Communication-Centric Capability Interactions in Vehicle Autonomy Interpreted via Deep Fuzzy Layers

Abstract

Autonomous vehicular communication networks are essential for modern intelligent transportation systems, enabling seamless data exchange among vehicles, roadside units, and centralized control systems. The performance of these networks depends on key parameters such as Random Access Memory (RAM), storage capacity, data transmission rate, and trust factor, which collectively determine the reliability and responsiveness of communication units. In dynamic traffic environments, accurately evaluating these capabilities is critical for ensuring safety, optimizing performance, and improving resource utilization. Traditional evaluation methods rely on manual inspections and threshold-based techniques that assess parameters independently. This approach fails to capture the complex interdependencies among system attributes, making it unsuitable for real-time vehicular scenarios and often leading to inaccurate capability estimation. To address this, machine learning (ML) regression models such as Decision Tree Regressor (DTR), Orthogonal Matching Pursuit Regressor (OMPR), and K-Nearest Neighbors Regressor (KNNR) have been applied. However, these models face challenges in handling nonlinear relationships, noise sensitivity, and generalization across diverse datasets. To overcome these limitations, this study proposes a hybrid Deep Fuzzy Regression (DFR) model that integrates Deep Fuzzy Encoding (DFE) with Random Forest Regressor (RFR) and Linear Regression (LR) in an ensemble framework. The DFE component effectively manages uncertainty and gradual feature variations, while RFR and LR enhance prediction robustness and stability. The system follows a structured pipeline including preprocessing, feature engineering, model training, and evaluation. Experimental results using MAE, MSE, RMSE, and R² demonstrate that the proposed model delivers more accurate and consistent predictions, improving communication efficiency and decision-making in autonomous vehicular networks.