bobbygarzagames.com
+ PERANNIMAE +
Combat Systems, AI, Prototyping, Iteration
Unreal Engine 5 | C++ | Blueprint
Project Summary
Built in Unreal Engine 5, PERANNIMAE is an experimental action RPG centered around historically inspired melee combat, intelligent enemy AI, and player agency. Players wield a sword, mace, and magical lance, each serving a distinct role within a combat system that rewards positioning, timing, and tactical decision-making over traditional damage-focused gameplay.
As the project's Technical Designer, I have focused on designing and implementing gameplay systems, prototyping mechanics, developing enemy AI, and validating design decisions through iterative playtesting. PERANNIMAE has also served as the foundation for multiple graduate research studies, allowing me to explore player perception, AI behavior, and combat design through structured experimentation.
Refining the Combat Loop
What would it feel like to gamify medieval combat?
This question remains the driving factor of PERANNIMAE's combat system. Historically, heavily armored combatants rarely defeated one another through direct sword strikes alone. Instead, battles often revolved around creating opportunities; breaking an opponent's balance, forcing them to the ground, and exploiting the vulnerable gaps in their armor. Rather than treating this as historical inspiration alone, I wanted to explore whether these ideas could create a satisfying gameplay loop.
The result was a combat system built around openings and positioning rather than traditional dodging and parrying. Players carry three primary weapons, a sword, mace, and a magical lance, each encouraging different tactical decisions. Armored enemies are most effectively disrupted with blunt force before becoming vulnerable to precise, sharp attacks, while lighter enemies reward different direct approaches.
Iterating through Prototypes
The earliest combat prototype strictly enforced these weapon relationships. Knock down armored enemies with the mace, and finish them with the sword, and do the opposite for unarmored foes. Players quickly understood the intended rules, but playtesting revealed an important problem: although the pattern was simple to understand, it did not align with player expectations of reality. Restricting damage entirely to the "correct" weapon made combat feel unnecessarily rigid and less satisfying than intended.
Early Combat Prototype
Rather than abandoning the core concept, I iterated on the implementation. Over multiple prototype builds and playtesting sessions, I redesigned the combat loop so every weapon remained useful while preserving the importance of creating openings. Knockdowns became moments of increased opportunity instead of mandatory requirements, allowing players to experiment more freely while still rewarding historically inspired tactics.
Second Combat Prototype
To further lean into and reward player expectations of combat and physicality, I implemented a momentum-based knockdown mechanic that evaluates the player's and enemy movement speed do apply a physical knockdown state to the enemies themselves. The faster the player moves, the more force they have to knock down an enemy with their body! If an enemy is being knocked over, it can also knock over or stun nearby enemies that it collides with! This allowed the player to directly create emergent, physics-based opportunities to create openings beyond their weapon choices!
The result was a combat system that better balanced player expectations with the original design vision. By expanding player options without sacrificing the project's core philosophy, the combat became more fluid, readable, and satisfying while maintaining the tactical identity that inspired the project in the first place.
Graduate Study: Creating an Intelligent Boss Fight
As part of my graduate research at Texas A&M University, I explored this question by designing and prototyping a boss encounter for PERANNIMAE. Rather than focusing on increasing the boss's mechanical difficulty, I wanted to understand what behaviors actually caused players to perceive an opponent as intelligent.
Working alongside a committee of professors, I designed multiple iterations of the encounter and conducted structured playtests with players from a variety of gaming backgrounds. Through observation, surveys, and post-playtest interviews, I evaluated which behaviors most strongly influenced the perception of intelligence.
Boss Battle Study
One of the strongest findings was that players consistently perceived the boss as more intelligent when it acknowledged and responded to their actions. The boss recorded specific player behaviors, commented on them through contextual dialogue, and acted on observations. This combination of observation, communication, and reaction proved significantly more effective than simply increasing combat complexity.
These findings continue to shape PERANNIMAE's AI systems beyond the boss encounter. Rather than treating intelligence as increasingly complex decision-making, I approach AI as a conversation with the player from the game world, where enemies demonstrate awareness by observing, communicating, and responding to player behavior in meaningful ways.
Reactive Enemy Behavior
For PERANNIMAE, I wanted even the standard enemies to feel responsive to the player's actions to further add to the immersion of the game world. While developing the game's AI systems, I focused on creating behaviors that communicated awareness, intent, and reaction, allowing players to understand how enemies perceived and responded to them.
A major focus of this system was creating readable interactions between the player and AI. Instead of enemies instantly attacking when they detected the player, I wanted moments of tension where players could recognize that they had been noticed and make decisions based on the enemy's response. This included behaviors such as enemies becoming startled at close range, alerting nearby enemies, and communicating their intent through animations and audio cues.
Enemy AI Showcase
To support these behaviors, I implemented custom animation logic that dynamically blends animations depending on simultaneous actions the enemy chooses to enact, or is affected by. Conditional, colliding states would communicate with my animation system, allowing for smooth, communicative, and consistent animations to truly highly such dynamic AI.