The basic idea is to do something along the lines of:
while not k(time, lat, lon, alt):
lat_dot, lon_dot, alt_dot = f(time, lat, lon, alt):
lat += lat_dot * dt
lon += lon_dot * dt
alt += alt_dot * dt
where
- f is a model (or a combination of, see below),
- k is a termination function.
Models, altitude profiles and termination functions must all be pure.
Besides being cleaner, it allows us to use more interesting integration methods without worrying about side effects evaluating the functions.
We principally deal with position represented as latitude, longitude and metres above sea level. While we do have to consider horizontal velocities in metres per second (e.g., when consulting wind data), we convert to latitude & longitude (or rather, “change in latitude per unit time”) as soon as possible since they will (probably) be simpler to work with. (“ASAP” is so that we—as much as possible—are only working in one coordinate system throughout the code.)
Time is represented as an absolute UNIX timestamp.
A model is a callable that looks something like this:
def f(time, lat, lon, alt):
# < calculation goes here >
return lat_dot, lon_dot, alt_dot
Return velocities predicted by this model (example function)
The latitude and longitude “velocities” (lat_dot & lon_dot) are “change in decimal degrees per unit time”; vertical velocity (alt_dot) is just metres per second.
| Parameters: | |
|---|---|
| Return type: | 3-tuple of floats: (lat_dot, lon_dot, alt_dot) |
…is specified via closures, i.e. we have a function that takes some configuration and returns the actual model function.
We want to be able to specify several models, and then “swap bits” out, or pick from a selection when setting up a flight. E.g., we might want to choose a combination of
For the majority of cases, a linear combination of the models we are interested in will suffice. Note that a function that linearly combines models is itself a model; see tawhiri.models.make_linear_model().
A termination condition decides if the prediction (stage) should stop. They are functions that look something like:
def k(time, lat, lon, alt):
return alt >= 30000
Note that a function returns True to indicate that the prediction should stop.
Decides if the prediction should stop (an example function)
Returns True if the prediction should terminate.
| Parameters: | |
|---|---|
| Return type: | bool |
Similarly to the ability to linearly combine models, we can “OR” termination functions together with tawhiri.models.make_any_terminator().
We want to chain stages of a prediction together: this essentially amounts to running several predictions, with the initial conditions of the next prediction being the final position of the last, and concatenating the results (see tawhiri.solver.solve()).
tawhiri.models contains a few “pre-defined profiles”, that is, functions that take some configuration and produce a chain of stages for a common scenario.
As an example, tawhiri.models.standard_profile() produces the chain containing two stages: