eGEN

A tool to specify energy-aware self-adaptive location-sensing strategies of smartphone applications. It contains a textual domain-specific modeling language and code generator that covers the domain-specific concepts of location-based applications.

eGEN Overview

eGEN is developed using Xtext and Xtend as an Eclipse plugin. eGEN is targeted for Domain Analyst and Developer. As shown in Figure, the use of eGEN consists of seven steps. eGEN is designed to give domain analysts options to assign values for critical battery level, and sensing-interval based on the application requirements. Further, the code generator of eGEN generates the Java code for developers that could be added to the existing Android repositories to make them battery-aware.

Language Description

Language Element	Category	Description	Allowed Values
BatteryState	Context	refers to the remaining battery percentage of the smartphone	Foreground or Background
BatteryLevel	Context	refers to the charging or discharging status of the smartphone	Charging or Discharging
AppState	Context	refers to the background or foreground execution of the application	Foreground or Background
Threshold_High	Context	used to assign High value for BatteryLevel	Any numerical value
Threshold_Medium	Context	used to assign Medium value for BatteryLevel	Any numerical value
SensingInterval	Feature	refers to the time difference between two subsequent location-sensing requests	Numerical value in milliseconds
DecreasingFactor	Feature	refers to the numerical value that will be used in BatteryAwareFunction	Any numerical value
BatteryAware Function	Feature	refers to the type of change (exponential or linear) in fixing the sensing interval	Numerical value
AdaptationPolicy	Keyword	refers to the definition of adaptation policy	Contains combination of Condition and Adaptation
Condition	Keyword	refers to the definition of context values	Contains the allowed context values
then	Keyword	A separator to differentiate context and feature	NA
Adaptation	Keyword	refers to the definiton of features	Contains all allowed feature values
AND	Keyword	Used to separate each context values and feature values	NA

Sample Adaptation Policy

AdaptationPolicy 01 {
    Condition {
        BatteryState = Discharging  AND
        BatteryLevel = Low AND
        Threshold_High = 63 AND
        Threshold_Medium = 46 AND
        AppState = Background 
    } then
    Adaptation {
        SensingInterval = 250 AND
        Decreasing_Factor = 20 AND
        BatteryAwareFunction = Linear 
    }
}

How to Use?

We have uploaded instructional videos on youtube.

Please watch the first youtube video to do the following:

Download the latest Eclipse IDE Installer

Download the Java JDK 17

Please watch the second youtube video to do the following:

Installation and Setup

> Install the IDE for DSL developers
> Create a Xtext Project
> Write DSL grammar 
> Write the necessary Xtend code

Use the framework to generate code

> Launch Eclipse Application to test the framework
> Create a sample project
> Write a sample adaptation policy 
> Generate Java templates containing code for self-adaptive location-sensing 

Exact Adaptation Policy Instrumented for Subject Applications

AdaptationPolicy 01 {
    Condition {
        BatteryState = Discharging  AND
        BatteryLevel =  High AND
        Threshold_High = 80 AND
        Threshold_Medium = 50 AND
        AppState = Foreground 
    } then
    Adaptation {
        SensingInterval = 3000 AND
        Decreasing_Factor = 10 AND
        BatteryAwareFunction = Linear 
    }
}
AdaptationPolicy 02 {
    Condition {
        BatteryState = Discharging  AND
        BatteryLevel =  Medium AND
        Threshold_High = 80 AND
        Threshold_Medium = 50 AND
        AppState = Foreground 
    } then
    Adaptation {
        SensingInterval = 4000 AND
        Decreasing_Factor = 20 AND
        BatteryAwareFunction = Linear 
    }
}
AdaptationPolicy 03 {
    Condition {
        BatteryState = Discharging  AND
        BatteryLevel =  Low AND
        Threshold_High = 80 AND
        Threshold_Medium = 50 AND
        AppState = Foreground 
    } then
    Adaptation {
        SensingInterval = 5000 AND
        Decreasing_Factor = 30 AND
        BatteryAwareFunction = Linear 
    }
}

Design rationale behind the Adaptation Policies

Android device at any instant of time can be subjected to the following exhaustive list of cases: 
    
    Case1: Battery is discharging and the subject application is in the foreground
    Case2: Battery is discharging and the subject application is in the background
    Case3: Battery is charging and the subject application is in the foreground
    Case4: Battery is charging and the subject application is in the background 

When the device is subjected to experimentation, the maximum battery consumption can be found in Case1 
followed by Case2, 3 and 4. The subject application trails were executed when the battery is discharging
and the application is in the foreground (Case1). A GPS sensing interval of 1 or 2 seconds helps in 
producing accurate distance measurement, and increasing the value to more than 10 seconds worsens the
accuracy of the distance measurement for the length of the track we considered (3km), hence for the
non-eGEN version of the subject application we have decided to set sensing interval to 5 seconds 
that can help in achieving decent accuracy. For eGEN versions of subject applications, We have designed 
simple adaptation policies that tend to reduce the battery consumption as the battery level decreases,
hence when the battery is high (greater than 80\%) the defined adaptation policy focuses more on 
producing accurate location than saving battery and this trend reverses as battery transitions to lower 
levels. The sensing interval behavior with respect to battery level of the device is plotted in the below 
figure. Since, we are confining to a GPS relaxation time of 5 sec (for non-eGEN) to not compromise on 
location accuracy, we decided to draft the adaptation policies such that the varying sensing interval
values when averaged for all the battery levels from zero to hundred would be close to 5000ms. 
The average sensing interval as calculated is 5666ms and we have shown in the below sections that 
savings have been reported through adaptation policies defined without losing the much of location 
accuracy when compared to Non-eGEN version.     

Sensing interval impacted by the change in battery level

Subject Application Selection Criteria

We have used F-Droid and Google Play Store to find out the subject applications. Specifically, we have searched for relevant apps with keywords such as gps logging, distance measurement, path tracker, location service, speed meter, location share, live location tracking, etc. Finally, we have selected the android applications that primarily rely on GPS for their operation. In addition, we have decided to choose applications that might use an accelerometer, geomagnetic sensors, etc., apart from GPS for location sensing. We have considered apps if it is open- source and the source code is available on code sharing platforms like GitHub.

Inclusion Criteria

The following inclusion criteria were applied to filter the relevant subject applications:

1. If the app is written in Native Android code
2. If GPS is primarily used for location sensing
3. If the source code repository published with an open-source license 
5. If the recent most commit is published less than two years
6. If the repository is well documented

Exclusion Criteria

We found 11 candidate applications by applying the above-mentioned inclusion criteria. Further, the following exclusion criteria were applied on the 11 candidate applications:

1. If the app's android support plugin is incompatible with our IntelliJ IDEA version used for instrumentation. 
2. If the app has missing dependencies or other build errors 
3. If it is not compatible with recent versions of Android (such as 10.0 or 9.0)
4. If the app activity crashes while installing and using the Android application  

We applied the exclusion criteria as mentioned above and removed the apps that satisfy at least one criteria. Finally, we found five subject applications that are suitable for instrumentation and conducting controlled experiments.

App Name	Play Store	F-Droid	GitHub URL
GPSLogger	Available	Not Available	https://github.com/BasicAirData/GPSLogger
OSMTracker	Available	Not Available	https://github.com/labexp/osmtracker-android
RunnerUp	Available	Available	https://github.com/jonasoreland/runnerup
OpenTracks	Available	Available	https://github.com/OpenTracksApp/OpenTracks
KinetiE-Speedometer	Available	Not Available	https://github.com/xyz-relativity/KinetiE-Speedometer

The above table shows the list of subject applications and their availability on Google Play Store, F-Droid, and GitHub. The subject applications except KinetiE-Speedometer have more than 5 contributors, 237 stars, and 74 forks. We have chosen KinetiE-Speedometer to verify the usefulness of eGEN generated for applications that do not rely on other position sensing sensors.

Caculating battery consumption in percentage

How to Contact?

For more information about the project and support requests, feel free to contact Kowndinya Boyalakuntla (cs17b032@iittp.ac.in),Marimuthu C (cs15fv08.muthu@nitk.edu.in) and Sridhar Chimalakonda (ch@iittp.ac.in). Please open an issue or pull request if you find any bug or have an idea for enhancement.

License

MIT