Runtime Execution

Reasonable Planning AI has two modes of Runtime execution. These modes are RPAI Native and Behavior Trees. Each of these modes may be present within the same project. The decisions of which system to choose should be determined by the AI designers of the project. Each mode is independent of the other. An agent cannot use both systems at the same time without effort invested in supporting two UBrainComponent components on a single APawn.

RPAI Native

This is the default reference mode of runtime execution within RPAI. It is completely abstracted from the AI Module while also offering full integration via AI Tasks. It can be added to any project at any time. All one does is add an RPAI Brain Component to an AIController or APawn.

RPAI Brain Component

The RPAI native method of execution is accomplished via the URpaiBrainComponent component - hereby known as the RPAI Brain. The RPAI Brain is a child of the Unreal Engine UBrainComponent class. It hooks into the existing UE AI module and game framework via the UBrainComponent::StartLogic(...), UBrainComponent::StopLogic(...), and UBrainComponent::PauseLogic(...) virtual functions. The RPAI Brain is also compatible with the AI Task subsystem built into Unreal Engine. This enables a flexible point of integration with other AI systems such as Smart Objects and Environment Query System (EQS).

---
title: RPAI Brain State Machine
---
flowchart TD
    Spawned(Pawn Spawned) --> BeginPlay
    GameStart(Game Begins) --> BeginPlay
    PauseLogic[[Pause Logic]] --> Paused[/Paused is now True/]
    UnPauseLogic[[UnPause Logic]] --> UnPaused[/Paused is now False/]
    UnPaused --> Tick
    BeginPlay[[Begin Play]] --> StartLogic
    StartLogic[[Start Logic]] --> Tick
    Paused --> Tick
    Tick[[Tick Component]] --> IsPaused
    Skip --> Tick
    IsPaused{Paused?} -- Yes --> Skip(Skip One Tick)
    IsPaused -- No --> ReasonGoal
    ReasonGoal[[Reason Next Goal]] --> HasNextGoal
    HasNextGoal{New Goal?} -- Yes --> PlanGoalActions
    HasNextGoal -- No --> Skip
    PlanGoalActions[[Plan Action Set for Goal]] --> HasPlan
    HasPlan{Any Actions?} -- Yes --> ExecuteNextAction
    HasPlan -- No --> Skip
    ExecuteAction --> Cancel(On Cancel)
    ExecuteAction --> Complete(On Complete)
    ExecuteAction(Tick Current Action) -- Not Paused --> ExecuteAction
    Cancel --> CancelRemaining{Cancel All?}
    CancelRemaining -- No --> ExecuteNextAction{Pop Action?}
    CancelRemaining -- Yes --> ReasonGoal
    Complete --> ExecuteNextAction
    ExecuteNextAction -- No --> Tick
    ExecuteNextAction -- Yes --> ExecuteAction
    EndPlay[[End Play]] --> StopLogic[[Stop Logic]]
    StopLogic --> Paused

Loading

Behavior Tree

Aside from RPAI Native execution, RPAI supports execution within a Behavior Tree. This is accomplished through two separate tasks defined in the plugin and an adapter component assigned to an AIController.

Behavior Tree Tasks

RPAI is also adapted to work with your existing Behavior trees. This is accomplished through inline definitions of a Reasoner, Planner, Goals, and Actions in two separate task nodes. The two nodes are pictured below:

TODO: Add Pictures

The two nodes are Determine Next Goal and Execute Planned Goal.

Determine Next Goal

This node accepts for configuration a Reasoner and a set of Goals. All are inline defined in the details panel. This node will return success if it outputs a Goal to the given blackboard key. Otherwise it will return failure and your Behavior Tree will move onto the next (right) node.

Execute Planned Goal

This node accepts for configuration a Planner and a set of Actions. All are inline defined in the details panel. This node will return success if it completes all planned actions. If it is unable to produce a plan or an action is cancelled that ends all other actions it will move onto the next (right) node.

State Blackboard Component

For a successful execution of the two tasks outlined above, an AIController must have a URpaiStateBlackboardComponent assigned. This component is used to configure which values to copy from the Behavior Tree Blackboard into the RPAI state object used for reasoning and planning. It supports copying all base types that Blackboard supports. It is therefore 100% compatible with your existing blackboard configurations.

Conclusion

Now that you are aware of the two modes of execution - RPAI Native and Behavior Trees - learn more about a third component of RPAI. This third component is the data driven framework for building RPAI Behaviors called Composer