Integrate Custom 3D Wind Field and Temperature Data into JSBSim

[gulugulu888]

I'm attempting to integrate a complex 3D wind field and temperature dataset into JSBSim. My dataset consists of:

Three 3D data files representing u, v, w wind speed components.
One 3D data file containing temperature data.

Each file is formatted as follows:

First 5 lines contain metadata (title, grid dimensions, starting coordinates, grid spacing)
Followed by data values arranged in x, y, z order

My objectives are:

Create a custom atmosphere model class in JSBSim that can read and utilize this data.
During simulation, use trilinear interpolation to calculate wind speed and temperature based on the aircraft's current position.
Ensure this custom model integrates seamlessly with JSBSim's existing architecture.

I've attempted to create a CustomAtmosphere class, derived from FGAtmosphere, which includes methods for loading the data and performing interpolation. However, I'm unsure about the best way to integrate this into JSBSim.
Specifically, I have the following questions:

Where in JSBSim should this custom atmosphere model be initialized and used?
Do I need to modify FGFDMExec or other core classes to support this new model?
How can I specify the use of this custom model in JSBSim's XML configuration files?
Are there recommended approaches for handling large 3D datasets to ensure simulation performance isn't impacted?

I would appreciate any guidance on best practices and integration methods. Thank you!

[ZHZJDSM]

I also encountered the same problem. In addition, I want to know how to write a proper script file to control the flight path of the aircraft

[seanmcleod]

In addition, I want to know how to write a proper script file to control the flight path of the aircraft

Create a separate discussion item for this.

[seanmcleod]

Couple of initial comments and thoughts.

So, in terms of integrating a 3D grid of wind vectors (u, v, w) into JSBSim this has been done in the past. One example is @agodemar's research project which used the CFD software OpenFOAM to generate the 3D grid of wind vectors behind a wind turbine and then fed the wind vectors into JSBSim while flying a light aircraft through this region.

Take a look at #74 for more details and a link to the research paper.

In summary there was an external process which received the aircraft's position for each simulation time step, which then interpolated the wind vectors from the 3D grid and then updated the current wind components in JSBSim via the properties atmosphere/wind-north-fps, atmosphere/wind-east-fps, and atmosphere/wind-down-fps. JSBSim ran in lockstep with the external process, i.e. it effectively waited for the wind data on each time step. It is also possible to update these wind components in JSBSim asynchronously via an external process, with the external process querying via the property tree for the aircraft's current 3D position and then updating wind properties, although in this case it won't be in complete lockstep.

Do you want to/need to run this in real-time, i.e. typically if you're going to have a human pilot in the loop? Or are you only interested in using this in batch mode?

In addition to @agodemar's approach another option is something along these lines:

fdm = jsbsim.FGFDMExec()
fdm.load_model('737')
fdm.run_ic()

while fdm.get_sim_time() < period:
   # Source and interpolate wind vectors
   (wN, wE, wD) = get_wind(fdm['position/long-gc-deg'], fdm['position/lat-gc-deg'], fdm['position/h-agl-ft'])
   fdm['atmosphere/wind-north-fps'] = wN
   fdm['atmosphere/wind-east-fps'] = wE
   fdm['atmosphere/wind-down-fps'] = wD
   fdm.run()

Another option I guess is for JSBSim to supply a mechanism for a user to supply a wind callback function like JSBSim does for determining the ground elevation via an optional ground callback function.

In which case then you would simply implement the wind callback function which would get called on each timestep, do your lookup and interpolation and return the wind components. Linking in your implementation with JSBSim.

In general the same options apply to setting the temperature, i.e. instead of implementing a whole new atmosphere model make use of one the options I've mentioned.

[gulugulu888]

Quote: “Do you want to/need to run this in real-time, i.e. typically if you're going to have a human pilot in the loop? Or are you only interested in using this in batch mode?”

Answer: My purpose is to get aircraft dynamics response in windshear, so I'm only interested in using it in batch mode.

[seanmcleod]

Have you decided which sort of approach you're going to try?

If you go with the Python approach you can make use of numpy/scipy interpolation methods.

[gulugulu888]

Python may be the optimal choice. Thanks for your help. I will try this.