| Cyber-Physical System(CPS),which interconnects physical facilities that cooperate with each other through wireless networks to achieve the goal of more complex functions and more efficient performance,is currently used in various fields of human life.Embedded systems have gradually become indispensable components of CPS due to their superiority of small size,low cost,low power consumption,easy customization,and multiple functions such as data collection,information processing,and wireless communication.Generally,system applications can be described as multiple subtasks that work collaboratively,and the data correlation between tasks is represented by a Directed Acyclic Graph,and then these subtasks are appropriately assigned to embedded system nodes with proper voltage/frequency,which can improve system performance.In addition,task reliability is a key factor affecting system performance,and nodes working in a network environment will also face reliability problems during data transmission and data processing,which can be improved by increasing node frequency or task duplication.At the same time,energy consumption and time delay are also important indicators to measure system performance,which can be adjusted through frequency selection and task mapping.This paper mainly studies the task mapping problem of the j oint optimization of energy consumption,delay and reliability of embedded nodes,which aims at optimizing energy efficiency under the premise of ensuring reliability and real-time constraints.The main work and contributions of this article are as follows:1)For the mapping problem of related tasks in embedded CPS,energy optimization under the premise of ensuring real-time and reliability is considered.In order to improve the reliability of tasks,a task duplication mechanism is introduced.When task reliability is lower than a certain standard,task duplication or Dynamic Voltage/Frequency Scaling(DVFS)can be used to improve its reliability.The task assignment problem is constructed as an Integer Non-Linear Programming(INLP)problem and then linearized by means of variable substitution to find the optimal solution.At the same time,in order to adapt to the fact that the large scale of the problem in practical applications makes it very time-consuming to find the optimal solution,a heuristic algorithm that can quickly solve this problem is proposed.The simulation verifies that the algorithm can greatly reduce the solving time while ensuring that the gap between the optimal and approximate solution is within an acceptable range.2)Since the routing between nodes can be completed through multiple paths in practical applications,this article studies the reliable task mapping problem based on multi-path routing,which is constructed to an energy optimizing task mapping problem under the constraints of real-time and reliability based on multi-path routing.In addition,two solutions are proposed for this optimization problem:one is to transform the task mapping INLP problem into an Integer Linear Programming problem and then use a linear solver to obtain the optimal solution,the other is to design an heuristic algorithm to find approximate solution.The simulation verifies that the multi-path task assignment problem is better than the single-path case in terms of the feasibility of the problem and the quality of the solution.In addition,in terms of solving time,the heuristic algorithm proposed in this article is very time-saving,even if the problem is large can it ensure a rapid solution.3)The multi-core system of embedded systems is developing rapidly,and this article also studies the reliable task mapping problem in embedded CPS based on homogeneous multi-core processors,and balances the energy consumption on each node as much as possible under the constraints of real-time and reliability.In addition,in order to assure task non-preemption,the task execution start time needs to be determined at the same time as task mapping,and this article considers two factors that affect the task execution start time:one is the execution end time of tasks assigned to the same core,and the other is the end time of the data transfer from the dependent task.In addition,this article designs a heuristic algorithm based on stratifying task graph for the reliable task mapping problem on embedded multi-core systems to reduce the solving time.Simulation experiments show that the task mapping problem based on multi-core processors is significantly better than single-core processors in terms of the result quality and feasibility. |
PDF Full Text Request