Massive Vehicle Optimizations and New Features
Hello! It's been a month since the last release. I try to keep a monthly release schedule but this month's work has ballooned into something large enough that a release at this very moment (even on the tester's branch) is not possible. However, there is good news and progress that I'm very excited to share.
In summary, vehicles are in the middle of a large refactor to fix blocking issues preventing more interesting maps, but has expanded to include big optimizations for guns, bullets, and their visual effects.
[h2]Pathfinding[/h2]
[previewyoutube][/previewyoutube]
The above video shows ~500 vehicles doing pathfinding on a map that is (theoretically) 100km across, at 200 FPS. Vehicles have never, and could never, drive across a bridge like this before.
The core feature I was interested in was pathfinding for the vehicles. At first, I was planning on writing this myself and applying it only at the formation level. However, a friend pointed me to the Unity pathfinding package. I was skeptical because of my previous experience with Unity packages, especially those marked experimental, but this proved to be shockingly effective and easy to use.
After some quick prototyping, I found out that this could not only be used for pathfinding, but as a replacement for the current, and very expensive vehicle physics. Put simply, rather than vehicles running their own kinda strange physics, they can instead just follow the navmesh which is a "good enough" approximation of the map, bypassing the need for expensive ground clamping and physics calculations.
With the integration of proper pathfinding, not only does this mean that vehicles can correctly path their way around mountains, buildings, and airfields, but their performance is dramatically improved. There were some very big scares in the implementation particularly with regards to the game's floating origin, but those hyper-critical issues have been resolved enough (navigation takes place in something like a parallel universe 😂) that this is a way forwards.
Worth mentioning too is that this also addresses many old reported issues of vehicles such as their shaky movement when going over uneven ground.
[h2]Gun/Bullet optimizations + Externalization of Data[/h2]
[previewyoutube][/previewyoutube]
In the process of updating the vehicles, I accidentally ended up doing that turret/bullet refactor I've been dreaming about. Bullets are pooled in a much smarter way, which in particular effects AAA, and all the many little bugs they had (such as disappearing tracers) are now fixed. This was a pretty deep rabbit hole that has resulted in new particle effects for most of the bullet based weapons along with new and improved sound effects, especially for bullet flybys. Please keep in mind that several effects are still WIP, since right now the tanks are using machinegun impacts.
This rework also allowed bullet data to be loaded from JSON like everything else. They can be edited by anybody, it's easier for me to make new ones, and most importantly, no longer rely on hardcoded and unmodifiable prefabs.
[h2]What does this mean going forwards?[/h2]
As anybody who has seen me talking about this on the MicroProse Discord would know, I am really excited about this.
[h3]1. The game can hit its vehicle targets[/h3]
The current arena demo, which featured ~120 vehicles, was already at the tipping point of what I was comfortable with performance-wise. There wasn't room to go bigger from there. My original target for active vehicle counts has been around 500 vehicles. While that isn't Falcon 4.0 numbers of vehicles (thousands...), I was never going for that. An upper limit of ~500 units on the map sizes I intend the game to have affords both the scale and "living battlefield" feel that I'm aiming for.
One of my primary goals for Tiny Combat was for the player to feel like they were instrumental in winning a larger battle taking place on the ground rather than isolated dogfights in the clouds. This is one big step closer to that.
[h3]2. Much more interesting maps[/h3]
The new pathfinding code comes with many new features I am so excited about, some of which I wasn't expecting to ever be solved. Bodies of water such as rivers and lakes can block unit movement. Vehicles can correctly drive over bridges. They will prefer to drive on roads when possible, and avoid driving over the runway! They can even drive through tunnels. Oh yeah, and they aren't restricted to flat ground anymore.
One of my biggest anxieties with the game has been map design and terrains. A combination of technical and design issues has made terrains the most difficult subject for me to approach because of both specific technical limitations regarding the game's style, and how vehicles could possibly be incorporated into them. Even something as simple as a base being on higher ground than another base was pretty much off the table until now, let alone having the AI navigate around mountains, through ravines, and over bodies of water.
A great weight has been lifted from my shoulders with this, allowing me to focus more on producing interesting maps, scenarios, and content to fly around (and through 👀) without isolating ground vehicles from the action. There's some dream ideas I've had for maps since the project began that are now possible and I thank you all for sticking with me through this journey. There's still work to be done before this can be re-integrated fully into Arena, but it's getting close.
In summary, vehicles are in the middle of a large refactor to fix blocking issues preventing more interesting maps, but has expanded to include big optimizations for guns, bullets, and their visual effects.
[h2]Pathfinding[/h2]
[previewyoutube][/previewyoutube]
The above video shows ~500 vehicles doing pathfinding on a map that is (theoretically) 100km across, at 200 FPS. Vehicles have never, and could never, drive across a bridge like this before.
The core feature I was interested in was pathfinding for the vehicles. At first, I was planning on writing this myself and applying it only at the formation level. However, a friend pointed me to the Unity pathfinding package. I was skeptical because of my previous experience with Unity packages, especially those marked experimental, but this proved to be shockingly effective and easy to use.
After some quick prototyping, I found out that this could not only be used for pathfinding, but as a replacement for the current, and very expensive vehicle physics. Put simply, rather than vehicles running their own kinda strange physics, they can instead just follow the navmesh which is a "good enough" approximation of the map, bypassing the need for expensive ground clamping and physics calculations.
With the integration of proper pathfinding, not only does this mean that vehicles can correctly path their way around mountains, buildings, and airfields, but their performance is dramatically improved. There were some very big scares in the implementation particularly with regards to the game's floating origin, but those hyper-critical issues have been resolved enough (navigation takes place in something like a parallel universe 😂) that this is a way forwards.
Worth mentioning too is that this also addresses many old reported issues of vehicles such as their shaky movement when going over uneven ground.
[h2]Gun/Bullet optimizations + Externalization of Data[/h2]
[previewyoutube][/previewyoutube]
In the process of updating the vehicles, I accidentally ended up doing that turret/bullet refactor I've been dreaming about. Bullets are pooled in a much smarter way, which in particular effects AAA, and all the many little bugs they had (such as disappearing tracers) are now fixed. This was a pretty deep rabbit hole that has resulted in new particle effects for most of the bullet based weapons along with new and improved sound effects, especially for bullet flybys. Please keep in mind that several effects are still WIP, since right now the tanks are using machinegun impacts.
This rework also allowed bullet data to be loaded from JSON like everything else. They can be edited by anybody, it's easier for me to make new ones, and most importantly, no longer rely on hardcoded and unmodifiable prefabs.
[h2]What does this mean going forwards?[/h2]
As anybody who has seen me talking about this on the MicroProse Discord would know, I am really excited about this.
[h3]1. The game can hit its vehicle targets[/h3]
The current arena demo, which featured ~120 vehicles, was already at the tipping point of what I was comfortable with performance-wise. There wasn't room to go bigger from there. My original target for active vehicle counts has been around 500 vehicles. While that isn't Falcon 4.0 numbers of vehicles (thousands...), I was never going for that. An upper limit of ~500 units on the map sizes I intend the game to have affords both the scale and "living battlefield" feel that I'm aiming for.
One of my primary goals for Tiny Combat was for the player to feel like they were instrumental in winning a larger battle taking place on the ground rather than isolated dogfights in the clouds. This is one big step closer to that.
[h3]2. Much more interesting maps[/h3]
The new pathfinding code comes with many new features I am so excited about, some of which I wasn't expecting to ever be solved. Bodies of water such as rivers and lakes can block unit movement. Vehicles can correctly drive over bridges. They will prefer to drive on roads when possible, and avoid driving over the runway! They can even drive through tunnels. Oh yeah, and they aren't restricted to flat ground anymore.
One of my biggest anxieties with the game has been map design and terrains. A combination of technical and design issues has made terrains the most difficult subject for me to approach because of both specific technical limitations regarding the game's style, and how vehicles could possibly be incorporated into them. Even something as simple as a base being on higher ground than another base was pretty much off the table until now, let alone having the AI navigate around mountains, through ravines, and over bodies of water.
A great weight has been lifted from my shoulders with this, allowing me to focus more on producing interesting maps, scenarios, and content to fly around (and through 👀) without isolating ground vehicles from the action. There's some dream ideas I've had for maps since the project began that are now possible and I thank you all for sticking with me through this journey. There's still work to be done before this can be re-integrated fully into Arena, but it's getting close.
