Koi Pond Interactive Simulation

An interactive boids simulation built in Unreal Engine as a part of larger slot car diorama environment. Runs natively on a Raspberry Pi 5 at 60fps.

Progress Videos:

Koi Pond - Interactivity Test

The first iteration of the touch screen input

Koi Pond - Avoidance & Hashing

Avoidance function and the spatial hashing system

Koi Pond - First Touch Preview

Previewing the touch input pipeline

Koi Pond - Alignment Iteration

Adding Alignment for the fishes

Koi Pond - Scatter!

Scatter behavior based on touch input

Process:

Ideation:

The client proposed a unique brief: design a slot-car environment as a birthday-party attraction featuring interactive elements, dynamic lighting, 3D‑printed buildings, and playable hazards. Working within the art direction of “Fantasy Japan Through the Ages,” the lake biome was planned to occupy a substantial portion of the table. Early in concept development I recognized the risk that this expanse would feel empty and passive for guests not actively racing. To address that, I designed a dedicated interactive feature for the lake zone intended to engage spectators and encourage group participation.

Constraints:

Developed under rapidly changing production constraints, this project required fast iteration, prioritization, and selective polish. My focus was to preserve the core experience: a quiet interactive pond with responsive surface motion and atmospheric depth. While several features were scoped down to meet the timeline, the pond remained an engaging supporting element during its soft launch.

The program needed to be computationally light, only running on a Raspberry Pi 5, while also retaining immediate touch feedback. This problem was solved with heavy optimization in areas where computation could be spared, such as 3D spatial hashing. This required an in-depth knowledge of how data is stored in memory, which I utilized to create index search functions.

Running the simulation and streaming it to the Pi was considered, but ultimately scrapped due to the latency that arose.

Highlighted Features:

• High optimization to run on limited hardware

• Interactivity with feedback responses.

• Boids simulation

• Spatial Hashing system

• Custom built touchscreen hardware designed to fit to map to the environment

Future Improvements:

• Fish feeding system that attracts fish

• Ripple-environment interactivity

• Refine environment to more accurately represent a pond

• Add system to allow party guests to texture their own fish

• Add environmental hazards to fish, such as alligators (Client’s request, sounds fun)

• Integrating the Unreal-native system into the game, allowing for multiple touch inputs at once.