Safety assessment of autonomous and connected vehicles by a model-based traffic simulation framework

Autoren

Mustafa Saraoğlu, Andrey Morozov, Klaus Janschek, Ifa/TU Dresden

Jahr

2019

Zusammenfassung

Safety analysis is an important topic in many different domains in engineering. One of the most important areas where safety assessment plays a major role is the autonomous driving. As a safety-critical domain, the automotive sector heavily needs the tools and techniques to assess the safety level in design phases. This paper presents a new simulation framework, MOBATSim, where simulation-based fault injection is used to test fault tolerance capabilities of autonomous driving systems in different abstraction levels such as component, model, and system level. The simulation framework is designed in MATLAB Simulink in a model-based fashion where the highly
customizable structure is extended with fault injection models. These fault injection models are then used for safety assessment by examining error propagation between low-level vehicle components and high-level systems. Then safety goals are defined for the high-level systems such as vehicles or the traffic. Simulation-based testing is done in accordance with a suitable driving scenario, in order to verify if these safety goals are met in the presence of low-level component faults. The functional safety assessment for these systems becomes an urgent necessity for the transition to the full autonomy. Testing these functions consisting of decision and control algorithms with a lot of variables and parameters in a unified manner is a daunting task. Threat assessment has to be made for vehicles to actively avoid hazardous situations. This requires the analysis of complex operational profiles such as routing, intersection management and collision
prediction in an environment where multiple vehicles are in different positions, and traveling at different speeds. There is a need for a comprehensive traffic simulation framework which models not only the functionality of the vehicles but also the interactions between them. As a solution, we offer a simulation framework based on MATLAB Simulink that allows the user to customize decision and control algorithms for the modeled autonomous vehicles and analyze their effects on the safety of the highlevel urban traffic environment. Also, the simulation-based fault injection is used to assess safety in the presence of various types of random faults of low-level vehicle
components.

Mitglieder-Login