Combinatorial test (CT) is a commonly used method to conduct comprehensive and efficient test. To further improve test efficiency, based on CT, this paper proposes an improved combinatorial test (ICT) method for model-in-the-loop (MIL) test of the autonomous parallel parking system (APPS). Different from CT, in the process to generate test cases, the ICT takes not only combination coverage but also importance degree into consideration. When selecting parameter value of the test case, the ICT tends to select the value with higher importance degree, on the premise of the combination coverage and the number of test cases remaining unchanged. Among the different values of the same parameter, these more likely to lead to bad parking performance are assigned higher importance degrees using Analytic Hierarchy Process (AHP). The experiment results show that, overall, for different values of the same parameter, the values with higher importance degree appear more frequently in the ICT test suite. Besides, compared with CT, the ICT further improves the test efficiency.