This paper focuses on a special type of course-of-action.Specifically, performing study on planning with the existence of qualitative preferences and functions on the actions and owns the inner inconsistence.The course-of-action that is taken into consideration is called ‘task-level’ course-of-action(COA), with abstracted action as basic element.The qualitative preferences in discussion include static preferences and temporal preference.The ob-jective of planning is a COA plan with satisfaction.Firstly, a unified formulated description is established for con-straints and preferences, based on which an algorithm for COA planning is developed.Furthermore, computational argumentation is utilized to exclude inconsistence in the set of preferences, to maximize the user’ s satisfaction for COA planning.The planning framework based on qualitative deduction is an effective add-in for conventional plan-ning scheme based on quantitative computation.The property of preference-decoupling makes itself adaptable to ap-plications in different domain.A case study on scheduling responsive imaging satellites is proposed to demonstrate the effectiveness of the scheme.%关注了一类典型行动序列,研究如何在动作集合上存在定性偏好,且偏好集合存在不一致性时开展规划。所考虑的行动序列问题称为任务级COA,以抽象层次的动作为基本要素,所考虑的定性偏好包括静态偏好和时序偏好,所讨论的规划目的是获得最大满意度的COA方案。首先建立了偏好与约束的归一化形式描述,在此基础上形成了COA方案设计算法;进一步,使用计算辩论技术排除偏好集合中的不一致性,形成用户接受度最高的COA方案。文中建立的以定性推理为基础的规划框架,实现了偏好解耦,能够适应不同的领域问题,是以定量计算为基础的传统规划算法的有效补充。通过快速响应卫星成像的COA案例,演示了算法的可行性。
展开▼