Declarative Logic-Based Pareto-Optimal Agent Decision Making

Area of Science:

• Artificial Intelligence
• Multi-agent Systems
• Logic in Computer Science

Background:

• Autonomous agents require selecting action sets adhering to behavioral constraints.
• Feasible Status Sets (FSS) represent action sets satisfying deontic logic constraints.
• Existing methods face challenges when multiple FSSs exist, hindering agent action selection.

Purpose of the Study:

• To introduce the novel concept of Pareto-optimal Feasible Status Sets (POSS).
• To develop methods for identifying and selecting optimal action sets for autonomous agents.
• To address the challenge of multiple feasible solutions in agent decision-making.

Main Methods:

• Formalization of Pareto-optimal Feasible Status Sets (POSS).
• Development of an algorithm to find POSS.
• Design of specialized, more efficient algorithms for monotonic or anti-monotonic objective functions.

Main Results:

• Demonstrated that checking for POSS is co-NP-hard.
• Developed a general algorithm for finding POSS.
• Proposed efficient algorithms for specific objective function types, validated through experiments.

Conclusions:

• The POSS concept provides a robust framework for autonomous agent decision-making.
• The developed algorithms offer effective solutions for identifying optimal action sets.
• Future work will explore strategies for managing multiple Pareto-optimal solutions.