1Research Scholar, Department of Computer Science and Engineering, Quantum University, Roorkee, Uttarakhand, India. Email: Sh.shilpy@gmail.com
2Professor, Department of Computer Science and Engineering, Quantum University, Roorkee, Uttarakhand, India. Email: mridula.cse@quantumeducation.in
3Associate Professor, Department of Computer Science and Engineering, Noida Institute of Engineering & Technology, Greater Noida, Uttar Pradesh, India. Email: drrashmisharma20@gmail.com
Title: A Neuro-Symbolic Agentic Digital Twin Framework for Causally Explainable Reckless Driving Behaviour Detection
Road traffic accidents remain one of the leading causes of fatalities worldwide, and reckless driving behavior is widely recognized as a primary contributor to these incidents. Despite significant advancements in artificial intelligence, most existing reckless driving detection systems rely heavily on correlation-based deep learning techniques such as Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and Transformer-based architectures. While these models can identify patterns in driving data, they often lack causal understanding, personalization, and explainability, which limits their effectiveness in real-world deployment. Furthermore, such systems typically treat reckless driving as a static classification problem rather than understanding the underlying behavioral intentions and contextual dynamics of the driver.
To address these limitations, this study proposes a novel framework called the Neuro-Symbolic Agentic Digital Twin (NS-ADT) for intelligent detection of reckless driving behavior. The proposed paradigm integrates neural computation with symbolic reasoning to provide a more comprehensive understanding of driving patterns. In particular, the framework models driving behavior as a continuous dynamic process rather than discrete events. A Neural Ordinary Differential Equation (Neural-ODE) is employed to learn the temporal evolution of driving dynamics, enabling the system to capture subtle changes in acceleration, braking, and steering patterns over time. In addition, the framework incorporates a Structural Causal Model (SCM) to infer causal relationships among various driving factors such as speed variation, lane deviation, environmental conditions, and driver actions. This causal reasoning component enables the system to perform counterfactual analysis, thereby improving interpretability and allowing the system to explain why a particular behavior is considered unsafe. Unlike conventional models, the proposed system also introduces a self-adaptive digital twin for each driver, which continuously updates its understanding of the driver's long-term behavioral patterns using online reinforcement learning.
Experimental evaluation demonstrates that the NS-ADT framework achieves higher detection accuracy, stronger robustness to behavioral drift, and better cross-dataset generalization compared to existing deep learning approaches. The results highlight the potential of agentic, causally-aware AI systems in advancing next-generation intelligent transportation safety solutions.
Keywords: Reckless driving detection, Agentic AI, Neuro-symbolic systems, digital twins.
Subject Classification: Primary 93A30, Secondary 49K15
How to cite this article: Sharma S, Singh M, Sharma R. A Neuro-Symbolic Agentic Digital Twin Framework for Causally Explainable Reckless Driving Behaviour Detection. Int J Drug Deliv Technol. 2026;16(16s): 398-416. DOI: 10.25258/ijddt.16.16s.43
Source of support: Nil.
Conflict of interest: None