Research on an Acceleration Test Method of Automatic Driving Based on Deep Neural Network

In order to realize the accelerated test of automatic driving simulation test under the consideration of scene coverage, this paper proposes an accelerated test method based on deep neural network. The neural network is used to approximate the deep relationship between each element of the test scene and its attributes and the performance of the module to be tested, effectively obtaining the performance of the module to be tested in each scene, so that the most meaningful scene can be quickly selected, Finally, the purpose of accelerated testing is achieved. Compared with the conventional traversal simulation, the experimental results show that the method proposed in this paper can effectively improve the test efficiency, quickly find out the dangerous scenes, and the more scenes need to be simulated, the more obvious the improvement.