1) 93.75% success rate in controlled conditions, 92.1% in a somewhat-realistic deployment scenario - too low for reliability. I wouldn't use something like that to trigger smart home automations.
2) Range hardcapped at ~1m due to how ultrasound works, you can't centralize detection. Their answer is to give everyone in the household a wearable receiver, which is eeeeeeeh idk, doesn't look consumer-friendly to me.
3) Paper suggests a mix of durable and consumable parts for the transmitter. Their numbers show that the 3d-printed PLA cantilever needs to be replaced every 900 cycles or so. Should work fine, but...
4) ...every transmitter pair needs to be tuned per-setup, every time. Not a plug&play in the consumer sense.