Indoor Positioning Technologies Compared: UWB, BLE, Wi-Fi, QR and Visual

Fixed anchors and a tag exchange very short radio pulses, and the system measures how long each one takes to arrive. Because the pulses are so brief, the timing is precise, which is why ultra-wideband reaches ten to thirty centimetres where Wi-Fi and Bluetooth reach metres. The trade is that every anchor must be installed and surveyed, and every asset must carry a tag.

Camera-based visual positioning is the most accurate of the general-purpose methods, at sub-5 cm median with orientation as well as position. Ultra-wideband follows at ten to thirty centimetres. But accuracy alone is the wrong way to choose: a marker is exact at the marker and useless between markers, and the most accurate system is worthless if it cannot see the thing you need to find.

The licence is rarely the number that matters. For anchor and beacon systems the cost is the device count, the mounting labour, the position survey and battery replacement across the deployment's life, and it scales with floor area and with the number of assets you tag. For a camera-based system the cost is one capture of the building, and most industrial sites have been scanned at least once already for engineering or facilities.

Yes, by inferring position from signal strength against a fingerprint of the building. It is attractive because the access points already exist. The limits are that accuracy is one to five metres, the fingerprint degrades as the space fills with people and stock, and it needs re-surveying when the environment or the network changes.

Yes, and most mature sites do. Tags on sealed and high-count items, vision on large fixed equipment, and on-device tracking for smooth motion between absolute fixes. The requirement that makes it work is a shared coordinate frame, so all of it resolves into one map rather than into separate systems somebody reconciles by hand.

With radio-based systems, yes, and that is the constraint people underestimate. The work scales with the number of things you want to identify, and a tag reports where the tag is rather than where the asset is once tags fall off or get swapped. Camera-based identification works from the asset itself, so nothing is attached and nothing falls off.