Autonomous Navigation of Forest Trails by an Industrial-Size Robot

Abstract

<abstract><title><italic>Abstract. </italic></title> A commercially available tracked vehicle was modified to make it capable of autonomous operation in a forest environment. The vehicle, of 1364 kg mass (3000 lbs weight), was powered by a 22 kW (30 hp) diesel engine coupled to a hydraulic drive. For autonomous operation, the vehicle was equipped with shaft encoders on the drive wheels, a magnetic compass, three ultrasonic sensors, and a stereoscopic camera. Autonomous operation was enabled by hierarchical fuzzy logic controllers. Two controllers were considered. One controller used shaft encoder counts on the drive wheels, ultrasonic waveform echoes, the magnetic compass for heading, and training data to navigate a forest path. The other controller added stereoscopic vision and dispensed with the training data to reactively navigate the autonomous outdoor vehicle. The vehicle was equipped with GPS, but GPS was only used as an independent measure of vehicle performance in the autonomous mode. Tests using both controllers showed that the autonomous vehicle was able to reliably navigate several forest paths with lengths ranging from 229 to 430 m (750 to 1410 ft) without relying on GPS for navigation. These tests indicate that an industrial-size vehicle may be capable of performing useful forestry functions in an autonomous mode.