This project contains a C# framework for action selection and behaviour arbitration. POSH#, or a sharper version of POSH, is designed for creating Behavior-based AI. Similar to the previous versions of POSH, POSH# is an extended form of a subsumption architecture layering modules of increasing complexity/capability on-top of each other, where higher levels are able to suppress lower levels. By doing so the complexity can be increased while remaining a stable system. The system is designed to separate logic and implementation of behaviour arbitration and utilize human creativity and design through the use human-readable and editable dynamic plans.
POSH# focuses on a lightweight and easy to integrate action selection system and is platform independent through the usage of mono.
Aim of the Project
The main aim of POSH# is make agent development as robust and intuitive as possible. By using the ABODEstar editor we allow for robust plan creation, even for people who rarely code. The design of a agent is crucial and should be as flexible as possible for advanced users. The one user group which is is however seldom taken into account are novice users or people not familiar with a specific language.
POSH is separated so that a designer can focus on creation behavioural plans and in parallel a programmer can implement and focus on the lower level elements, namely the actions and senses, usable in the designers plan.
To make it short POSH# is aimed at asynchronous, iterative development of artificial agents. The current version of POSH# can be integrated into Unity3D to allow the development of android, web or desktop applications.