PRUDENT: A Sequential-Decision-Making Framework for Solving Industrial Planning Problems

Planning and control are critical problems in industry. In this paper, we propose a planning framework called PRUDENT to address many common issues and challenges we are facing in industrial applications, including incompletely known world models, uncertainty, and very large problem spaces. This framework considers planning as sequential decision-making and applies integrated planning and learning to develop policies as reactive plans in an MDP-like progressive problem space. Deliberative planning methods are also proposed under this framework. This paper describes the concepts, approach, and methods of the framework.