BraGenBrain

BraGenBrain (Braitenberg Vehicles Genetic Algorithms Simulation of Neural Development) is a simulation suite for examining neural development in embodied-cognition settings. It uses a genetic algorithm to evolve populations of Braitenberg-style agents across epochs, searching for neural configurations that produce adaptive sensorimotor behavior.

BraGenBrain originated as the software outcome of Stefan Dvoretskii’s Google Summer of Code 2019 work with the Orthogonal Research Lab under the parent project Modeling Neural Development with Braitenberg Vehicles.

Approach

The simulation runs epochs of many agents in parallel and applies a genetic algorithm to drive variation and selection on their neural controllers. This evolutionary loop lets the system search the space of developing nervous systems rather than hand-designing them, treating neural development as an optimization process embedded in an agent–environment interaction.

Implementation

• Distributed as an executable JAR (BraGenBrain-{VERSION}-XXX.jar).
• In-code documentation of core functions collected under doc/code.
• A Documentation.ipynb Jupyter notebook walks through major features and how to run simulations.
• Architecture Decision Records under doc/adr capture key technical choices made during development.

Resources

• Repository (Stefan/ subdirectory)
• INCF 2019 tutorial presentation by Stefan Dvoretskii — Stefan’s final GSoC project presentation and rehearsal for an INCF 2019 tutorial featuring BraGenBrain.
• Originating GSoC 2019 project archive