Evaluating the Impact of Real-Time Vision-Based Lane and Vehicle Detection Accuracy on High-Density Connected Autonomous Vehicle Platoon Stability in Urban Environments

• Abbas Ahsun

  Texas University

  Author

Connected Autonomous Vehicle (CAV) platooning represents a transformative paradigm for urban mobility, offering potential improvements in traffic throughput, energy efficiency, and road safety. However, the stability of high-density platoons in complex urban environments remains critically dependent on the accuracy of perception systems, particularly vision-based lane and vehicle detection. This research investigates the relationship between real-time detection accuracy and platoon stability metrics—specifically spacing error variance, string stability propagation, and disturbance recovery time—using a modified OverFeat CNN architecture operating at >10 Hz frame rates . Through controlled simulation experiments incorporating urban traffic scenarios with varying detection accuracy levels (ranging from 89.4% to 97.2% mean average precision), the study demonstrates that detection accuracy below 92.5% induces significant string instability, resulting in spacing error amplification of up to 340% across a 10-vehicle platoon. The findings establish quantitative accuracy thresholds for stable platoon operation and provide a replicable evaluation framework for vision-based CAV perception systems. The research contributes to both autonomous vehicle perception literature and connected vehicle control theory, with practical implications for perception system design requirements in urban CAV deployments.

• pdf

Copyright (c) 2026 Abbas Ahsun (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.