WB vs Bluetooth vs GPS: which follows better
“Why not just use Bluetooth?” We get asked this a lot. The answer depends on which Bluetooth you mean, and whether the follower needs a bearing or just a rough distance.
GPS: the wrong tool for the job
GPS tells you where you are on Earth. A follower needs to know where you are relative to it. Different problem.
Standard GPS gives you 3 to 5 meters of accuracy in the open. In a parking structure or under trees, it degrades to 10 or 20 meters. Moving Baseline RTK pushes this down to 2 to 10 centimeters without a fixed base station by differencing two rovers against each other, but two RTK-capable modules cost $400 to $1,000, need an independent radio link between them, take 30 seconds to minutes to converge after a cold start, and still do not work indoors or under dense canopy.
GPS also gives you a position in world coordinates, not a bearing from the cart. To get a heading you would need to fuse in a compass or IMU, and compass headings drift badly near metal structures or under tree cover.
GPS works fine for tracking a car across a city. It does not work for a cart that has to hold 1.5 m behind you at walking speed.
Bluetooth: the honest comparison
Bluetooth is not one thing. The newest versions are much better than the RSSI tracking most commercial followers still ship with.
Bluetooth RSSI (the old way)
Most commercial “follow me” products today use Received Signal Strength Indication. The tag broadcasts a Bluetooth signal, the cart measures how strong it is, and maps signal strength to distance. The physics: signal attenuates with distance, so weaker means farther.
The problem is that everything else also attenuates signal. A wall. A tree. A person walking between the tag and the cart. A change in the angle of your body relative to the antenna. Humidity. The same distance can read at 3 bars or 7 bars depending on what is in the room.
RSSI distance estimates drift 1 to 3 meters in practice. One minute the cart thinks you are 1.5 m away, the next it thinks 4.5 m, and you have not moved. There is no angle information either. The cart has no idea whether you are in front, beside, or behind it.
Bluetooth 5.1 AoA / AoD (the current middle ground)
Angle of Arrival places a multi-antenna array on the receiver (the follower) and measures phase differences of the incoming signal across antenna elements. Angle of Departure flips this: the array sits on the transmitter. For a follower that needs to track a person, AoA is the relevant method. This gives you an angle estimate of 1 to 5 degrees in good conditions, which is comparable to single-array PDOA on UWB.
AoA is genuinely useful. It can tell the cart which direction you are. But it requires a calibrated multi-antenna array on the receiving side, which means the follower needs purpose-built hardware. It is not a software upgrade to an existing product.
Distance estimation in AoA systems is still often paired with RSSI, so the ranging accuracy is typically 0.5 to 2 meters. Better than raw RSSI, but not centimeter-level.
Bluetooth 6.0 Channel Sounding (the new contender)
Channel Sounding arrived in Bluetooth 6.0 (2024) and it is the first real improvement. It uses two methods. The first is Phase-Based Ranging (PBR): the tag and the follower exchange tones across multiple frequency channels in the 2.4 GHz band, and the carrier phase slope across those hops gives distance. The second is Round-Trip Time (RTT), which measures packet time-of-flight as a cross-check. PBR is what delivers the sub-meter accuracy; pure RTT with BLE’s 2 MHz bandwidth would only resolve to a few meters. Because PBR measures phase rather than signal strength, it is far less affected by environmental changes than RSSI.
Early implementations report sub-meter accuracy in line-of-sight conditions, and some vendors claim 10 to 30 centimeters in controlled settings. This is approaching UWB territory.
Limitations: both ends need Bluetooth 6.0 chips. The update rate is 5 to 10 Hz right now, versus our 20 Hz, which matters when the follower has to react within a quarter second. And the accuracy numbers come from chip vendor datasheets, not from anyone running this through a golf course.
So CS is a real contender for distance tracking. For angle you still need a separate AoA array, which adds cost and parts.
We will keep watching BLE Channel Sounding. If the update rate improves to 20 Hz and independent tests confirm centimeter-level accuracy through body blocking, it becomes a viable alternative to UWB for the distance half of the pursuit loop. For now, UWB remains the safer bet for outdoor following where the tag is close to a human body.
UWB: what it gives you
UWB sends nanosecond-scale pulses across a 500 MHz or wider channel. The pulses are short enough that arrival time can be measured with sub-nanosecond precision. That maps directly to centimeter-level ranging.
Two measurements come out of a single UWB exchange:
TWR (Two-Way Ranging) measures distance. The tag sends a poll, the anchor responds, and the round-trip time gives distance. Accuracy is 10 to 30 centimeters in line of sight. This is 10 to 100 times better than RSSI.
PDOA (Phase Difference of Arrival) measures angle. Two antennas on the anchor receive the same pulse at slightly different times. The phase difference gives the bearing, accurate to a few degrees within a 120-degree field of view.
Put together you get (r, theta): distance and bearing in the cart’s own frame, every 50 ms. No coordinate transforms, no compass fusion, no base station.
Side by side
The numbers that actually matter:
GPS BLE RSSI BLE 5.1 AoA BLE 6.0 CS UWB TWR+PDOA
Distance accuracy 3-5 m 1-3 m 0.5-2 m 0.1-0.5 m 0.1-0.3 m
Angle information No No Yes (1-5 deg) No Yes (2-5 deg)
Update rate 1-10 Hz 1-10 Hz 10 Hz 5-10 Hz 20 Hz
Works indoors No Yes Yes Yes Yes
Works under trees Degraded Fading Fading Fading (PBR mitigation)* Yes (attenuation)
Body blocking N/A Severe Severe Severe* Significant**
Cost (tag+follower) $50-500 $5-15 $30-80 $15-40 $30-100
Maturity Mature Mature Growing Early Growing*BLE 6.0 CS body-blocking data is limited to chip vendor claims, not independent field tests. UWB under trees handles multipath well (its nanosecond pulses separate first path from reflections); the concern is foliage attenuation reducing SNR, not multipath confusion. BLE at 2.4 GHz suffers Rayleigh fading in the same conditions.
**Body blocking is a physics problem at all frequencies, not a UWB-specific weakness. A human body attenuates 2.4 GHz (BLE) by 15-20 dB and 6.5 GHz (UWB) by 15-25 dB. The consequence for BLE RSSI is that a 1.5 m distance reads as 5 m or more. For BLE AoA, line-of-sight phase is lost and reflected paths give false angles. For UWB, packets drop or show NLOS delay, but the dual-anchor layout and outlier rejection make it more manageable. It is still UWB’s biggest practical weakness.
Why we still picked UWB
All of these have failure modes. The question is which ones you can engineer around.
GPS has no signal indoors or in forests. Nothing to fix. There is no signal.
BLE RSSI falls apart the moment the environment changes, which is always. You cannot calibrate your way out of physics.
BLE CS looks promising on paper. Nobody has published field test data yet. And it has no angle output.
UWB drops frames when a body blocks the signal. But a dual-anchor layout plus outlier rejection plus the pursuit loop’s smoothing filter absorbs most of that. A few bad frames do not crash the cart. It holds its last good distance, waits for the signal to come back, and catches up. That is a solvable problem. “There is no signal” is not.
UWB is not perfect. Its failure mode is just the only one we can fix.
What we actually run
Our system uses UWB TWR for distance and PDOA for bearing, with a dual-anchor layout covering 360 degrees around the cart. Each anchor has a 120-degree PDOA field of view; in the transition zones between them, the system falls back to dual-TWR trilateration, using both distance readings to triangulate the tag position through the side gaps. The first article covers the ranging and angle math. The second article covers the pursuit loop that turns those measurements into smooth following.
The lidar layer acts as a safety envelope, independent of the UWB tracking layer. UWB answers “where is the person.” Lidar answers “is the path clear.” When UWB loses the person behind a tree and the lidar sees a person-sized obstacle at the same spot, the safety layer takes priority and stops or steers the cart. Keeping the two layers separate means a UWB dropout triggers conservative behavior, not a crash.
That combination, UWB for tracking and lidar for safety, holds up on a parking lot, a golf course, and a trail with trees and people and a cart full of tools that cost more than the cart.
Which one should you pick
BLE CS is worth watching. UWB works today. GPS is not built for this. RSSI Bluetooth should be retired.
Pick wrong and no amount of software fixes it.