DWM1001 expected precision

We want to use DWM1001C but we have particular needs, and we are not able to find precise information regarding the scenario we are facing.
The goal is to localize a slow moving object (up to 1m/s max) in a 3m x 3m area with 4 (or 3) accurately placed anchors. The module datasheet reads a precision up to 10cm which is not enough for our purpose. Does anyone has an application case close to ours ? What is the precision ? We would like <5cm (dream would be <1cm) is it possible ?

Thanks !

With the standard code as supplied you will have trouble reaching that accuracy. It may just be possible but it will depend a lot on the environment and exactly how you set the system up.

The underlying technology is certainly capable of <5 cm accuracies if you apply the correct positioning algorithms but you may need more than 4 anchors and it will require some work. The decawave firmware is a compromise aimed to be flexible and do most of what most people want but unfortunately that means it isn’t as good as it could be if you optimize things for a specific application.

1 Like

Yes, this is possible assuming other constraints you have don’t prevent it.

They key to doing it would be to use TWR based location (which is more accurate with low anchor counts than TDoA), and to take lots of readings and smooth them. This will get you high precision, then you can convert that to high accuracy with an error mapping.

We see about 2.5 cm standard deviation on TWR for a single hit for our devices. If you do a burst of 25 of them and average, that will result in about 0.5 cm standard deviation for the raw range results. Then you can compute the location.

The main limitations are that you have enough power and low enough update rate to support taking multiple measurements.

In our slotted multirange system we are working on, a 4 anchor measurement to one tag will take about 1.4 ms, or 700 measurements (range to each anchor, 2800 total range values) per second. If you average 25, that’s about 25 net updates per second to allow from some network management time.

There will be some fixed distortions due to antenna variances and other factors. To convert the precise measured location to an accurate one, you have to create an error mapping. The better the antennas are, the less natural error there will be.

Net result would be 25 Hz measurements based on 625 Hz raw locates, with a std dev of about 0.5 cm, thus >99% of readings will be within 2 cm of a repeatable position. Higher updates per output and more anchors improve the results.

When someone asks, “How accurate is UWB?”, I say “How much money have you got?”. You can always put an anchor every meter and make it more accurate.

Mike Ciholas, President, Ciholas, Inc
3700 Bell Road, Newburgh, IN 47630 USA
+1 812 962 9408

When someone asks, “How accurate is UWB?”, I say “How much money have you got?”.

“Accuracy is a function of price” was a favorite saying of my manager at Trimble over 15 years ago. He was talking about GPS at the time but it’s true for just about any positioning technology.