Traffic Operations AI Tools
TrafficOps uses AI-assisted workflows through product APIs, platform tools, AI Engine forecast and simulation services, and review workflows. The goal is to help operators understand congestion risk, compare scenario options, prepare action plans, and keep decisions traceable.
Use this page when an implementation team needs to understand which TrafficOps surfaces can be called by the application or an approved Agent workflow, which outputs require review, and how to combine TrafficOps with DFS, simulation, Inspector, and ECM.
Tool layers
| Layer | Primary users | Access boundary | Output type |
|---|---|---|---|
/api/v1/trafficops/* product APIs | TrafficOps UI and approved backend workflows | TrafficOps module enablement and product permissions | Dashboards, queues, snapshots, flow KPIs, incidents, equipment state, rules, reports, workflow actions, and simulation records |
| Platform tools | Approved Agent workflows | Tool framework permissions and tenant-scoped knowledge graph access | analyze_traffic_flow and predict_wait_time outputs for checkpoint review |
| AI Engine TrafficOps services | Product services and controlled backend jobs | Service authentication and project configuration | Forecasts, proactive alerts, monitor state, what-if comparison, optimization, calibration, and scenario analysis |
| DES/DAG simulation | Operations planners and simulation reviewers | Simulation permission, engine availability, and scenario input validation | Run result, queue metrics, throughput, wait time, bottlenecks, replay/export data, and scenario comparison |
| DFS | Data owners and integration teams | DFS permissions for connectors, mappings, datasets, and quality review | Prepared arrival, queue, incident, equipment, staffing, schedule, and layout datasets |
| ECM and workflow | Operations owners and reviewers | Document and workflow permissions | Incident reports, what-if documents, action-plan status, and review evidence |
Runtime tool availability should be discovered in the target environment before exposing an Agent workflow.
Recommended AI-assisted flow
Platform tools
The platform tool framework currently exposes these TrafficOps tools:
| Tool | Purpose | Inputs | Output to review |
|---|---|---|---|
analyze_traffic_flow | Analyze traffic flow at a checkpoint using checkpoint identity and related lane context. | checkpointId, optional period. | Checkpoint ID, related lane count, period, traffic summary, bottleneck lane, and recommendations. |
predict_wait_time | Predict wait time for a checkpoint or lane over a selected horizon. | checkpointId, optional laneId, optional horizonMinutes. | Prediction timeline, wait-time values, confidence, model identifier, checkpoint ID, and lane ID when supplied. |
Use these tools for review and decision support. The caller should preserve the checkpoint, time window, input parameters, source data state, and reviewer decision.
AI Engine services
TrafficOps AI Engine services support forecasting, scenario comparison, optimization, calibration, and simulation:
| Service area | Example endpoints | Review focus |
|---|---|---|
| Forecast and pattern analysis | /ai/trafficops/forecast, /ai/trafficops/patterns, /ai/trafficops/proactive-alerts | Forecast window, checkpoint, SLA threshold, demand pattern, alert trigger, and source freshness. |
| Monitor workflow | /ai/trafficops/monitor, /ai/trafficops/monitor/scenarios, /ai/trafficops/monitor/approve, /ai/trafficops/monitor/dismiss | Current risk state, proposed action, reviewer decision, and action-plan status. |
| What-if comparison | /ai/trafficops/what-if, /ai/whatif/run, /ai/whatif/batch, /ai/whatif/submit | Baseline, overrides, engine type, comparison metrics, confidence, and run status. |
| Optimization | /ai/trafficops/optimize, /ai/optimize/trafficops | Budget, cost per staff, primary KPI, Pareto options, and selected recommendation. |
| Calibration | /ai/trafficops/calibrate, /ai/trafficops/calibration/status, /ai/trafficops/calibration/drift, /ai/trafficops/calibration/history | Calibration scope, data days, drift state, scheduler state, and model update history. |
| DES/DAG simulation | /ai/des/run, /ai/des/batch-run, /ai/des/runs, /ai/des/runs/{runId}, /ai/des/runs/{runId}/export, /ai/des/runs/{runId}/replay.ndjson | Scene type, scene ID, simulation time, replications, staff schedule, failure config, replay, and export package. |
| Package scenes and road network | /ai/packages/{package_key}/scenes, /ai/packages/{package_key}/roads/*, /ai/packages/{package_key}/whatif/*, /ai/packages/{package_key}/simulate | Package enablement, scene definition, route graph, preset, simulation inputs, and result interpretation. |
| Layout import | /ai/cad-import/*, /ai/packages/{package_key}/dxf/import | Imported layout, recognized entities, layer hints, simulation readiness, and review owner. |
Only enable services that match the project scope and data readiness. Keep scenario assumptions and result interpretation with the review record.
Data preparation
AI-assisted TrafficOps depends on a source package that can be reviewed:
| Data | Why it matters |
|---|---|
| Checkpoint and lane identity | Connects dashboards, tools, snapshots, routes, simulations, and action plans to the same operating object. |
| Arrival and queue history | Provides demand, wait-time, and throughput context for forecast and what-if analysis. |
| Staffing and schedule data | Drives capacity, shift planning, optimization, and action feasibility. |
| Incident and density records | Provides safety, congestion, and replay context for response planning. |
| Equipment health and configuration | Connects service-time assumptions, lane availability, sensitivity, and maintenance planning. |
| Layout and route graph | Supports scene review, CAD/DXF import, DES/DAG simulation, and road-network analysis where enabled. |
| Rules and SLA targets | Defines thresholds used by alerts, monitor workflows, reporting, and action approval. |
Use Data Fusion Services and Getting Started with DFS before exposing automated review.
Scenario and simulation guidance
Use what-if and DES/DAG simulation when a team needs to compare options before changing operations:
| Scenario type | Typical inputs | Output to review |
|---|---|---|
| Staff change | Checkpoint, staff count, cost per staff, shift window, target KPI. | Average wait, P95 wait, throughput, queue size, and cost tradeoff. |
| Lane or checkpoint capacity | Lane count, service time, direction, open/closed state, failure assumptions. | Bottleneck movement, utilization, wait distribution, and lane impact. |
| Vehicle or cargo incident | Incident location, affected lane, duration, demand window. | Queue growth, expected recovery, affected throughput, and response option. |
| Density surge | Density thresholds, floor-plan grid, selected time window. | Red/amber cells, incident link, response notes, and replay evidence. |
| Multi-region disruption | Region, route, event timeline, resilience assumptions. | Regional KPI movement, timeline, and cross-region impact. |
Simulation output supports planning. It should be reviewed with local constraints, staffing rules, and field owner judgment before action.
Suggested Agent answer structure
When an AI assistant summarizes TrafficOps risk, use a stable format:
| Section | Content |
|---|---|
| Scope | Tenant, site, facility, checkpoint, lane, direction, time window, and selected flow. |
| Current state | Throughput, queue length, average wait, P95 wait, utilization, incidents, density, and SLA state. |
| Evidence | Source records, lane snapshots, arrival history, staffing schedule, simulation inputs, incident notes, and tool outputs. |
| Options | Ranked staffing, lane, incident, or maintenance options with expected impact and assumptions. |
| Handoff | Action plan, incident task, report, work order, or request for field verification. |
Validation checklist
- The workflow runs under the correct tenant, site, checkpoint, lane, direction, and time window.
- The caller has permission for the read, simulation, report, workflow, or write action.
- Checkpoint, lane, arrival, queue, staff, equipment, incident, and route IDs resolve to the same operating model.
- Forecast or what-if output includes input parameters, time window, scenario assumptions, and confidence or run metadata when available.
- Recommendations name the action owner and stay in review until approved.
- Any lane, staff, incident, or maintenance action is recorded with decision status and feedback.
- Reports, exports, replay files, or ECM documents are stored with the operating review package when used in handover.