Research On Multi-core Static Task Allocation Technology For Unmanned Express Car In Park

Under the background of the high cost of manual delivery in terminal logistics and the development of intelligent vehicles,the application of unmanned express car to complete the "last mile" delivery has become a new attempt to replace manual delivery.In order to complete autonomous driving and delivery tasks,unmanned delivery vehicles often use multi-core processors to run programs.Since the real-time nature of the program affects the driving safety and user experience,it is very important to make reasonable use of multi-core resources to achieve high parallelism and low latency execution of system programs.In this paper,the system program of unmanned express car is developed by using Simulink modeling method for the campus environment.Aiming at the problem of unbalanced utilization of processing cores in Simulink’s implicit task division,which cannot meet the real-time performance of the system,by studying the system task characteristics and the multi-objective static task allocation algorithm,the optimal assignment of the unmanned express car program is realized on a low-cost processor.In the aspect of system task characteristic analysis,by analyzing the data dependencies of each operation process of the system,a directed acyclic graph model of the system tasks of the unmanned express car system is constructed to realize the visualization of complex programs,and then the system tasks are decomposed into computing nodes and communication nodes.Aiming at the problem that the operation of each node is affected by the complex architecture of multi-core processors,it is impossible to accurately predict the execution time,through static analysis of the execution process of the calculation and communication nodes,and considering the worst execution situation of the node,an indirect estimation method of execution time based on the maximum instruction execution cycle and the maximum number of bytes occupied by communication data is proposed.When assigning tasks with the goal of making full use of multi-core resources to meet the real-time performance of the system,aiming at the problem that the distribution targets cannot be combined and difficult comparison of allocation schemes due to the different dimensions of the analysis results of the communication and computing nodes and the functional relationship with time cannot be determined,a solution that evaluating the allocation scheme independently from multiple dimensions,and then using the dominance relationship to compare is proposed.On this basis,by comparing and analyzing the existing task allocation methods and conducting related experiments,it is determined that the multi-objective particle swarm algorithm is used to solve the task allocation scheme,so as to generate a Pareto front composed of non-dominated allocation schemes.The distribution scheme is tested on the existing platform,and the results show that the task allocation optimization scheme can balance the utilization of each processing core and meet the real-time performance of the system,finally the effectiveness of the task allocation optimization scheme on the low-cost computing platform is verified through offline and real vehicle tests.