at various speeds over multidifferentiated terrains broadly classified as sand, grass, asphalt, and gravel.Todays unmanned ground vehicles (UGV) must operate in and increasingly general set of circumstances. These UGV’s must be able to provide stable controlability on whatever terrain they may encounter. A UGV might begin a assignment on asphalt but quickly be required to negotiate sand, mud, or even snow. Traversing these terrains can affect the performance and controllability of the vehicle. Like a human driver who can feel his vehicles response to the terrain and make changes to compensate, a UGV that could autonomously determine the current terrain type could also make necessary changes to its control strategy. This article focuses on the development and application of a terrain detection technique based on terrain induced vehicle vibration. Using IRobots ATRV Jr mobile robot equipped with various sensors, we extract the frequencies of the vehicles terrain induced vertical acceleration. These frequencies are input to a fast online terrain classification algorithm based on the Probabilistic Neural Network. We present terrain classification results for the robot traveling at various speeds over multidifferentiated terrains broadly classified as sand, grass, asphalt, and gravel.