Advanced Driver Assistance Systems (ADAS) must be tested thoroughly before they can be deployed for series production. Bench test has been widely used owing to its advantages of less manpower cost, more controllable scenarios, etc., comparing to road test and field test. However, there’s no satisfied systematic approach to automatically generating high-efficiency and full-coverage scenarios for ADAS because of its integration of human, vehicle and traffic. Most of the design methods still use the typical scenarios specified by ISO and other standards which are commonly used in field test, or the critical conditions selected from the traffic database. In this paper, an automatic virtual test scenario generation method based on combinatorial theory for bench test is presented. Firstly, considering the fact that the tested device is easier to malfunction under complex scenarios, an evaluation index measuring the complexity of scenario is proposed by using the method of Analytic Hierarchy Process (AHP). Based on the existing combinatorial test cases generation algorithm, the proposed complexity index is used to evaluate the effectiveness and guide to generate more effective test cases. Furthermore to improve the test efficiency and ensure higher complexity of test scenario, a clustering method has been used to combine the discrete test cases into continuous scenarios. The effectiveness of the proposed method has been validated by applying to a Lane Departure Warning System (LDWS). The results show that: (a) A more complex scenario is easier to find the faults of a system; (b) The scenarios generated by the proposed method can detect the system malfunctions more efficiently with a more compact test suite; (c) The generated scenarios can achieve full coverage of the specified N-dimensional combination.