EGR 555 Mechronics Innovation

May 20, 2025

From The Course (Motivated by) EGR555 Mechatronics Device Innovation Arizona State University

Autonomous emergency braking using computer vision 2025 CODE

(SOURCE CHATGPT SEARCH)

Autonomous Emergency Braking with YOLOv8 and LiDAR Fusion

This project integrates YOLOv8 for object detection with 2D LiDAR for distance measurement. The system calibrates both sensors to complement each other’s strengths — YOLOv8 identifies objects, while LiDAR calculates distances. The fusion enables accurate collision risk assessment, achieving a 100% success rate in 10 test scenarios. The code is available on GitHub. GitHub

2. V-CAS: Vision Transformer-Based Collision Avoidance

V-CAS employs the RT-DETR vision transformer model for object detection, DeepSORT for tracking, and an adaptive braking mechanism. It computes a composite collision risk score using vehicle accelerations, distances, and brake light signals from multiple camera streams. Implemented on the Jetson Orin Nano, V-CAS achieved over 98% accuracy with an average proactive alert time of 1.13 seconds. arXiv

3. Dual-AEB: Combining Rule-Based and Multimodal LLMs

Dual-AEB integrates a multimodal large language model (LLM) with a rule-based AEB system. The LLM enhances scene understanding, while the rule-based system ensures rapid braking decisions. This hybrid approach aims to improve adaptability and response times in diverse driving scenarios. The source code is available on GitHub. arXiv

Contact for Business in Delhi and India