Optimizing traffic light cycles by combining local analysis (NVIDIA Jetson, Metropolis, DeepStream)

We are currently developing an innovative project which leverages NVIDIA AI technologies for dynamic traffic light management at urban intersections. The project focuses on optimizing traffic light cycles by combining local analysis (NVIDIA Jetson, Metropolis, DeepStream) and global traffic data (Google Maps, Green Light).

To successfully implement this project, we seek your recommendations for the most suitable NVIDIA equipment and cameras. The key components of the project include:

Real-time video analysis for detecting vehicles, pedestrians, and priority vehicles.
Edge computing processing using Jetson devices.
Integration with traffic APIs and optimization of traffic light cycles based on real-time data.
We plan to equip at least 5 intersections with cameras and edge devices, ensuring low latency and high precision in data processing. Could you please provide guidance on the following:

Recommended Jetson devices (e.g., Xavier, Orin, Nano).
Types of cameras compatible with NVIDIA platforms (e.g., AI-enabled IP cameras).
Software licenses or SDKs needed for Metropolis and DeepStream integration.
Any additional resources we should consider for successful implementation.
Our goal is to maximize system efficiency while minimizing resource consumption and CO₂ emissions, in line with the project’s objectives. We look forward to your response and recommendations to move forward with implementation.

Thank you in advance for your support!

Can you refer this blog for traffic insights with JPS: Generate Traffic Insights Using YOLOv8 and NVIDIA JetPack 6.0 | NVIDIA Technical Blog?
Please check here for deploy Yolov8 on JPS 2.0: DeepStream Perception — Jetson Platform Services documentation