Thursday 30 September 2021

One Year Of Aeronauticals

 So it has been an ENTIRE YEAR (from 1st October 2020) of working on my aeronaut/aeronauticals/dogfighter game that was loosely inspired by Ace Combat with 1930s aircraft and a lot more tightly inspired by Descent: The Great War space combat simulator but with gravity.

How it started (after a month or so anyhow) ...

And, how it's going ...

 
So ... that's been a bit of a change then ...
 
First up, I've created a brand new class of object, aeronautVehicle, that derives off vehicle class, which derives off rigidShape which is the thing that controls the physics, and shapeBase class ... which everything and it's dog seems to derive off if you want it be any form of a dynamic object.
 
This meant that I had to work out how an aircraft actually flew worked. Obviously this would be incredibly boring so instead I worked out how an exciting arcade style flight model worked basing it on ideas and physics values that I mostly pulled out of my bum.
 
Actually that's not entirely true, I did detailed research for months previously, basing everything on real world values and compiled lengthy text files on how the whole thing would work; eg: hitpoints were based on empty aircraft weight divided by length per area element (wings, tail, etc); and thrust was based off maximum take-off weight divided by kilo-watts of power per engine (or something like that) with modifiers for total aircraft area versus air friction... eg: fat aircraft would lose speed faster than smaller ones, etc, etc.

Similarly gun damage was based off calibre, calibre per 10mm modifier, number of guns, and some other stuff, with more guns increasing both damage and overheat timing.

The real problem was creating an Ai system to fly the aircraft against the player. Not only did they not have to crash into anything --- which ended up being much harder than you'd think --- they also needed to act as if they were a real pilot, so I ended up making 12 (?) evasive maneuvers that they could pull when neccessary. Stalling ended up being a huge issue depending on speed and attitude of the aircraft and tightness of a turn. I ended up creating a class of super-Ai pilots who I never stood a chance against, and had to dial the whole OP back, discovering that the margin between genius Ai flyer and absolute idiot was very small.

Damageable elements reduced performance of the aircraft's speed, turning, rudder, etc, and the whole thing had to modeled in 3D and textures to give the player good visual feedback as to what is happening to their aircraft.

PBR was a thing ... a thing which I did not know what ORM stood for and so ended up putting the rough and occlusion maps into the wrong channels because that is how they are set out in the materials editor.

My grasp of networking certainly improved, I had managed to create a co-op mode - based on seperate instances on the same box because I only have the one machine - and I soon realized that I wasn't supposed to just fill the whole of pack/unpack with all of the read/write data because much of that was on the server anyway. Trailing definately need to be on both or either Ai/players would trail and the others wouldn't depending on how they were handling their aircraft. Huge amounts of pack/unpack were commented out and the weird jittering my AI aircraft had been doing in debug mode stopped.

I created a crude lobby system were the level loaded without the mission starting except a briefing to give the player the details of the mission via a seperate file and to choose their aircraft and paint schemes. The player would then wait in this psuedo-lobby for other clients to connect or start a co-op multiplayer mission alone and newly connected players could drop in and out. Debriefing required a bit more voodoo as shifting large amounts of player-mission contextual descriptors - that's text - required some fancy thinking to avoid abusing huge numbers of server to client calls - something that I had been very keen on working smoothly and with the least possible amount of bandwidth.

 Creating missions was a series of timed and triggered events based on what the player and Ai pilots were doing. This required lots of scripting due to the lack of physical triggers which a player could walk into on a more normal character based game of the types I had previously demoed or developed.

And so that is kinda of it. After a year of development I have a fully functioning dogfighting game that plays halfway between an arcade shooter and simulator with a mission complete with playable roster of aircraft, briefing, debriefing and drop in and drop out online co-op mode. So naturally it's time to relegate this to the backburner and go return to my previous game project (and really tidy up it's networking) as I want something that can at least be released for Early Access early(ish) next year and Freespace-style dogfighter will require the most work whilst topdown catgirl mayhem is around halfway complete but needs "fluff" adding for the player's ease of life for the first few levels.