Authors : Navin Kumar Sehgal

Volume/Issue : Volume 11 - 2026, Issue 7 - July

Google Scholar : https://tinyurl.com/2hwh5kpr

Scribd : https://tinyurl.com/ytf9fux4

DOI : https://doi.org/10.38124/ijisrt/26jul923

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

Abstract : Autonomous driving systems must operate safely and reliably under diverse traffic densities, varying weather conditions, and dynamic road environments while maintaining real-time performance and low computational cost. However, many existing autonomous driving frameworks rely on expensive LiDAR sensors, high-performance computing hardware, and cloud-based processing, limiting their practical deployment in cost-sensitive applications. This paper proposes a Hybrid AI Framework for Autonomous Driving Across Diverse Traffic and Weather Conditions that integrates Segment Anything Model 2 (SAM2) for semantic scene segmentation, transformer-based multimodal sensor fusion for comprehensive environmental perception, reinforcement learning (RL) for adaptive decision-making, and a neural vehicle controller for continuous steering, throttle, and brake control. To reduce system cost and computational complexity, the framework employs a low-cost sensor suite comprising an RGB camera, automotive radar, ultrasonic sensors, GPS, accelerometer, gyroscope, odometer, magnetometer, temperature sensor, humidity sensor, and vibration sensor, eliminating the need for expensive LiDAR systems. The entire architecture is optimized for deployment on the low-cost NVIDIA Jetson embedded edge-computing platform, enabling real-time processing with reduced latency, lower power consumption, and improved operational efficiency. The proposed framework is designed to handle challenging edge-case scenarios, including sudden obstacles, adverse weather, dense traffic, and low-visibility conditions, by dynamically adapting sensor fusion and driving policies. Experimental evaluation demonstrates that the proposed approach achieves over 90% perception and decisionmaking accuracy, while maintaining stable vehicle control, smooth steering, adaptive throttle regulation, and timely braking across varying traffic and weather conditions. The results further indicate significant improvements in driving safety, vehicle stability, collision avoidance, and computational efficiency compared with conventional autonomous driving approaches. The proposed hybrid architecture provides a scalable, cost-effective, and practical solution for next-generation intelligent autonomous vehicles operating in real-world environments.

Keywords : Autonomous Driving, Artificial Intelligence, Segment Anything Model 2 (SAM2), Transformer-Based Sensor Fusion, Reinforcement Learning, Edge Computing, Semantic Segmentation, Vehicle Control, Intelligent Transportation Systems.

References :

1. A. Kirillov et al., "Segment Anything," in Proc. IEEE/CVF Int. Conf. Computer Vision (ICCV), Paris, France, 2023, pp. 4015–4026.
2. N. Ravi et al., "SAM 2: Segment Anything in Images and Videos," arXiv preprint arXiv:2408.00714, 2024.
3. A. Vaswani et al., "Attention Is All You Need," in Advances in Neural Information Processing Systems (NeurIPS), vol. 30, 2017, pp. 5998–6008.
4. V. Mnih et al., "Human-Level Control Through Deep Reinforcement Learning," Nature, vol. 518, no. 7540, pp. 529–533, 2015.
5. J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov, "Proximal Policy Optimization Algorithms," arXiv preprint arXiv:1707.06347, 2017.
6. T. P. Lillicrap et al., "Continuous Control with Deep Reinforcement Learning," in Proc. Int. Conf. Learning Representations (ICLR), 2016.
7. H. Caesar et al., "nuScenes: A Multimodal Dataset for Autonomous Driving," in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, 2020, pp. 11621–11631.
8. A. Geiger, P. Lenz, C. Stiller, and R. Urtasun, "Vision Meets Robotics: The KITTI Dataset," International Journal of Robotics Research, vol. 32, no. 11, pp. 1231–1237, 2013.
9. H. Li et al., "BEVFormer: Learning Bird's-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers," in Proc. European Conf. Computer Vision (ECCV), 2022, pp. 1–18.
10. Y. Bai et al., "TransFusion: Robust LiDAR-Camera Fusion for 3D Object Detection with Transformers," in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR), 2022, pp. 1090–1099.
11. A. Dosovitskiy et al., "An Image Is Worth 16×16 Words: Transformers for Image Recognition at Scale," in Proc. Int. Conf. Learning Representations (ICLR), 2021.
12. S. Thrun et al., "Stanley: The Robot That Won the DARPA Grand Challenge," Journal of Field Robotics, vol. 23, no. 9, pp. 661–692, 2006.
13. A. Howard et al., "Searching for MobileNetV3," in Proc. IEEE/CVF Int. Conf. Computer Vision (ICCV), Seoul, South Korea, 2019, pp. 1314–1324.
14. NVIDIA Corporation, NVIDIA Jetson AGX Orin Developer Kit User Guide, Santa Clara, CA, USA: NVIDIA Corporation, 2024.
15. D. Silver et al., "Mastering the Game of Go with Deep Neural Networks and Tree Search," Nature, vol. 529, no. 7587, pp. 484–489, 2016.