Higher-Fidelity Modelling and Simulation of the CAN Protocol Stack

Abstract

This thesis details a higher-fidelity, scalable simulation tool and model for message response time and bus utilization rate analysis for the Controller Area Network (CAN) protocol stack. This tool achieves higher fidelity than existing commercial and academic simulation tools by including details of the stack implementation that are often neglected, such as receive and transmit hardware buffer availability and usage policy (i.e., which messages are able to be copied to which buffer resources), and the buffer polling or queueing policies. Key details of these features have been identified by a thorough examination of CAN stack behaviour, taking into account the physical considerations of commercial CAN implementations. Inclusion of these details in the simulation can produce better accuracy by exposing certain priority inversion scenarios. Scalability is achieved by using a transaction-based modelling approach and modelling transmissions at the protocol level rather than the physical/bit level. The tool requires minimal user interaction, and system level model generation is automated using an AUTOSAR XML system description file (ARXML format) to specify network topology and message information (transmitter, receiver(s), period, length, etc.), and an Excel spreadsheet file (XLS or XLSX format) to specify node hardware/software implementation details (buffer resource details, polling loop rates, main control loop rates, etc.) as inputs.