Evaluation of a centralized architectural concept for the ECU communication, in a distributed real-time embedded system environment

Abstract: A modern Scania vehicle has a large number of Electronic Control Units (ECUs) that communicates via CAN in a distributed real time embedded system environment. The magnitude, complexity and modular nature of the system produce a challenge to design and optimize. Investigated in this thesis is the powertrain system which consists of five ECUs and one CAN bus, which has a very high communication load. The hypothesis is that centralization can simplify the system and concentrate all information on a single point creating a Centralized Database. The goal of this master thesis is to investigate the value and cost of a centralized database concept, by focusing on five research questions.Investigating the options in scope, the Coordinator seems to be the best candidate since it already has a centric role in the system. By introducing a centralized database in the system the communication model can change to a simpler model that the COO is subscribed to all ECUs and all ECUs are subscribed to the COO. An ECU Database is needed to include all information on all ECUs and the optimization parameters in order to create and select profiles that designate the communication parameters for the ECUs in the network. That means that all messages should be included in the centralized database since the communication model requires all communication to go through the COO. Hence the packing of the signals can change to benefit the message payload utilization which now is low in many of the messages. Additional process at the centralized database can evaluate and derive the best values to transmit to the ECUs. Such an implementation will raise the bus load due to the extra messages and transmissions that are required as well as extents the message transmission time due to the extra process time at the centralized database. Simulations using the SCANLA software, which calculates the worst case response times, were performed on the original design given by Scania under 500 kb/s CAN bus bandwidth. The results showed that the bus load was up to 90%. Then including the extra messages and transmissions that would result from introducing a Centralized database, the bus load was above 100% thus the simulation failed. This proves that such an implementation is not feasible under these parameters. Raising the CAN bus bandwidth, the following simulations were under 750 kb/s and 1 Mb/s that resulted at 90% and 67% CAN bus load respectively, including all extra messages and transmissions, which proves more promising for such an implementation.