Towards Chaos Engineering for Fault Injection Testing of Internal Automotive Systems - Quantifying Disturbance Tolerances in a Centralized Automotive Architecture

Abstract

As the automotive industry becomes more and more reliant on software-defined functionality, the vehicle’s internal communication system has to develop to keep up with the ever-increasing demands. A recent development is the centralization of the internal architecture, focusing most of the internal computation on one powerful central computer as opposed to across several small ones distributed throughout the vehicle. This introduces a single point of failure into the system and while automotive systems are built to be robust, the potential effects of such a failure should be investigated preemptively. To face this challenge, this thesis investigates a method to enable Chaos Engineering - negative testing in the production environment - in the automotive domain. A Device-In-The-Middle fault injection system was developed and implemented into the core automotive system along with several fault models, enabling the disturbance of traffic flowing between the vehicle control unit and its connected gateway units, to quantify the disturbance tolerances of the system. Additionally, the throughput of the Device-In-The-Middle when exposed to increasing data rates was measured and compared to data rates expected of a vehicle in operation. By systematically applying different disturbance magnitudes to repeated test case executions aimed at validating the core system, the system’s approximate disturbance tolerances, along with some deviations from expected system operation, were found and analyzed. The combined tolerance and performance results indicate that, with some further development and latency optimization, the system has the potential to work as a chaos testing method in the automotive software testing process.