RvEYoda

I'd like to believe both ED AFM and BMS5 have better flight models. For the BMS 5 case we have real F-16 pilot comment on it and for ED AFM, well, ED has more than just 1 real world operator testing these things ;). Again I dont have any data comparison, only "feeling".

Yes, it was poor naming perhaps :P Update: Back button bug confirmed Unable to reproduce the F11 bug. I was running 1024 windowed mode. Any recommendations on what settings I should use? I have been pressing F11 for 2 minutes and using fast rotate (in a server with 30 players) After 5 minutes memory usage (normal RAM) is still at 410 MB

The auto-snap has a certain "radius" around each target from where it can snap. If two targets are flying very close to each other, then it can be tricky to bug both. (But not impossible) check you have all required dlls

yes it is possible, Gadroc has already done it. It is on top by default and you can set your control buttons around it to not steal focus when pressed.

Steps to reproduce please (otherwise I cannot report to devs). See if you can figure out a foolproof way of creating this crash, like you did with the F-11/spin bug :) (i am still not home so cannot confirm yet) I'll need details like "start attached mission, press escape, press Q and quickly press..." I expect to send 2 major crash reports when I get home, in about 3 hrs

Good catch. That does sound like a memory leak. When I get home I will try to reproduce this. If I get the same results I'll ask the developers and let you know their answer (if I am permitted to relay it :)) In the mean time I ask more people to try exactly this method and report back!

1,2 agreed. What about something like this (terribly unpractical and doesn't take all effects into account, but might be possible to do ;)): You put the F-16 on mount #1. Then you cut off an elevator from the main aircraft model but keep it in the same place, just on a separate mount #2. Now you run your tests but measure only the forces/moments on mount #2 Perhaps defeats its own purpose, but I think it would be fun to see :)

That is fine. I am thinking of hosting a separate forum+chat+etc for this project.

OK. I will do that. Btw do you know of any proposed FEM solutions to the problem? I always thought it would be an idea to reduce the complexity of whatever equations used and instead use more separate lift/effect zones on the aircraft.

Never said I wanted to, but if someone says "this is true" to me I say "hang on" until I'm showed why. I aint going to be satisfied until I know hwo to implement it :) (which you basically showed here)

Example: You have been told you need to investigate the shape of a function depending on 2 varaibles: F = F(A,B). You are also given a range of A and a range of B You are also given access to an abstract tool, which given a value of A and a value of B gives you a value of F. Now you want to make a heightmap plot or something simple to describe the function. You have no more knowledge about the system except you have been told it has two inputs. The case is not meaningless. Variables A,B,C,D,E,F,G,H *enters* [bLACK BOX] *produces* pair of shoes. For example [bLACK BOX] = wind tunnel :) Like you have shown the case of NS is NOT = [bLACK BOX], but again, NS is just a case. I agree with you, properly stated tables will probably yield far better results if the mdoel is complex and the needed simplifications to solve it in real time will significantly reduce the solution accuracy. ------------ Afaik falcon 5.5 fm are just data edits, but I could be wrong :P

Perhaps you could call it that but I assume no shape at all about hwo the function looks. The question is not "if", the question is "how much". Of course NS is a model (still I have never seen it) F=ma is a model, world model hehe :P.

For the case of solving your model as stated, yes, for the general case no. Again, like I wrote I have not studied your model, so i can only draw conclusions about the general case. If I come across a completely unknown system of 8 paramers I'd have to do it my way, which assumes no model at all. I have no disagreements with what you say here.

Nobody has stated anything else, but if someone claims he can table solve/store an 8D completely unknown equation with good precision, I will not sleep until he shows me what ram he has got :P.

As I've been saying all along...Again, I merely gave one of my examples where the they were equal, some of my other examples were of non-equal. (Here comes the "kick-ass algorithm for finding parameter importance out of empirical data" :P) And this answers why you can do it. You were able to derive an expression of the equation which made it no longer "completely unknown", so it is no longer an entirely empirical table of Cm, but rather empirical tables of its individual sub-functions. Yes, you were able to separate teh equations into smaller functions of maximum 3 parameters, thus it is easily feasable to use data tables. No disagreement here :) What I'm saying, is that if you had an completely unknown function of 8 variables, there would be no guarantee you could split it, and you would worst case end up with an 8 dimensional table that likely would be too large. Again in this case, your function isn't completely unknown (the opposite case of what I've been talking about all along), so it isn't an 8D table. This is not an assumption, this is the general case. You could say "an example". The general case is never "incorrect", but can in special cases be simplified. In our case you have just proven that we DO have knowledge about this function, enough knowledge to simplify it into several sub-functions and making the table approach possible.

I need everyone interested and ideas from all squads, so welcome aboard if you are interested!

I have been given information and sample source code from ED on how to download the master server list. I have also been allowed to share this, though I will be restrictive since it can be abused. What can we do with this? We can create a server browser with chat and a built in mod manager. How does this help? This browser will show the same server list as the ingame one, but users will automatically be asked if they want the browser to auto-download and install the required modifications or "default files" to play on a server. This ensures everyone is on a level playing field but also prevents cheating, since every server can have their own integrity check profile, without the hassle of pilots having to switch files before joining another server. Consider it an automated version of the debated "community mod pack". Better yet: Servers can have new aircraft, missiles, sensors, views, flight models YOU NAME IT profiles, and it will all go seamlessly between all integrity checked servers - and everyone will have the same. (It also solves the problem with Tacview working, but live-tacview wont) Best of all: It wont break anything for those that want to integrity check the absolute stock setup, in fact it will help them, since the browser will "auto-revert" when finished playing on a modded server (or when joining a standard one). When? I hope everyone reading this that are interested will ask around for people knowledgable in C/C#/C++ programming (.NET also!), especially network programming and GUI programming. We will need to develop this from scratch and make it very simple, so we need to set up a team for this project so we can get it finished ASAP :).

What assumption, I dont know which one. You say I made an incorrect assumption but I dont know what is incorrect, not even what is the assumption. Compute a variable, hang on here, what variables, we were talking about functions made up of independent variables right? In order to calculate a function of N independent variables you cannot use N/50, NASA cannot change that. It is math. You may say that some have less impact on the function due to their impact of the overall value compared to the less, but that is just dancing around the question. The question is whether flight models can be developed to take into account multi-variable solutions of the constants mentioned before using only data tables. I browsed through the document but see only 2-dimensional data tables at first glance. I dont think this document has anything to do with our case. Strictly mathematical question Are you saying Cn is an unknown function of 8 independent variables which are all in tables or not? Example: If an unknown mathematical function such as F = F(A,B,C) is to be tested empirically and the results stored in tables (in order to describe the function for a specific regieme), then you need to create a 3 dimensional table, such as double[N_a][N_b][N_c]. Nothing can change that, because we have said the variables are all independent, and the function is completely unknown. Consider: A 1-variable function F=F(A). To numerically explain this function you need a 1-dimensional data set. A 2-variable function F=F(A,B). To numerically explain this function you need a 2-dimensional data set. An N-variable function F=F(A_1, A_2....,A_N). To numerically explain this function you need an N-dimensional data set to include the effects. Other Example: Falcon calculates (afaik) some CL (is this a lift coefficient of some sort) from a function CL(AoA,Mach). This function is a 2D linear interpolation of a 2-dimensional data table. However the real CL value changes with a lot more parameters than just AoA and Mach, but because working with only 2 parameters is easy, generates a reasonable low data amount and requires low amount of computations it can be used to reasonably describe basic flight characteristics of an aircraft. Now assume a new situation: We wish to create a new flight sim based on data tables. We come up with a kick-ass interpolation algorithm and another kick-ass method to determine what parameters a function mostly relies on and which ones need the highest data resolution (step distance). Let us say we want to know a basic parameter (such as some force on the aircraft) F and our algorithms say we have F = F(A,B,C,D,E), so a function of 5 parameters mostly. Now what wind tunnel data do we need? Now we must go back and see what our second kick-ass algorithm did and get the parameters' individual required resolutions. Let us say they are 5,6,7,8,9 entries respectively. So we once again ask our fictional kick-ass algorithm to tell us where to best place our measurements, we run them for all the combinations possible. This generates 5*6*7*8*9 data points, which we store in a data table. This data table is in fact our source for how to determine the function F while flying the sim together with our kick-ass interpolation algorithm. The total memory required for this function was about 60 kB when using floats to store the data. Now increase the parameter count of F and you run into trouble. Lastly You cannot say that a function an unknown function F(A,B,C,D) can be simplified to F(A,B,C,D) = F1(A,B) (+,-,/,*,...) F2(C,D), because then it is no longer unknown. You have assumed a lot about the function and you can no longer claim that your model is entirely data based. My Question to you What is the highest dimensional data table you would need for your sim?

I may have some really good news soon about the possibilities of an auto-modman server browser

Stutters unfortunately, but thanks for the idea. I strikes me now that we dont need to export the textures. All we need to do is to use the built in multi monitor capability :P

I already wrote you can of course distribute them freely, I merely mentioned 11 for the case where you use a uniform distribution... How am I incorrect in mentioning an example, I don't understand. Let's go back to your original post (I'm sorry but I cannot understand what your post above means): Cn................................................ .......α, β, r, p, δh, δa, δr, δlef The function..........................................The independent variables that the function depends on To be able to Map from variable space to the resulting function space using table data, each independent variable is given one dimension. Then we must give each dimension a set of steps (of course step range can vary by alot). Then we must calculate how many table entries we get, it is N_entries = N_α * N_β * N_r *N_p *N_δh * N_δa *N_δr *N_δlef My question to you, is how many steps do you believe the different N_[]s require independently. Then the total memory usage for the lookup table becomes N_bytes = N_entries * sizeof(entry)

I never said you required 1GB of data. I said with 1GB of data you can store 11 steps per dimension in an 8-parameter function, if you want to have the values calculated from 8 dimension tables. And this is what i've said all along... no in fact, I never used any "fly on rails" The question is not whether the principle works. The question is can we store enough data in RAM so we can use efficient enough interpolation to represent the flight regiemes interesting ? For example, in Falcon terms, how many alpha breakpoints would you consider enough? Now make the same for the remaining 7 parameters and assign more points to those parameters you consider more important. Multiply them all (like 3*4*5*6*7*8*25*36 = N_table_entries) Now multiply by number of aircraft you wish to have this type of flight model and multiply by how many such functions you need, you will quickly run out of space if you are not careful.

Nice, I was suggesting such a flight model also, but they have to be careful so that they also remember that each flight surface also affects the others. I cannot give you specifics, but it happens that the commercial sims have more accurate flight models than the military ones ;). Maybe I was wrong, but when I flew it (;)) and those I spoke to I got the impression that table data was not used in the same way, but instead the flight parameters were calculated in real time like ED's AFM.

I think Mav-jp already stated that X-plane doesn't use table data but solves simplified equations instead. I am eagerly awaiting his answer on how he would use 1 GB of FM space. I think it should be possible, but not so simple as he has made it sound before. Basically he needs to get the avg points per dimension somewhere around 10, then use some nice interpolation (and possibly extrapolation) on that data.

No I would like to have Mav-Jp explain how he would use the 1GB of data to create a table based flight model. I am not saying it isn't possible. But what I am saying is that it likely isn't going to give you what you want if you use AF's linear interpolations.