In view of the influence of uncertain situations on Unmanned Surface Vehicle (USV) autonomous navigation, an approach to evaluate USV autonomy level based on multi-attribute decision making theory is presented. The method is to build evaluation model by giving a problem description of decision and supplying objects design. Using the model to construct a directed assignment graph, the assessment result of autonomy level is the maximal utility node. Advantage of USV navigation application in the simulation system, it proved that the approach can effectively decide the autonomy level for USV under uncertain situations. So by combining human cognition with USV adjustable autonomy, the unmanned system's overall performance is improved under uncertain situations.