Weighted Observation Information Reduction In Uncertain Planning

In uncertain programming based on model checking, but because of the uncertain factors, some of the state in performing an action after will reach some random state, only to look at the state of these to identify a specific position. So it is necessary to look at the planning process, and these observations are observation information. With the development of information society, the scale of the problem is more and more big, the information involved is becoming more and more widely. At the same time the scale of the planning model is also very big, including observation of the number and type of information is also considerable. The vast majority of real programming problem, observation information to obtain the required price is different, and in the execution of a plan solution, not all the observation information is meaningful, therefore in order to reduce the cost in the process of execution and the observation of the large amounts of weighted value information about Jane is very important.This thesis absorbs the previous scholars quite observation information with weight reduction aspect of study, systematically the basic idea of reduction of observation information, distinguish the arriving according to uncertain movement state and uncertain movement itself the pros and cons of two kinds of uncertainty sets are obtained and applicable conditions. According to the above problem, for the first time on the characteristics of uncertainty in the field of single agent and agent weighted values of observed information reduction problem has made a study respectively.In this thesis, the research is in the planning solution and all of the known observation information were observed under the situation of. Under the condition of single agent, design and defines all the need to distinguish between the state on, integrating the concept of the greedy algorithm is that each search choose observation variable in a specified order, and the branches in the process of the search application threshold method, reduce a lot of unnecessary search, finally obtaining the total cost minimum observation sets, achieved the purpose of reducing transaction cost. The experimental results show that the algorithm can efficiently get involved in the weighted value of the optimal observation set, to reduce in the implementation of the planning overhead contribution is obvious.In the case of multiple agent, will extend the concept of the single agent to apply to more agent; Among agent, although are logically independent, but by sharing knowledge, task and so on to achieve collaborative work, this is the main difference of single agent. Designed an algorithm based on the productive capacity of the table, the comprehensive combination of heuristic function and greedy thought, the experimental results show that the algorithm is a good way to observation of the agent case more information about Jane.