Research On Reliability Goal Assurance Of Functional Safety Towards The New Generation Automotive Electronic System

Due to the inherent heterogeneity,interaction,and diverse nature of automotive embedded systems,it requires the joint and tight interaction between the cyber(networked computational)and physical worlds,making the new generation automotive electronic systems a typical automotive cyber physical system(ACPS).As a highly safety-critical system,ACPS has system-level safety and reliability requirements.At the same time,the functional safety proposed in response to the risk of automotive functional failures and the functional safety standard ISO 26262 specifically targeted at automotive electronic systems have raised the safety requirements of the automotive electronic systems to a new level.Therefore,assuring the reliability of the automotive electronic systems is a long-term research focus and difficulty.Based on this,from the perspective of functional safety of the new generation automotive electronic systems,this paper aims to assure the reliability goal of a distributed automotive function based on the DAG scheduling model,and proposes effective reliability goal assurance methods under non-fault-tolerant and fault-tolerant manners,respectively.The main works are as follows:First of all,aiming at the complex system structure of the new generation of automotive electronic systems and the the complex relationship between the tasks of distributed automotive functions,the heterogeneous computing units and distributed automotive functions should be abstracted and described by more accurate models.At the same time,for the reliability goal scheduling problem based on DAG model,we categorize and summarize the scheduling methods and elaborate the progress of reliability goal scheduling research.Secondly,we propose a method called Reliability Goal Assurance using Geometric Mean(RGAGM)under non-fault-tolerant manner.The algorithm introduces the mathematical definition of the geometric mean,and uses the pre-allocation mechanism to transfer the reliability goal of function to that of each task.On the basis of satisfying the reliability goal of each task,we assign the task to the ECU with the minimum resource consumption cost.The RGAGM algorithm utilizes the central tendency of the geometric mean,thereby solving the problem that the existing algorithms cause a waste of resources for pre-assigning maximum reliability values to unassigned tasks,and fail to satisfy the reliability goal of function for pre-assigning minimum reliability values to unassigned tasks.Finally,we propose a method called Geometric Mean-based Fault-tolerant Reliability Pre-assignment(GMFRP)under fault-tolerant manner.This algorithm is designed to use fault-tolerance mechanism to minimize the response time the system while assuring the reliability goal of a distributed automotive function.GMFRP also introduces a mathematical geometric mean and defines pre-assigned reliability value for each unassigned task,and then confirms the reliability goal of each task.Then,GMFRP iteratively assigns the backups of each task to the ECU with minimum earliest finish time until assuring the task's reliability goal.The GMFRP method utilizes the central tendency of the geometric mean to solve the problem of unbalanced reliability values pre-assigned to the high-priority and low-priority tasks.Experiments on the real-life automotive function and the randomly generated distributed automotive functions by the task graph generator show that compared with the existing algorithms,the above algorithms can meet the certificated reliability goal of a distributed automotive function and optimize the function's target performance metrics.Therefore,the two methods are effective reliability goal assurance method.