- Setting up the beacons in a corridor of size 30m x 2.35m.
- Creating the reference map by getting the RSS at predefined locations. Here, we took 20 measurements.
- Developed an interpolating observation model. (k-NN and simple GP regression: https://www.robots.ox.ac.uk/~mebden/reports/GPtutorial.pdf)
- Pair the observation model with a dynamic motion model (e.g. quasi-constant model)
- Build a PF with the above models
- Evaluate the results by using the model on the test data.
- Beacons used were from kontact.io.
- Product link: https://store.kontakt.io/our-products/27-beacon.html
- Measurements app: Indoor Atlas proprietary app SenseServe.
- Mobile platform: Android 6.0.1
- Mobile device: Nexus 6P
- Not considering the wifi data as the time-stamp recorded is wrong.
- So, we will be using the constant turn - constant velocity model with states being location(x coordinate and y coordinate), velocity (v), angular velocity (omega)
- It was decided to use the coordinated turn model.