Authors : Grace Angel P.; M. M. Harshitha; Dr. Girish Kumar D.

Volume/Issue : Volume 11 - 2026, Issue 4 - April

Google Scholar : https://tinyurl.com/a7bcfs5t

Scribd : https://tinyurl.com/y7vprmtf

DOI : https://doi.org/10.38124/ijisrt/26apr722

Note : A published paper may take 4-5 working days from the publication date to appear in PlumX Metrics, Semantic Scholar, and ResearchGate.

Abstract : The complexity and safety-critical nature of autonomous driving demand rigorous and enhanced testing methodologies beyond standard driving scenarios. Traditional end-to-end self-driving models often lack the transparency needed to debug failure cases and analyze system behavior under edge conditions. This research presents RoadSense, a modular, web-integrated architecture designed specifically for simulation-enhanced testing and performance analysis of self-driving car policies within the Udacity Simulator environment. The system decomposes the control pipeline into four distinct, sequential AI models: the Input Model, the Processing Model, the Decision-Making Model, and the Output Model. The architecture allows for targeted testing and failure-tracing across the stages of perception and control. RoadSense is deployed on a web-based dashboard, offering real-time visualization, control parameter injection for enhanced stress testing, and detailed log tracing for every command cycle. Evaluation demonstrates the system's ability to not only achieve reliable autonomous navigation (with a high Track Completion Rate) but, critically, to provide clear, traceable logs that isolate performance bottlenecks, validating its effectiveness as an enhanced testing and verification framework for modular autonomous driving policies.

Keywords : Autonomous Driving, Simulation Testing, Modular Architecture, Verification and Validation, Udacity Simulator, Behavioral Cloning, Failure Analysis.

References :

1. Bojarski, M., et al. (2016).End to End Learning for Self-Driving Cars. NVIDIA Corporation. arXiv:1604.07316.
2. Huang, Y., & Chen, Y. (2020).Autonomous Driving with Deep Learning: A Survey of State-of-Art Technologies.arXiv preprint arXiv:2006.06091.
3. Krizhevsky, A., Sutskever, I., and Hinton, G. E. (2012).ImageNet deep CNN-based classification Networks. The model generates a confidence score indicating the reliability of predictions. (NIPS).
4. Morga-Bonilla, S. I., et al. (2024). Behavioral Cloning Strategies in Steering Angle Prediction: Applications in Mobile Robotics and Autonomous Driving. MDPI - Applied Sciences, 15(11).
5. Shashank, M. S., et al. (2021). Behavior Cloning for Self Driving Cars using Attention Models. International Journal of Innovative Science and Research Technology, 6(12)..
6. Dam, A. T. Q. (2021). Self-driving Car with Udacity Simulator. Independent Study Thesis 9361, The College of Wooster.
7. Xie, X. (2025). Artificial Intelligence-Driven Autonomous Vehicles: Current Developments and the Future Prospects. Computers and Artificial Intelligence, 2(3).
8. Zhao, Y., et al. (2025). A Survey of Decision-Making and Planning Methods for Self-Driving Cars. Frontiers in Neurorobotics.