Software test items seeing increasing functions that are becoming more complicated
Challenges identified in combining reliability and safety alongside cost reductions
With increasing demand for further improvements in safety and convenience of vehicles, the scale of the development of software to realize this is steadily expanding. The number of software test items has in turn been rapidly increasing, and it is necessary to conduct them within limited timeframes and costs in a more rational manner. NTT DATA MSE (“MSE”) takes advantage of AI automation platforms and applies AI automation testing to large-scale software evaluation operations for vehicle-mounted devices. By having AI carry out conventionally personal operations such as the preparation of test items and priority management of test implementations, we intend to promote efficiency.
Building low-cost and high-quality AI models, thereby preventing post-release defects
To realize AI automation testing, a model is created by inputting product specifications into AI from humans. MSE properly designs a model that matches vehicle specifications/configurations and an AI algorithm, and by establishing a mechanism to build a low-cost and high-quality model, makes it easier to introduce AI inspections to projects. In addition, by providing information on similarities between product functions, it is possible to identify not only defects but also similarity issues. In the testing of vehicle-mounted devices, a wide range of tests are conducted based on vehicle-development phases and from differing viewpoints of tests. We provide test completion criteria for individual tests to AI to ensure that test purposes are achieved, thereby realizing cost reductions and the prevention of post-release defects.
Detection of defects with low frequency of occurrence also possible
Contributing to further quality improvements in customers’ products by expanding the range of applications for AI testing
By applying AI testing to the evaluation of large-scale vehicle-mounted systems, a coverage rate of about ten times that of man-made items has been realized, which leads to quality improvements in customers’ products. In addition, AI testing can also detect defects with a low frequency of occurrence (probability of occurrence: 20%), effectively preventing post-release defects. Thanks to the automation of the model creation mechanism by utilizing standardization, efficiency can be promoted to cost performance that is twice as good as that of man-made models. Therefore, expanding the range of AI testing applications will lead to further quality improvements in customers’ products.