theheraldproject / herald-analysis Goto Github PK

• 1cm step data covering range 0-350cm
• Improve segmentation algorithm to be less wasteful with data around segmentation boundaries
• Time synchronisation between devices to further reduce waste

• Publishing experiment data
• Update to Edison robot script to match high resolution data capture procedure, and reduced audio signals

[Who] As a Developer / Tester

[What] I need to frequently run regression tests, and gather test logs on many real devices.

[Value] In order to gain assurance that the software operates correctly across target devices. This is a repetitive, labour intensive, and time consuming task.

Describe the potential solution you'd like

After deployment of software on target device, use a central test server to remotely control start/stop of devices, and also collect test logs from all devices. The test server will enable scheduling of these actions to automate the existing test process.

Describe alternatives you've considered

Using an existing mobile app test automation solution such as Appium, but it is unclear if these solutions function correctly for primarily backgrounded app with little/no UI interactions, and also provide control over BLE settings.

It's getting more and more difficult to find issues and their root causes. Create a tool to find known failure modes, as well as unknown mode log file activity correlation, to aid rapid diagnosis.

• New reference data sets
• Automated extraction of reference data set from raw phone log data

• Current top level README points to out of date RSSI results page
• Changing links to new page

Would it be possible to add a few pre-generated graph/ images to the README file to illustrate to the reader what the analysis reports will provide.

Also could make the reporting issues a link to https://github.com/adamfowleruk/ping-analysis/issues.

Copy the calibration analysis R script from herald-calibration and use here to perform analysis across each test run from the calibration robot to determine formulae for distance conversion for known handsets.

• R script updated to handle situations where a phone detected nothing
• Required for cross-protocol testing where the legacy protocol cannot detect HERALD but HERALD can detect legacy protocol

Publish transmit power for Android devices (iOS devices do not offer this information)

Placeholder for publishing new releases

• Get distribution charts showing at auto configured size and height
• Ensure all charts aren't too big in image or file size
• Run for both phone A and phone Bs data on each run
• Spot missing result files and recreate intelligently (not all files, all the time; not single test only)

• Document observations, investigations, and findings for 1cm resolution RSSI-distance data

• Optional time shift is currently set to 1 hour by default
• Locale of TestAnalysis.R and Android test phone logs are likely to be the same
• Recommend setting default to 0 hour (disabled) as current default shifts Android phones log timestamps by 1 hour leading to confusing chart results

• New data captured on updated test environment
• Latest data include test run in noisy environment
• Improved WEKA interoperability with updated batch CSV to ARFF converter

Parse the raw contacts.log over a period of time to generate analyses useful for self calibration work.