Image Credit: Jon Rawlinson
Locator 4.0 has just landed
This is a very exciting moment for the indoo.rs team, lots of planning, research, coding and more than a few tears have been shed getting this out to you, the public.
As you are all aware the SLAM Engine™ was only just released but that was just the tip of the iceberg, that was a middle release, one that needed it’s own moment to shine.
That said, 4.0 is based on simplifying the indoor mapping process with some fantastic new usability improvements.
So let’s start with the huge announcement
We’ve implemented a new Kalman filter that improves the accuracy of the locator by up to 15%.
Yes, you are reading that correctly! This means that every single map, every single building, every single recording is now 15% more accurate as we have implemented our own painstakingly researched and programmed Kalman filter. So it great news for all indoo.rs users present or future.
Let’s get technical for a minute
Here is a word from the research department about our Kalman filter
The main innovation as compared to a normal Kalman filter is that we can handle multiple locations with different delay/latency. This allows us to create a low latency high accuracy filter.
The filter tracks the heading and velocity to predict the latent positions at current time. The PDR is used in the filter to help control the velocity/heading estimates. This makes the new filter more responsive to velocity changes than the old one. In principle we can use any step detection algorithm for this input as long as it has low latency and produces step (time, heading) for the steps.
Step Kalman filter
* new kalman filter able to fuse locations with different latency from multi-tap radio signal buffer
* includes step detection from PDR buffer for corrections to speed and heading
Before we had a “combiner” which did the latency thing. The main difference is that the Kalman filter does this job with better velocity estimation (from the PDR).
Now back to the great features
To simplify the whole mapping process, which remember has been sped up amazingly with SLAM Engine (sorry but we are very proud of that), we have added the ability to pre-draw your paths.
This is not to be confuse our dot-on-rails feature which is known in the GPS world as Map Matching. This is when you are using the viewer to navigate through the building where the path is automatically snapped to that path you have drawn for smoother navigation.
A quick way to describe it is that you draw a route and then follow it to make sure you map the entire building.
While that is a great thing in itself, there more usability improvements that have been included into the pre-drawn paths feature.
Once you have drawn your map designating where you will record your ground truths, you no longer have to manually scroll the map with you as you walk. If you have taken measurements before, you will understand how much of a skill that is.
Now you just press start recording and walk to the marked spot on the map and confirm ground truth. This is the drum-roll moment, when you confirm the ground truth, the phone vibrates so you know that ground truth has been confirmed as recorded.
To sum that up the pre-drawn path brings three major benefits to the user:
- Increased time efficiency,
- Enhanced usability, and
- Stress free recording
All part of the measure twice cut once, or in this case, plan, review and record.
All the new features in 4.0
- We’ve implemented a new Kalman filter that improves the accuracy up to 15% on existing maps.
- Mobile Toolkit has a new feature to draw the recording path in advance and walk through it.
- Fixed an issue with Mobile toolkit in which the mobile toolkit sometimes did not record all ground truths for recordings.
- iOS option to ‘alwaysDownloadNewBuildingVersion’ is now set to false by default in SDK. This was causing problems with bundled maps.
- Several bug fixes and stability improvements