摘要
现实世界中,动作的执行通常都要耗费一定的代价,且由于外界环境的干扰,动作执行后的结果具有不确定性。针对这一问题,对不确定状态转移系统的动作赋予权值,使用概率分布表示状态转换的随机性,提出了强循环规划解的期望权值,并且设计了求最小期望权值强循环规划解的方法。该方法的主要思想是使用深度优先搜索求出规划问题的所有强循环规划解,再将强循环规划解分别转换成以状态到目标状态的期望权值为变元的线性方程组,最后使用高斯消元法解方程组,从而找出最小期望权值强循环规划解。
In real world,the action＇s execution often takes a cost.Due to the interference of the external environment,the result of an action＇s execution is uncertain.To solve this problem,we added weight to the action in system,used a probability distribution to show the stochastic state transition.A method of solving strong cyclic planning with minimal expectation weight was designed based on the proposed concept of expectation weight for strong cyclic planning.Mainly,this method applies depth-first search for getting all strong cyclic plannings.Then,it uses Gaussian elimination to slove the problem after converting the plannings into linear equations with variables in expectation weight.
出处
《计算机科学》
CSCD
北大核心
2015年第4期217-220,257,共5页
Computer Science
基金
国家自然科学基金(61272295
61105039
61202398)
湘潭大学智能计算与信息处理教育部重点实验室
湖南省重点学科建设项目(0812)资助
关键词
不确定规划
概率分布
最小期望权值强循环规划解
深度优先搜索
高斯消元法
Uncertainty planning
Probability distribution
Strong cyclic planning with minimal expectation weight
Depth-first search
Gaussian elimination
