Calibration, Cleaning and Sensor Health

A sensor can be healthy electrically and still deliver unusable evidence because it is dirty, blocked, misaligned or out of time. Calibration, cleaning and self-diagnosis are therefore part of the safety function.

Geometric calibration

Perception software needs the position and orientation of every sensor relative to the vehicle and to the other sensors. This lets it project a camera pixel, radar return or lidar point into a common coordinate frame.

Windshield replacement, bumper repair, suspension changes, body damage or a disturbed mounting bracket can alter that geometry. Even a small angle error grows with distance. The result may be a lane boundary shifted sideways, an object track that does not line up across sensors, or a braking target assigned to the wrong path.

Some vehicles calibrate dynamically while driving; others require a workshop procedure, targets or controlled conditions. A successful calibration routine means its internal acceptance criteria were met. It does not prove that every mounting or environmental problem has been eliminated.

Time synchronization

A moving scene also requires accurate timestamps. If a camera frame and radar observation are combined as though they occurred simultaneously when they did not, the same vehicle may appear at different positions. The error grows with relative speed and latency.

Production systems compensate for known processing delay and ego motion. They also need to detect clock faults, delayed messages and stale data. Timing is especially important for sensor fusion.

Contamination and obscuration

Different sensors fail differently:

Cameras lose contrast or detail through mud, salt, water, ice, condensation and glare.
Lidar can receive strong returns from droplets or a contaminated cover and lose useful range.
Radar can be attenuated or distorted by coverings, ice, water films or unsuitable repairs over the radome.
Ultrasonic transducers can be blocked by dirt, snow, ice, paint or accessories.
Thermal cameras still need a clear optical path and can lose contrast when target and background temperatures converge.

Sensor heaters, washers, hydrophobic surfaces, air jets and favorable placement can reduce exposure. They cannot guarantee a clean view in every condition. Waymo describes active cleaning and weather management as part of its all-weather system work at Waymo: creating an all-weather Driver.

The vehicle must know what it cannot see

A safety-relevant system should monitor signal quality and compare sensors, motion estimates and expected scene structure. Detection methods include image contrast checks, blocked-view classifiers, radar or lidar return statistics, calibration consistency and plausibility checks across modalities.

Self-diagnosis has limits. A sensor may produce plausible but wrong data, or several sensors may share the same obstruction. The system therefore needs defined responses at several confidence levels: continue normally, reduce performance, warn the driver, disengage, or execute a fallback maneuver.

Tesla's owner documentation lists dirt, weather, glare and low light among conditions that can degrade camera-based features. See Tesla owner's manual: camera limitations and troubleshooting. Its service documentation also emphasizes camera calibration after relevant work at Tesla service manual: camera calibration. These warnings are useful evidence about failure mechanisms, not a comparison with other manufacturers.

Repair and maintenance matter

Crash-avoidance sensors are often mounted in windshields, grilles, mirrors and bumpers: the same parts commonly replaced after minor damage. The Insurance Institute for Highway Safety found that post-repair problems with crash-avoidance features were not rare among surveyed owners whose vehicles had relevant repairs. See IIHS: crash-avoidance features and post-repair problems.

Owners and repairers should follow the exact procedure for the vehicle, hardware revision and software version. A visually correct bumper or windshield installation can still require calibration. Paint thickness, aftermarket accessories and nonstandard brackets can affect some sensors.

Testing degraded sensing

Testing only a clean, correctly aligned vehicle in good weather measures the best case. A credible validation program also covers partial obstruction, dirt, rain, fog, snow, low sun, night, vibration, calibration drift, sensor loss and recovery.

NHTSA research has examined the safety implications of ADAS sensor degradation. See NHTSA vehicle safety research: ADAS sensor degradation. The relevant outcome is vehicle behavior: missed hazards, false warnings, false braking, unstable control, driver notification and fallback. A perception confidence score alone is not enough.

What drivers should take from this

Keep sensor areas clean using the manufacturer's instructions.
Treat persistent “camera blocked,” “radar unavailable” or calibration warnings as functional limitations.
Ask whether sensor calibration is required after glass, bumper, mirror, suspension or alignment work.
Do not assume another sensor will compensate unless the vehicle documentation explicitly says the function remains available.

Sensor maintenance is not cosmetic. It protects the measurement chain on which the assistance or automated-driving function depends.