TERRAIN CLASSIFICATION FOR TRAVERSABILITY ANALYSIS FOR AUTONOMOUS ROBOT NAVIGATION IN UNKNOWN NATURAL TERRAIN

Abstract

Autonomous robot navigation in unknown natural terrain is the key requirement of a robotic system such as a rover as the natural terrain is unpredictable. For safe and robust navigation a robot must possess onboard intelligence to perceive the terrain ahead so that it can avoid hazardous areas by discriminating the navigable regions for traversal thereby optimizing its speed and increasing its efficiency. This is accomplished by realtime assessment and classification of terrain for traversability analysis. In this paper terrain classification system is developed for its traversability analysis for the robot operating in natural unknown terrain. The proposed method classifies the unknown terrain describing its suitability for navigation by extracting the textural features from visual imagery of terrain data. Clustering technique is employed to classify the features derived from statistical and transform based texture analysis. And further we have presented a fast and optimum algorithm for path planning of robot on the assessed navigable terrain. Images from NASA’s Mars exploration rover and those obtained from real time natural terrain have been tested to validate the proposed approach.