Aeronautical and marine casualty statistics indicate that the human being, when under stress or at times of peak load, can be a poor co-ordinator of the information available to him, particularly when that information is from a number of different source:, as is often the case in modern ships. Integration and co-ordination of information and its useful application in a closed loop feedback system can reduce the probability of accident as has already been demonstrated in the case of automatic landing systems for aircraft. This thesis describes the development of a digital filter/estimator for use in conjunction with an optimal controller in the automatic guidance of large ships in the approaches to a port. A non-linear mathematical model of a ship is developed and validated by comparison with data from an actual ship. The model is then used in digital computer simulations of the passage of a twin screw car ferry into the Port of Plymouth. The simulations show that the control and guidance system is capable of safely navigating the vessel along the predetermined track through noisy measurements of position, course and speed, A reduced non-linear digital simulation model is then used in the design of a minimum variance filter suitable for installation in a physical model of the car ferry. Tests with this physical model confirm the earlier full scale digital computer simulations, showing that a minimum variance filter is capable of giving very good estimates of the measured states, even though the measurement subsystems are unable to give accurate information because of noise. In the event of a malfunction of one or more of these measurement systems it is shown that the filter continues to give good estimates of all the states.