Subscribe to AllThingsTalk newsletter

Get updates on our developments and the Internet of Things in general

Thanks for subscribing to our newsletter.


To complete the subscription process, please click the link in the email we just sent you.


OK

Subscription Confirmed


Your subscription to our list has been confirmed.

Thank you for subscribing!



OK

If you didn't already know, localisation services are core for developing context aware IoT apps, and we have another top secret project to share with you today.

A Machine Learning study we're working on, concerns algorithms for intelligent geo-services, to automatically identify places of interest, common routes and estimated arrival times. All of which contribute to the AllThingsTalk Smart Data components offering.

Our escapades into geo-services started with geofencing features, some of which found their way into SmartLiving, however we weren't satisfied, so we took them back to the drawing board. So what is the problem with geofences?


  1. Radial fences monitor unnecessarily huge areas if not configured in a granular way (A 25km fence around your home may span an entire city!)
  2. Polygon fences work better, however accuracy is not optimal for Just-In-Time events
  3. Responsiveness is key, however sensors must work at higher frequencies which is bad news for battery life
  4. Geotrigger fences tend to fire at dumb times (You don't want heaters to turn on as you arrive home, nor do you want notifications about forgotten keys once you've left)

The solution

We believe in sensor fusion techniques (Aggregating different data sources) and predictive insights, combined to configure sensors to adapt autonomously to support dynamic frequency readings (Only read sensors when the context is valid). The new services being developed provide three important features:


  1. Understand and identify places of interest and common routes for things you're tracking
  2. Predict which place of interest a thing maybe travelling towards
  3. Estimate arrival time based on predicted common routes a thing maybe taking

OK, it all sounds very fancy but what can it actually do, and what does it solve for you? Here's a short list of use cases and automation rules you could configure:


  • You'll no longer need to manually enter places of interest or define common routes (after the learning period has passed)
  • Get real-time accurate arrival times for items which are being delivered using common routes
  • Intelligent returning home automation rules: When I'm returning home, then turn on the heating, when and only when, I do not stop at the store or any other point of interest
  • Notifications rules only when you need them: When my children leave for school, then send notifications, only if they deviate from the common route
  • Notify others when you're running late: When I'm running late for a meeting, then notify other attendees of my estimated arrival time

Are you working on geo-awesome applications for Smarter track and trace solutions? We'd love to hear more and see how we could help you, get in touch info@allthingstalk.com.