Research On Reliability-constrained Multi-objective Testing Resource Allocation

As software systems become larger and more complex,users' requirements for software reliability are also increasing.In actual software testing,the software project manager should not only consider the user's requirements for software reliability,but also take into account the testing resources(usually expressed as testing time)and testing costs.In recent years.,how to allocate the limited testing time effectively to optimize system reliability,testing cost,and testing time at the same time has become a research hotspot in the field of search-based software engineering.The traditional multi-objective testing resource allocation is mostly limited to testing time constraints without considering the user's requirements for software reliability,resulting in many low-reliability testing resource allocation solutions.This not only wastes a lot of computing costs,but also has no reference value for the software project manager.Therefore,from the perspective of users' requirements for software reliability,this dissertation studies the reliability-constrained multi-objective testing resource allocation problem.The main research contents of this dissertation are as follows:(1)The research background and significance of testing resource allocation are briefly discussed,and the previous studies on the multi-objective testing resource allocation problem are analyzed in detail.(2)The basic knowledge of the reliability model and cost model in the testing resource allocation problem is introduced.In addition,the multi-objective optimization problem,the evaluation criteria of the solution,multi-objective evolutionary algorithms,and the existing constraint handling techniques are discussed.(3)From the perspective of users' reliability requirements,a mathematical model of Multi-Objective Testing Resource Allocation Problems(MOTRAP)with reliability constraints is constructed.Then,the new lower bounds of the testing time invested in each module are deduced on the basis of the pre-specified reliability.This theoretical results can clarify which modules are more economical for the growth of system reliability.In addition,a reliability-constrained multi-objective testing resource allocation algorithm,called NSGA-II-RC,is developed based on the Nondominated Sorting Genetic Algorithm II(NSGA-II)and several enhanced constraint handling techniques.The proposed constraint handling techniques are used in population initialization,crossover,and mutation,and can theoretically ensure that infeasible solutions can be repaired to be feasible with a satisfactory reliability.The comparative experimental results show that the proposed model and algorithm can make the population quickly evolve in the feasible region of the desired reliability,and can provide the software project manager with more and better testing resource allocation schemes.To meet with the different needs of users and the software project manager at different testing stages for system reliability,a multi-stage multi-objective testing resource allocation model with dynamic reliability constraints is constructed.Then,a dynamic reliability-constrained multi-stage multi-objective testing resource allocation algorithm,named MS-DRC-GDE3,is developed based on the third generation generalized differential evolution(GDE3),parameter estimation,population re-initialization,weighted normalized optimal solution selection.The comparative experimental results show that the proposed model and algorithm can dynamically adjust the allocation of testing time in different testing stages according to the previous testing results of the module,which can provide a certain reference for the software project manager and facilitate more informed planning of testing stages.