Dynamic multi-objective software project scheduling method based on Q learning memetic algorithm

Abstract

The dynamic multi-objective software project scheduling method based on the Q-learning memetic algorithm provided by the present invention comprises the following steps: (1) reading input information; defining optimization objectives and setting constraints; (2) initializing the parameters of the algorithm; The agent perceives the initial state in the project environment; determines the global and local search strategies in the static multi-objective memetic algorithm; (3) at the beginning of the project, generates an initial scheduling plan and generates a return value; (4) in the project implementation In the process, the rescheduling method is adopted; the agent perceives the current state of the project environment, updates the Q value in the state-action pair table according to the reward value, and determines the global and local search strategies in the dynamic memetic algorithm based on the selection mechanism; based on the dynamic memetic algorithm A new scheduling scheme is generated in the new environment, and a return value is generated. The invention can learn the characteristics of the project environment, and quickly and efficiently realize the dynamic scheduling task in the software project.

Description

technical field [0001] The invention relates to the technical field of software project scheduling, in particular to a dynamic multi-objective software project scheduling method based on a Q-learning memetic algorithm. Background technique [0002] Software project scheduling means that the project manager determines the sequence of tasks to be developed in sequence based on the estimated workload of the tasks, and allocates human resources to meet the constraints of task priority, task skill constraints, and software engineers not being able to work overloaded. Under the premise of optimizing the performance indicators of software projects. Software project scheduling is a multi-objective optimization problem, because there are multiple optimization objectives that need to be considered at the same time, such as cost, construction period, and software engineer satisfaction. [0003] In some emerging software development fields, such as cloud computing, mobile communication...

AI Technical Summary

Problems solved by technology

[0007] 1) Most of them only consider the static development environment, and they assume that all the information in the project is known in advance and fixed. Obviously, when the development environment of the actual project changes dynamically or there are uncertain factors, it is generated according to the static method. The schedule for is no longer applicable

[0008] 2) The processing method for multiple optimization objectives is relatively simple

[0009] 3) The performance of solving large-scale software project scheduling problems is low

When faced with a large-scale software project scheduling problem with many tasks and a large number of developers, due to the huge search space and the lack of a search and guidance mechanism for autonomous learning and effective use of problem characteristics, the existing scheduling methods only rely on fixed search operators to find Therefore, there are problems such as slow convergence speed, premature maturity, weak local search ability, etc., and it is difficult to effectively obtain the optimal scheduling solution

Images