This paper describes a behavior-based method of controlling an autonomous skid-steer robot operating in an unknown environment. We introduce a new modular, neural-fuzzy system called a threshold fuzzy system (TFS). Supervised training, using error backpropagation, is used to find optimal parameters of the TFS. In this paper, a TFS controller is developed for a skid-steer autonomous vehicle system (AVS). Several hundred simulations are conducted and results for the AVS are compared (favorably) with a traditional neural network approach.