Smart AI

Our Smart AI integrated solution

• Traffic detection: Live detector and signal state data describe current demand
• Smart-AI CPU: A second dedicated CPU inside the ITC-3 controller evaluates the traffic state every 100 ms
• Safe action: A second dedicated CPU inside the ITC-3 controller evaluates the traffic state every 100 ms
• Better flow: Signal timing adapts continuously to actual demand at the intersection

Local, safe and operationally transparent

• Local operation: Runs inside our ITC-3 controller without an internet dependency
• Safe integration: Operates within configured controller logic, safe modes and national requirements
• Dashboard & logs: Dedicated dashboards support monitoring; logs follow the standard ITC-3 approach
• Plan-based control: SMART-AI can be enabled or disabled through normal plan changes

How Smart AI works

1. Observe: Read detector inputs, signal states, and demand
2. Evaluate: Compare permitted actions using its learned policy
3. Act: Apply the selected action through the ITC-3 controller
4. Learn: Use the result to improve traffic flow continuously

Smart AI. Designed for real traffic

• Self-learning optimisation: Learns from traffic and improves control without relying on fixed timing plans
• Real-time decisions: Evaluates the intersection and selects an action every 100 milliseconds
• Local edge operation: Runs inside the ITC-3 controller; no cloud connection or internet is required
• Controlled deployment: Enable or disable SMART-AI by plan change and retain standard safe modes

Adaptive intelligence built into the ITC-3 controller

Multiple ai models - one optimised decision

SMART-AI combines multiple specialised AI models, each designed for a specific traffic-engineering task. The models continuously exchange information to optimise traffic operation in real time. All models exchange data and update their predictions every 100 ms.

• Incoming traffic prediction: Predicts approaching demand before vehicles reach the intersection
• Dynamic traffic cost: Assigns different costs to moving and stationary vehicles to optimise overall flow
• Turning-movement prediction: Predicts the percentage turning left, right, or continuing straight
• Signal switch-point optimisation: Selects the optimal moment to change stages while balancing efficiency, delay and safety
• Lane-geometry modelling: Uses lane layout and permitted movements to improve decisions
• Platoon speed estimation: Estimates platoon speed and arrival time to support coordinated progression