Add topographic lift library and initial example implementation to SITL #28756
Conversation
Signed-off-by: Ryan Friedman <[email protected]>
Signed-off-by: Ryan Friedman <[email protected]>
Signed-off-by: Ryan Friedman <[email protected]>
|
Check It will need to be optional if it's not trivial flash. Linking to an external library or mavproxy module to feed in higher fidelity/accuracy wind estimates is totally an option. It could be a LUA script with TCP/UDP that can interract with AP to feed in the topographic lift estimate, Lua is faster than mavproxy. Longer term, we could have an AP server that can provide soaring estimates directly to the vehicle in flight if we want. Tridge is keen on improving SITL's capabilities and has other projects such as describing the vehicle in JSON format (wing size, etc). |
| } | ||
|
|
||
| // hXY are the heights of the 4 surrounding grid points | ||
| int16_t h00, h01, h10; |
There was a problem hiding this comment.
const these and define as part of declaration
There was a problem hiding this comment.
Like the PR from yesterday:) #28734
| @@ -422,6 +429,7 @@ class AP_Terrain { | |||
|
|
|||
| // cache the home altitude, as it is needed so often | |||
| float home_height; | |||
| Vector2f home_gradient; | |||
There was a problem hiding this comment.
| Vector2f home_gradient; |
|
|
||
|
|
||
| return true; |
There was a problem hiding this comment.
| return true; |
| return false; | ||
| } | ||
|
|
||
| [[maybe_unused]] const float veh_height_agl = loc.alt * 1E-2 - terrain_height_amsl; |
There was a problem hiding this comment.
We have a define for unused stuff
| return terrain->gradient(loc, grad); | ||
| } | ||
|
|
||
| bool AP_TopoLiftSinglePointEstimate::get_wind_at_current_loc(Vector3f& wind) { |
There was a problem hiding this comment.
| bool AP_TopoLiftSinglePointEstimate::get_wind_at_current_loc(Vector3f& wind) { | |
| bool AP_TopoLiftSinglePointEstimate::get_wind_at_current_loc(Vector3f& wind) | |
| { |
etc
| @@ -0,0 +1,3 @@ | |||
| #ifndef AP_TOPO_LIFT_ENABLED | |||
There was a problem hiding this comment.
| #ifndef AP_TOPO_LIFT_ENABLED | |
| #include <AP_HAL/AP_HAL_Boards.h> | |
| #ifndef AP_TOPO_LIFT_ENABLED |
| wind_ef.z += get_local_updraft(position + home.get_distance_NED_double(origin)); | ||
| auto const curr_pos = position + home.get_distance_NED_double(origin); | ||
| wind_ef.z += get_local_updraft(curr_pos); | ||
| wind_ef.z += get_topo_lift(location); |
There was a problem hiding this comment.
Definitely want this parameterized; don't want more CPU generally - even in SITL :-)
There was a problem hiding this comment.
Should I parameterize get_local_updraft to establish the pattern?
| // Wind in the topographic frame. X points east, Y points north. | ||
| // East wind will be [speed, 0] | ||
| // North wind will be [0, speed] | ||
| const Vector2f wind_topo(cosf(dir) * speed, sinf(dir) * speed); |
There was a problem hiding this comment.
| const Vector2f wind_topo(cosf(dir) * speed, sinf(dir) * speed); | |
| const Vector2f wind_topo{cosf(dir) * speed, sinf(dir) * speed}; |
| const AP_AHRS &ahrs = AP::ahrs(); | ||
|
|
||
| return ahrs.wind_estimate(wind); |
There was a problem hiding this comment.
Should we really be using AP::ahrs() here? Your simulation shouldn't really depend on what NavEKF3 is up to - if your fix is wandering your simulation is going to follow that wandering vehicle rather than represent the true lift value the simulated vehicle should be experiencing.
I'm hoping to have a set of AHRS backend results available from the SITL backend which might allow for using the "perfect" AHRS for determining the simulatedwind and a stnadrd backend's results when using on a real vehicle.
Purpose
Demonstrate slope soaring in SITL using a dead-simple lift estimator.
longer term
I want to investigate the idea of feeding in much more complex wind data from CFD models using this interface.
Can ArduPilot be extensible to heavy models that run ONLY on Linux, and perhaps can we feed in data to AP's FDM from another library (python), or can we link in shared libraries that use things like the STL and Boost just to get fancy CDF models?
AP SITL is awesome, but it doesn't have all the capabilities. I don't expect this to run on embedded.