The performance of unmanned surface vehicles (USVs) is defined as the ability to complete specific tasks in specific environments within a given time scale as a result of the cooperation of multiple technical aspects. However, the traditional optimization method that forcus on a single part in the system provides limited effect on improving the performance of USVs.

Based on the features of autonomous system of USVs, two main forms of the intelligent evolution of USVs are conducted from the perspective of algorithms: the evolution of algorithm functions and evolution of algorithm parameters respectively. In this case, a machine learning-based intelligent evolution method is proposed. An automatic USV control system which satisfies the requirements of intelligent evolution is then designed and tested in a sea trial.

The obstacle-avoidance task in the sea trial proves the capability and feasibility of the proposed method.

The machine learning-based intelligent evolution of USVs is an effective way to continuously improve the performance of USVs, making it a worthy research topic with high application value.