Telemetry In Mobile and Ubiquitous Computing

1. Types of telemetry

1. Obvious sensors like accelerometer
2. Hidden sensors like touching screen or clicking buttons
3. Internal measures like CPU utilization.
   It’s practically that could be used as collecting data for device.

2. Ad and DisAd

Ad:

1. telemetry collects data in the background and minimize the influence on observing objects , a person for instance.
2. Telemetry collects data continuously as long as the device is on.
3. Telemetry provides fine-grained(detailed) data for any tiny bit of fluctuations.

DisAd:

1. Telemetry data is noisy for the sensors are always detecting data and cause the signal to be buried.
2. Telemetry could lack the construct validation in some cases. (is not widely applicable)
3. Privacy doubts: 1. People are not told how telemetry is collecting their data and how these data would be used. 2. There is power asymmetries between data collectors and people been observed, this caused surveillance for bosses and companies over common people.

Sensors in Android

The combination of virtual sensors for aggregated scenarios $SensorManager$ and real sensors for raw data $SensorEvent$

AWARE

Mostly for collecting users data. Require accessibility for OS.

Clients : (Opearting System) IOS, Android, macOs devices
Server: (Stack) Linux, PHP, MySql
Ad: Could store the data on local machine for security.

Plug-in Models:
Could create new sensors for specific activities.
Ad: Better chance for getting the data without too much noise.

MQTT(Message Queue Telemetry Transport)

Machine-to-machine protocol, using publish-subscribe model, takes up smaller resources on CPU, memory and network traffic. Therefore, mostly used in IoT devices embedded in Android System.(Only supported by AS)
Beaten by HTTP on ubiquity but superior in limited resources.

Duty Cycling

The data is not free. For saving battery energy as least, telemetry could sample data on a interval. This is for cases where critical data is predictable or there is a specific timing for interest.
Yet not for continuously dynamic events.
Adapted duty cycling:
Change the sampling rate dynamically depending on whether an event is detected or not…
The rate increment vary. If in the case where more events are expected, rapid change of rate is needed.