Workload Analysis of Active Safety Processor

Abstract

Active Safety (AS) is an important feature to protect drivers and passengers in vehicles, however, the growing complexity of AS functions increases the CPU workload of the AS processor. This thesis attempts to 1) identify the signals that cause more CPU workload than others and 2) predict this CPU workload. In order to achieve the purpose, regression models of CPU workload in real-time contexts are proposed. Based on these models, different feature selection techniques are used to rank the contributions of signals. As a result, given the sampled input signals to all the software modules that are embedded running on the CPU, a model for predicting CPU workload in a real-time context is proposed. Signals and the associated contribution coefficients are listed using different feature selection methods and criteria. Volvo Car Corporation will be able to trace back to the functions that generate signals with high contribution coefficients according to the list of signals, and improve the corresponding AS functions to decrease the CPU workload.