Zetheros
Developer
Hey Exanimates,
We're now mostly moving on from general AI changes, there will need to be some testing and tweaking, so we're looking at other things that can be tested at the same time. Specifically we're getting into the pathfinding system updates. We briefly mentioned this before as some of the planned changes feel necessary for the upcoming content, but they are already relevant, will improve the game in general and support a lot more in future.
The key change to the pathfinding is basically making it a lot more 3D. While the current pathfinding is basically a 2D map, the new system is a 3D map based on detailed voxel data. This properly supports multiple levels (verticality) and allows AI to understand how deep a drop is and potentially support future climbing and jumping actions, or even crawling. 3D objects in different orientations are described much more accurately (imagine an upside down table), which can also be used to build ramps or bridges to create new paths. Beyond these new features, this is a more stable approach that should result in less getting stuck or AI failing to recognise dangerous drops and so on.
From the AI perspective we're also including support for preferred paths based on "terrain types" and the like, this would for example encourage characters to walk over a paved path or stairway rather than the rougher ground near it. This behaviour and general path choices can change based on urgency or emotional states. Another important change we're working towards is high level logic that considers things like doors, and eventually interaction with them.
Finally we want to improve performance by offloading pathfinding work from the main physics and AI. Pathfinding is generally not very time critical, as in it really doesn't matter if it takes AI a few more milliseconds to figure out exactly where they're going next. This means we can put expensive pathfinding work in the background, freeing up resources for physics and having more active characters at once.
We don't need to make all these changes at once, even a basic version of the new system represents an immediate improvement over what we have, and removes a lot of headaches from content creation and dealing with problematic cases. Importantly this new system has been carefully designed to support everything we might need it do in the future, we can build on this without worrying about running into technical limitations that could invalidate or complicate our work.
Have a great week!
-the BM team
We're now mostly moving on from general AI changes, there will need to be some testing and tweaking, so we're looking at other things that can be tested at the same time. Specifically we're getting into the pathfinding system updates. We briefly mentioned this before as some of the planned changes feel necessary for the upcoming content, but they are already relevant, will improve the game in general and support a lot more in future.
The key change to the pathfinding is basically making it a lot more 3D. While the current pathfinding is basically a 2D map, the new system is a 3D map based on detailed voxel data. This properly supports multiple levels (verticality) and allows AI to understand how deep a drop is and potentially support future climbing and jumping actions, or even crawling. 3D objects in different orientations are described much more accurately (imagine an upside down table), which can also be used to build ramps or bridges to create new paths. Beyond these new features, this is a more stable approach that should result in less getting stuck or AI failing to recognise dangerous drops and so on.
From the AI perspective we're also including support for preferred paths based on "terrain types" and the like, this would for example encourage characters to walk over a paved path or stairway rather than the rougher ground near it. This behaviour and general path choices can change based on urgency or emotional states. Another important change we're working towards is high level logic that considers things like doors, and eventually interaction with them.
Finally we want to improve performance by offloading pathfinding work from the main physics and AI. Pathfinding is generally not very time critical, as in it really doesn't matter if it takes AI a few more milliseconds to figure out exactly where they're going next. This means we can put expensive pathfinding work in the background, freeing up resources for physics and having more active characters at once.
We don't need to make all these changes at once, even a basic version of the new system represents an immediate improvement over what we have, and removes a lot of headaches from content creation and dealing with problematic cases. Importantly this new system has been carefully designed to support everything we might need it do in the future, we can build on this without worrying about running into technical limitations that could invalidate or complicate our work.
Have a great week!
-the BM team
