Using Millimeter-Wave Radar to Evaluate the Performance of Dummy Models for Advanced Driving Assistance Systems Test

Abstract

Abstract With the rapid development of intelligent and connected vehicles, the experimental road test for Advanced Driving Assistance System (ADAS) is dramatically increasing around the world. Considering its high cost and hazardous situations, simulation test based on dummy model is becoming a promising way for ADAS road test practice to reduce the experiment expanses. This study proposed a methodology for the evaluation of the performance of human and dummies with distinct designed materials based on the data extracted from the Doppler effect of millimeter-wave radar. Echo data of 8 different angles from 0 to 360 degrees, with the interval of 45 degrees, at the same distance between the test object and the signal source are collected. Meanwhile, the echo energy is collected for correlation modeling and analysis among groups. By evaluating the performance of human and dummies via statistical analysis, a close correlation was found which results verified the substitutability of the dummy for ADAS experiment test. The correlation coefficient between human and dummies ranges from 0.75 to 0.93. The support vector machine (SVM) model was developed and fitted to predict the echo energy in diverse environments. The Mean Average Error is 5.42 to 11.42 in the training and testing datasets while Root Mean Square Error is 0.43 to 0.90. The methods developed in the study can simulate the real ADAS road test environment and support future experimental research.