Controller area network with offset scheduling: Improved offset assignment algorithms and computation of response time distributions


Tezin Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Mühendislik Fakültesi, Elektrik ve Elektronik Mühendisliği Bölümü, Türkiye

Tezin Onay Tarihi: 2018

Öğrenci: AHMET BATUR

Eş Danışman: KLAUS VERNER SCHMİDT, ŞENAN ECE SCHMİDT

Özet:

The Controller Area Network (CAN) is the most widely-used in-vehicle communication bus in the automotive industry. CAN enables the exchange of data among different electronic control units (ECUs) of a vehicle via messages. The basic requirement for the design of CAN is to guarantee that the worst-case response time (WCRT) of each message is smaller than its specified deadline. Hereby, it is generally desired to achieve small WCRTs that leave sufficient slack to the message deadline. In addition, it has to be noted that it might be very unlikely that a message experiences the WCRT when being transmitted on CAN. That is, instead of only considering the message WCRT for the design of CAN, it is beneficial to determine the actual response-time distribution of each message, which indicates the probability of experiencing a certain response time. In order to achieve small WCRTs, the idea of offset scheduling has been introduced. In this setting, messages on CAN are released with offsets in order to avoid message bursts that lead to undesirably large response times. In order to use offset scheduling efficiently, it is required to assign a suitable offset to each message. To this end, a load distribution (LD) algorithm is proposed in the existing literature. The first contribution of this thesis is the development of new algorithms for the offset assignment on CAN. Evaluating different example scenarios, the thesis shows that the proposed algorithms outperform the existing LD algorithm in most of the cases. As the second contribution, the thesis studies the computation of response time distributions. First, an algorithm for determining the exact response-time distribution of each message on CAN is proposed. Since this algorithm comes with a high computational complexity, it cannot be applied if there are too many messages on a CAN bus. Moreover, experimental results show that the response time distribution depends mostly on the initial phasing of the nodes. Therefore exact response time distribution as computed is not observed in the measurements. In response to this observation, the thesis proposes the computation of a local response time distribution and develops and implements a weak synchronization method which bounds the phase shift between the nodes. The resulting computed local response time distribution shows a very tight match with measured response time distributions.