[Suggestion] Tracking based on movement profile

[detrophy]

👋
I've got an idea swirling around my head of a feature(set) that will increase / decrease GPS accuracy & time and distance between each recorded geopoint based on:

• Start of recording

  • Short moments of movement (with guardrail)
    Time is set manually, default value ~30sec (?) of constant movement before starting recording.
    First geopoint is recorded at ~10m accuracy GPS.
    Examples:
    • Walking from [the room] to [the exit] and start being outside (hiss at the sun while looking into the sky, because being a vampire and quickly turning into ash!) to go somewhere / go to the vehicle and start driving (with / on) it.
    • Walking from [the room] to [another room / outside place connected to building] and [doing something], not going to some far away place.
  • Coming to a stop for some time
    Depending on historical speed, increase time between geopoints
    Examples from "long time to time increase" to "short time to time increase":
    • A car will stop for pedestrians, other cars, red lights and other things. This can take some time.
    • A bike will also stop for all those things, but has no issues with traffic blocks and user is able to walk for some time
    • A pedestrian has almost no hindrances besides red lights.

• End of recording

  • Reaching the destination
    Reaching an already visited place / area and reside there for more than 30 minutes stops the tracking. Within the 30 minutes, time between geopoint is increased to 30 seconds, distance to at least 50m.
  • Charging within a known space
    Most of us charge the phone while either at home and at night, or at work. There is absolutely no need to track the phone while bound by a physical connection. -> Can be overwritten by condition for "constant movement" or "Android Auto connection". I know some people who walk around with a power bank connected to their phones all the time.

• Changes in acceleration / deceleration, lateral forces, speed (with guardrail)

  • Train and plane
    They tend to go into one direction for quite some time at around the same speeds. No good reason to stay on high frequency recording.
    If the vehicle is either decelerating or turning to some degree, frequency of points should be higher and the distance between points shorter to get a higher resolution of the change in direction and resulting changes in speed.
    A limit is needed. There is no need to record geopoints of a plane or train, with 200km/h within 25m of geopoint distance.
  • Car
    Cars are some kind of a mixed bag depending on where they drive, but the faster they go, the greater the distance between each geopoint can be.
    In cities, they go quite a lot slower, reducing the distance between geopoints is great for resolution of speed and changes in direction.
  • Pedestrian or bike
    They tend to quite more often change direction and speeds.
    Like with trains, planes or cars, but on a smaller scale.
  • Recognising changes
    To quickly change accuracy, time and distance in accordance with the movement of the phone it will help to know the direction the phone is facing. This could be possible with the built in sensors.
    • pocket, hand, bag or neck strap as an example. All those have the phone face a different direction which may or may not change, as the user may pick it up and / or use it

• Places / areas (I haven't understood the exact function of areas yet, so I will include it) or connected WiFi / visible WiFi

  • If we're at home, there is no need to waste battery. -> Overwritten by condition for "constant movement" or "Android Auto connection", since some WiFi tend to have quite the coverage.
    Thought: Tracking may then also be started when doing launrdy, vacuuming the place. A guardrail may be needed.
  • If we're at work, it depends (a toggle would be necessary) where were working.
    If you're in the woods, construction work or a driver, having good accuracy could be a good thing (still constantly hissing at the sun!).
    If you're an office worker, not so much.

• Android Auto connection

  • Movement profile is "car", start tracking 10 seconds after the car is constantly moving
    Value would need to be user changeable, to reflect special needs.
    Examples:
    • Moving the car between parking spots
    • Moving car out of the garage for changing tyres, clean, maintain and other things
  • If Android Auto connection is not available anymore, automatic switch to "walk"

[sunstep]

Hey, firstly, thank you for the support. It is truly appreciated!

This suggestion is a mouthful, so I'm going to try to take this in small steps.

From what I am reading, it sounds like these are ideas to make auto mode tracking more efficient. Along with some additional profile behaviors that affect the tracker (such as android auto, wifi etc.).

These are great ideas! And I'll definitely take them into account while developing the app. In this comment I will focus especially on the start and end of recording sections, and the rest will probably come later.

I am in fact currently working on efficiency in the auto mode tracker. My biggest focus right now are stopping battery consumption while being stationary and ensuring the tracker doesn't get randomly killed by the OS (which is very unexplainable to me why it happens on the current beta version).

Stopping battery consumption while not moving is the most important one, the reason being is that any battery used while you don't move is wasted battery. While I am still working on this, my current solution for this seems to be using no location at all while being stationary. There is an API which detects user movement, which is very cheap and helpful to make the tracker jump to a more active tracker mode. Instead of letting the app itself figure out when the user starts moving, we rely on events in the system.

Detecting 30s of constant movement is not the most efficient approach as it needs GPS context to be able to detect this, this is basically the issue I have been fighting lately, which is using GPS just to detect the fact that you're moving, which wastes battery while you're not moving. My goal is 0 - insignificant battery usage by not needing any GPS at all to start recording. The GPS should only be involved when actually tracking.

Currently I have set the tracker to go to sleep after 1 minute of inactivity, this would be the point where it stops recording to save battery. This value might change depending on how reliably the activity recognition to start tracking works but by putting the tracker to sleep fast, we avoid using the location for too long which is another counter measure against high battery usage.

I am currently also experimenting with a second mode in between active and passive which is monitor mode, which gives you more time to move before going back to passive, it should be a mode that is intended for stop/ go traffic and should be more responsive than passive mode. Monitor mode would last about 3 minutes before going back to passive mode.

Once start and end of tracking works reliably enough to properly detect if you're actually travelling or are stationary, I am willing to look at the next ways to make the tracker more intelligent such as the ways you suggested.