Dickmanns 1988b: The 4D-approach to machine vision using integral spatio-temporal models is applied to road vehicle guidance. Both for road parameter estimation in a look-ahead range and for vehicle state estimation relative to the road, dynamical models are given. A system-architecture for active machine vision in road vehicle guidance is discussed. Based on the recursive, feature-based state estimation, numerically efficient situation recognition and process control has been achieved. With 0.1 second cycle time, speeds up to 60 mph and distances up to 24 km (maximum available run length) have been driven autonomously on an empty standard highway with the 5-ton test vehicle VaMoRs.