We describe the application of fuzzy logic to the control of a robot intended to navigate forest paths. Robots could improve the safety of forest operations by removing the operator from the vehicle and also reduce costs by automating these operations. The controller was organized into three modules, each of which was capable of providing a heading recommendation for the robot. Sensory information was provided by shaft encoders and three ultrasonic rangefinders. The controller required a set of sensory data acquired from one training run down the path. Two of the modules provided heading recommendations from the shaft encoder and ultrasonic rangefinder data, respectively. A third module arbitrated the recommendations of the first two modules and provided a final heading recommendation. Decisions made by the operator during the training run were used in the arbitration to weight the recommendations of the individual modules. It was observed that the combination of modules performed better than the individual modules alone. Despite the simplicity of the controller, a test vehicle was able to consistently navigate 152 m (500 ft) down a forest path. It was concluded that the approach is feasible, and further development of the control algorithm is warranted. © 2004 American Society of Agricultural Engineers.