- 期刊
Study of the Sheep-dog Game Problem based on Differential Equations and Neural Network Deep Learning Models
摘要
In this paper, by analyzing the old sheep-dog game problem, establishing differential equation model, neural network deep learning model, hierarchical analysis method, decision tree and other models and algorithms to analyze the problem respectively, and using matlab to solve the problem to get the solution of the problem about sheep-dog game. According to the restrictions of the article for the model, the differential equations of motion of the sheep and the equations of motion of the dog are established respectively, and to simplify the calculation, the model is analyzed by using the sheepdog isotropy in the special make case, and finally the optimal strategy of the dog is the isotropy with the sheep, so that the dog chases the sheep with the optimal strategy. On this basis, it is found that it is more favorable for the dog after the sheep are fully isotropic, and a critical model between the radius ratio and the speed ratio is established. When training the sheep, finally the sheep to the dog's movement route is analyzed based on the kinematic optimal decision knowledge, the movement direction can be divided into two categories combined with the principle of neural network deep learning, and finally can optimize the trend of machine learning and the probability of success in the learning process.