Planning and Choosing: Augmenting HTN-Based Agents with Mental Attitudes

Abstract

This paper describes a new agent framework that fuses an HTN planner, through its underlying conceptual model, with the mental attitudes of the BDI agent architecture, thus exploiting the strengths of each. On the one hand, the practical and proven ability to reason about actions that is the strength of HTN planning fleshes out the option generation function in the inference loop of the BDI model; on the other hand, the mental attitudes make explicit the knowledge that plays an essential role in plan selection, an important aspect that is not considered in the traditional formulation of the planning problem. The result is a coherent framework that allows for the design and implementation of activitycentric rational agents.

Construction and Application of the Operational Plan Ontology Model

“…International Conference on Communications, Information Management and Network Security (CIMNS 2016) the relationships among the various elements are included. Axiom: The inspection and judgment rules and operation rules in various fields are included [7].…”

I-Room: a Virtual Space for Intelligent Interaction

“…The incorporation of BDI into the <I-N-C-A> model, with beliefs corresponding to constraints and intentions to nodes, and with desires manifest in the decision-making processes, both allows a processcentric account of agent systems and provides a model for implementing and deploying rational intelligent agents within this system. 6 • Shared or overlapping ontologies and associated vocabularies are the basis for formal and informal communication and mutual comprehension among participants. The use of intelligent systems and services during the collaboration will determine the extent to which these ontologies and their use need to be formalized and made explicit.…”

Learning hierarchical task network domains from partially observed plan traces

“…HTN planning has been successfully used in a number of real-world applications [19,61,62,13]. It requires as input both methods that describe how to decompose a task into subtasks and the preconditions under which they are applicable and action models that describe how actions alter the world [66,45].…”