AI and Machine Learning Architectures for Level-4 Autonomous Vehicle Control: A Safety-Aware Engineering Perspective
DOI:
https://doi.org/10.63282/3050-922X.ICAILLMBA-119Keywords:
Artificial Intelligence (Ai), Machine Learning (Ml), Level 4 Autonomous Vehicles, Operational Design Domain (Odd), Deep Neural Networks (Dnns), Dynamic Driving Task (Ddt), Level 4 (L4) Model Predictive Control (Mpc), Reliability And Robustness In Av Systems, Iso26262Abstract
Level-4 autonomous vehicles are made to drive themselves in certain operational design domains, which means they need to be able to see, predict, plan, and control the vehicle in a wide range of traffic situations. Recent progress in artificial intelligence and machine learning has made perception more accurate, multi-sensor fusion better, and decision-making more powerful. However, adding learning-based parts to safety-critical vehicle control makes it harder to make sure they are strong, predictable, and work in the real world. This paper offers a system-level examination of current AI and ML methodologies employed in Level-4 autonomous driving, focusing on practical engineering limitations rather than theoretical assurances. We look at both classical and hybrid control strategies, as well as learning-based methods for perception and prediction. Safety-aware reference architecture is suggested that keeps learning-based intelligence separate from deterministic control execution and adds independent supervision, redundancy, and fallback mechanisms. Comparative analysis shows that hybrid architecture offers the best balance between safety and performance. The findings indicate that architectural design is fundamental to attaining dependable and scalable Level-4 autonomous vehicle control.
References
[1] SAE International, “Taxonomy and Definitions for Terms Related to Driving Automation Systems,” 2021. https://www.sae.org/standards/content/j3016_202104/
[2] Thrun, S., Montemerlo, M., et al., “Stanley: The Robot that Won the DARPA Grand Challenge,” Journal of Field Robotics, 2006. https://onlinelibrary.wiley.com/doi/10.1002/rob.20147
[3] Pendleton, S. D., et al., “Perception, Planning, Control, and Coordination for Autonomous Vehicles,” Machines, 2017. https://www.mdpi.com/2075-1702/5/1/6
[4] Goodfellow et al., Deep Learning, MIT Press, 2016. https://www.deeplearningbook.org/
[5] Grigorescu et al., “A Survey of Deep Learning Techniques for Autonomous Driving,” J. Field Robotics, 2020. https://onlinelibrary.wiley.com/doi/10.1002/rob.21918
[6] Kuutti et al., “A Survey of Deep Learning Applications to Autonomous Vehicle Control,” IEEE T-ITS, 2021. https://ieeexplore.ieee.org/document/9312467
[7] Paden et al., “A Survey of Motion Planning and Control Techniques,” IEEE Trans. ITS, 2016
[8] https://ieeexplore.ieee.org/document/7490340
[9] Rajamani, R., Vehicle Dynamics and Control, Springer, 2012.
(Classic reference for vehicle control fundamentals; strengthens Sections 2.4 and 5.) https://link.springer.com/book/10.1007/978-1-4614-1433-9
[10] Shalev-Shwartz, S., Shammah, S., and Shashua, A., “On a Formal Model of Safe and Scalable Self-Driving Cars,” arXiv preprint arXiv:1708.06374, 2017. https://arxiv.org/abs/1708.06374
[11] Leveson, N., Engineering a Safer World: Systems Thinking Applied to Safety, MIT Press, 2011. https://mitpress.mit.edu/9780262016629/engineering-a-safer-world/
