Abstract
This paper proposes a plan structure for excution in a hostile dynamic world. A plan in a dynamic world can not be specified completely in advance as the world is subject to unexpected changes anytime. Agents do not have sufficient time to plan in a hostile world. Thus, problem solving in a hostile dynamic world mainly consists of executing pre-designed plans selected on-the-fly from a library of abstract plans. The abstract plans in a hostile dynamic world have to have several features that plans in other domains (such as in RAP [Firby 95]) may not have. In this paper, we propose a recipe structure to encode such plans for execution in hostile dynamic worlds. Agents executing such recipe structures translate the recipe into plans (as mental attributes), and execute them. We also present the details of an implementation of this execution strategy where we translate the recipe incrementally into a progressively abstract plan structure.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have