Blood and Sand - AI Combat System Design

Combat System in Action

This demo showcases the designed enemy behaviors in action: the Berserker's relentless aggression, the Warrior's tactical defense, and the Cleric's supportive healing - all working together to create strategic, engaging combat encounters.

My Design Journey

Creating Blood and Sand's AI combat system was about solving a core design challenge: How do you make AI enemies that feel intelligent, distinct, and fun to fight? My approach focused on giving each enemy archetype a unique personality through their combat behaviors, creating a rock-paper-scissors dynamic that keeps players engaged and thinking tactically.

My Core Design Philosophy

Distinct Personalities

Each enemy needed to feel like a different opponent, not just a stat variation

Tactical Variety

Players should need different strategies for each enemy type

Group Dynamics

Enemies should complement each other when fighting together

Emotional Impact

Combat should feel visceral, strategic, and rewarding

The 3.5km combat arena where all AI behaviors come together

Multi-enemy Encounter

Tactical Positioning

Dynamic Combat Flow

Player Character

Designing the Enemy Archetypes

Each enemy type was designed to solve a specific gameplay problem and create unique tactical challenges

The Berserker

Relentless Aggressor

Design Challenge

Create constant pressure that forces players to stay mobile and think quickly under stress.

My Solution

No Shield: Pure offense - vulnerable but dangerous
Constant Aggression: Never stops attacking once engaged
High Damage Output: Punishes players who get caught
Predictable Pattern: Readable but relentless

"The Berserker teaches players to maintain distance and timing"

The Warrior

Tactical Defender

Design Challenge

Create a methodical opponent that rewards patient, strategic play over button mashing.

My Solution

Guard Stance: Blocks incoming attacks strategically
Random Attack Windows: Unpredictable openings to exploit
Counter-Attack Focus: Punishes reckless aggression
Defensive Positioning: Controls space effectively

"The Warrior forces players to learn patience and timing"

The Cleric

Support Healer

Design Challenge

Add a priority target that changes the entire flow of multi-enemy encounters.

My Solution

Evasive Behavior: Runs from direct confrontation
Smart Healing: Only heals damaged allies (<100% HP)
Positioning AI: Stays behind other enemies for protection
Threat Assessment: Prioritizes healing the most damaged ally

"The Cleric creates tactical depth in group encounters"

Technical Implementation Deep Dive

Testing Behavior Tree Logic

My approach to behavior testing focused on validating that each behavior tree executes correctly, produces consistent results, and handles edge cases appropriately.

Key Testing Methodologies:

State Transition Testing: Verified all transitions between idle, combat, and ability states function correctly
Priority Validation: Confirmed higher-priority behaviors correctly interrupt lower-priority ones (Cleric healing vs. attacking)
Contextual Testing: Validated AI responses to player position, ally health, and environmental changes
Edge Case Identification: Discovered scenarios where behavior trees might deadlock or produce unintended results

Berserker Behavior Testing

Test Case: Can See Player? → Chase → Attack → Repeat

Validated that simple aggression loop executes consistently without breaks or unexpected pauses across all combat scenarios.

Master Behavior Tree

Combat Sub-tree

Movement Sub-tree

System Validation & Integration Testing

Testing a multi-component AI system required validating that all components work together correctly and communicate as intended.

My Validation Approach:

Component Integration Testing: Validated perception, decision-making, and execution components work together
Communication Validation: Tested enemy-to-enemy communication for threat sharing and coordination
Modular Testing: Verified new abilities and attacks integrate without breaking existing systems
Configuration Testing: Validated parameter changes produce expected behavioral modifications

Tested AI System Components

AI Controller (Integration Tested)
├── Perception Component ✓
├── Behavior Tree Component ✓
├── Combat Component ✓
└── Movement Component ✓

Group Coordination Testing

Validated Cleric communication with allies for healing prioritization and confirmed Warriors correctly call for backup under appropriate conditions.

Advanced Behavior Tree

Decision Making Sub-tree

Performance Benchmarking & Stress Testing

Performance testing validated that the system maintains 60+ FPS even under stress conditions with multiple concurrent AI agents and complex calculations.

Performance Test Cases:

LOD System Validation: Verified distant enemies correctly use simplified behaviors without visual glitches
Frame Rate Testing: Benchmarked CPU load distribution across frames with staggered sight checks
Update Rate Verification: Confirmed non-combat AI reduces update frequency without noticeable behavior changes
Memory Testing: Validated object pooling prevents memory leaks during extended play sessions

60+ FPS Verified

20+ Concurrent AI Tested

Gameplay Feel Validation

Tested AI timing, attack telegraphs, and visual feedback to confirm combat feels responsive and fair. Validated that players can distinguish between earned victories and unfair defeats.

Combat Flow & Feel

Before and after polish passes - notice the improved visual clarity and spacing

Validating Combat Feel & User Experience

Beyond technical validation, testing needed to confirm combat creates the intended player experience. User experience testing validated that emotional and gameplay goals were met.

Validated: Visceral Impact

Tested hit feedback timing and visual/audio cues to confirm every attack feels weighty and consequential. Validated player satisfaction through repeated playtesting.

Validated: Strategic Depth

Confirmed through testing that players discover optimal strategies organically. Verified the "aha!" moment when targeting priority (Cleric first) becomes clear.

Validated: Tension Escalation

Tested encounter progression from single enemies to coordinated groups. Confirmed difficulty curve creates mounting tension without frustration.

Validated: Earned Victory

Verified AI behaviors create fair challenge. Testing confirmed players attribute victories to skill and defeats to strategic mistakes, not unfair mechanics.

"The moment when everything clicks"

Player facing multiple enemy types, using terrain and tactics to overcome the challenge

Key Testing Insights

01

Simple Behaviors Require Thorough Testing

The Berserker's straightforward aggression still needed extensive testing. Simple doesn't mean edge cases don't exist—validation confirmed consistency across all scenarios.

02

Test For Player Experience, Not Just Functionality

Validating AI timing against human reaction patterns was critical. Testing revealed that frame-perfect AI creates frustration—balance testing ensured fairness.

03

Group Testing Reveals Hidden Issues

Individual behaviors that passed solo testing showed edge cases in multi-enemy scenarios. Integration testing was essential to catch group coordination bugs.

04

Performance Testing Is Continuous

Performance benchmarking couldn't be deferred—it was tested throughout. Stress testing with 20+ agents validated optimization maintained smooth gameplay.

Technology & Tools

Core Development

Unreal Engine 5

C++

Blueprint Visual Scripting

AI Systems

Behavior Trees

AI Perception

Navigation Mesh

State Machines

Development Pipeline

Perforce Version Control

Custom Profiling Tools

Debug Visualization

Want to Discuss AI Combat Design?

I'd love to talk about the specific design challenges, architectural decisions, and lessons learned from building this AI combat system. Every project teaches something new about creating engaging, intelligent opponents.

Let's Talk AI Design View More Projects Download Resume