Removing the need for GPS with AI, Trilateration and the Helium Network
We’ve been busy at Trackpac, looking into the future of trackers to discover what is possible. It’s our goal to have the smallest, longest battery life trackers to enable tracking of anything at a low cost.
LoRaWAN is the perfect low power, long distance network for asset tracking, its reduced power draw over sim card based trackers is impressive, but generally still needs a lot of battery to support the GPS chip.
GPS can be power-hungry because it requires a lot of processing power to determine the location of a device using satellite signals. The GPS receiver in a device must constantly process signals from multiple satellites to determine the device’s location, which requires a lot of computation.
There are ways to make GPS less power-hungry, such as by using more efficient GPS hardware and software, and by using power-saving techniques like turning off the GPS receiver when it is not needed. However, these measures can sometimes come at the cost of reduced accuracy or performance.
What if there was a better way? With the Helium Network we know where the hotspots are located, couldn’t we use that information to work out location based on signal strength (RSSI), Trilateration?
What is Trilateration?
Trilateration is a method of determining the position of an object based on the distances from that object to three or more known points. It can be used to determine the location of a point on a two-dimensional map or in three-dimensional space, such as with GPS.
How is Trackpac planning to use Trilateration?
Using signal strength of LoRaWAN uplinks (RSSI) and knowing where the hotspots that heard the tracker are (all helium hotspots have their location asserted on chain) we can use mathematics and our RF knowledge to work out a location from these known points.
We have trackers live in the field collecting both GPS location and any hotspots that have heard their broadcast — we have all packets enabled on helium meaning that any hotspot that hears the broadcast is reported, not just the first. We have a dataset of nearly 2 million records of locations based on GPS and a list of hotspot that heard it.
We feed this into Tensor-flow (AI) to create a prediction engine that’s very accurate by scoring its predictions based on the GPS data collected.
It can easily spot miss-asserted hotspots and ignore the data, and learn how to improve its results by using this data set and any future uplinks from GPS devices we receive.
How accurate is it? Can we see some examples?
Yes! A live demo is set to follow this article but here are some examples.
The Circle is our prediction, the Marker is the GPS results. This is an early beta and we can refine this even more.
Why not use the LoRa Cloud?
Semtech have the LR1110 chip, an innovative way of listening for WIFI broadcasts and then using data from services such as Google street view, where one of their camera cars drives around listening for wifi broadcasts and tagging that SSID with GPS data to say that SSID is based here.
The LR1110 sends the data containing which WiFi SSID’s its heard to a service like this and gets a prediction back based on the data their cars collect.
This data is served at a premium, increasing costs, we want the opposite, to lower costs.
For the same reasons HiveMapper exists to provide quicker updates to map providers, the data that street-view cars provide can be stale and not be updated for long periods of time.
We’ve also found when you’re in a field with just one WiFi connection available the location can wander quite far, we’ve heard stories of LR1110 based trackers being used to track cattle that have apparently wandered into peoples houses but never left the field.
Whilst devices exist with both the LR1110 chip and GPS as a backup, in practice we’ve seen a failure to swap over leaving huge gaps in location history.
What are your plans for this predicted location system? What can it mean for asset tracking?
We can have blockchain confirmed proof of location (some call it proof of presence). This can provide data that enables smart contracts to automatically complete when an object arrives at a location or prove an object/person was present and is where it’s meant to be.
For Trackpac it means we can build smaller trackers with longer battery life. Small rugged devices with a single SIP (system in package) chip, antenna and coin cell battery that offers great battery life for a low price. How low? Less than $20 for small production batches and in bulk even lower.
It also means we can score and improve traditional GPS systems, reducing GPS drift and providing better accuracy when GPS satellite connections are low. You’ll see this implemented into Trackpac’s services shortly.
Where can I find out more? Can I use this predicted location system outside of Trackpac?
Get in touch, hello@trackpac.io or join our Discord. Whilst this system exists as a proof of concept at the moment with a demo to follow, we aim to roll it out as a service. We want people to be able to use it on chain so we are aiming to make it available via API3.
I want to start using Trackpac, where can I buy trackers?
The Browan tab object locator is the perfect pocket sized tracker, find stockists for Trackpac here — https://trackpac.io/tabs-vendors.html
For the more industrial use case, checkout the Digital Matter Oyster 3 a rugged long life tracker — https://trackpac.io/oyster3-vendors.html
