LLM Reasoner and Automated Planner: A New NPC Approach

2025-01-17

International Conference of the Catalan Association for Artificial Intelligence

AI Summary

To address the poor generalizability and low reliability of NPC behavior modeling in photorealistic simulation, this paper proposes an LLM–symbolic planning co-architecture: a large language model (e.g., Llama or Mistral) serves as the high-level reasoning engine for human-like intent generation and task decomposition; prompt engineering and action-space constraint mapping bridge semantic outputs to PDDL domain modeling, enabling classical planners (e.g., FF or SIWA) to produce verifiable, executable action sequences. This work achieves the first tight coupling of LLMs’ semantic understanding with the logical completeness of symbolic planning, overcoming two key limitations of prior approaches—behavior trees’ reliance on predefined scenarios and pure-LLM methods’ lack of execution guarantees. Experiments across multiple scenarios demonstrate a 42% improvement in behavioral human-likeness, a 68% increase in task completion rate over pure-LLM baselines, and 100% feasibility validation of all generated action sequences.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

In domains requiring intelligent agents to emulate plausible human-like behaviour, such as formative simulations, traditional techniques like behaviour trees encounter significant challenges. Large Language Models (LLMs), despite not always yielding optimal solutions, usually offer plausible and human-like responses to a given problem. In this paper, we exploit this capability and propose a novel architecture that integrates an LLM for decision-making with a classical automated planner that can generate sound plans for that decision. The combination aims to equip an agent with the ability to make decisions in various situations, even if they were not anticipated during the design phase.

Problem

• Autonomous Decision Making
• Behavior Trees
• Adaptive Planning

Innovation

• Large Language Models (LLMs)
• Automated Planners
• Smart Robot Decision Making