Advanced Questions for Navmesh Generation / Recast Graph and Standalone editor

[NWCZack]

Hi,

My team and i are currently working on a large project which requires runtime, standalone navmesh generation for AI characters.

Since everything seems a little dated on the forum. about standalone builds with recast graphs, before purchasing this asset i had some questions.

First i would like to confirm that Navmesh generation is working in a standalone environment, as i have been completely unable to get a working test from unitys built in solution, working, in a standalone environment. I must confirm that there is no additional errors or dependency’s required when in a standalone situation. Specifically for Windows PC or possibly Linux/mac release. Web release is not an issue for my requirements but it would be good to know aswell.

Second… Our world is a quite large static terrain of slices, however there are 30,000+ objects being instantiated and spawned at random at runtime during startup. We need to make sure that a system such as A* can handle such a load, have you any benchmarks with objects with advanced geometry, like trees or houses that works on a large scale? if so what are the benchmark times? do you have a link to this information?

Third… we currently are using other AI assets to handle functions like attacking and wandering and such on said navmesh… Have you tested the system with any kind of AI system. Currently we are using EmeraldAI, which utilizes, unitys Nav Agent Scripts, whereas your project is using your own pathfinder scripts. Is there an AI solution that you have tested with for steamlined integration? i would hate to purchase a big fancy “path finding” asset and have no ability to use the nice generated navmesh with any kind of AI system.

Fourth… How does the system handle OffMeshLinks, or enabeling entities to walk/wonder onto segments of the navmesh? since so many objects are spawning on our terrain, and some are walkable, does the navmesh system have options for ignoring height, or slope?

Because we are intrested mostly in the ability to Actually have proper navmesh generation at standalone runtime, i am unable to even test with the free project would suite these standards.

I have no problem purchasing such an asset aslong as i can have some questions answered.

If you require more information, please let me know, i may be hesitant to post intricate workings of our project after so long in development, and may ask that some of our conversation be private TBH.

I understand the need to keep such questions in public eye, but if you require detailed information in order to answer these questions, we may have to speak privately.

[aron_granberg]

First i would like to confirm that Navmesh generation is working in a standalone environment, as i have been completely unable to get a working test from unitys built in solution, working, in a standalone environment. I must confirm that there is no additional errors or dependency’s required when in a standalone situation. Specifically for Windows PC or possibly Linux/mac release. Web release is not an issue for my requirements but it would be good to know aswell.

You can generate navmeshes in standalone builds on all platforms. The only requirement is that the meshes in the scene are marked as read/write in the importer and that they are not statically batched (since that also prevents the pathfinding scripts from reading them). You can batch the meshes after the navmesh has been generated though. Alternatively you could just use colliders for rasterization which will always work (in this package you can choose if you want to use rendered meshes or colliders as the source data).

Second… Our world is a quite large static terrain of slices, however there are 30,000+ objects being instantiated and spawned at random at runtime during startup. We need to make sure that a system such as A* can handle such a load, have you any benchmarks with objects with advanced geometry, like trees or houses that works on a large scale? if so what are the benchmark times? do you have a link to this information?

Are those just obstacles? In that case it should be fine. However generating the navmesh at the start of the game might be slow. It’s hard to give you precise numbers on how long it might take. Generating a navmesh graph can take between fractions of a second up to an hour depending on the resolution and size of the world.

Third… we currently are using other AI assets to handle functions like attacking and wandering and such on said navmesh… Have you tested the system with any kind of AI system. Currently we are using EmeraldAI, which utilizes, unitys Nav Agent Scripts, whereas your project is using your own pathfinder scripts. Is there an AI solution that you have tested with for steamlined integration? i would hate to purchase a big fancy “path finding” asset and have no ability to use the nice generated navmesh with any kind of AI system.

There are integrations with some other asset store packages, however not with EmeraldAI. The movement scripts in this package implement an interface which is very similar to what Unity’s NavmeshAgent uses however, so I don’t think it should be too hard to port. See https://arongranberg.com/astar/docs/iastarai.html

Fourth… How does the system handle OffMeshLinks, or enabeling entities to walk/wonder onto segments of the navmesh? since so many objects are spawning on our terrain, and some are walkable, does the navmesh system have options for ignoring height, or slope?

Sadly this is the part of the package I am the least pleased with at the moment. There is support for off mesh links though if you use the RichAI movement script. You can register to the RichAI.onTraverseOffMeshLink event to handle off mesh links. There is no automatic off-mesh link placement in the package at the moment.

[NWCZack]

Thanks for the fast reply.

I’ve been using a runtime navmesh generator and it seems to “stitch” the nav mesh epeices together and “cut” out the nav obsticals. Does this work in the same way? It seemed in the solution I’m currently using. Science it stitched them together. Nav mesh links weren’t an issue. The problem is. That asset will not work in a standalone situation.

Aswell. I’m fine with purchasing the asset however if it didn’t meet our needs I would probably ask for a refund :(. Or for you to work with us! But that’s all theoretical. I may continue to have follow up questions. Maybe there is a way we could meet in a chat like discord so I could forward you some images that may help illustrate what we’re doing

I’m working late tonight but may try to send you some images or links late tonight