This paper describes the navigational control scheme used in the CO-Operative Locomotion Aide (COOL Aide), an intelligent walker designed to assist the elderly or the disabled with normal, and routine walking tasks. Navigation is achieved through a shared control architecture that recognizes the goals of both the human user and the walker. The control system is based on a synthesis of heuristic logic that exploits a dynamic model of walker system that can detect sliding and loss of walker stability. The model is used to predict the user’s intended path, based on the history of information collected from the walker’s sensors. Sensor information consists primarily of the forces and moments the user exerts on the walker’s handles during the natural assisted walking process, as well as the user’s local environment. Based on the model’s prediction, the walker’s state, and the walker’s environment, the control system can confirm or overturn the hypotheses of user’s intent it put forward and can influence the walker’s heading if the system believes the user will not reach the perceived intended goal unassisted. This paper discusses the model’s use in the shared control scheme and the mechanism for detecting/handling errors in the model’s predictions.