-
Notifications
You must be signed in to change notification settings - Fork 3
PurpleRobot
Please go to settings and set these properties:
-
You should do this in a slightly reversed order, to make sure that no data is sent outside of our system (unless you are ok with that...)
-
User ID: anything, not used
-
General data upload settings
- Accept all SSL certificates: false
- HTTP upload endpoint: As listed in the server page,
https://data.koota.cs.aalto.fi/post/purple/$device_id
(make sure it is https) - Only use wifi connection: true (if you want)
-
Probes configuration:
- Enable probes: on, then go through and manually disable every probe, then turn on the ones you want.
- See below for information on specific probes that you may want to try out.
- Here are our suggested probes for general use.
- Hardware sensor probes: Location (frequency: 30 min), Step counter.
- Device Info&Config: Battery probe, Screen Probe, Device in Use.
- External device probes: Wifi Probe (sampling frequency: every 5 min)
- You may experiment with any other probes you would like, but consider the tradeoffs between data collected, privacy, and battery usage.
-
JSON uploader settings ==> Enable JSON uploader: on
- This is what starts sending data.
-
Configuration URL: blank
-
Refresh interval: Never
-
Logging: don't enable logging (in particular, don't enable network logging, the app seems to be able to upload them, and thus they collect and make an error message.) If someone wanted to investigate this...
You should try out different probes, and add information on them to this page. For each probe, we should know a) what raw data looks like b) privacy considerations (what identifiers are in it) c) data quantity d) distributions of typical data.
- BluetoothProbe makes a huge amount of data, most of it redundant. If on, it is over 90% of the data usage, and a lot of it is internally redundant. With postprocessing to remove redundancies, this can be made much smaller.
- The WifiProbe is (maybe) 9% of data usage, if it is on. But this is OK.
- StepCounterProbe has timestamps that are not in unixtime. They seem to be increasing at the rate 1/s, but the origin is wrong. Perhaps the are normalized by boot time, which can be found in RobotHealthProbe.