The Ground

Neural Simulation Playground (NSP 3D) is an interactive simulation playground where users can explore 3D virtual environments through spiking neural circuits based on the Drosophila connectome. Users can experience the entire process of visual input from a virtual camera being processed through 1,854 neurons and translated into real-time flight behavior. The playground features LIF and ALIF neuron models and supports comparative experiments across 10 different maze environments under varying conditions. Meta Muse Spark was used throughout development for code generation, debugging, and experiment analysis. During development, several flight-control challenges emerged, including camera-mode-dependent obstacle avoidance bias, upward flight drift, and neural firing saturation. These issues were investigated and addressed through condition-based batch experiments using Multi-Lab, an Ax optimization loop, and guidance from Meta. The project aims to make neural simulation accessible to everyone, allowing users to experiment firsthand and explore the connection between perception and behavior.