Assemble the network state
Normalize live sensor values, incident reports, road topology, control infrastructure, and historical traffic patterns into a shared operational model.
Traffic as behaviours.
A prototype component that combines probabilistic congestion forecasts with configurable behaviour logic. It structures network-wide conditions into traceable recommendations for operator review.
Human in the loop Vendor neutral Explainable by design
Network intelligence / A12 East Test mode
Network state
2,940 veh/h · 67 km/h
Detected
Stopped vehicle
Calculated test state
High congestion likelihood
Congestion probability
+5 min73%
+10 min60%
+15 min46%
+30 min31%
Candidate measure
Protect lane 1 and reduce approach speed.
Illustrative output · operator review required
Test parameters
Adjust the inputs to recalculate the illustrative forecast.
Test visualization only. Probabilities are generated by a simplified demonstration formula and are not outputs from the trained prototype model.
01
Why this
Traffic events are connected network events.
A stopped vehicle can affect lane availability, approach speeds, queue formation, adjacent routes, and can lead to a chain reaction of incidents.
Operators have to correlate these dependencies across several systems and apply network-specific procedures. Existing automation remains responsible for local, deterministic control; the prototype investigates how a separate supervisory layer can add predictive context and behaviour-based evaluation.
How it works
Configured and validated before the first operational cycle.
The runtime cycle is only one part of the system. Network layout, behaviours, dependencies, and validation rules are defined first and tested against the possible operating states.
Our scopePre-operational engineering
Completed by us before the system enters cyclic operation
1 · Network configuration
Model the road as behaviours
Define the physical layout, sign locations, distances, topology, operational settings, and the behaviours that determine how the network should respond to an event.
2 · Scenario validation
Test the complete rule space
Generate and evaluate the possible traffic and device scenarios. Invalid, incomplete, or conflicting rule outcomes are flagged for review before the configuration is used operationally.
≈250,000scenarios for an illustrative road with 300 traffic signs*
*The actual number depends on sign distances, topology, behaviour settings, and the permitted state combinations.
Configuration validated
External scopeExisting monitoring platforms
Performed only by the customer’s and vendors’ existing systems
3 · Data collection
External platforms monitor the field
Before our runtime cycle starts, established third-party systems collect and monitor the available operational data. This monitoring remains outside our system.
Traffic counts
Speed
Occupancy
Incidents & events
Device status
Network topology
Normalized interface boundary
Our scopeBehaviour engine runtime cycle
Steps 4–8 are performed by us and repeat continuously
4
Retrieval
Receive data from vendor-specific interfaces and normalize it into a common data model.
5
Preparation
Prepare the normalized state for the trained prediction model and the behaviour-based network evaluation.
6
Recommendation creation
Use the forecast either as an input to the recommendation logic or as operator information, depending on configuration.
7
Application & distribution
Resolve the network-wide measures and return approved decisions to the relevant connected systems.
8
Audit & trend
Store decisions and actions where required, then use the history for model training and network-specific fine-tuning.
↶
Continuous cycle
8 → 4After audit and trend processing, the system returns to step 4 and retrieves the next network state.
02
Capabilities
A supervisory layer for the decision chain.
The prototype explores how heterogeneous inputs can be evaluated within one consistent network and behaviour model.
Estimate near-term development
Apply AI-based analysis to evolving traffic patterns and calculate congestion probabilities for several forecast horizons.
Evaluate operational behaviours
Resolve forecast conditions against configurable behaviour rules, topology, dependencies, and operational constraints.
Expose the decision trace
Present the input state, evaluated rules, constraints, and resulting measures for operator review and approval.
Illustrative operating sequence
Stopped-vehicle detection and response evaluation.
This example shows how an observed event can be linked to forecast network effects and a set of measures for operator review. Values are illustrative.
09:42:18 · Incident detected
Stopped vehicle · Lane 1
A12 eastbound · km 18.4
Current flow
2,940 veh/h
Average speed
67 km/h
Queue direction
Upstream ↗
AI forecast
Congestion probability
+5 min
60%
+10 min
40%
+15 min
28%
+30 min
14%
Evaluated response
Proposed incident protection and approach stabilization
- Close and protect lane 1
- Reduce approach speed in two stages
- Monitor queue growth and adjacent route load
Operator approval requiredTrace available
03
Architecture
Supervisory evaluation outside the real-time control loop.
The concept is intentionally separated from deterministic field execution, PLC cycle times, and safety interlocks.
Works with existing traffic-management environments
Stays outside hard real-time PLC cycle requirements
Preserves field-level interlocks and safety logic
Keeps every recommendation subject to operator approval
Behaviour-Based Traffic Engine
Predict · reason · recommend
Explainable recommendations
Traffic management & operator systems
Review · approve · coordinate
Validated commands
PLC & roadside control
Execute · interlock · protect
Working prototype
Ready for technical discussion
A functional prototype is available for technical discussion. It is intended to validate the behaviour model, interfaces, and operator workflow against a defined motorway environment.
Request a technical demo
Review the prototype and its assumptions.
A technical walkthrough covers the forecast representation, behaviour evaluation, decision trace, system boundaries, and possible integration points.
Submitting opens a prefilled email in your mail app. You can review, edit, or add anything before sending.