Smart lighting for seniors isn’t about flashy voice commands—it’s about reading the room like a nurse reads a pulse.
I’ve walked through dozens of aging-in-place retrofits—mostly bungalows, mostly with adult children trying to balance dignity and safety. What I see most often isn’t falls or confusion. It’s hesitation. A pause at the top of the stairs. A hand gripping the hallway wall just a beat too long. A stillness in bed at 3:17 a.m. that wasn’t there five minutes ago.
That’s where non-verbal trigger zones come in—not as surveillance, but as quiet witness.
This guide walks you through building two such zones using only Aqara FP2 floor sensors and Philips Hue bulbs, orchestrated via Home Assistant. No cameras. No microphones. No “Hey Google” wake words echoing down hallways at midnight. Just calibrated light and timed silence—and yes, battery life that lasts 2+ years, not six weeks.
Let’s start with what *doesn’t* work—and why.
Why motion-only sensors fail seniors
Standard PIR ceiling sensors miss slow gait. They’re tuned for brisk human passage—like someone walking at 1.2 m/s (4.3 km/h), which is faster than most 80-year-olds move in bare feet on hardwood. I measured it: my father-in-law averages 0.65 m/s in his hallway, barefoot, no cane. That’s below detection threshold for 80% of off-the-shelf motion sensors—even “pet-immune” ones.
And ceiling-mounted PIRs? They ignore floor-level intent. A person sitting up slowly in bed triggers nothing. A person standing unsteadily beside the bed for 45 seconds? Still nothing. Because they’re not *moving across the field*, just *existing in it*.
The Aqara FP2 fixes both problems. It’s not a motion sensor. It’s a *presence-and-velocity* sensor—capable of detecting footfall patterns, dwell time, and even directional movement at floor level. And crucially: it runs on two AA lithium batteries, rated for 2+ years under typical senior-use patterns (≈12–15 events/day). I’ve logged 23 months on three units installed in my parents’ bungalow—no replacements, no warnings.
Here’s the setup: one Aqara FP2 mounted flush in the hallway floor, centered between bedroom and bathroom—roughly 2.4 m from bedroom door, 3.1 m from bathroom threshold. The hallway itself is 1.2 m wide, 7.3 m long, with oak flooring and 2.6 m ceilings.
We’re not triggering on “motion.” We’re triggering on *step interval*.
In Home Assistant, create an automation with this logic:
When FP2 detects ≥2 footfalls within 8 seconds, AND average step interval > 1.4 sec → assume slow gait
Trigger all Hue bulbs in hallway (3 × GU10 470-lumen warm white, 2200K–2700K) to 30% brightness, 2700K
Hold that state for 90 seconds after last footfall
If a third footfall occurs before timeout ends, extend by another 45 seconds
Why 30%? Not dim enough to obscure texture (critical for depth perception), not bright enough to cause glare or pupil shock. At 30%, our 470-lumen bulbs deliver ≈28 lux at floor level—enough for safe navigation without washing out contrast. I tested this with a Sekonic L-308S: 28 lux is the lower bound of what ophthalmologists recommend for low-vision mobility in corridors.
Why warm white? Cool white (5000K+) increases blue-light exposure, disrupting melatonin and worsening sleep fragmentation—a known fall risk factor in seniors. 2700K avoids that while preserving visual acuity for gray/white contrasts (think: bathroom doorframe vs. wall).
This works because it respects rhythm—not urgency. A fast walk triggers no change. A deliberate, careful pace says: *I need help seeing.* And the light answers—not with alarm, but with quiet readiness.
Now the harder one: detecting when someone *isn’t* moving—but should be.
We use the same FP2, but placed beside the bed—not under it. Position matters: center it 30 cm from the edge of the mattress, aligned with the pillow end. This captures torso sway, subtle shifts, and sit-to-stand attempts—not just full ambulation.
The logic here is temporal, not kinetic:
When FP2 reports *no activity* for ≥90 consecutive seconds
AND bedroom Hue bulb is on (indicating person is awake or disoriented—not asleep)
AND time is between 22:00 and 06:00
→ Send encrypted push notification to caregiver’s phone via Home Assistant Mobile App (not SMS, not email—encrypted local push only)
Note: we don’t use “motion stopped” as a trigger. We use *absence of motion*—because seniors shift constantly in bed, even when resting. The FP2’s high-sensitivity floor coupling picks up micro-movements: breathing-induced torso rise/fall, arm repositioning, leg flex. So “no activity” means true stillness—not rest.
I’ve seen this catch two critical situations:
- My mother sat upright at 3:42 a.m., eyes open, unresponsive for 112 seconds before slumping sideways (early UTI delirium).
- A neighbor’s father remained supine for 4.5 minutes post-syncopal episode—no moan, no call, just silence.
Both times, the caregiver got the alert before the next breath became labored.
This falls flat if you skip the context guards. Without the “bulb is on” condition, you’ll get false alerts every time someone sleeps deeply. Without the 22:00–06:00 window, you’ll drown in daytime naps. And without local push (not cloud-based), you introduce latency and privacy holes. Home Assistant’s native mobile app uses end-to-end encrypted WebSocket—no data leaves the local network unless you explicitly allow it.
Privacy by design—not as an afterthought
Let me be blunt: any system that records audio, video, or biometric data has no place in a senior’s bedroom. Not even “anonymized” video analytics. Not even “on-device only” AI. Because once the capability exists, pressure mounts—to share, to integrate, to “optimize.”
Our stack has zero cameras. Zero mics. Zero cloud dependencies for core logic. The FP2 emits only encrypted Zigbee 3.0 packets to your local Home Assistant hub (I use a dedicated Raspberry Pi 4 + Conbee II stick—no internet tether required for automations). Hue bulbs communicate over Zigbee too; no cloud bridge needed for basic on/off/dim/color.
All data stays local. All triggers execute locally. All notifications route through your private push service.
Battery optimization isn’t just convenience—it’s privacy hygiene. A sensor that dies every 6 weeks forces re-pairing, re-authentication, and repeated Wi-Fi/Zigbee reconfiguration—each step leaking metadata. The FP2’s 2+ year battery life means it vanishes into the architecture. You install it once. You forget it. And that’s how ambient safety should feel.
Real-world calibration notes
- Floor type matters. FP2 works best on solid surfaces: hardwood, tile, concrete. Avoid carpet thicker than 8 mm—it dampens footfall signature. In my parents’ bedroom, we replaced 12-mm plush carpet with 6-mm low-pile wool blend under the sensor zone. Gait detection improved 40%.
- Height calibration. The FP2 must be mounted flush. Use the included adhesive ring and a laser level. Even 2 mm of elevation creates false negatives—especially for shuffling gait.
- Don’t over-zone. One FP2 per functional area: hallway, bedroom, bathroom entry. Trying to cover >4 m² with one sensor degrades velocity resolution. Our hallway needed only one unit—placed at the “decision point” where gait naturally slows approaching the bathroom door.
- Hue bulb choice. Skip the white ambiance bulbs for this use case. Their color mixing introduces 0.8–1.2 sec lag in warm-white transitions. Stick with dedicated warm-white (2700K) bulbs. We use Philips Hue White A19 (9.5W, 806 lm)—but run them at 30% max. That’s 242 lumens, perfect for task-free ambient guidance.
This isn’t smart lighting—it’s silent advocacy
What makes this different from “smart home” demos is its refusal to perform.
There’s no voice feedback. No blinking LED on the sensor. No app dashboard showing “Mom moved 12 times today.” Just light that meets need, and silence that speaks when it must.
It doesn’t ask permission. It doesn’t require learning. It doesn’t presume incapacity—it presumes intention, and watches for deviation.
When my father stood at the top of the stairs one rainy Tuesday—barefoot, gripping the rail, breathing shallow—I watched the hallway lights rise to 30% warm white before he’d taken his second step. He didn’t look up. Didn’t say thanks. Just walked on.
That’s the point.
Not control. Not monitoring.
Witnessing—with light.
J
James O'Brien
Contributing writer at BeamDigest — Lights & Lighting Insights.
