This thesis describes a program of research aimed at the creation of an unmanned ground vehicle. In this research the Nottingham Off-road Vehicle Architecture (NOVA) was developed along with the ARP (Autonomous Route Proving vehicle. NOVA is a control architecture for a vehicle with the role of autonomous route proving in natural terrain. The ARP vehicle was constructed to demonstrate this architecture. NOVA includes all the required competence for the ARP vehicle to be deployed in unknown outdoor environments. The architecture embodies systems for vehicle localisation, autonomous navigation and obstacle avoidance. The localisation system fuses data from absolute and relative localisation equipment. GPS provides the absolute position of the ARP vehicle. Relative position information is derived from wheel encoders and a pose sensor. NOVA uses a probabilistic technique known as a particle filter to combine the two position estimates. NOVA maintains a local obstacle map based on range data generated by the perception sensors on the ARP vehicle. Analysis is performed on this map to find any untraversable terrain. A local path planner then selects the best path for the vehicle to follow using the map. Decisions made by the path planner are recorded to allow the vehicle to backtrack and try another path if NOVA later finds the chosen route is blocked. NOVA has been extensively tested onboard the ARP vehicle. Results from a series of experiments are presented to validate the various parts of the architecture.