A Deep Learning-Based Test Sequence Automatic Generation Method for Automatic Train Operation in High-Speed Railway System
Pengfei Ning, Tao Tang, Li Zhu · 2021
The automatic train operation (ATO) system for high-speed railway (HSR) can further reduce operation costs, and increase railway transport capacity, which is the future development direction. Testing is an essential technical method to ensure the functional correctness of the ATO system in HSR. Test sequences guide the entire test work, and their quality has a significant effect on test quality and efficiency. Testers connect test cases into test sequences based on train operation control system principles and test experiences. However, the process is redundant and complicated. Aiming at efficiency and automation of test sequences generation, we propose a test sequence automatic generation method based on the encoder-decoder model with an attention mechanism. Concretely, the model can learn the representation of the source text through an encoder component, and the attention mechanism allowing it selectively search for a part of the source feature to predict a target test case during encoding process. Experiments are conducted by using the Beijing-Shenyang HSR test sequences to train the neural network model. We find that the proposed model can transform the natural language into test sequences effectively and achieve better precision performance in the test dataset. Finally, we verify the feasibility of this method by comparing and analyzing the differences between the sequences generated by the model and the test sequences manually edited quantitatively and qualitatively.