Planning in a commonsense framework
作者:Carlos Pinto-Ferreira, João Pavão Martins
摘要
The ability to make plans is a necessary condition for general intelligent behaviour. To plant everyday activities, an agent has to be capable of representing and reasoning about change. However, as it is impossible exhaustively to represent conditions for performing actions and their corresponding effects, artificial systems have to reason without complete knowledge. On the other hand, the fact that in the macroscopic world most characteristics of things persist over time has to be introduced in the formalization of change, to avoid stating explicitly what characteristics of the world are unaffected by every action. These are the main difficulties in planning in a commonsense framework, as far as formalization aspects are concerned. However there are others, of a more pragmatic nature (e.g. managing search). As a motivation, an example of formalization is given together with a brief explantion of the most known approaches to commonsense reasoning: circumscription, default theories and autoepistemic logic.
论文关键词:Neural Network, Artificial Intelligence, Complex System, Nonlinear Dynamics, Main Difficulty
论文评审过程:
论文官网地址:https://doi.org/10.1007/BF00168959