A Technical Analysis from the Kingston Doors Engineering Team
The Problem: The Entrance as a Friction Point
For decades, the automatic sliding door has been a study in electromechanical compromise. A standard IR sensor array creates a crude, conical detection zone. The door must wait for a human to breach this virtual plane, then react with a predictable, fixed-speed cycle. This creates a fundamental lag—a moment of hesitation where the user must acknowledge the door’s mechanism. In high-traffic scenarios, this leads to the “dance”: individuals adjusting stride, groups fragmenting, and throughput bottlenecking. The entrance remains a conscious transaction, a point of friction and energy transfer between conditioned interior and exterior environments. Our engineering mandate for the 2026 platform was not to make the door faster, but to make the entrance event disappear.
Structural & Envelope Specifications: The Physical Foundation
Before addressing the sensory and cognitive layers, the physical assembly must be re-engineered for silent, reliable, and robust operation. The 2026 platform is not a sensor package bolted to a standard door; it is a unified system.
| Component | Specification & Rationale |
|---|---|
| Leaf Construction | 3.0mm 6063-T6 aluminum extrusion, thermally broken with 34mm polyamide bar. Rationale: Provides the rigidity required for 2400mm clear height with minimal deflection (<1/500 of span under 80 Pa wind load), while the thermal break eliminates condensation bridges and reduces HVAC coupling. The mass (approx. 28kg/m²) is optimized for rapid, controlled acceleration. |
| Glazing | 12mm laminated low-e glass (0.05 emissivity) with a fritted pattern at head and sill. Rationale: Lamination provides inherent safety and damping for vibration. The low-e coating is spectrally tuned to be transparent to the 77GHz FMCW radar bands and 850nm LiDAR used by the sensor suite, while reflecting infrared heat load. |
| Sealing System | Triple-stage perimeter seal: Outer brush seal, primary hollow-cell EPDM (Ethylene Propylene Diene Monomer) with Shore A 60 hardness, and internal magnetic compression seal. Rationale: EPDM provides superior weather resistance, ozone stability, and elastic recovery over a -40°C to +70°C operational range. The staged design achieves an air infiltration rating of ≤0.5 cfm/ft² at 75 Pa, decoupling the interior environment. |
| Drive & Control | Dual 3-phase BLDC motors with integrated absolute encoders, driven by a distributed CAN FD network. Rationale: CAN FD allows high-speed (5Mbps) deterministic communication between the motor controllers, the central processing unit (CPU), and all ambient sensors. Torque is vector-controlled, enabling speed profiles that are adaptive to crowd density and environmental conditions. |
The Cognitive Layer: Sensor Fusion & Predictive AI
The mechanical system is the actuator. The intelligence is provided by a three-tier sensor architecture feeding a dedicated neural processing unit (NPU).
1. Long-Range Ambient Sensing (6-15m)
A ceiling-mounted 77GHz millimeter-wave radar array provides initial tracklets. It measures approach velocity, vector, and group clustering in all weather conditions. Concurrently, a wide-angle 4K optical flow sensor (not a camera) mounted in the header processes anonymized pixel-group movement patterns, feeding a trained model that distinguishes a person walking toward the door from one walking parallel.
2. Short-Range Intent & Biometric Verification (1-6m)
Here, low-power ultrawideband (UWB) transceivers in the door frame communicate with secure mobile credentials or employee badges, providing cryptographically signed identity and precise location within 10cm. For higher-security applications, a 3D time-of-flight (ToF) sensor performs passive gait analysis. Our model, trained on over 500,000 anonymized gait cycles, can identify an authorized individual with 99.2% accuracy based solely on their walk kinematics—no facial recognition required. This is continuous, hands-free authentication.
3. Micro-Adjustment & Safety (0-1m)
Four solid-state LiDAR units (two in each jamb, two in the header) create a dynamic, real-time 3D point cloud of the threshold. This detects a child crouching, a trailing suitcase wheel, or a hand reaching back. The NPU fuses this with capacitive proximity sensors embedded in the door edge seals. The safety-rated PLC (IEC 61508 SIL 2) maintains a redundant, independent check on this data.
The NPU runs a temporal convolutional network (TCN) that models the predicted trajectory of all entities in the sensor envelope. It does not simply detect presence; it calculates a probabilistic path. The door’s motion profile—acceleration, peak speed, and deceleration—is generated in real-time to intersect the user’s path at the exact moment the clearance space is optimal, creating a seamless flow.
Critical Engineering Trade-Offs
This integration forces explicit, quantified compromises.
Latency vs. Predictive Accuracy: The AI model’s prediction horizon is configurable (default: 1.2 seconds). A longer horizon allows smoother acceleration but increases the risk of false positives (initiating a cycle for someone who then turns away). We balanced this by implementing a dual-threshold system: the door begins a low-power, silent “pre-stage” movement at 0.8s confidence, committing to full retraction only at 0.95s confidence within 0.5m. This reduces unnecessary cycles by 73% over standard sensors.
Data Richness vs. Privacy: The sensor suite generates vast data. Our architecture is designed for on-device processing. Raw optical flow, radar, and gait data are converted to anonymized vector trajectories within 50ms, then discarded. No personally identifiable information leaves the door’s CPU unless explicitly configured for audit logs in high-security mode. This is a non-negotiable hardware design constraint.
Sealing Integrity vs. Sensor Transparency: The EPDM seal and low-e glass are excellent environmental barriers but challenge sensor placement and signal transmission. We engineered dedicated, hermetically sealed waveguide apertures for the radar and LiDAR, filled with dielectric polymers matching the refractive index of the glazing. The seals are molded around these ports, maintaining the continuous barrier. This increased unit cost by 12% but was essential for performance.
Computational Power vs. Thermal Load & Reliability: The NPU consumes 28W under peak load. This heat must be dissipated in an enclosed header subject to solar gain. We used the aluminum door leaf itself as a heat sink, with a thermally conductive pad coupling the CPU to the 3.0mm aluminum. This passive system keeps junction temperatures below 85°C in ambient temperatures up to 50°C, ensuring a >100,000 hour MTBF for the electronics.
Conclusion: The Invisible Interface
The 2026 platform is not merely an upgrade. It is a paradigm shift from a reactive mechanical barrier to a proactive environmental interface. By fusing millimeter-wave radar, anonymized biometrics, and predictive AI within a rigorously engineered physical envelope, we have successfully decoupled the act of entry from the user’s cognitive load. The entrance, as a distinct event, is erased. What remains is a dynamic, adaptive aperture—a responsive part of the building’s skin that facilitates flow, enhances security, and fortifies the environmental boundary. The result is not a door that works better, but a doorway that ceases to be an obstacle. This is the culmination of applied systems engineering: complexity embedded to produce profound simplicity.