Simulation Analysis of Multi-level Intention Estimation Agent Behavior Strategies in a Collaborative Task

Abstract

In multi-agent behavior strategies, it is often assumed that agents estimate others’ intentions in a hierarchical manner, known as multi-level intention estimation. This study analyzes the effects of such strategies through simulation using a cooperative task involving a small group of agents and multiple targets. In the proposed model, each agent recursively estimates intentions―for example, “What is the other agent intending? ”(first level) and“How does the other agent perceive my intention? ”(second level) ― and selects its actions accordingly. By varying both the behavioral strategies and the depth of intention estimation across multiple scenarios, we quantitatively evaluate their impact on coordination success rate and task completion time.