More Anchors, Sharper Positions — Positioning That Doesn't Shy Away from Compute (The Science of Accuracy, Part 3)

Figure skating scoring holds an old piece of wisdom: take the scores from a panel of judges, drop the highest and the lowest, and average the rest. It is a safeguard that keeps one or two extreme scores from swaying the result. With only two judges, the safeguard is impossible. It works precisely because there are many.

Positioning follows exactly the same logic. And yet, surprisingly, many systems deliberately choose designs that reduce the number of judges. Today's story is about that paradox — and about a design that goes the other way.

Ⅰ. An Industry That Turns Down the Signal — the Paradox of Convention

When a tag transmits a strong signal, more anchors receive it. As the number of receiving anchors grows, the number of combinations to evaluate explodes. With 4 anchors there are 5 combinations to consider; with 8 anchors, 126; with 12, 715.

So rather than shoulder that computation, a good number of commercial systems choose to deliberately lower the UWB transmit power. Fewer receiving anchors means lighter computation. But as we saw in the previous part, positional sharpness comes from the diversity of anchor directions. Cutting down receiving anchors amounts to throwing away the raw material of accuracy with your own hands. It is the same as sending the judges home.

Ⅱ. Going the Other Way — Maximum Power, Maximum Reception

ORBRO's positioning architecture took the opposite direction. It runs UWB at the theoretical maximum transmit power so that as many anchors as possible receive the signal. And it treats the increased combinations not as a burden but as raw material.

The method is a two-round review.

Preliminary round — geometry screening (GDOP). The system computes the GDOP of every combination the receiving anchors can form and filters out the geometrically unfavorable ones first. Combinations with anchors clustered on one side, or anchors too close to each other, are eliminated here.

Final round — verification screening (residuals). Each combination that passes the preliminary round estimates a position, and the system then checks how well that estimate agrees with the actual measurements. It is like solving a math problem and plugging the answer back into the equation to verify it. The better a combination's verification holds up, the greater the say it is given.

Ⅲ. The Power of the Majority — Why the Median Is Robust

The final step is aggregating the estimates of the surviving combinations. Here, a weighted median is used instead of an average. An average can be dragged far off by a single extreme value (the way one resident with a seven-figure salary pulls up a neighborhood's average income), while a median holds its ground even with a few extremes mixed in.

Why does this matter in the field? Because the troublemakers from the previous part — echoes (multipath) and blockage (NLOS) — contaminate the estimates of some combinations. If there were only two or three combinations, a single contaminated one would ruin the result. But within the statistical consensus of dozens to hundreds of combinations, the voices of the contaminated few are structurally drowned out. The more anchors, the thicker this shield becomes. That is why "maximum power, maximum reception" is not a mere spec race but a statistical design principle.

Ⅳ. Where the Computation Happens — an AI Accelerator Inside the Site

One question remains: where do you run a computation that evaluates hundreds of combinations every time a single tag moves — and do it in real time for hundreds of tags?

ORBRO's answer is not the cloud but the AI accelerator of an edge computer installed on site. Because every computation completes on site, there is no cloud round-trip latency, and positioning continues even if the internet goes down. Location data never leaves the site, which also suits industrial environments with strict security requirements. This architecture supports simultaneous positioning of 1,000 tags.

In short: instead of avoiding computation and cutting the raw material, place an engine on site that can handle the computation and grow the raw material — that is what ORBRO means in its technical white paper by "reinterpreting receiving anchors as a statistical asset."

In Closing

Good positioning is not the art of picking out one good signal; it is the art of gathering many signals and reaching a wise consensus. In the next part, we will look at how a confidence grade is attached to the coordinates produced this way — which is also the secret behind safety systems whose alarms do not misfire.

If you are wondering which positioning configuration fits your site's scale and structure, please contact ORBRO.