Fleet Sensors and Perception

cleaning · calibration · weather · KPIs

Fleet operators do not design the sensors, but they decide how clean and well aligned they are. Sensor health is a depot workflow, and it sets how many vehicles are in service.

Sensor health in the depot dwell window: a vehicle arrives, is charged and cleaned, then gets a sensor check; if it passes it goes back to service, and if it fails it is pulled from service. What degrades sensing: dirt and road spray on lenses and lidar windows; rain, snow and fog on cameras and lidar; low sun and glare on cameras; and sensor misalignment after impacts or repairs. Fleet KPIs: sensor-fault downtime, remote assists per mile and weather dropouts. SENSOR HEALTH IN THE DEPOT DWELL WINDOW Arrive Charge + clean Sensor check Pass to service Fail pull vehicle WHAT DEGRADES SENSING Dirt, spray lenses, windows Rain, snow, fog cameras, lidar Low sun, glare cameras Misalignment after repairs FLEET KPIs Sensor-fault downtime Remote assists per mile Weather dropouts
Sensor design
OEM
Sensor health
Fleet operator
Check point
Every dwell
Hub
Autonomy

Finding

For an autonomous fleet, dirty or misaligned sensors cost more uptime than sensor failures do

A camera smeared with road spray or a lidar window coated in grime degrades perception long before any part fails. The vehicle then slows, calls for remote assistance or stops, and each event takes a vehicle out of service. Cleaning and inspection during the charging dwell is the cheapest uptime lever a fleet operator has.

Robotaxi developers now build cleaning into the vehicle: Waymo's sensor domes have washers and wipers, and Zoox and Baidu's Apollo RT6 integrate cleaning nozzles for cameras and lidar. That handles light contamination on the road; the depot handles the rest. The workflow pairs automated washing with the charging stop, runs a sensor self-test before release, and pulls any vehicle whose sensor-health score falls below threshold.

Repairs are the other risk. Replacing a windshield, bumper or mirror moves cameras and radar, and the vehicle must be recalibrated before it returns to driverless service. Fleets that track remote-assist events by vehicle, route and weather can find recurring contamination and misalignment faster than the OEM can. See Autonomous Fleet Depots for depot layout and Fleet Safety and Teleoperation for escalation rules.

Roles

Who owns what

Sensor responsibilities: OEM versus fleet
AreaOwnerFleet lever
Sensor selection and placementOEM or AV developerSpecify operating climate and duty cycle in procurement
Calibration and fusion logicOEM or AV developerRequire recalibration procedures after windshield, bumper or mirror repairs
Perception modelsOEM or AV developerReport weather and lighting conditions where autonomy drops out
Cleaning and contaminationFleet operatorClean on every dwell; track recurrence by vehicle and route
Fault responseFleet operatorPull vehicles when sensor-health thresholds are crossed
Weather limitsSharedAgree thresholds for suspending service and for remote-assist coverage

Modalities

Sensors and their fleet implications

Sensor modalities for fleet operators
SensorRoleFleet implicationDepot practice
CamerasSignals, signs, classificationMost sensitive to dirt, rain and glare; onboard washers and air jets helpClean every dwell
Lidar3D geometry and depthWindow contamination raises remote-assist load; heated, washable windows are now standard on robotaxisClean and inspect every dwell
RadarRange and velocity in bad weatherRobust, but bumper damage and repairs can misalign itCheck alignment after any front or rear repair
UltrasonicTight maneuvers in depots and yardsShort range; mud and ice block itVisual check
IMU, GNSS, wheel sensorsLocalization and motionGNSS jamming near some sites; wheel-speed faultsMonitor fault codes

Sources

Sources and scope

Scope
Fleet Autonomy Sensors and Perception
Sources
Waymo, Zoox and Baidu vehicle disclosures; ElectronsX fleet operations analysis
Updated
October 2026