摘要
通过人工神经网络,实现了短道速滑仿真系统中智能体(agent)的决策过程。将智能体的当前状态作为神经网络的输入,智能体要采取的动作作为神经网络的输出,从而实现智能体的决策。神经网络的训练采用有监督学习的误差反向传播(BackPropagation)算法,样本采集自仿真系统运行时使用者的输入。通过所述方法,能够保证智能体在高速的比赛过程中较少犯规,并可以达到在体能受限的情况下取得较优秀成绩的目的,还能够模拟某一特定运动员的滑行特性和决策习惯。
Decision-making of Agent in short track speed skating is resolved through artificial neural network, by the method of making the state of agent as the input of neural network aod the action as the output. The back propagation supervised learning algorithm is chosen as the way of training neural network, and samples come from the input of user when the system is running. Through the method in this thesis, the agent is hardly to foul and can get good grades when the physical fitness is limited and even simulates sliding properties and decision-making habit of the particular athlete in short track speed skating.
作者
张时铭
黄剑华
李长朋
ZHANG Shiming, HUANG Jianhua, LI Changpeng (School of Computer Science & Technology, Harbin Institute of Technology, Harbin 150001, China)
出处
《智能计算机与应用》
2011年第1X期28-31,34,共5页
Intelligent Computer and Applications
基金
基金项目:国家科技支持计划项目资助(2009BAK57807).
关键词
BP神经网络
短道速滑
仿真系统
智能体
决策过程
BP Neural Networks
Short Track Speed Skating
Simulation System
Agent
Decision-making