Managing AI Sensory Data with State Awareness

In this lesson, you will learn how to filter incoming sensory data based on your AI’s current state to ensure your character only reacts to stimuli that are relevant to their immediate goals.

The Problem of "Sensory Noise"

By default, the AIPerceptionComponent registers every stimulus it detects, whether your AI is currently idle, patrolling, or in the middle of a high-stakes combat sequence. If your AI "hears" footsteps while it’s busy executing a complex tactical maneuver, it might erroneously switch targets or interrupt its flow. To build truly intelligent agents, you must gate your perception processing against your current Blackboard state.

Implementing State-Aware Filtering

The most effective way to manage this is to use a custom Decorator or a validation check within your OnPerceptionUpdated logic. Before your AI updates its Blackboard keys with the new stimulus, it should verify if the current state allows for that sensory input.

For instance, if your AI is in a Stealth state, you might want to ignore "Sight" stimuli from distant enemies but prioritize "Hearing" stimuli.

void AMyAIController::OnTargetPerceptionUpdated(AActor* Actor, FAIStimulus Stimulus)
{
    // Retrieve current state from Blackboard
    FName CurrentState = Blackboard->GetValueAsName("AIState");

    // Logic Gate: Only process "Sight" if we are in "Alert" or "Combat" states
    if (Stimulus.Type == SightConfig && CurrentState == "Idle")
    {
        return; // Ignore sight while patrolling peacefully
    }

    // Otherwise, proceed to update Blackboard keys
    UpdateBlackboardForStimulus(Actor, Stimulus);
}

Why State Awareness Matters

If you don't filter your data, your AI becomes reactive rather than goal-oriented. By checking the AIState before committing a stimulus to the Blackboard, you prevent "sensory ping-pong"—where the AI constantly flips between two tasks because it can't decide which stimulus is more important. This approach forces the AI to be "deaf" or "blind" to certain events until it has transitioned into a state where it is logically prepared to handle that information.

Key Takeaway: Never pass raw sensory data directly to the Blackboard; always use your current state as a filter to ensure your AI acts with intentionality, not just reflex.