Implementing F-35's Flight Model and Flight Control System in DCS - Technical Discussion

[DummyCatz]

Given the upcoming DCS F-35A module and after reading through some LM's technical papers about its performance and flight control system designs, I'd like to discuss what I believe might be the most challenging aspects of implementing the F-35's unique flight model and control laws, particularly its NDI (Nonlinear Dynamic Inversion) system.

From what I understand, this would be the first implementation of an NDI-based flight control system in DCS. The key challenges I see are:

1. Onboard Model (OBM) and aerodynamic model. The F-35's NDI relies heavily on a detailed onboard model (OBM) of the aircraft's mass properties and stability/control characteristics. How ED is going to handle the complex aerodynamic modeling, especially in the high AOA and transonic regime is crucial, where you have:

• Nonlinear control effectiveness
• Strong control surface interactions
• Complex vortex systems and flow separation effects
• Transonic roll-off phenomena

2. Real-time Control Solution Computation. Unlike traditional scheduled-gain FBW systems, the NDI system computes control solutions on the fly. Does the DCS engine architecture allow for this kind of real-time computation without impacting performance?

3. Effector Blender Implementation. The effector blender needs to:

• Compute optimal control surface allocation in real-time
• Handle control priority shifts (like yaw vs. pitch priority for horizontal tails)
• Manage control surface interactions at high AOA
• Deal with control surface saturation and overflow conditions

4. Auto Recovery Systems. The integration of:

• Automatic Pitch Rocker (APR) for deep stall recovery. (Yes, the F-35 have deep stalls just like the F-16)
• Anti-spin modes
• Complex mode transition logic
• Adaptive filtering systems for yaw rate control

It's also interesting to see how ED will handle the transition between different control modes (like the shift from traditional roll command to yaw rate command at high AOA) while maintaining the "feel" of the real aircraft.

This would be a unique kind of FM and FCS to witness in DCS.

Cheers and take a read at https://arc.aiaa.org/doi/10.2514/5.9781624105678.0525.0574 (F-35 High Angle of Attack Flight Control Development and Flight Test Results)

Edited January 21 by DummyCatz

[DummyCatz]

Some point of view to think about:

While we'll always be dealing with approximations in DCS, the challenge here isn't just about fidelity - it's about creating a basic NDI architecture that captures the fundamental behavior of the real system. The F-16/F-18 approximations can be built on known control structures, while the F-35 would require essentially designing a new NDI implementation from scratch.

The F-16 and F-18's FBW simulation take benefits from the well-known and public control laws (e.g. NASA TP1538 and DTIC ADA189675 for F16 and NASA TM107601 for F18) with well-documented PID controllers and by-the-book gain schedules and filters, that you can even 100% replicate the logic according to the diagram. But yet, look at the current and previous state of FCS problems in DCS F-16 and F-18 would give you an idea of how hard it is to implement even a PID-based control structure.

The F-35's NDI presents a fundamentally different challenge compared to the traditional PID-based control architectures used in those traditional FBW aircraft. Without access to the actual onboard model data or knowing how the effector blender optimizes control allocation, creating even a rough approximation would require making many assumptions.

Edited January 17 by DummyCatz

[Cepheus76]

In addition to what was said before, I'd like to ask that since the F/A-18s flight model is said to be modeled less accurately in the high alpha regime lest important deductions about its real capabilities can be made, how many deliberate inaccuracies are then to be expected in the F-35? 

Edited January 17 by Cepheus76

[MudMoverGSH]

And on the hardware side, there is the Active Inceptor System as well (active stick and throttle).
Per the 2008 ppt("F35 AIS Krumenacker SAE 081016PPT"):
 

Quote

 

•Each Axis(*pitch, roll, throttle) contains duplex 28V electric motor drives
connected to grip interface by mechanical linkages
• Stick Axes contain mechanical springs for backup
mode

AIS Modes & Fault Accommodation
• Each Inceptor has three primary control modes:
– Active: sensed grip force is used to actively position the inceptor
according to the programmed force vs. position characteristics
• Flight Control Laws use inceptor position as pilot command
– Passive: motor drives disengaged, stick springs provide fixed
linear force gradient, throttle has fixed friction & no detents
• Used upon unrecoverable error with motor drives
• Flight Control Laws use stick forces and throttle position as pilot
command
• Both stick axes will maintain like mode (if one axis downgrades
passive mode, other axis will be place passive mode)
– Jammed: inceptor position is fixed
• automatically detected by software
• Flight Control Laws use inceptor force as pilot command
• jammed throttle requires some Control Law reconfiguration

How Is Active Capability Currently Used?
• Throttle:
– Variable aft & forward end-stops (e.g. STOVL mode is different
from CTOL mode)
– AB gate (when STOVL system is not deployed)
– Launch gate (CV only)
– STOVL center detent (zero commanded acceleration)
– STOVL on-ground power braking force gradient
– Back-drive
• Auto-Throttle Approach (all variants)
• STOVL Decel-to-Hover

How Is Active Capability Currently Used?
• Stick:
– Tailored STOVL pitch force
characteristics
• wingborne vs. jetborne
variations
• forward soft-stop(s) for
vertical landing sink speed
– Pitch force feedback at
higher AOA
– Roll force tailoring: left vs.
right
– Roll force tailoring: CTOL
vs. STOVL
– Increased force breakout
for CV launch (pitch & roll)

 

Hope some FFB support can be included, at least for the primary mode (the active mode), i.e. high AOA feedback, breakout force, range limiter/travel stop, stick force differences btwn each direction, auto throttle etc. And later, support for the HOTAS characteristic of STOVL, CV if those variants are released in the future.
A lot of manufacturers are now developing FFB devices, including VPForce, Winwing, Moza, and FliteSim. Viripil and VKB might also join the battle as well by the time the F-35 is released(Virpil officially says they are working on it but still need quite a while). Additionally, Winwing is working on motorized throttles.

Edited January 18 by MudMoverGSH

[NineLine]

18 hours ago, Cepheus76 said:

In addition to what was said before, I'd like to ask that since the F/A-18s flight model is said to be modeled less accurately in the high alpha regime lest important deductions about its real capabilities can be made, how many deliberate inaccuracies are then to be expected in the F-35? 

That is a fair question but not sure it can be answered just yet. I expect to see development reports on different aspects as we progress, we will look at those when the time comes. Thanks.

[ThePops]

On 1/17/2025 at 5:31 AM, DummyCatz said:

NDI (Nonlinear Dynamic Inversion)

Just a note. Nonlinear Dynamic Inversion is a general method used in controlling nonlinear systems. What NDI does is to make this control simpler, not harder. In practical applications, the problem with NDI is it relies on a very good and detailed physical model. This is problematic because data in regions with high nonlinearities, like in a stall for instance, is by nature fuzzy (turbulent and chaotic).

While NDI is good in theory, in practice it breaks down because it just isn't robust enough in the regime you want to improve in the first place. I'm not aware of any robots and drones that actually use NDI today. What is used (if not just PIDs) is INDI (Incremental Nonlinear Dynamic Inversion). The "incremental" stands for measuring accelerations in small time increments. This adds the needed robustness that NDI lacks by measuring the input needed (acceleration as far as drones and robots are concerned). It's what makes NDI usable in practical applications.

That's my understanding. How Lockheed with their $billions of research and development have solved this, I have no clue. One point here is that small drones with their onboard Arduino or whatever is perfectly capable of implementing INDI. They react much more suddenly due to their small size, than an F-35 weighing several tons. The needed computing power is not particularly terrifying IMO.

It will be cool to see how ED solves this though. 

[DummyCatz]

5 hours ago, ThePops said:

Just a note. Nonlinear Dynamic Inversion is a general method used in controlling nonlinear systems. What NDI does is to make this control simpler, not harder. In practical applications, the problem with NDI is it relies on a very good and detailed physical model. This is problematic because data in regions with high nonlinearities, like in a stall for instance, is by nature fuzzy (turbulent and chaotic).

While NDI is good in theory, in practice it breaks down because it just isn't robust enough in the regime you want to improve in the first place. I'm not aware of any robots and drones that actually use NDI today. What is used (if not just PIDs) is INDI (Incremental Nonlinear Dynamic Inversion). The "incremental" stands for measuring accelerations in small time increments. This adds the needed robustness that NDI lacks by measuring the input needed (acceleration as far as drones and robots are concerned). It's what makes NDI usable in practical applications.

That's my understanding. How Lockheed with their $billions of research and development have solved this, I have no clue. One point here is that small drones with their onboard Arduino or whatever is perfectly capable of implementing INDI. They react much more suddenly due to their small size, than an F-35 weighing several tons. The needed computing power is not particularly terrifying IMO.

It will be cool to see how ED solves this though. 

Thanks for the excellent write up. The F-35 IRL is indeed using INDI, as the engineers had to deal with significant challenges from onboard aerodynamic model (OBM) mismatches, especially in high AOA as well as transonic regions.

In DCS, you could theoretically have a naive NDI system using the exact same aerodynamic model that's driving the simulation - creating a perfect match between the OBM and the simulated aircraft behavior. This would be "cheating" in the sense that it eliminates one of the fundamental challenges that makes real NDI implementation difficult. But this raises some other interesting questions about simulation fidelity:

• Should ED deliberately introduce model mismatches to more accurately represent the real system's limitations?
• Could using the same model for both simulation and control (OBM) make the aircraft behave more like an idealized mathematical model than a real machine, so that the handling characteristics is in a way unrealistically and overly optimistic, with the resulting control allocations being too "precise" that doesn't match the real F-35's handling qualities?

5 hours ago, ThePops said:

The needed computing power is not particularly terrifying IMO.

The most computational cost probably comes from an iterator, like the pseudo-inverse iterative calculation of the Effector Blender (EB), as explained in https://arc.aiaa.org/doi/10.2514/6.2018-3516

Compared to the traditional control surface mixer, whose control logic will only be running one time per simulation cycle, the EB logic will be running several times per simulation cycle, due to the iterative solver.

 

It will indeed be cool to see how ED will handle this.

Edited January 18 by DummyCatz

[Bingo911]

I'm sure that the F-35 and things like Dynamic Campaigns will stretch DCS in ways that will make the sim grow in positive ways. Thanks for the informative read in here.

[ThePops]

Good info, but it all appears to be behind a paywall. At least it shows that info is indeed available, even for something as obscure and complex as the flight control system.

Anyway, good info about the control system DummyCatz. I didn't know the F-35 used a (I)NDI system. I thought this was a drone, robot, missile thing.

[Muchocracker]

On 1/17/2025 at 4:24 AM, Cepheus76 said:

In addition to what was said before, I'd like to ask that since the F/A-18s flight model is said to be modeled less accurately in the high alpha regime lest important deductions about its real capabilities can be made, how many deliberate inaccuracies are then to be expected in the F-35? 

This is evidently not true anymore considering the update some months ago that allows it to reach upwards of 60 AoA now.

[Cepheus76]

1 hour ago, Muchocracker said:

This is evidently not true anymore considering the update some months ago that allows it to reach upwards of 60 AoA now.

So does a Tomcat in a flat spin. That the DCS flight model of the F/A-18 allows such high AoAs doesn't necessarily imply that the FM exploits that capability to its full extent.

[DummyCatz]

Regarding the notion of deliberately hiding classified aspects of high AOA modeling - this would be detectable. Flight dynamics are deeply interconnected, and any artificial limitations, inaccuracies or simplifications would create detectable inconsistencies. Knowledgeable users can identify these discrepancies using publicly available information, as demonstrated through existing flight model bug reports that rely solely on public sources. (e.g. some of my F-18 FM reports.)

The core challenge remains implementing both the complex high AOA aerodynamic modeling with its asymmetric moments, and the NDI-based flight control system. However, we do have some useful reference points. The AIAA papers provide stability derivative and control power trends (Cm, Cn, Cl diagrams) that, even without absolute values, still reveals the general shape of those coefficients of aerodynamic moments. So we can know the relative control effectiveness and aircraft static stability across AOA ranges. For example the lateral-directional stability derivatives of the F-35 from Fig 6 of the AIAA paper quoted above, that you know when the Cn-beta dips into negative values, the AOA should be somewhere around 25-30 deg.

What makes this particularly interesting is how the NDI system essentially defines the aircraft's maximum allowable performance envelope. We know the system aims to maximize performance within safety boundaries, prevent departures while minimizing interference, and deliver consistent handling qualities. This suggests that if you can accurately model the basic aerodynamic behavior using scaled/estimated data from public trends, as well as the NDI control architecture's fundamental principles with departure prevention and departure recovery modes, you could potentially achieve a reasonable approximation of the aircraft's actual performance characteristics, since the NDI system would naturally optimize control within whatever envelope you've defined. While exact values remain classified, the general behavior and performance could be reasonably reconstructed through careful implementation of the documented NDI principles and known aerodynamic trends.

Edited January 21 by DummyCatz

[Df555]

I don't know if you've seen this news, but just in case I decided to post it here: https://www.forbes.com/sites/petersuciu/2024/07/10/military-documents-related-to-f-35-lightning-ii-leaked-on-social-media/
Is it really true? Of course I don't know. But given the announcement that happened, I think there is a possibility. It happened quite recently, as you can see. We will never know the truth anyway, and if I had these documents, I wouldn't tell you anything either . I want to believe that the F-35 will be a little more realistic than we think.

Also, to further support my words, I leave a link to a comment from Indiafoxtecho, which was discussed earlier on this forum:

 

Edited January 23 by Df555

[Temetre]

Am 22.1.2025 um 15:41 schrieb Df555:

I don't know if you've seen this news, but just in case I decided to post it here: https://www.forbes.com/sites/petersuciu/2024/07/10/military-documents-related-to-f-35-lightning-ii-leaked-on-social-media/
Is it really true? Of course I don't know. But given the announcement that happened, I think there is a possibility. It happened quite recently, as you can see. We will never know the truth anyway, and if I had these documents, I wouldn't tell you anything either . I want to believe that the F-35 will be a little more realistic than we think.

Also, to further support my words, I leave a link to a comment from Indiafoxtecho, which was discussed earlier on this forum:

Article says it contains F-15EX engine maintenance documentation, but not what it contains about the F-35. Makes me doubt its a big deal?

The usual stance is also that ED cannot use leaked classified documentation, even if its easily available. Thats eg why the F-18 is ~5% or so off from the real performance charts, which they werent able to use.

[Df555]

3 hours ago, Temetre said:

Article says it contains F-15EX engine maintenance documentation, but not what it contains about the F-35. Makes me doubt its a big deal?

The usual stance is also that ED cannot use leaked classified documentation, even if its easily available. Thats eg why the F-18 is ~5% or so off from the real performance charts, which they werent able to use.

ED can still replicate the F-35 much more accurately if they actually have something.
Even if you try to make the module absolutely accurate and have all the documents, you can't do it. Because it's a game, many things will have to be greatly simplified one way or another.
People are still trying to argue that the F-35 can't come close to the real, all the data is classified, etc. But in reality, if you want to keep it secret, then selling the F-35 was not the best idea. This article is still just a rumor, but the leaked documents are not that surprising. Plus, let's not forget that the 5th generation is the current generation, not the next one, even if not everyone can afford it. You can't expect its characteristics to be protected so seriously.

Edited January 25 by Df555

[Cato]

I was just watching a video about the F-35's Auto Ground Collision Avoidance System (AGCAS). The presenter (a former F-16 pilot) was in the sim at LM and put the aircraft in an inverted dive. AGCAS rolled the aircraft over and pulled the aircraft up before it hit the ground. Amazing piece of technology that apparently has saved numerous lives. Will be interesting to see if this feature gets put in the DCS version of the F-35.  

[EchoOneOne]

11 hours ago, Cato said:

I was just watching a video about the F-35's Auto Ground Collision Avoidance System (AGCAS). The presenter (a former F-16 pilot) was in the sim at LM and put the aircraft in an inverted dive. AGCAS rolled the aircraft over and pulled the aircraft up before it hit the ground. Amazing piece of technology that apparently has saved numerous lives. Will be interesting to see if this feature gets put in the DCS version of the F-35.  

Auto-GCAS is available on Block 3F and above. It was integrated in 2019 only for the CTOL version, so I assume we will not see it on our Block 2B.