Development and testing of a risk-based control system for autonomous ships

This paper presents a method for designing and verifying a control system with risk-based decision-making capabilities to improve its intelligence and enhance the safe operation of autonomous systems. The decision-making capabilities are improved, compared to existing control systems, using a Bayesian Belief Network (BBN) that is derived from the systems theoretic process analysis (STPA) as a foundation for an online risk model, which represents the operational risk for an autonomous ship. Combined with an electronic navigational chart (ENC) module to get accurate information about the environment, this enables the ship to operate in a safe and efficient manner. In addition, the control system is verified against safety and performance requirements using a formal verification method, based on temporal logic and Gaussian processes. The proposed methodology is tested in a case study where the system's behavior is compared with an existing conventional (manned) ship on experimental data from two routes along the coast. The case study shows that the performance of the Supervisory Risk Controller (SRC) with respect to the autonomous ship speed and maneuvering is similar to how the existing ship is operated. This means that the proposed methodology shows promising results with respect to developing autonomous ships with control systems and leads to intelligent and safe behavior.