EECH Hover Hold & Alt Hold Mods

Hi Guys, I'm currently looking at correcting the problems with Hover Hold and Altitude Hold in the sim.

As you know, Hover Hold works most of the time, as long as your vertical velocity is low. If your velocity is higher, it tends to oscillate up and down more than necessary. Altitude Hold on the other hand is mostly useless, it doesn't work over anything but flat terrain. The goal of flying "Nap of the Earth" wasn't achieved by the original programmers. I think we can fix both subsystems.

I've dug through the code and found the code for both control subsystems. It's located here: \aphavoc\source\gunships\dynamics\common\co_force.c
Search in the file for "Auto hover calculations", it's about a third of the way through the file.

After digging through the Hover Hold section, I can see that it attempts to control the helicopter using mostly Proportional Control. The proportional code has two different ranges, one for higher velocity and another for low vertical velocity. Proportional Control is inherently unstable, it overshoots horribly and tends to oscillate and never settle. To correct this they've added in a section after the Proportional Control which adds in a damping component artificially, nudging the system with a slight acceleration opposite to the current velocity. When the vertical velocity is high the two sections are unbalanced, causing the system to oscillate. It will eventually settle, but you shouldn't have to wait so long. It should work better than this. There is also a short code section which levels out the helicopter by countering the Pitch and Roll with the cyclic. These motions are damped by gravity elsewhere in the dynamics system and are inherently stable. My original goal wasn't to fiddle with the Hover Hold, but now that I see how it's implemented I think we can improve it quite a bit.

I haven't looked at Altitude Hold in depth yet, that's my next task. It appears to look ahead of the helo one second and attempts to use the altitude ahead of the helo as part of the calculations. That probably helps some, but since it's most likely all Proportional Control, it doesn't work. I've watched the sim while flying over changing terrain and I've seen it oscillate up and down significantly. On one run I tried to lock Altitude Hold in at 300 ft. The terrain dropped away and the helo dropped all the way down until it slightly touched the ground, then swung upwards to almost 600 ft before coming back downwards again. It almost hit the ground again on round two as well. That much overshoot is what you get with 100% Proportional feedback control. Completely useless. We can fix it.

I've modeled the dynamics roughly in a spreadsheet and implemented a PID control loop (Proportional-Integration-Differentiation loop). I've optimized the coefficients to see what works, so I have a starting point for tuning the control loops. Once I get through digging through the Alt Hold section of the code, I'll look at ways to fix the control loops and see what I can do with it.
Thor

Hello Javelin.
I am happy that you are trying to improve something in eech.
I do not know if you saw this thread, especially thealx post.they can be helpful for you.
Maybe if you finish improving the hover you will look at flight model.It would be great, good luck



http://simhq.com/forum/ubbthreads.php/topics/4388796/1

Hi XIII,
I've poked around in the code a little bit and I saw where all of the helos are based on one general purpose dynamics engine. All of the parameters for each helo type are ratio'd to the base model, which I believe is based on the Comanche. I haven't compared the parameters between types, though. It could be the ratios aren't ranged widely enough to give you a feel for the weight differences between each model. It also may not account for control differences between different styles of helicopters (single rotor vs coaxial rotors). I can take a look at it when I get some time after I finish fixing the Alt Hover code, that's what bugs me the most right now.
Javelin

Thanks a bunch, looking forward seeing the result Thor, that sim deserves it

Marc..

Nice to have another modder around. Do you mind if I ask what your coding background is? I've learned C just from working on this sim, and not all that well. I get stuck at bits quite often, which is the reason my own Blackhawk project hasn't progressed much.

My Background? I'm a Mechanical Engineer who has been writing code for personal projects on and off for 30 years. I've mostly programmed in Delphi but I have done a little bit with C here and there. The Electrical Engineer I work with helps me out out when I need it, he's a very strong C programmer. Most of the stuff I've programmed has been analytical graphical stuff, working at stock market projection, but I also worked on coding a 3D sim up from scratch in OpenGL which included a physics modeled helicopter. I was working up to a Mechwarrior-5 sim, but then Piranha Games took it on so I stopped. It was too much for one person anyway. I should have used someone else's 3D rendering engine instead of coding one up from scratch, but I had fun doing it. I had some really unique features in it, like a continuously variable level of detail module for rendering complex items (no one has ever done that before!) an octagonal skybox, procedural scenery placement, stuff like that. I ran out of steam when Piranha stepped in with Mechwarrior Online.

I went through the Altitude Hold code last night. It expends a lot of energy calculating the helo's position one second in the future, but they made a mistake in the physics. The acceleration term (1/2)at^2 was implemented incorrectly, they forgot to multiply by 1/2. This throws the projection off by a bit, but that wasn't the worst of it. The control is purely proportional like I thought, using a coefficient of 5.0 which is pretty high. This time they didn't damp it at all, the altitude is left to cycle up and down at will. I modeled the controls roughly in a spreadsheet and found a solution with a proportional coefficient of 1.0, and a damping coefficient of 2.5. It takes a lot of damping to keep the helo from overshooting and potentially hitting the ground during fast nap of the earth flying. I also found that throttle control isn't enough, you need to pitch the nose up and down in order to maneuver severely enough to follow the terrain closely. Over 30 to 50 knots I kicked in a bit of Pitch control to help it maneuver. Hopefully EECH will respond similar to my model and it won't take much tweaking to get it working smoothly.

After flying for a bit last night, I suspect the Auto Pilot is directly adjusting the helo's altitude, I don't think they used control inputs to adjust the flight path at all. It moves too cleanly and the control inputs don't quite match a physics based solution. I'll take a look just for my own satisfaction when I get some time. Since it's working just fine I don't intend on changing it, I'm just curious.

Taking on an entire helicopter like the Blackhawk is pretty adventurous! To customize it all the way will take some work, there are bits and pieces of code to rework all over the place. The 3D models shouldn't be too difficult, just a lot of hours to create and texture. There might even be 3D models out there you could use to jump off from. Are you going for clickable cockpits? That's a chore I wouldn't want to take on. I'm not sold on clickable cockpits. It means you can't just jump into a new helicopter type and fly it right away. I haven't got the time to learn a dozen different cockpits, I'd rather have a simpler basic control set that works for all of the helo types so I can fly anything in the sim without a steep learning curve every time.
Javelin

Flight model and everything associated with it is the weakest part of the eech from the beginning.
any improvements will be everyone's dream.The more that you are a specialist.

Uhmm... I just took a look into the Blackhawk project and Wow, you guys are really taking it to the next level! It looks like a ton of work you're doing, getting all the instrument displays working.

The flight model is more my thing, I enjoy messing with the physics. I'll take a look under the hood when I get time and see what I can do. The Apache feels sluggish to me, and the Comanche behaves oddly when I'm near a hillside. Something's not quite right.

not only blackhawk, mi24 also

Nice! Adding completely new helos is really cool.

I looked through the dynamics code though, and almost all of the parameters are the same for every helicopter type. The only parameters that got updated were the the helicopter weight, fuel weight, and max torque. The developer got lazy and didn't bother to fill in all the parameters even though the code is set up to handle it. Basically, every helo uses the Comanche parameters even though they might be heavier or lighter and handle differently. I think all we have to do is find the correct numbers and enter them into the tables, recompile, then things will start feeling like they should. It will unbalance the multiplayer portion of the game, but I don't think any of us really care about that. It's supposed to be a sim and real life is unbalanced.

Does anyone know if there is a reference online like "Jane's All The World's Aircraft" or have access to the data we need?
Javelin

Yeah, it has been a lot of work, and taken a long time so far. The 3D part I find easy-ish, but the coding is the tough part. Most of the code can be re-used for the instruments. But I'm not at the point of having to try and get some unique instruments working, and it's proving tricky. Sockzien is also doing a model,. based on the newer digital cockpit version of the Blackhawk. But I like the "old fashioned" ones, and based it on the special forces model, as it's the only real world model capable of laser targetting and firing Hellfires. The current in game AI model cheats.

Please notice that completely different code used for different flight modes ("flight_model" parameter of EECH.ini). By default - each helicopter has it's own functions, but for future improvements sake - they was combined in common functions that used only for flight_model=2. Different DYN files used as well (default folder is \common\data\dyn, new - \common\data\dynamics). I believe it will be fair to leave default FM as is, as some people still like it. Different question is fixing issues - it will be helpful anyway. But as long as you are talking about physics improvements, you may need to stick to the fm=2 as it was made for such purposes. That is just recommendation, it's your decision anyway.

Thanks thealx, I haven't dug deep enough to really have a good handle on things yet. I wasn't aware there were two different Flight Models in the code.
How do you switch between the two different flight models?

I did notice one thing yesterday... I didn't see any code for modeling the rear horizontal and vertical stabilizers anywhere.

in eech.ini you have flag.

Right! I need to look in there and see what else I'm missing out on...

OK, I was already using FM2, but there are some other interesting settings in there.

animate_helicopter_controls ( entity *en )

That would be the keyframe animation code. I was looking through the physics code, which doesn't seem to have a section for calculating the forces on the airframe from the horizontal and vertical stabilizers.

The horizontal stabilizer tends to level the helicopter out at higher speeds, the vertical stabilizer straightens out the helicopter and reduces the need for the tail rotor at higher speeds. It also affects the helo when you try to fly directly sideways, it tends to swing the tail in behind, opposite to the direction of motion. The older Huey Cobra's had an large tail fin, making it quite difficult to fly directly sideways without using a lot of pedal action.

Javelin-I am a total amateur, but it's great to read your posts

There is no such calculations, as I'm aware. Even animation of stabilizers and rotor blades was implemented several years ago, but no one had any plans to improve flight model deeply. Originally game wasn't designed as serious simulator, and devs was focused on gameplay mostly - don't expect too much. You can find option in the eech.ini that show all forces applied to the helicopter, and makes it easier to understand how simple dynamics scripts are.

Hi guys, I did some sideways flight testing today, it seems to want to behave properly. It doesn't feel snappy like the sim I programmed a number of years ago though, it feels a bit sluggish and mushy. Somewhere in the physics code there is an artificial restorative force of some kind I need to find. I think I can add in something a bit more realistic for the stabilizers once I figure out where the restorative force is hidden away. The developers did a great job on modeling all the characteristics of the two rotors from what I can see, they just didn't have the knowledge to take on things like feedback control or the stabilizers properly.

My next step is to set up Watcom and figure out how to compile this beast, then I can get to work fixing the Hover code.

Spent some time digging through the dynamics code, overall it doesn't look too bad. All the needed bits are in there. The implementation of the tail boom drag force is odd, it isn't a calculated aerodynamic force, it's an artificial restoring force that is either on or off. That can be upgraded once I figure things out. I also noticed a couple of minor errors, the gravity force is applied at the rotor hub, not at the center of mass like it should be. The weight of the helicopter tends to pull the helo back to vertical over time, proportional to the angle. They've added in two artificial restoring forces that serve a similar function, but the artificial forces are not proportional to the moment arm like a weight force would be. I need to look at the math more closely to see how hard it would be to fix this. The aerodynamic drag force has the same problem, it acts at the rotor hub, not the center of the helo body like it should.

I turned on the diagnostic plot of the forces on the helos in the ini file and watched the system react while flying. I noticed that the aerodynamic forces are not implemented correctly on all of the helo types, some of the user-created models need some adjustments to the dynamics. Some of the physical constants need to be updated and some of the forces need to be turned on or off so they are not based on an incorrect dynamics model. Then the different types of helicopters will have a different feel when you fly them and behave correctly.

I didn't see any evidence that the rear horizontal stabilizer is modeled, but that isn't a very significant issue. The main rotor down wash flow is so complex on a helo it's hard to really say what the stabilizer will do under all flight conditions. The Comanche is different, with the stabilizer way up high out of the down wash, it could be modeled pretty easily on the Comanche. When it's located in the middle of the tail boom it's a difficult call. Mainly it's there to help level the helo out when it's moving forwards at a high speed. Leaving it out really isn't a big issue. Some helos have a horizontal stabilizer and some don't, so even the design engineers can't decide which is better.

I have Watcom working, that wasn't difficult at all.

it's good to hear that everything is working
A long time ago thealx mentioned that he plans to work the engine depending on temperature or altitude.

I did see a block of code which calculates the air density based on altitude for the aerodynamics, but I'm guessing it isn't applied to the turbine engines. I haven't seen anything yet that uses temperature.

Yeah, that's my function. there was some linear calculations that allow to fly over 30 000 feet, so it was easiest way to limit ceiling by linking lift power to density. Not very accurate maybe, but good enough at the moment.
I had plans to implement temperature and wind calculations in 3D but never started. Currently only wind is generating in two dimensions, which affect particles, helicopters (controlled by players) and some projectiles.

Ah, so it is linked to the engines already. Nice.

I see in the normal dynamics code where gravity is subtracted out in bas_dyn.c then again in co_dyn.c which can't be right. Is there a description anywhere of the program flow? Which module is the traffic cop that directs how the program executes and which dynamics model gets used? I see three dynamics models, a simple vector based one, a common dynamics model, then an advanced dynamics model for each helicopter type.

Maybe I should stop digging and get to work on fixing the Hover/Altitude Hold instead. Grin.

I remember reading a while ago about open source flight simulator physics engines. Do you think it would be possible, and worth the effort, of implementing one of those? Or do you think the fixes you can make would be good enough?

An open source flight physics engine? It might be useful to look at it, you can always learn something new from other people's work. The open source engine would have to include a large number of helicopter specific features to match what's in EECH already. It would be quite a chore to replace the existing engine in EECH, though. From what I've dug into so far, I think we can make a few minor changes here and there and have a pretty robust physics sim. It's mostly there already. It might be more interesting to add in a module to handle fixed wing VTOL aircraft like the Harrier...

New flight physics for airplanes?
it will be unnoticeable for the player who is doing missions imo

Right you are, not of interest to most players. Graphics improvements seem to be the most desired features. I'd like to see the terrain rendering/texturing go up a notch or two.

it's true.for the past two years EECH has been using dx9, but there is not any graphical effect that offers dx9.

OK, I went through all of the Advanced Dynamics settings files, pulled the numbers into a spreadsheet, and gathered up all the real numbers I could find for comparison to actual values. A lot of the dynamics values for a large number of the helos are the default numbers from the Apache which is why things don't feel quite right. A large heavy chopper shouldn't maneuver just like a nimble Apache. Some efforts were made by you guys to adjust some of these values, but I think we can improve that by quite a bit. The number of rotor blades has never been used, the distance to the tail rotor has never been updated, the default maximum main rotor rpm may be incorrectly set, things like that can be updated. I still need to dig into the dynamics equations to make sure i understand how all of the numbers are being used, and check the units for proper use. For example, all of the Power values are in horsepower, where all of the rest of the numbers are metric. I need to check and make sure the developers really meant to use hp and not kw's.

After some additional digging I've come to the conclusion that the developers really did intend on using hp for power levels, not kw's. Updating all of the numbers (especially the Inertial values) in the setup files is about as far as I can to go in correcting the physics, without the developer's raw equations it's pretty difficult to go much further than that. I'm still going to attempt to fix the location of the gravity force, though, and maybe a drag force or two.

it's not a flight model but ...out of curiosity, you can increase these parameters in eech.ini.to the values I have given. I love this inertia head.




g-force_head_movement=13.0 # amount of head movement caused by gravitational force (wideview only) (n = Gs, 1.0 = normal, 0.0 = off) (default = 1.0)
cockpit_vibrations=5.1

Wow, 13 g's. That's a LOT!

All Right! Stable-Hover-Hold is working!

I tested most of the helos and did a bit of tuning and it's working well. You hit Shft-H and it first stops the vertical motion, then pulls you back to the altitude where you hit the button. I tested it at the max climb rate, the max falling rate, and everything in between. When it locks in at an altitude it holds it there to within a meter. I suppose I could loosen up the loop and let it bob up and down a bit. What do you guys think of that?

Altitude-Hold is next.

Well done, you are get in the scripts really quickly. Small altitude deviation sounds like a good idea, if it will not cause problems.

Test bobup in another simulator.

The only time you would notice the deviations would be hovering next to the ground, at altitude you won't even notice it. The chase camera bobs around a bit already, though. Right now I have it over damped, there is a slight bit of drift, but it locks in pretty tight.

Yes, I'm getting familiar with the code, the dynamics portion anyway. The developer has done some screwy things, but I'm figuring it out. The hardest part so far was tracking down the air_density problem. The developer has three physics models in the code, vector mode, a basic flight model, and flight model #2. The air_density parameter changes in each one and one of the mods missed the alterations in FM2. I had to fix that before the hovering code I wrote would work.

I haven't tried the bob-up feature yet, is it working or not? I could write one if it isn't working. All I'd have to do is input a step change into the target hovering altitude. My Hover-Hold isn't slow like the video, it moves really quickly. It would surge upwards and stop at the target altitude within a second or two, then wait for you to hit the key again to drop back down.

OK, I re-read the manual and there is not an automated bob-up feature. Bob-up engages a display screen and lets your wingman know what you're doing. Do we want an automated bob-up function? I could set one up using a second press of the "O" key and use it to toggle up/down motion. You would use it in Hover mode.

for me-yes please

As we having Christmas days here, I have a request too.
As you know, autopilot is presented in the game but works pretty ugly - control of your helicopter is taken by AI and it's flying like on the rails. What will be great, and what actually I need for Hind project - waypoints following automatics, that uses same controls as a player. It will be not so accurate, but good enough to move from the base to the engagement area, and same time will act more natural. Technically, it should be not a replacement of current autopilot (which could be named "pass controls to the second pilot"), but addition to the existing hovering modes.
Maybe you will be interested as it will use same code that you are changing now.

Wow great idea.unnatural ai autopilot is one of the worst things in this game.

Well, I suppose I can add it to my list to look at, I haven't dug into any of the autopilot code. I did notice it's a bit over-aggressive and rides on rails.
My current list includes the following now:

Fix the Air_Density problem **DONE**
Fix Hover-Hold **DONE**
Fix Altitude-hold **Halfway Done**
Add in auto-Bob-Up **Started**
Update all of the physics parameters for all of the helos **Started**

Go through all of the FM2 helos and fix various physics problems
Take a look at the auto-pilot to see if I can fix it

Great work!

Good to have a skilled dev helping with this game (I don't count myself in that group. My dev skills with C are not that great)

Thanks guys. Quick question, does the current Mi-24 Hind have a heads up display?

No, at least not one that shows flight information. I think there might be an aiming reticule, but haven't flown the Hind much.

I started adding one that could be switched on/off via the ini file as a "cheat", but I didn't complete it.

That's right, only targeting mark appear when player switching on turret gun.
Different story is Clickable Cockpits mod - same layer also used for instruments manipulation.

Roger. I need to look into the Clickable Cockpits mod.

I've got the Altitude-Hold most of the way working, I stayed up way too late last night. Currently I have two problems I still need to fix. It has a tendancy over time to keep speeding up until it exceeds the maximum allowed velocity, and it's having trouble negotiating V-shaped valleys at the bottom transition. I think I know how I'm going to fix both, it just takes time and more testing. It's not going to be capable of handling all terrain extremes, you will have to keep an eye on it and pull it out of hitting vertical walls, diving into deep, steep sided narrow canyons, things like that. It's really pretty easy to fly with it on, all you have to do is steer and keep it away from vertical walls.

The Razorworks guys cheated and directly modified the position variables for the helo to make the autopilot work over ALL terrain features. That's why it behaves so strangely.

Yessss!!!

Marc..

That's fantastic
Marc..

OK guys, I have a functional question for you. I have the pop-up maneuver working. I'm calling it a pop-up maneuver because I had to use the R-SHFT-P key to activate it, the EECH event handler wouldn't let me reuse the O key. "P" for pop-up.

Option A: Right now it toggles. Press the key once and it pops you up 20 meters, press it again and it brings you back down.

Option B: We could change the way it functions and have R-SHFT-P pop you up in 20 meter increments, hit it twice to go up 40 meters.
And we would use the R-CTRL-P to bring you back down in 20 meter increments. Harder to coordinate with your fingers, but more flexibility in heights.

Which Option do you like? Vote now and vote often.


P.S. I have the Altitude-Hold function working too, I'm testing it and making adjustments to the settings. It works most of the time, but it has trouble with vertical walls. You're going to have to steer around them. I'll put a 1.16.2alpha executable up for you guys to test with in a day or two with all of my upgrades.

Javelin I think everyone use some kind mapping keyboard software, so it's not a problem do what you think is best.

"1.16.2alpha executable "
do you have any description of what is in the changes? I do not have access to the repository.I suspect that the cockpit lighting, something else?

Option B

Marc..

This would be my preference.

Cheers!

Option B for me.

Some additional thoughts though, would it be possible to have a key that makes the pop-up enabled, then use some key combination of Arrow Up / Arrow Down to change the height?

Is 20 metres increments set, and could it be too much? I'm thinking of things like hiding behind a line of trees, would popping up 20 metres leave you too exposed? I've not used pop-up too much, so don't have experience to think about.

B

Right now I have a combination of both A & B working. I'm using A to pop up and down, and for B I modded the Alt-J and Alt-K adjustments to Alt-Hover to work for ALL hover modes, and increased it's adjustment to 10 ft or 10 meters. The 1 ft thing wasn't working for me. As it stands now, you can get either 10 m or 20 m hops out of it, depending upon which keys you press.

Generally, I go into Shft-H Hover mode, use Alt-J or Alt-K to adjust my height, then Shft-P to pop up and down.

I'm also testing the Alt-H Altitude Hold and Alt-J and Alt-K are handy to adjust the control height for my nap-of-the-earth flying mode testing.

New status list:

Fix the Air_Density problem **DONE**
Fix Hover-Hold/Stable-Hover-Hold **DONE**
Fix Altitude-hold **DONE**
Add in auto-Bob-Up **DONE**
Update all of the physics parameters for all of the helos **DONE**

Go through all of the FM2 helos and fix various physics problems **Next**
Take a look at the auto-pilot to see if I can fix it