3. Occupant classification and position sensing

An airbag that deploys with full force into an empty seat, a rear-facing child restraint, or an occupant sitting 80 mm from the instrument panel can turn a life-saving device into an injury source. Occupant classification and position sensing exist to answer two questions before the restraint control module (RCM) commits to a deployment decision: who or what is in the seat, and where are they relative to the airbag module. The RCM fuses these inputs with crash severity (longitudinal/lateral acceleration, delta-v) to select suppression, low-risk deployment (LRD), or full-power deployment.

Four sensing families dominate production vehicles: weight (occupant weight) sensing, capacitive (electric field) sensing, camera-based detection, and seat track position sensing. Each has distinct physics, failure modes, and regulatory roles, and modern systems typically combine at least two of them.

Weight sensing: load cells and bladder systems

Weight-based occupant classification uses either four strain-gauge load cells mounted between the seat pan and the seat frame, or a pressure bladder under the cushion. Load cells measure vertical force at each corner; the classifier sums them to estimate occupant mass and computes the center of gravity (CG) from the four readings. A typical passenger seat might read, for example, 0 kg (empty), 5–25 kg (child restraint or small child), and above approximately 45–50 kg (adult, 5th percentile female and larger) — thresholds are calibration-specific and must never be assumed from memory.

The CG calculation is the diagnostic gold standard. With corner loads FL, FR, RL, RR and known seat geometry (front–rear spacing Lx, left–right spacing Ly), the longitudinal offset from the seat center is:

x_cg = ((FR + RR) - (FL + RL)) / (FL + FR + RL + RR) * (Lx / 2)
y_cg = ((FL + FR) - (RL + RR)) / (FL + FR + RL + RR) * (Ly / 2)

A concentrated load (knee pressed into the cushion, a bag of groceries) produces a CG far from the expected occupant centroid, which is why CG gating suppresses false "adult" classifications. Practical checks: verify each corner reads within tolerance under a known test weight, confirm the sum tracks the applied mass, and confirm CG stays inside the seat envelope. Common failure symptoms include a persistent "passenger airbag off" telltale with an adult seated (drifted load cell or connector corrosion), intermittent classification changes over bumps (loose seat-frame bolts altering preload), and a CG that jumps between samples (damaged bladder or wiring harness chafe).

Capacitive sensing: presence and proximity

Capacitive occupant detection places one or more electrodes in the seat cushion and measures the change in capacitance as a body (which is conductive and grounded through clothing and the vehicle) approaches. The sensing chain is: electrode → guard ring → capacitance-to-digital converter → classification algorithm. A guard electrode driven at the same potential as the sense electrode shields against the seat frame and nearby metal, improving signal-to-noise ratio.

Capacitive sensing is fast (millisecond response) and can detect proximity, not just presence, which makes it valuable for out-of-position (OOP) detection and for distinguishing a rear-facing child restraint (low coupling, high ground isolation) from an adult. However, it is sensitive to moisture, spilled liquids, and aftermarket seat covers, and it cannot reliably estimate mass. In practice, capacitance is used as a complementary channel: it confirms presence and proximity, while weight sensing provides the mass estimate.

Camera-based detection

An interior camera (typically near the rearview mirror or overhead console) with an infrared illuminator captures the passenger compartment and runs a classifier to estimate occupant presence, posture, and head/chest position. Modern systems output a bounding box plus keypoints (head, shoulders, chest) and a distance estimate to the airbag module. Camera-based sensing is the only channel that directly observes posture — leaning forward, arms on the dashboard, child standing — which is exactly the OOP condition that weight and capacitance cannot resolve.

Engineering considerations: IR illumination is required for night operation; the classifier must handle occlusion (blankets, pets, cargo), and the system must fail safe if the lens is blocked or the camera faults. Camera outputs are typically fused with seat track position to convert image-space distance into a physical standoff distance from the airbag module.

Seat track position sensing

Seat track position is measured by a Hall-effect or potentiometric sensor on the seat rail, reporting fore–aft position in millimeters or as discrete zones (e.g., forward/mid/rear). It matters because airbag standoff distance depends on seat position: a seat moved fully forward places the occupant closer to the module, increasing injury risk from a full-power deployment. The RCM uses track position to select deployment energy and, combined with occupant classification, to decide suppression versus LRD.

How the channels combine

The figure below shows the fusion logic: each sensor contributes a distinct piece of evidence, and the RCM resolves them into a deployment state. Note that no single channel is authoritative — the decision is a weighted combination with fault handling.

Practical fusion and verification

A representative fusion rule set (illustrative, not a specific OEM calibration) might read: if weight < child threshold and capacitance indicates low coupling, classify as child restraint and suppress; if weight indicates adult but camera reports head-to-module distance below a standoff threshold, select LRD; if seat track is fully forward, reduce deployment energy regardless of class; if any channel reports a fault, fall back to the safest default (suppression or LRD) and illuminate the telltale.

Verification follows the same logic. Using a seat with known calibration, apply test masses at known positions and confirm the weight channel's mass and CG outputs; place a grounded test body at several distances and confirm capacitance proximity response; position a test mannequin at nominal and OOP postures and confirm the camera classifier's keypoint output; move the seat through its full track range and confirm position reporting. Then check the fused output at each combination — the RCM should transition between suppress, LRD, and full deploy exactly at the calibrated boundaries, and should enter the fault default when any sensor is disconnected.

The key engineering insight is that these four channels are complementary by design: weight answers how much, capacitance answers how close and how coupled, camera answers what posture, and seat track answers where the seat is. The RCM's job is to reconcile them into a single, fault-tolerant deployment decision — and the technician's job is to verify each channel independently before trusting the fused result.