January 2022
·
167 Reads
Due to the increasing complexity and interconnectedness of different components in modern automotive software systems there is a great number of interactions between these system components and their environment. These interactions result in unique temporal behaviors we call underlying scenarios. The signal data from all system components, which is recorded during runtime, can be processed and analyzed by observing changes in their runtime. Different system behaviors can be characterized by dividing the whole data spectrum into appropriate segments with consistent behavior, classifying these segments, and mapping them to different scenarios. These disjunctive scenarios can be analyzed for their specific behavior which may divert from the expected average system behavior.We state the emerging problem of data segmentation as follows: divide a multivariate data set into a suitable amount of segments with consistent internal behavior. The problem can be divided into 2 subproblems: ”How many segments are present in the data set?”, and ”What are good segmentation indices for the underlying segments?”. The complexity of the problem still needs to be assessed, however, at this point we expect it to be NP-hard, as both the number of segments and the segmentation points are unknown. We are in search of appropriate metrics to quantify the quality of a given segmentation of a whole data set. In this paper, we discuss the segmentation of multivariate data, but not the classification of segments into scenario classes. In the following, we investigate segmentation algorithms for solving the subproblem of finding suitable segmentation indices by constant amount of segments. The algorithms are investigated towards effectivity and efficiency by applying them to a data set taken out of a real system trace provided by our automotive partners.