Enemy AI Prototype Study

UE5 Enemy AI System Prototype: Technical Breakdown

This project's primary objective was to conduct an in-depth study of complex, reactive enemy AI development within Unreal Engine 5 to enhance foundational gameplay programming skills. The result is a dynamic AI system featuring multiple enemy archetypes, layered perception, and sophisticated environmental awareness.

Core Decision Architecture (Behavior Tree & Blackboard)

The base AI utilizes a Simple State Machine managed through a Behavior Tree, centered around a Blackboard Component for data centralization. This setup enables the AI to transition smoothly between states (e.g., Idle, Patrolling, Combat, Healing). The core combat loop was engineered to be fluid:

• State Transitions: Reactive boolean variables on the Blackboard control high-level state entry and exit.

• Action Flow: The system executes sequential actions such as equipping a weapon, locking on via controller rotation, pathfinding toward the player, executing an attack, and returning to a wait state.

Advanced Movement & Sensor Integration

To support sophisticated enemy pathfinding, the system incorporates dedicated navigation tools:

• Patrol Systems: Utilized Spline Components to define fixed, predictable patrol routes, offering a robust method for managing deterministic AI movement in level design.

• AI Perception: Implemented the AI Perception Component to integrate sight, sound, and damage sensing. Detection triggers an immediate transition from the Patrolling state to the Combat state.

• Animation Synchronization: Resolved animation blending issues (e.g., character sliding while preparing to equip a weapon) using Event Dispatchers to ensure all movement actions are correctly synchronized with the completion of the required animations.

Environmental Query System (EQS) Implementation

The Environment Query System (EQS) was leveraged to enable proactive, data-driven decision-making, allowing the AI to dynamically react to the environment and player location:

• Tactical Positioning: Designed EQS queries to sample surrounding terrain and identify the optimal range for the current attack pattern (e.g., finding the best distance for a melee attack).

• Dynamic Cover Generation: Implemented a healing state where EQS is used to quickly and reliably find viable line-of-sight-blocking cover points on the map, allowing the enemy to temporarily retreat and heal before re-engaging.

Modular AI Design (C++ Inheritance)

To demonstrate scalability, the system was built using Object-Oriented Design (OOD) principles:

• Class Inheritance: A specialized Mage enemy was created as a Child Blueprint Class of the core AI, inheriting all base behaviors (patrol, perception, healing logic).

• Archetype Specialization: The Mage customized the Combat state with unique abilities, including a targeted teleport attack, and implemented an ability selection logic that dynamically chooses between basic energy blasts and a powerful barrage based on the current distance to the player.

This framework ultimately delivers a dynamic and reactive combat experience, providing a foundational and flexible architecture for developing diverse enemy behaviors in future projects.