Research On Several Key Technologies For Network Architecture Of The New Generation Of Automotive Electronic System

Automotive electronic system has become the main driving force of innovation and competition for car industry, and it is also one of the most important approaches for developing new car models and improving a car’s performance. In recent years, the stricter economic and societal requirements that are put forward on car make it develop from a close and weakly coupled system to an open, highly networked, extremely heterogeneous and strongly coupled new generation of automotive electronic system. But the rapid increasing complexity brings big challenges for the design of the new generation of automotive electronic system, therefore how to gurantee that the system can meet the requirements from the real-time and reliability and can be implemented efficiently, and then keep the safe and reliable operation and decrease the system’s development cost has attracted much attention from the car industry.Network is the basis to realize the interaction and cooperation among the ECUs and the feedback control within the car. Along with the complicating of the new generation of automotive electronic system, its network architecture that shows the nature of complexity, heterogeneity and hierarchy becomes the main source of system’s complexity. And also considering the fact that the new generation of automotive electronic system is a cyber-physical system in essence, network is promoted to the same status with that of computation and control, this paper proposes to design the new generation of automotive electronic system with Automotive CPS’s point of view. It means that the main focus should be given to the problems existing in the design of the network architecture, which are related with the real-time performance and extensibility. CAN and FleRay, which are the two most important real-time networks that are employed in the critical subsystems relating to the driving and traffic safey of the car, are chosen as the research object, and beginning from the message scheduling problem, some optimization algorithms are proposed to realize the message’s schedulability analysis and schedule optimization, and optimization on system’s resource utilization. The target is to gurantee the safe and reliable operation and the resource-efficient implementation of the automotive electronic system, so that the challenges that are caused by system’s complexity can be relieved to some extent for the new generation of automotive electronic system.Based on the problems that are mentioned above, this research is mainly focused on the following aspects:1. Gateway-interconnected CANs is a key component of the network architecture for the new generation of automotive electronic system, and it is also an important manifestation of its characteristics of hierarchy. But the current research about the CAN message’s schedulability analysis is restricted to the case of the single CAN, therefore this research proposes a safe schedulability analysis algorithm for messages that are transferred on the gateway-interconnected homogeneous CANs. The proposed algorithm is based on the schedulability analysis algorithm raised for messages in single CAN, and combining two new concepts that are "Busy Sequence" and "The Minimum Distance Constraint" with the proposed two arriving order searching algorithms, WCRT of the non-gateway type messages can be analyzed. For gateway type messages that need to be scheduled in two CANs successively, this research takes a "first divide and then transform" approach to transform their schedulability problem into the same situation with that of non-gateway type messages. Through the theory proving and the comparison analyses based on the simulated message set and the real message set provided by automobile company, it verified the safety and the effectivess of the proposed algorithm.2. Heterogeneity is a main feature for the network architecture of the next generation of automotive electronic system, and it is not only reflected in the employed different kinds of networks such as CAN and FlexRay, but also in the gateway-interconnected CANs with different bandwidths. For example, there is gateway-interconnected high-speed CAN and low-speed CAN in BMW7. To support the integration of heterogeneous networks and guarantee the safe and reliable operation of the electronic functions communicating with gateway-interconnected heterogeneous CANs, this paper extends the proposed schedulability analysis algorithms for messages that are transferred on the gateway-interconnected homogeneous CANs in the aspects of the impact of the bandwidth’s heterogeneity on message’s execution time and on the analysis of the interfering delay for messages, and then the schedulability analysis for messages transferring on the gateway-interconnected heterogeneous CANs can be finished. Through the comparison analysis that is based on the real message set provided by automobile company, it verified the effectivess of the proposed algorithm.3. Aiming at the disparity between the high extensibility requirement from the network architecture’s design for the new generation of automotive electronic system and the low extensibility support of the FlexRay schedule for its static configuration approach, this research proposes an extensibility-aware message scheduling algorithm for the static segment of FlexRay so that the uncertainty from the signal’s size increase can be accommodated. Based on the extensibility requirement’s analysis for the message scheduling result of the static segment of the FlexRay, it first gives an uncertainty model to formally describe the growth of signal’s size, and it also gives the definition about the extensibility, the extensibility policy and the extensibility evaluation indexes. And then, two basic message scheduling algorithms that are suitable for signal sets of different scales are proposed for signal packing, and combining the basic message scheduling algorithm with the two extensibility improving steps that are contributing to the distribution improving of the left time spaces inside the assigned slots, it realized the accommodation to the size increase of the signals included in the message scheduling result. The comparison analyses that are based on the simulated signal sets and the real signal set provided by the automobile company showed the effectiveness of the proposed algorithm on improving the slot cost and modification cost.4. Considering the need for improving the real-time performance for the network architecture of the new generation of automotive electronic system, and the complexity of its network parameters’configuration and the need to integrate the design of the computation system and the network system for FlexRay contributing to its time-triggered property, this research proposes a new message scheduling algorithm that can optimize the slot assignment to improve the functional-level’s real-time performance from the system level’s point of view. Starting from the viewpoint of CPS with the design integration of the computation system and the network, it first gives a scheduling unit model to formally describe the synchronization relation between the task and the signal, and with the aid of the schedulability analysis theory, the feasible slots for each scheduling unit and each ECU can be analyzed successively. Next, according to the average WCRT of all scheduling units that are included in the ECU when it assigned each feasible slot, there is a two dimensional table that includes the above analyzed average WCRT is constructed. Finally, it uses the optimal brand and bound algorithm to search inside that table to find the slot assignment result that can realize the minimization of the sum of the exectution paths’end-to-end WCRT included in the corresponding electronic functions. It firstly proved that under the given assumptions, the proposed slot assignment method realized the minimization of the sum of the end-to-end WCRT of all execution paths incuded in the system. And then through the comparison analyses with the other slot assignment methods based on the simulated and the real signal set and task set, it furtherly verified the effectiveness of the proposed slot assignment method on improving the function level’s real-time property.