Reproduce “ASTM :Autonomous Smart Traffic Management System Using Artificial Intelligence CNN and LSTM” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Christofel Rio Goenawan
An autonomous smart traffic management system developed at KAIST utilizes YOLOv5 for real-time vehicle detection and an RNN-LSTM for 12-hour traffic prediction. This AI-driven approach achieved a 50% increase in congestion flow rate and a 70% reduction in vehicle pass delay when evaluated against conventional systems in CARLA simulations.
Key insights
- Optimal hyperparameter tuning for YOLOv5 significantly improves detection accuracy, achieving a 0.4034 mAP for tuned hyperparameters compared to approximately 0.2852 for other configurations.
- RNN-LSTM models can accurately predict future traffic density for up to 12 hours, capturing both increases and decreases in vehicle numbers with low error (e.g., RMSE of 2.232 vehicles).
- Integrating real-time AI-driven perception (CNN) with predictive AI (LSTM) enables proactive, adaptive traffic signal management that outperforms reactive conventional systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.