How Accurate Does Your RTLS Really Need to Be?

Accuracy is often one of the first numbers discussed when buying a Real-Time Location System (RTLS).

Can it locate something to 5 meters? One meter? 30 centimeters? 10 centimeters?

It seems reasonable to assume that the system with the smallest number must be the better system. Sometimes it is. But for many RTLS applications, buying more accuracy than the process needs adds cost without adding much value.

If the purpose of your RTLS is to tell you which room contains a hospital bed, knowing its position to within 10 centimeters probably does not improve the answer. If you need to know which workstation a product has reached, identifying the correct station matters far more than calculating its exact coordinates. Even tool locations in a factory rarely need to be less than a couple of meters (if you are within arm's length from an asset but still can't see it, you may have other problems).

The right question is therefore not:

How accurate can the RTLS be?

It is:

How accurately does my process need to know where something is?

The answer to that question can have a major effect on the technology, infrastructure, and cost of the system you choose.

First, what does accuracy actually mean?

Accuracy describes how close a measured position is to the object's actual position.

Suppose an asset is sitting at a known point (the ground truth) on a factory floor. The RTLS reports that it is 40 centimeters away from that point. The position measurement therefore contains an error of approximately 40 centimeters.

Repeat the measurement hundreds of times and the error will change. One measurement may be 20 centimeters away, another 60 centimeters, another 35 centimeters.

This is why statements such as "30-centimeter accuracy" need more information.

Does the system achieve an error of 30 centimeters or less 50% of the time? 90%? 95%? 99%?

A more useful specification might therefore state:

95% of location measurements have an error of less than one meter.

Now both the size of the error and how consistently the system achieves it have been defined.

Accuracy and precision are not the same thing

Accuracy is also frequently confused with precision.

Accuracy tells us how close measurements are to the correct location. Precision tells us how closely repeated measurements agree with each other.

Imagine an asset that remains stationary while an RTLS takes 100 measurements. If those measurements form a tight group 2 meters away from the asset, the system is precise but inaccurate. It consistently gives almost the same answer, but that answer is consistently wrong.

Another system might scatter measurements around the correct location. Its average position could be correct, but individual measurements vary widely. It may therefore have reasonable accuracy when averaged over time but poor precision.

For an RTLS application, both can matter. A system controlling whether an asset has crossed a boundary, for example, can create problems if its reported position continually jumps backward and forward across that boundary, even when the asset has not moved. More advanced systems may use hysteresis (a software filtering technique) to prevent "ping-ponging" or rapid flickering of an asset's status between being "inside" or "outside" a zone due to such signal noise and measurement errors to mitigate the coordinate jitter along the boundary.

There is also a third term worth knowing: resolution. Resolution describes how small a change the system can report. A software interface displaying coordinates to the nearest centimeter does not mean that the RTLS is accurate to one centimeter, just that a centimeter is the smallest unit of measurement available to the system.

Start with the business decision, not the technology

Most organizations do not install RTLS because they need coordinates data or like shiny new tech. They install it because they need to answer questions.

• Where is the tool I need?
• Which production stage has this assembly reached?
• Has this vehicle entered the correct bay?
• Is this piece of equipment still inside the hospital?
• Did this asset leave a controlled area?
• Is a worker inside a location where an alert should be raised?

The accuracy requirement comes from the spatial difference between the answers that matter.

Suppose two production areas are 20 meters apart. A location error of 2 meters may make no difference because the system can still identify the correct area. Put two workstations 1.5 meters apart and the same performance may no longer be enough.

This means accuracy is not an abstract technology requirement. It is a property of the process you want to monitor.

Sometimes a zone is all you need

Many RTLS applications do not need a precise x-y position at all.

They need zone-level location.

Consider a maintenance operation with 30 tool cabinets across a facility. If the business problem is reducing the time technicians spend searching for equipment, knowing that a tool is in "Hangar 2, Tool Store B" may solve the problem. The exact position within the room may add little value. Likewise in correctional facilities - when an officer raises a duress alarm, they need help in the room or cell where they are located. If help arrives to within 1 meter, but it's on the other side of a concrete wall, that's an issue.

The same principle applies to warehouses, hospitals, factories, construction sites, yards, laboratories, and many other environments.

An RTLS that reliably distinguishes between meaningful business areas may provide more value than one that produces a more accurate coordinate but costs substantially more to deploy.

More accuracy normally has a cost

Achieving finer location accuracy generally requires the system to extract more information from the radio signals it receives.

Depending on the technology, this may involve more accurate timing measurements, wider bandwidth, antenna arrays, synchronized infrastructure, additional reference points, denser infrastructure, more calibration, or stronger requirements for receiver geometry.

The environment also becomes more important.

A system trying to distinguish between positions 10 meters apart has more room for measurement error than one trying to distinguish between two objects separated by 30 centimeters.

Reflections, blocked signals, antenna orientation, people, machinery, walls, and other sources of RF error that may have little effect on a zone-level system can become significant when the required position error becomes smaller.

Higher update rates can add another cost. If an application requires both frequent updates and fine location measurements, tags and infrastructure may need to communicate and process measurements more often. This can affect network capacity and, for battery-powered tags, battery life.

There is therefore rarely a free jump from "good enough" accuracy to much finer accuracy.

You may not need the same accuracy everywhere

One of the most useful questions when designing an RTLS is whether the accuracy requirement needs to apply uniformly across the whole site.

Often it does not.

Imagine a 50,000-square-meter manufacturing facility. Across most of the building, the requirement may simply be to determine which production area contains an asset. Meter-level or even zone-level positioning might be enough. At a handful of inspection stations, however, the system may need to distinguish between products separated by less than a meter.

Designing the entire 50,000 square meters around the requirement of those few inspection stations could increase infrastructure and deployment cost for little benefit.

A better architecture may use different levels of location performance in different parts of the facility.

This might mean broad location coverage across most areas, with finer positioning around work cells, doors, storage positions, inspection points, or other locations where the business process requires it.

Different levels of positioning granularity may be achieved by the density of receiving sensors/locators (more sensors = more measurement data = ability to discard erroneous measurements) or using hybrid technologies, moving from RFID in the warehouse, to BLE on the factory floor, then to UWB on the production line. Some vendors can do this within one system using hybrid tags (devices that combine multiple location technologies in a single device).

Accuracy can also change with the application

The same tracked object may require different location accuracy at different times.

Consider a tool moving through an aircraft maintenance facility. When it is in storage, knowing the correct tool room may be enough. During maintenance, knowing which aircraft bay contains it may be enough. When checking whether tools have been removed from an aircraft before the aircraft leaves the bay, much finer location or presence detection may be required.

The question is therefore not necessarily "How accurately must I track this tool?"

It may be:

"Which location decisions do I need to make about this tool, and what accuracy does each decision require?"

That can lead to a different system design more optimized for performance and cost.

Be careful with accuracy specifications

When comparing RTLS products or technologies, a single accuracy number should be treated as the beginning of the discussion rather than the end.

Ask what the number actually represents.

Useful questions include:

• What percentage of measurements achieve the stated accuracy?
• Was the result measured in a lab, an open test area, or an environment similar to mine?
• Does performance change near walls, machinery, people, shelving, or other obstructions?
• Is the number based on average error, median error, or a specified percentile?
• Does it apply across the entire coverage area?
• What infrastructure density is required to achieve it?
• What happens when one or more reference points cannot see the tag?
• Does performance change with tag orientation or mounting?
• How frequently is that position updated?

A requirement such as "one-meter accuracy" leaves many of these questions unanswered.

A specification such as "95% of location updates must be within one meter of the known position throughout defined production areas under normal operating conditions" provides a much stronger basis for comparison.

Buy the accuracy your process needs

There are applications where centimeter-level location is justified. If a system must distinguish between objects positioned close together, control a process based on exact position, or detect movement across a narrow boundary, fine accuracy may be essential.

There are many other applications where it is not.

The objective of an RTLS project should not be to produce the most accurate coordinates possible. It should be to provide location information that allows the business process to work reliably.

Before comparing RTLS technologies, map the decisions the system needs to support. Determine how far apart the locations that matter actually are. Decide how often the location system can return the wrong answer without affecting the process. Then determine where finer positioning is genuinely required.

You may discover that you need 30-centimeter accuracy in one part of the building, two-meter accuracy somewhere else, and simple room or zone detection everywhere else. That is not compromising the RTLS. It is specifying the system around the problem you are paying it to solve.