↓40%
Wait time reduction
↓21%
Emissions cut
V2X
Vehicle-to-everything
360°
Pedestrian + car data
Smart connected intersection network diagram Four intersections linked by data streams, showing cars, pedestrians, and real-time data propagation between nodes Cloud AI engine real-time inference Intersection A 12 cars 3 peds → passing data Intersection B 8 cars 6 peds ← receiving Intersection C 5 cars 9 peds ped priority Intersection D 15 cars 2 peds vehicle flow Stop Caution Go Data link Ped link
1
Sense — intersection A collects data
IoT sensors, cameras, and inductive loops at intersection A continuously capture vehicle count, queue length, pedestrian crossing requests, wait times, and micromobility (bikes, e-scooters). Edge devices pre-process the data locally to reduce latency.
2
Broadcast — data packet sent to next node
As vehicles leave intersection A, a V2X (Vehicle-to-Everything) data packet is transmitted over 5G to intersection B. The packet includes: current congestion score, pedestrian density heatmap, average vehicle speed, and emergency vehicle flags.
3
Predict — AI pre-adapts the next intersection
The cloud AI engine (using deep reinforcement learning, e.g. DDPG) receives the packet and computes the optimal signal phase for intersection B before the vehicles arrive — giving green corridors to high-volume streams and extending pedestrian crossing times when pedestrian load is elevated.
4
Adapt — intersection B adjusts its signals
Intersection B's controller updates signal timings in real time. Pedestrian signals are extended if ped count from A was high (multimodal awareness). Emergency vehicles get preemptive green corridors. Each intersection node is autonomous — if the cloud is unreachable, it falls back to local AI scheduling.
5
Cascade — the data propagates across the network
The insight from A flows to B, B's data flows to C and D, creating a city-wide green wave. The central dashboard updates traffic authority operators in real time. Historical data builds a digital twin of the city for long-term urban planning and CO₂ reduction targets.
Pittsburgh · CMU Surtrac Decentralized node architecture
Each intersection runs its own AI scheduling node that passes insights to neighboring lights — exactly the architecture your idea describes. The system incorporates pedestrian app data, GPS from e-bikes, and connected vehicle routes as multimodal inputs.
↓26% travel time · ↓40% idle time at red lights · deployed at 50+ intersections
2025 · H-ATLM system Deep reinforcement learning for signal timing
A hybrid adaptive traffic light management system using the DDPG reinforcement learning algorithm dynamically optimizes green/red phase durations based on real-time sensor input, integrating cameras and inductive loops. Pedestrian accessibility features are a key extension target.
↓50% congestion · ↑149% throughput · ↓84% clearance time
NYC · V2X Pilot Vehicle-to-everything communication backbone
New York City equipped 3,000 vehicles and upgraded 450+ signals with V2X technology, enabling real-time pedestrian alerts and speed compliance improvements. This is the communication layer your inter-intersection data transfer would rely on.
+16% speed compliance · enhanced pedestrian alert system
2025 · Passable system Computer vision for pedestrian + incident detection
Using YOLOv8 deep learning on traffic camera feeds, this system detects pedestrians, cyclists, and road incidents in real time, dynamically adjusting signal timings based on live vehicle density and vulnerable road user counts — directly addressing the pedestrian data layer of your concept.
Real-time incident detection · wireless driver alerts · centralized dashboard
Beijing · DRL system Environmental impact of connected signals
A deep reinforcement learning traffic system in Beijing demonstrated measurable environmental benefits beyond congestion reduction, validating the sustainability case for cross-intersection data sharing.
↓25% CO₂ emissions during peak hours
Ask anything about the connected traffic system — the AI will answer in context.