An Accurate Intelligent Plan Library for Belief-Based Desire Prioritization to Intentions in BDIx Agents

Abstract

The emergence of 6G technology is poised to revolutionize wireless networks through advanced Device-to-Device (D2D) communications, offering ultra-reliable, low-latency, and
high-throughput connections. However, the increased complexity and density of 6G D2D networks necessitate autonomous, distributed, and dynamic decision-making capabilities to ensure efficient and reliable operations. Belief-Desire-Intention eXtended (BDIx) agents present a promising solution by enabling intelligent selection and execution of communication strategies based on real-time environmental assessments and internal states. This paper introduces an Plan Library designed to enhance BDIx agents’ ability to prioritize desires grounded in their beliefs, thereby facilitating robust intention formation tailored to the dynamic requirements of 6G D2D communications. Our approach
integrates advanced machine learning and fuzzy logic techniques to optimize desire prioritization and intention management.

We evaluated six models— Adaptive Neuro-Fuzzy Inference System PlanLibrary (ANFIS PL), Random Forest, Support Vector Machine (SVM), Fuzzy Logic, Gradient Boosting, and K-Nearest Neighbors (K-NN)—across metrics including Accuracy, Precision, Recall, F1-Score, Computational Time, Memory Usage, and CPU Utilization. Results demonstrate that ANFIS PL and Random Forest achieve superior performance while maintaining efficient computational resource usage. The proposed Plan Library approach significantly enhances BDIx agents’ capability to prioritize and execute communication strategies autonomously, ensuring optimal performance and adaptability in 6G D2D networks. This advancement underscores the pivotal role of intelligent agent architectures in next-generation wireless communication systems, paving the way for more resilient and efficient network operations.