View on Asset Store (opens in new tab)
Package details
- Latest version
- 1.3.1
- Last updated
- First released
- Unity version
- 2022.3.62 or later
- Package size
- 306.9 MB
- Files
- 1,705
- Favorites
- 13
- Views in the past week
- 23
- Requires
Price tracking
- Last checked
- Tracked since
- Lowest recorded
- $30.00 on
Keywords
DeepNavigation - Mapless Navigation Framework
DeepNavigation is a mapless navigation framework based on deep reinforcement learning.
Affiliate link · Purchases support this site
Price history
Last 6 months
USDAsset preview
1 / 12
DeepNavigation - Mapless Navigation Framework screenshot 1
About this asset
DeepNavigation is a mapless navigation plugin. It completely eliminates NavMesh, WayPoint, and A* algorithm, utilizing deep reinforcement learning to enable agents to autonomously learn to move in complex environments, avoid dynamic obstacles, and reach target points.
Whether your character is walking on the ground, flying in the air, or swimming underwater, DeepNavigation can provide unified, smooth, and adaptive navigation capabilities.
Features:
1: No need to construct or bake Nav Mesh, Way Point or Path Node.
2: No need to write or optimize A* or Dijkstra path search algorithms.
3: Pure neural network driven, dynamic decision-making.
4: Modular design, allow custom reinforcement learning modules.
5: Supports multiple movement modes within a unified framework, including ground, air, and water.
Applicable project types:
FPS, RPG, RTS, MOBA, SLG, Air or underwater games.
Target user:
Indie developer, Level designer, AI researcher.
From the publisher's Asset Store listing.
Technical details
Details:
1 :Developed based on TinyMind, it is entirely written in C#.
2: A complete toolchain covering training, testing, and runtime navigation.
3: Based on 2D vision and 1D vector perception input.
4 :Complete ground and aerial navigation demos to enable you to quickly learn how to use the framework.
5 :The trained model file is less than 2 megabytes.
6: The extensible framework allows you to customize rewards and perception inputs.
System Requirements:
Memory requirements: 32GB for training, 8GB for runtime.
Supported Development Platforms: Windows Mac.
Supported Target Build Platforms: Windows, Mac, IOS, Android.
Why choose DeepNavigation?
"Traditional navigation is static modeling of the world, whereas DeepNavigation enables agents to learn to move in dynamic environments"
Should I replace NavMesh with it?
If NavMesh performs your task well, you should continue to use it.
Dependencies:
DeepNavigation is developed based on TinyMind, so you first need to install it, the URL is:
https://assetstore.unity.com/packages/tools/ai-ml-integration/tinymind-deep-learning-framework-374274
The demos utilizes the following free resource packages, which are already included in the package, so you don't need to download them separately.
- Character Controller SUPER ( https://assetstore.unity.com/packages/tools/game-toolkits/character-controller-super-135316 )
- Drone Free Pack ( https://assetstore.unity.com/packages/tools/physics/free-pack-117641 )
- Green Dungeon ( https://assetstore.unity.com/packages/3d/environments/dungeons/green-dungeon-33935 )
- Autumn Level Set ( https://assetstore.unity.com/packages/3d/environments/fantasy/autumnal-nature-level-set-69314 )
In addition, GizmoExtensions is used in the code to draw debugging data. It is licensed under MIT. For details, please refer to Third-Party Notices.txt.
Compatibility
| Unity version | Built-in | URP | HDRP | Custom |
|---|---|---|---|---|
| 2022.3.62f2c1 | Compatible | Compatible | Compatible | Compatible |
As listed on the Asset Store.