Research On Multi-skilled Staff Scheduling Optimization For New Product R&D Project Portfolio

With the development of science and technology, people have higher and higher requirements for the function and performance of new products, which enhances the complexity of new product research and development activities, and results in that the traditional theory of single project management cannot meet the requirements of management for complex R&D activities in an enterprise. Therefore adopt portfolio theory and method to manage new product research and development activities become a new trend for enterprises. In new product development project portfolio, R&D staff is a key factor for the success, and the existing project scheduling theory on staff scheduling problem for project portfolio is very lack. Further, the development of computer and information technology has greatly promoted human learning efficiency and ability, which makes more and more multi-skilled staff in enterprises as a common phenomenon. Therefore, it is meaningful in theory and valuable in practice to research the problem of multi-skilled staff scheduling for new product R&D project portfolio.Taking the IT industry as the background, this thesis researches the problem of multi-skilled staff scheduling for new product R&D project portfolio.The problem of multi-skilled staff scheduling in new product R&D project portfolio optimization belongs to the crossing field between the resource-constrained project scheduling problem and skilled staff scheduling problem, and it is a type of multi-skilled project scheduling problem. Research on the problem abroad is at the initial stage and domestic research is still relatively rare. Due to the fact that a new product R&D project portfolio usually contains multiple projects and R&D staff may grasp multiple skills with heterogeneous levels, the optimization model of multi-skilled staff scheduling may include many constraints and the solving space is very large. Besides that, optimization modeling has a tendency to be multi-objective and stochastic, and it increases the difficulty of problem modeling and solving. Since traditional heuristic algorithm for this kind of problem has a lower efficiency, adopting meta-heuristic algorithm to solve the problem is required from the theoretical and realistic perspectives.From the two dimensions of skill level changing or not and staff number changing or not, this thesis proposes four models of multi-skilled staff scheduling for new product R&D project portfolio, designs the solving algorithms according to the characteristics of different models, and tests the feasibility and validity of each algorithm by instances of new product R&D from IT industry. The main contents are as follows:(1)Aiming to the condition of invariable staff skill level and invariable staff number,this thesis builds the single objective 0-1 integer linear constraint programming model of multi-skilled heterogeneous staff scheduling for new product R&D project portfolio,and makes R&D cycle or R&D costs as the target respectively. It also designs a heuristic serial schedule generation scheme and from that it designs genetic algorithm with integer coding to solve the model considering that one emoployee is assigned to one skill required in each project. At the same time, it uses discrete particle swarm optimization (PSO) algorithm to compare the performances, and combines an instance to verify the feasibility and effectiveness of the model and algorithm.(2)Aiming to the condition of variable staff skill level and invariable staff number,this thesis builds the multi-objective mixed integer nonlinear constraint programming model of multi-skilled heterogeneous staff scheduling for new product R&D project portfolio. Considering two different models based on learning effects and learning-forgetting effects, the goals of the models are to maximize skill value increments, to minmize R&D cycle and to minimize R&D costs. Considering the maximization goal of skill value increments have an advantageous for talent team construction of enterprise. A rapid non-dominated sorting genetic algorithm (NSGA- II)and a multiple species Pareto ant colony optimization algorithm (P-ACO) are designed to find the solution of the two models respectively. The models and algorithms are verified through an instance. At the same time, performances of the two algorithms are compared, and results show that algorithm performances can meet the demand of decision making for medium scale problems.(3) Aiming to the condition of invariable staff skill level and variable staff number,considering random turnover, the thesis builds the stochastic multi-objective 0-1 integer linear constraint programming model of multi-skilled heterogeneous staff scheduling in new product R&D project portfolio. The expectation objecitives are to maximize the weighted sum of personnel training time, to minimize R&D cycle and to minimize R&D costs. According to the characteristics of the model, an adaptive pareto sampling algorithm (APS) based on markov monte carlo (MCMC) sampling technology and the NSGA-? algorithm is designed to solve the model, and an instance is used to verify the feasibility of the model and the algorithm.(4) Aiming to the condition of variable staff skill level and variable staff number,considering variable random turnover probability, this thesis builds the stochastic multi-objective mixed integer nonlinear constraint programming model of multi-skilled heterogeneous staff scheduling for new product R&D project portfolio. Considering that staff skill level changes with the effect of learning and forgetting, this thesis calculates the turnover probability of each employee at each time interval to simulate the turnover process of staff by linear interpolation method. The APS algorithm is designed to solve the model, and an instance is adopted to verify the feasibility of the proposed model and algorithm.