4 ms·

I wish all video games had this level of scientific depth underlying them.

by dforrestwilson 7y ago

keyle 7y ago
You'd be surprised how deep some of the AAA studio go in developing complex systems for physics, rocks or grass!
- nineteen999 7y ago
  Unreal 4.23 just integrated a beta implementation of "Chaos Destruction" - pretty crazy stuff. https://docs.unrealengine.com/en-US/Engine/Chaos/ChaosDestruction/index.html https://docs.unrealengine.com/en-US/Engine/Chaos/ChaosDestru... https://www.youtube.com/watch?v=u3ktiewcLpo https://www.youtube.com/watch?v=u3ktiewcLpo
dahart 7y ago
As an ex console games developer, I'd say they absolutely do. There may be a difference in the fidelity & geometric complexity of the output, and it's not all a singular focus by one person or as pretty as the trees made by Grove8, but the methods being used even in mobile games are building on decades of compounding scientific work. Procedural geometry and shading models are just the easy ones to point at, but generally speaking, physics, AI, design, gameplay, and art are all compounding as well.
- derefr 7y ago
  For the physics and art, maybe. But how about NPC AI in e.g. a market-arbitrage game, that shows known psychological biases that can be exploited? Or a damage system in an FPS where people/creatures get weaker and slower as they get hurt, and leave blood trails to find them by if they’re not careful? Or how about an item crafting system that’s just regular chemistry, requiring complex glassware setups, catalysts, ice baths, filtering and distilling, fume hoods, etc.? Games sure don’t seem to favor fidelity when it comes to the “game” part. :)
  - dahart 7y ago
    I’m not sure I understand what you mean. What you’re talking about is gameplay ideas that involve the player doing realistic things. What I was talking about is game engines that use scientific research in the construction of game engine features. You’re bringing up an entirely different concept. And I agree, and even said before, that fidelity doesn’t match these film style assets. The whole point is that looking at fidelity is not the way to gauge how deep the underlying techniques are.
    - derefr 7y ago
      Enemy AI, damage systems, crafting systems, etc. are still—at least potentially—game engine components. They’re thought of as “application layer” components of individual games because games don’t tend to share any code for these, rather coming up with their own implementations (usually just some finite-state machines.) But if you had an in-engine system for doing realistic chemistry (maybe in bulk, on a voxel or particle level, on the GPU) then that’d be less of a gameplay idea, and more a module to drop into a game’s physics, no? Sort of like soft-body dynamics is a module to drop into a physics engine. It’d just be part of the game’s physics that’d constrain the way that a crafting system must be implemented if you’re using it; sort of like how using kinematics engines constrains the way you implement vehicles. If the entire game is full of tracked atmospheric gases; and the heat of these gasses can convect into nearby solids/liquids; and the gasses can also condense if they get too cold, and things can crash out of liquid solutions if the liquids get colder, etc.; then inherently you’re going to be able to build things like a distillation apparatus under such physics; and inherently you’re probably not going to be able to build a Minecraft-like arbitrary-crafting system without going “out-of-engine” in a difficult way (because now you have to worry about how popping things into existence affects air pressure—you’ll be fighting the “chemistry engine” every step of the way to building it!) And, also, you’ll get side-considerations, like it being hard not to have a weather system in games using such an engine, because each zone of bounded size inherently becomes a terrarium where the “ceiling” can form condensation. (And there would have to be such a ceiling—in fact, a whole physically-simulated “world box”—otherwise the gas escapes! Unless you want an Earth-sized spinning zone to keep your atmosphere sucked down...) Again, you’d have to fight the engine to use it but not have realistic weather. It’d be a lot easier to “trick” the engine by just putting materials into the game that would have the same real properties that would create an illusion of there being no weather in a real terrarium (i.e. a really big air conditioner, sitting “outside” the terrarium-edge-box, like the ceiling in The Truman Show.) Or, in the sense of a damage system, I’m not really talking about the gameplay concept of a damage system, but rather choosing to implement player-controllable avatars (and mobs, and everything else “living” in the game) in terms of a scripted or player-controlled “mind” commanding a faithfully biologically-emulated “body” to move (think “body as kinematically-simulated robot, that has a movement AI, and accepts goal-pathing instructions; player control/NPC AI as source of such pathing instructions.”) And, since the body would be physical, entities would only be able to do what their bodies are able to do—such that if an entity’s body gets damage to a crucial muscle or organ, its body is now less able to follow commands to run, however much the control-layer sending those commands wants it to run. Sure, from a top-down sense, that’s a “choice of damage system”; but from a bottom-up sense, that’s a physiology modelling engine that could be put into any game engine and used for any number of things (from a game about trying to stab your opponent in their vital organs, to a game about poisoning, to a Wii Fit-like game that tries to model which of your muscle-groups will tire first, to...) Consider why we have a “physics engine” in the first place. 2D games back in the 80s and 90s mostly just implemented one-off physics at the application layer; but then, soon after entering 3D, we wanted real physics, and there’s only one “real physics” to model, so we started reusing and generalizing our logic for “real” physics out into physics engines, and started building the “game physics logic” for games that wanted mostly-realistic physics, as just a policy layer on top of such a generic real-physics engine. That’s what I’m talking about here: there are other such “engines” we could have, which would be scientifically-sound models of reality that players could predict the behaviour of based on their knowledge of the real world, where in-game systems could then be implemented the way most 3D games’ physics systems are: as configurations of, and scripting hooks into, a few polished, flexible, generic models.
fenwick67 7y ago
You may enjoy this talk, the devs for Far Cry 5 procedurally generated quite a lot of their world, placing plants based on the ground slope, sun, whether they were under power lines, elevation etc. https://youtu.be/NfizT369g60 https://youtu.be/NfizT369g60
SolarNet 7y ago
The issue is software development time and the tools available. Game developers can make trees that look like this, no problem. And they do study the underlying science. The issue comes with building a system to generate it asynchronously, to load and unload the data in a timely manner, to optimize around common use cases, to plan out memory allocation, and so on. Yes your computer could run it, but game engines are too simple to support the level of complexity to use these kinds of tools dynamically (or even at all sometimes). Easier to just glom it all into a static world made by artists that are supported by tools like these, and ship it. To draw an analogy, it's like trying to manage a cluster of a thousand containers across a hundred interrelated applications across a machines with only docker compose when one really needs something like Kubernetes.
hyperpallium 7y ago
Most of them use speedtree, original tech was developed for a golf game (trees on the course), they eventually did the trees for Avatar. https://wikipedia.org/wiki/SpeedTree https://wikipedia.org/wiki/SpeedTree
acomjean 7y ago
They’re starting to get quite sophisticated. For example the birds of the game “red dead redemption” inspired an Audubon article https://www.audubon.org/news/birding-its-1899-inside-blockbuster-american-west-video-game https://www.audubon.org/news/birding-its-1899-inside-blockbu...

I wish all video games had this level of scientific depth underlying them.

keyle 7y ago

You'd be surprised how deep some of the AAA studio go in developing complex systems for physics, rocks or grass!

nineteen999 7y ago
Unreal 4.23 just integrated a beta implementation of "Chaos Destruction" - pretty crazy stuff. https://docs.unrealengine.com/en-US/Engine/Chaos/ChaosDestruction/index.html https://docs.unrealengine.com/en-US/Engine/Chaos/ChaosDestru... https://www.youtube.com/watch?v=u3ktiewcLpo https://www.youtube.com/watch?v=u3ktiewcLpo

dahart 7y ago

As an ex console games developer, I'd say they absolutely do. There may be a difference in the fidelity & geometric complexity of the output, and it's not all a singular focus by one person or as pretty as the trees made by Grove8, but the methods being used even in mobile games are building on decades of compounding scientific work. Procedural geometry and shading models are just the easy ones to point at, but generally speaking, physics, AI, design, gameplay, and art are all compounding as well.

derefr 7y ago
For the physics and art, maybe. But how about NPC AI in e.g. a market-arbitrage game, that shows known psychological biases that can be exploited? Or a damage system in an FPS where people/creatures get weaker and slower as they get hurt, and leave blood trails to find them by if they’re not careful? Or how about an item crafting system that’s just regular chemistry, requiring complex glassware setups, catalysts, ice baths, filtering and distilling, fume hoods, etc.? Games sure don’t seem to favor fidelity when it comes to the “game” part. :)
- dahart 7y ago
  I’m not sure I understand what you mean. What you’re talking about is gameplay ideas that involve the player doing realistic things. What I was talking about is game engines that use scientific research in the construction of game engine features. You’re bringing up an entirely different concept. And I agree, and even said before, that fidelity doesn’t match these film style assets. The whole point is that looking at fidelity is not the way to gauge how deep the underlying techniques are.
  - derefr 7y ago
    Enemy AI, damage systems, crafting systems, etc. are still—at least potentially—game engine components. They’re thought of as “application layer” components of individual games because games don’t tend to share any code for these, rather coming up with their own implementations (usually just some finite-state machines.) But if you had an in-engine system for doing realistic chemistry (maybe in bulk, on a voxel or particle level, on the GPU) then that’d be less of a gameplay idea, and more a module to drop into a game’s physics, no? Sort of like soft-body dynamics is a module to drop into a physics engine. It’d just be part of the game’s physics that’d constrain the way that a crafting system must be implemented if you’re using it; sort of like how using kinematics engines constrains the way you implement vehicles. If the entire game is full of tracked atmospheric gases; and the heat of these gasses can convect into nearby solids/liquids; and the gasses can also condense if they get too cold, and things can crash out of liquid solutions if the liquids get colder, etc.; then inherently you’re going to be able to build things like a distillation apparatus under such physics; and inherently you’re probably not going to be able to build a Minecraft-like arbitrary-crafting system without going “out-of-engine” in a difficult way (because now you have to worry about how popping things into existence affects air pressure—you’ll be fighting the “chemistry engine” every step of the way to building it!) And, also, you’ll get side-considerations, like it being hard not to have a weather system in games using such an engine, because each zone of bounded size inherently becomes a terrarium where the “ceiling” can form condensation. (And there would have to be such a ceiling—in fact, a whole physically-simulated “world box”—otherwise the gas escapes! Unless you want an Earth-sized spinning zone to keep your atmosphere sucked down...) Again, you’d have to fight the engine to use it but not have realistic weather. It’d be a lot easier to “trick” the engine by just putting materials into the game that would have the same real properties that would create an illusion of there being no weather in a real terrarium (i.e. a really big air conditioner, sitting “outside” the terrarium-edge-box, like the ceiling in The Truman Show.) Or, in the sense of a damage system, I’m not really talking about the gameplay concept of a damage system, but rather choosing to implement player-controllable avatars (and mobs, and everything else “living” in the game) in terms of a scripted or player-controlled “mind” commanding a faithfully biologically-emulated “body” to move (think “body as kinematically-simulated robot, that has a movement AI, and accepts goal-pathing instructions; player control/NPC AI as source of such pathing instructions.”) And, since the body would be physical, entities would only be able to do what their bodies are able to do—such that if an entity’s body gets damage to a crucial muscle or organ, its body is now less able to follow commands to run, however much the control-layer sending those commands wants it to run. Sure, from a top-down sense, that’s a “choice of damage system”; but from a bottom-up sense, that’s a physiology modelling engine that could be put into any game engine and used for any number of things (from a game about trying to stab your opponent in their vital organs, to a game about poisoning, to a Wii Fit-like game that tries to model which of your muscle-groups will tire first, to...) Consider why we have a “physics engine” in the first place. 2D games back in the 80s and 90s mostly just implemented one-off physics at the application layer; but then, soon after entering 3D, we wanted real physics, and there’s only one “real physics” to model, so we started reusing and generalizing our logic for “real” physics out into physics engines, and started building the “game physics logic” for games that wanted mostly-realistic physics, as just a policy layer on top of such a generic real-physics engine. That’s what I’m talking about here: there are other such “engines” we could have, which would be scientifically-sound models of reality that players could predict the behaviour of based on their knowledge of the real world, where in-game systems could then be implemented the way most 3D games’ physics systems are: as configurations of, and scripting hooks into, a few polished, flexible, generic models.

fenwick67 7y ago

You may enjoy this talk, the devs for Far Cry 5 procedurally generated quite a lot of their world, placing plants based on the ground slope, sun, whether they were under power lines, elevation etc. https://youtu.be/NfizT369g60 https://youtu.be/NfizT369g60

SolarNet 7y ago

The issue is software development time and the tools available. Game developers can make trees that look like this, no problem. And they do study the underlying science. The issue comes with building a system to generate it asynchronously, to load and unload the data in a timely manner, to optimize around common use cases, to plan out memory allocation, and so on. Yes your computer could run it, but game engines are too simple to support the level of complexity to use these kinds of tools dynamically (or even at all sometimes). Easier to just glom it all into a static world made by artists that are supported by tools like these, and ship it. To draw an analogy, it's like trying to manage a cluster of a thousand containers across a hundred interrelated applications across a machines with only docker compose when one really needs something like Kubernetes.

hyperpallium 7y ago

Most of them use speedtree, original tech was developed for a golf game (trees on the course), they eventually did the trees for Avatar. https://wikipedia.org/wiki/SpeedTree https://wikipedia.org/wiki/SpeedTree

acomjean 7y ago

They’re starting to get quite sophisticated. For example the birds of the game “red dead redemption” inspired an Audubon article https://www.audubon.org/news/birding-its-1899-inside-blockbuster-american-west-video-game https://www.audubon.org/news/birding-its-1899-inside-blockbu...