Receiving force and moment from external app and Sending update of aircraft rotation and position to external app

[nmfarras]

Is it possible for jsbsim to communicate with external app via network like UDP or other protocol? It still a concept, but i want to create system where the physics come from let says SIMULINK and it can update the aerodynamic coefficients or forces and moments to jsbsim unreal and jsbsim will send back the position and rotation to the SIMULINK. Is it possible?

[agodemar]

@nmfarras what do you mean by "the physics" comes from an external app?

Probably, at the generic time $t$ you have the external aerodynamic force $\vec{F}_A (t)$ and moment $\vec{\mathcal{M}}_A^{(P)}(t)$ — mind the definition of the pole $P$, make sure it is the aircraft CoG — and want to pass them to JSBSim to propagate the vehicle's state to $t+\Delta t$.
If this is the case, I would force JSBSim to calculate a zero force and a zero moment internally and pass $\vec{F}_A (t)$ and $\vec{\mathcal{M}}_A^{(P)}(t)$ as "external" actions.

[nmfarras]

I kinda consider something like that too. I consider doing something like this for example

<!-- Property that can be written -->
<property value="-.6000E-01"> aero/calc/CnDr <property>

<!-- input the value into jsbsim -->
<function name="aero/coefficient/CnDr">
    <product>
        <value>1</value>
        <property> aero/calc/CnDr <property>
    </product>
</function>

but direct as forces and moment like Lift instead. Then is it possible? Also if it is possible, any idea the if i want it to still feel responsive, at which rate the data should be updated?

[seanmcleod70]

Yes, I was going to suggest doing it the way you've shown it.

I've done something similar in the past. In my case I didn't want to calculate the moment or force externally, rather I wanted to be able to modify the stability coefficient from an external program while JSBSim was running. So I made all the coefficients properties.

    <aerodynamics>

        <property value="0.75">    aero/stability_coefficients/Cm-alpha </property>
        <property value="-27.0">   aero/stability_coefficients/Cm-q </property>
        <property value="-16.0">   aero/stability_coefficients/Cm-alpha-dot </property>
        <property value="1.0">     aero/stability_coefficients/Cm-d-elevator </property>

        <property value="-0.09">   aero/stability_coefficients/Cl-beta </property>
        <property value="-0.4">    aero/stability_coefficients/Cl-p </property>
        <property value="0.09">    aero/stability_coefficients/Cl-r </property>
        <property value="1.0">     aero/stability_coefficients/Cl-d-aileron </property>
        <property value="0.01">    aero/stability_coefficients/Cl-d-rudder </property>

        <property value="0.26">    aero/stability_coefficients/Cn-beta </property>
        <property value="0.00">    aero/stability_coefficients/Cn-p-xx </property>
        <property value="-0.35">   aero/stability_coefficients/Cn-r </property>
        <property value="0.00">    aero/stability_coefficients/Cn-d-aileron-xx </property>
        <property value="-0.2">    aero/stability_coefficients/Cn-d-rudder </property>

Then modified the moment functions to reference the property rather than having a hard-coded value for the coefficient in the function itself.

        <axis name="PITCH" unit="LBS*FT">
            <function name="aero/coefficient/Cmalpha">
                <description>Pitch_moment_due_to_alpha</description>
                <product>
                    <property>aero/qbar-psf</property>
                    <property>metrics/Sw-sqft</property>
                    <property>metrics/cbarw-ft</property>
                    <property>aero/alpha-rad</property>
                    <property>aero/stability_coefficients/Cm-alpha</property>
                </product>
            </function>

In my case I didn't need to continuously update the coefficient property for every time step, only very periodically in order to demonstrate handling qualities for a particular flight task.

However in your case you'll need to update the forces and moments for each JSBSim time-step, ideally figuring out how to run in lock-step.

And as @agodemar mentioned you need to be aware of the moment reference point (pole). Specified in the JSBSim's FDM like this:

jsbsim/aircraft/737/737.xml

Lines 39 to 43 in 30a7261

	<location name="AERORP" unit="IN">
	<x> 625 </x>
	<y> 0 </y>
	<z> 24 </z>
	</location>

JSBSim then uses the difference (arm) between this AeroRP and the current cg to calculate additional moments based on the current net force and this arm.

jsbsim/src/models/FGAerodynamics.cpp

Line 288 in 30a7261

vMoments = vMomentsMRCBodyXYZ + vDXYZcg*vForces; // M = r X F

So you need to know what the moment reference point is for the moments in Simulink to make sure they're in sync with the JSBSim's model.

The other thing is that you will obviously also need to have the mass, cg and moments of inertia duplicated in the JSBSim FDM. Which is different to Christoper Lum's demos of using FligthGear, X-Plane etc. for visualization, since in that case there is no need for any JSBSim FDM in FlightGear since all the calculations are done in Simulink and Simulink is only sending the pose (position and attitude) for each timestep to FlightGear for rendering purposes.

If you have an existing Simulink model then it's going to have details in it regarding mass, moments of inertia, cg, moment reference point etc. which you need to duplicate in the JSBSim FDM. Not impossible, just some extra work and something you need to keep in sync.

[seanmcleod70]

Hmm, what are you really wanting/needing JSBSim to do? Reading between the lines I'm guessing you really just want to use it to generate outside visuals in conjunction with Unreal?

If you just want outside visuals from Simulink are you aware of the Simulink FlightGear component? In which you perform all the 'physics' including calculating the accelerations, then integrating to get velocities and then integrating again to get pose (position and attitude) in Simulink and then the pose state on each time step is sent to FlightGear which acts purely as an image generator to generate the scene given the aircraft's pose.

Do you have a particular need or preference for Unreal as opposed to FlightGear?

[seanmcleod70]

Yep, although what Chris Lum is demonstrating is exactly what I was suggesting originally, i.e. let Simulink do all the calculations and simply feed (lon, lat, alt, heading, roll, pitch) from Simulink to FlightGear acting as a plain image generator.

So your requirement is a bit different since you want to make use of Unreal specifically and want JSBSim to handle ground interactions.

I did reply above regarding your planned mechanism to get the forces and moments from Simulink into JSBSim. However, thinking about your requirement now for JSBSim to handle ground interactions I see an additional complication for the Simulink approach.

Who is calculating the inputs to the aerodynamic force and moment calculations being performed in Simulink? Is Simulink also calculating and keeping track of alpha, beta, altitude, attitude and velocities? With JSBSim duplicating that calculation and state tracking?

If so what happens when the aircraft's gear impact the ground that JSBSim knows about but Simulink doesn't? JSBSim will sum the gear forces and moments with the aerodynamic forces and moments to get the net forces and moments that it then integrates and then calculates alpha, beta, dynamic pressure etc. But on landing now these input states into the aerodynamic calculations will now diverge.

Unless your plan is to actually also have state data like alpha, beta, dynamic pressure etc. fed back on each timestep from JSBSim to Simulink so that it uses them as input to it's aerodynamic forces and moments calculations?

[nmfarras]

Who is calculating the inputs to the aerodynamic force and moment calculations being performed in Simulink? Is Simulink also calculating and keeping track of alpha, beta, altitude, attitude and velocities? With JSBSim duplicating that calculation and state tracking?

As mentioned, the Simulink need these variable from JSBSim. So both JSBSim and Simulink will feed each other information. I think i do need to consider the state initialization and the update frequency from each. And of course what @agodemar and you suggest regarding the frame of reference of JSBSim and Simulink should be the same.

If so what happens when the aircraft's gear impact the ground that JSBSim knows about but Simulink doesn't? JSBSim will sum the gear forces and moments with the aerodynamic forces and moments to get the net forces and moments that it then integrates and then calculates alpha, beta, dynamic pressure etc. But on landing now these input states into the aerodynamic calculations will now diverge.

For this, for ground effect, i am thinking to have the height above the ground (AGL) as input when implementing this. While the interaction between aircraft and the ground, i thought it can be decoupled as longa as i am using the velocity, position, and rotation of the aircraft calculate by JSBSim. So i only need to calculate aerodynamic forces or so i thought.

Unless your plan is to actually also have state data like alpha, beta, dynamic pressure etc. fed back on each timestep from JSBSim to Simulink so that it uses them as input to it's aerodynamic forces and moments calculations?

Yeah i plan to use these for the Simulink calculation to ensure it used same condition as JSBSim. I can extract these parameters and sending it to Simulink right?

[seanmcleod70]

As mentioned, the Simulink need these variable from JSBSim.

aerodynamic coefficients or forces and moments to jsbsim unreal and jsbsim will send back the position and rotation to the SIMULINK

I had only seen you mention position and rotation in terms of state data going from JSBSim to Simulink.

Yeah i plan to use these for the Simulink calculation to ensure it used same condition as JSBSim. I can extract these parameters and sending it to Simulink right?

So you'll need to include a lot more state going from JSBSim to Simulink, in addition to the ones I've mentioned already you'll also need to include the angular velocities, p,q,r, so that aerodynamic moments that include them, e.g. like roll and pitch damping have access to them.

But yes, all of them should be available each time step from JSBSim, you just need to ensure you include the particular set of states that are required and transmit them back to Simulink.

[agodemar]

I'm afraid I still don't get the point of this discussion.
If one has Simulink, which is well equipped with all the tools to propagate the state of the aircraft, and to model the environment, why one should mix it with JSBSim?
Simulink and FlightGear or Unreal engine aren't enough?

[nmfarras]

Honestly I do think it is not a well thought concept. But at least I kinda get the gist of what JSBSim can do. Thank you for @agodemar @seanmcleod70 and @dershow for giving more insight and advices about this.

[seanmcleod70]

@nmfarras mentioned:

• I can use aircraft models purely modeled in JSBSim
• or others generated externally model (via Simulink or another custom one). The system supports feeding externally computed aerodynamic forces and moments into JSBSim, allowing JSBSim to focus solely on propagating the state.

As for the use of Unreal Engine, I've already developed a 3D environment there, and I want the simulation to consider real-time interaction with terrain and visual elements already created in that environment.

like this

But yes, trying to split the state calculations and updates between Simulink and JSBSim is going to be tricky.

My question, regarding a customer/client bringing a Simulink model that they want to fly in your engineering simulator is whether the ground interactions that you need to handle are simply with runways or do you really also need to model interactions between the aircraft and 3D structures like buildings etc.?

I'm sort of assuming that often the Simulink model that a client brings may already have a ground interaction component in terms of modelling gear forces etc. and quite possibly a much more detailed and accurate gear contact model than the one provided by JSBSim. In which case wouldn't it make more sense to leave everything in Simulink in this case, and simply make sure that the DEM you use in your Unreal based simulation visuals matches the DEM, or supply your DEM to the Simulink model? In which, in this case your simulator would simply be providing the visuals, just like the examples you shared of Chris Lum using FlightGear and X-Plane as purely image generators.

[dershow]

I agree. The “difficult" thing that JSBSim does well is finding the forces and moments from a given state (position, velocity, control position, altitude etc.). It then is able to account for these forces and moment, and keep track of all the reference frames etc. Just integrating those forces and moments, once you have appropriate forces, to update the state is fairly easy. It’s also something that Simulink can do well also. So, it’s really not clear what you want JSBSim to do and what you want Simulink to do for your project. If you want both to calculate forces and both to integrate those forces into a new state, you will have to “trust” one as the correct state for your simulation. Then you can use that to drive graphics (with Flightgear, or Unreal, or whatever else you want) Adam