A Research On Solution Methods For Static Resource-task Assignment Problems In Hardware Environment Of A Microcontroller Unit

With the development of artificial intelligence,autonomous AI devices will be of great importance in the application in the future.One of the essential abilities of AI devices is that they can automatically assign limited resources to designated tasks to maximize efficiency.Focusing on the characteristics of some small automatic AI products that are inconvenient to equiped with regular computers,the thesis conducts research on the algorithm of the static resoure-task assignment problem with microcontroller unit.Focusing on the static resoure-task assignment problem in specific projects,the thesis establishes two kinds of mathematical models for different target functions.Based on the survey of the application of microcomputer,the thesis chooses STM32F103 as the solver of this research.Three kinds of basic algorithms are improverd,and two kinds of international standard cases are selected to test algorithms in the thesis.By comparing the algorithm solutions to the optimal solution according to the current literature,the thesis concludes that the improverd algorithms are effecttive.The research in the thesis contains three results.Firstly,the thesis proposes the pruning enumeration algorithm based on direct enumeration.Through branch enumeration and the search for the optimized loose upper bound,the result of pruning has been improved and shows that the algorithm is suitable for the exact solution of small-scale problems.Secondly,the thesis proposes a modification strategy for exchange based on the result of the greedy algorithm.Optimization has been carried out through loop interchanges of initial results and the effects of two rounds of loop exchanges and that of three rounds of loop exchanges are compared.The result shows that the algorithm is suitable for the approximately optimal solution of large-scale real-time problems.At last,a greedy exchange algorithm based on the simulated annealing algorithm.Simulated annealing decision has been conducted based on the modified exchange algorithm of greedy result and the global optimization performance of the algorithm during exchange has been improved.The result shows that the algorithm is suitable for the approximately optimal solution of low real-time problems.The solution proposed by the thesis could be applied in the task assignment of resources when the hardware is restrained.The tests of the research cover all sizes of the international standard test sets and suitable solutions of certain hardware conditions could be selected according to the size.The modified algorithms could compute accurate results rapidly and could serve as an important reference to the research on solutions to task assignment of resources with microcontroller unit.