SLAM Crowd Engine – A deeper look

Nick Stein, 
/wp-content/uploads/2016/09/cephomba-light-speed-image.jpg

indoo.rs SLAM Crowd Engine™ – The Revolution hits the crowd

indoo.rs SLAM Crowd Engine is the second phase of indoo.rs’ SLAM Revolution! In this phase we extend the SLAM Engine by adding the crowd element to it. This brings self-correcting radio maps that adapts to the environment and improve over time. By leveraging data collected by the crowd – our navigating users, we remove the need for manual maintenance measurements and through analytics receive up to date indicators of the system performance.

indoo.rs SLAM Crowd Engine: Pwoer of the Crowd - gifsoup.com

 

Project indoo.rs SLAM Crowd Engine

Accurate radio positioning requires up to date radio maps with information about radios every 1-3 meters throughout the building. Before SLAM Engine each of these points took around 1 min to measure by manually visiting each point. With SLAM Engine trajectories are measured and data is interpolated instead. This not only speeds the mapping process up by a factor of 10x, but also creates smoother more stable maps. Still as a building changes over time, re-measurement is required every 6-12 months.

To understand the revolution it is necessary to understand what existed before. Manual Fingerprinting, the painstaking method of creating a radio data map through individual measurements being taken throughout a building. Each recorded fingerprint would take 15 seconds to make. That was solved with the release of the well documented SLAM Engine but obviously, buildings change, store layouts are optimized and that leads to having to re-map the venue.

The release of indoo.rs SLAM Crowd Engine will stop the re-mapping issue and build off of the initial map to make improvements and updates from crowd-sourced data. This enables Indoor Navigation installations by indoo.rs to be dynamic and ever improving. Therefore the ease of testing new layouts, or even different setups for temporary events, will be easy and the indoor navigation will be as accurate as before with a static setup.

How we move the Crowd

To harness the crowd, the research team has had to get creative. No solution like this has ever been created before, maybe dreamed about but never pursued and implemented into a real world solution.

  • The Crowd data builds off the initial recordings with all the original ground truths and adds the trajectories from the crowd recordings. This then creates a new iteration of the radio map and this process can be repeated as many times as possible. There has to be a part of the data that remains constant but that is not much, for instance the load bearing walls would be enough, to enable the update.
  • The mobile recordings are cut into segments with a single floor and continuous data. SLAM then fits the positions and steps to a highly precise trajectory in each segment using GraphSLAM. Data for each radio point is interpolated to the radio map position. The results are all the radio points and segments combined into an up to date building radio map.
  • The indoo.rs SLAM Crowd Engine algorithm is well suited for parallel execution. Recordings can be sliced in parallel, slices can be SLAM’d in parallel, the radio map for each Beacon can be interpolated in parallel; all to increase the efficiency and efficacy of multiple, simultaneous recordings.
  • Single threaded execution is already faster than real time for recordings, but with larger data sets parallelization is needed for scaling to be a reality. The Crowd recordings are done on the indoo.rs Navigation SDK as opposed to the mobile toolkit, which is where the initial recording is done.
  • In the cloud our Crowd collector front end receives and commits recordings to the data store.
  • Finally, with all of this extra data that we are receiving through the crowd our analytics can be further improved as the analytics engine leverages high quality SLAM trajectories for visual analytics and data export.

 

Where can you find out more

Download our indoo.rs SLAM Crowd Engine info sheet or meet us in person on our roadshow which is kicking off next week with two events first Tech Crunch Disrupt in San Francisco, followed by ION GNSS+ event in Portland, where you will be able to meet with our CRO Thomas Burgess and talk about all the phases of SLAM. Later in the Roadshow he will be giving a Keynote speech at IPIN in Madrid

indoo.rs SLAM the revolution continues

This is only phase 2. Follow our journey with the crowd all the way to SLAM Crowd Learning which will be released in 2017 but more information will be available very soon.

This is a leap forward for the industry

SLAM Crowd Engine: Leap forward in technology

This is a technology that of course was initially planned to give indoo.rs an extra advantage but has evolved into something that will change the industry as we know it. Scalability has always been a prime problem with Indoor Navigation as installations are time consuming. As we continue on the path to automating the installation and maintenance of maps (radio data maps), scalability will become a reality.

 

Download the SLAM info-sheet for more information

For more on indoo.rs solutions visit our main website indoo.rs

* Header image from cephomba