Research On The Task Allocation Methods In Automotive Electronic Systems

With the development of automotive electronics, the number of automotiveelectronic control unit (ECU) in a car is increasing, modern high-class automobilesemploy up to70ECUs, which interconnect into a complex distributed system byseveral bus systems. The ECU communicates with each other, working in the mannerof coordination to complete the tasks and functions. According to the designmethodology of AUTOSAR, the function of the ECU is statically configured by taskallocation at the time of the development of the software, which is a combinatorialoptimization problem that has been proved to be Np-hard. How to design feasible andheuristic task allocation methods that can optimize the system reliability and keepload balancing has become a research focus.According to the characteristics of automotive electronic systems and the highlynetworked, intelligent development trend, the paper explored the task allocationmethods target to optimize system reliability and keep load balancing from the twoaspects of static and dynamic task allocation respectively. The main work is asfollows:Firstly, summarized the architecture of AUTOSAR and the methods of softwaredevelopment, then an abstract model of task allocation was proposed by analyzing thecharacteristics of software development under the AUTOSAR environment, whichwould be used to study the problem of static and dynamic task allocation, additionally,the restraints that should be satisfied by the task allocation problem were defined.Secondly, aimed to optimize the reliability of the system, an improved particleswarm optimization algorithm was proposed. By redesigning the particle’s position,speed and fitness function based on the basic particle swarm optimization algorithm,the combinatorial optimization problem is successfully solved through the algorithm.Meanwhile, by embedding a new mutation operator in the iteration process, itovercomes the deficiencies of the PSO algorithm easily falling into local optimum tosome extent. Experiments show that this method has a good effectiveness andfeasibility.Finally, studied the problem of dynamic task allocation to ECUs, based on theanalysis of task execution model and the definition of QoS about the node resource,the paper proposed a dynamic task allocation method based on the QoS of the node’s resource. Experiments show that this method can nearly keep load balancing acrossECUs in automobile, which will help improve the system resource utilization andreliability.