Skip to content

Airside Perception-SLAM Commissioning Playbook

Last updated: 2026-05-09

Purpose

This playbook defines the commissioning sequence for airside perception, SLAM/localization, mapping, calibration, runtime monitors, dashboards, and operational response. It starts after a site has an approved concept of operations and ends when the site owner, safety lead, fleet operations, and runtime owners accept autonomous operation for a defined route/ODD slice.

Entry Criteria

ItemRequired before vehicle commissioning
Site authorityAirport sponsor/site owner approval, safety management process, stakeholder communication plan
CONOPSRoute, mission type, speed, ODD limits, supervision mode, MRC, stop authority
Regulatory scanFAA/CAA/airport guidance, airside rules, insurance, SMS integration
Safety baselineHazards, risk controls, safety-case claim IDs, incident reporting workflow
Runtime baselineSigned software/model/map/calibration/config manifest and rollback set
Data permissionsLogging, upload, privacy, retention, and access control approval

Phase 1: Site Survey and Instrumentation

TaskAcceptance check
Survey route, stands, depot, charging, emergency access, no-go zonesRoute graph and protected zones reviewed by operations and safety
Identify weak-feature, GNSS-shadow, glare, wet-surface, jet-blast, FOD, and pedestrian hotspotsHotspots become ODD slices and test waypoints
Establish ground control and map datumSurvey package and uncertainty model stored
Validate communications and time infrastructurePTP/GNSS/PPS or site clock plan documented
Install observability tagsVehicle/site/route/map/calibration IDs join in telemetry

Phase 2: Vehicle and Sensor Readiness

TaskAcceptance check
Verify sensor serials, firmware, timestamp mode, mounting, cleaning stateVehicle inventory matches compatibility manifest
Apply baseline calibration packageIntrinsics, extrinsics, time offsets, TF tree hash stored
Run stationary and route calibration checksResiduals inside commissioning threshold
Validate diagnostics graphCritical sensor, timing, calibration, localization, perception, map, and runtime nodes present
Run local stop/MRC checksVehicle executes controlled stop under monitor red state

Phase 3: Map Capture and Map QA

TaskAcceptance check
Capture multiple route passesDay/night and operational traffic slices represented where in ODD
Build map bundle and semantic layersPermanent, movable-static, dynamic, hazard, unknown, route/geofence layers separated
Review temporary assetsCones, chocks, parked GSE, aircraft, barriers classified or overlaid with expiry
Run source-session timing and calibration checksRed/unknown sessions excluded from automatic map publication
Run map publication gatesNo false-free-space, frame, tile, overlay, or route mismatch blocker

Phase 4: Closed-Course and Controlled Route Tests

TestGate
Localization start/relocalizationPose valid inside route threshold before motion
Free-space fixturesAircraft/GSE/person/FOD/cone/chock fixtures are not marked traversable
Unknown-object fixturesNovel/unknown objects trigger conservative action
Timing/calibration perturbationMonitors detect fault before unsafe planner consumption
Wet/night/glare/adverse slice if in ODDMetrics pass by slice, not only aggregate
Emergency stop and remote-assist workflowOperator response time and evidence preservation pass

Phase 5: Shadow Mode

RequirementPass rule
ExposureRepresentative route windows, traffic states, weather/lighting, and operational tempo
Metric envelopeLocalization residual, intervention proxy, unknown/OOD rate, free-space disagreement, latency, diagnostics state inside baseline
Operator workflowAlerts reach the right role and produce correct acknowledgement/action
EvidenceEvery shadow run joins vehicle, route, map, calibration, build, bag/MCAP, and operator notes
ExitSafety lead and site owner accept remaining restrictions

Phase 6: Canary Operations

StepPromotion gate
One vehicle, low-risk windowNo P0/P1 perception-SLAM alert; manual override available
One operational routeMission success and safety metrics inside shadow envelope
Extended operating windowsNo unresolved disagreement cluster, calibration drift, or unknown-object cluster
Multiple vehiclesFleet dashboards and incident joins scale without missing fields
Steady stateMaintenance, map update, calibration, OTA, and alert runbooks exercised

Stop-Work Triggers

TriggerAction
False-free-space in protected zoneStop autonomous testing; preserve evidence; safety review
Pose red without vehicle responseStop testing; runtime assurance defect
Unknown/OOD object suppressed in route corridorStop promotion; model/monitor/planner review
Wrong map/calibration/model activeStop dispatch; compatibility manifest audit
Diagnostic graph missing critical nodeTreat site evidence as incomplete until fixed
Airport operations change route/stand rulesRe-run impact analysis and affected tests

Handover Package

ArtifactOwner
Commissioning report and route/ODD approvalDeployment lead
Active manifest and rollback setRelease manager
Map QA and source-session provenanceMapping owner
Calibration package and drift thresholdsCalibration owner
Safety-case delta and residual risk decisionSafety lead
Alert runbook, escalation roster, and training recordFleet operations
Dashboard screenshots/sample event joinsFleet SRE
Maintenance and recalibration SOPMaintenance lead
  • 70-operations-domains/deployment-playbooks/deployment-playbook.md
  • 70-operations-domains/deployment-playbooks/multi-airport-adaptation.md
  • 70-operations-domains/airside/operations/ground-control-instructions.md
  • 50-cloud-fleet/map-operations/map-publication-gates-airside-hygiene.md
  • 40-runtime-systems/software-operations/sensor-calibration-fleet-ops.md
  • 50-cloud-fleet/observability/perception-slam-alert-runbooks.md

Sources

Public research notes collected from public sources.