An algorithm has been developed for detecting a collision-free path for an Automated Guided Vehicle (AGV) when it is required to navigate in an unstructured environment. Real time implementation issues are discussed and investigated experimentally using an ultrasonic sensor system. Concepts of obstacle matrix and path penalty are used in the guidance of the vehicle in a work area discretized as matrix cells. Two search methods namely, the arc searching method and obstacle crawling method, are developed for the AGV to facilitate navigation. The algorithm is capable of admitting the appropriate search method depending on the nature of the obstacles in the work area. Case studies using real time data from ultrasonic sensors demonstrate an excellent level of performance for the algorithm.