AINA-2017

This project was created by Ales Omerzel for the purpose of the research paper:

Pournaras, E., Nikolic, J., Omerzel, A. and Helbing, D., 2017, March. Engineering Democratization in Internet of Things Data Analytics. In 2017 IEEE 31st International Conference on Advanced Information Networking and Applications (AINA) (pp. 994-1003). IEEE.

This code was made publicly available upon request for the reproducibility of results presented in the research paper. The repository was created on 23.4.2020 together with the README.md file and the LICENSE addition. The original code developed in 2016 (and its Git history) that produced results of the research paper is cloned to this repository.

Dependencies

• Weka library: https://weka.wikispaces.com
• GraphView-4.0.1: https://github.com/jjoe64/GraphView

Note Dependencies are already part of the module and can be found under AINA-2017/nervousnet/lib.

Installation

One can download this repo and import AINA-2017/nervousnet as module to get it into existing Android project. However, if you don't have a project yet, follow the instructions below.

Note Instructions are given using Android Studio 2.1.2 (build 143.2915827) on MacBook Pro Retina 2014.

1. Start Android Studio and create new project File -> New.
2. Insert any application name and click Next.
3. Select minimum SDK to API 21: Android 6.0 and click Next.
4. Select Add No Activity and click Finish.
5. Click File -> New -> Import Module ...
6. Search for AINA-2017 repo that you cloned to your local machine and select root directory AINA-2017/nervousnet and click Finish.
7. Connect your Android 6.0 (or later) phone with USB to your computer and make sure your phone is running in developer mode with permission for installation via USB.
8. Now click run. Project will start building.
9. Select your phone as deployment target when Select Deployment Target window pops up.
10. Be aware of any notification on your phone about installation. Allow access if needed.
11. Application will get installed on the phone and start automatically. It is ready for use.

Usage

When application is installed and is successfully running, press Start button to start collecting sensor data.

Pull Data

All sensor data is stored in SQLite database. There are multiple ways to access this database. One way is by using the following command to transfer whole database to your working directory:

adb backup -f ./data.ab -noapk ch.ethz.coss.nervousnetgen
(echo -ne "\x1f\x8b\x08\x00\x00\x00\x00\x00" ; dd if=data.ab bs=1 skip=24) | gzip -dc - | tar -xvf -

How to Explore Fetched Database Data?

Our recommendation is DB Browser for SQLite.

Configuration

Sensors

Application supports light and accelerometer sensor data collection.

Sensors configuration can be found in nervousnet/src/main/assets/sensors_configuration.json. Light sensor configuration is the following:

{
    "sensorName" : "Light_v2",
    "androidSensorType" : 5,
    "parametersNames" : [
      "Lux"
    ],
    "parametersTypes" : [
      "double"
    ],
    "metadata" : [
      "not supported yet"
    ],
    "androidParametersPositions" : [ 0 ],
    "samplingPeriod" : 500000
  }

Field sensorName holds name for sensor and can be anything. androidSensorType defines sensor type according to Android documentation. In this case value 5 represents light sensor.

Then names for each sensor parameter and type are specified in parametersNames and parametersTypes, respectively. androidParametersPositions field is used to map and order values from the sensor output to the parameters specified in parametersNames. For example, accelerometer sensor outputs three values x, y and z axis and such configuration gives us freedom to order axis in the database in a way we want.

One of the most important fields is samplingPeriod which is frequency of sensor data collection in microseconds.

---

Note Re-install application when applying new configuration.

Virtual Sensors

Application supports also virtual sensors. Virtual sensor is similar to a regular sensor and it can combine parameters of multiple sensors. In the configuration can be specified what are the parameters from each sensor and what is parameters' order.

{
  "sensors" : {
    "Accelerometer_v2" : [
      "accX",
      "accY",
      "accZ"
    ],
    "Light_v2" : [
      "Lux"
    ]
  },
  "samplingRate" : 60000,
  "slidingWindow" : 86400000,
  "virtualSensorName" : "VirtualSensorACCLight"
}

Note In order to include sensor in a virtual sensor, it must be configured in nervousnet/src/main/assets/sensors_configuration.json.

Accelerometer_v2 refers to the sensor in nervousnet/src/main/assets/sensors_configuration.json and ["accX", "accY", "accZ"] are selected parameters to be included in a virtual sensor (in this case all of them). Light_v2 is the second selected sensor for the same virtual sensor with its parameter Lux.

slidingWindow is a time period which is used to aggregate all collected data in that period and compute a possible state.

samplingRate is a frequency of moving sliding window in microseconds and each time window moves new possible state is computed.

Possible state is a set of centroids computed by kMeans on collected data in a window.