Skip to main content
An Introduction to Developmental Embodied Neurosimulation
Bradly Alicea, Daniela Cialfi, Rishabh Chakraborty, Stefan Dvoretskii, Sam Felder, Ziyi Gong, Akshara Gopi, Ankit Gupta, Avery Lim, and Jesse Parent
Presenting author:
Bradly Alicea
Developmental Embodied Neurosimulation is an approach that allows us to explore the intersection of the brain, embodied behavior, and the role of development in the emergence of brain and behavior. Taken collectively, this approach provides us with a unique window into Neuro-AI. Neurodevelopmentally-inspired agent-based approaches provide a means to both understand the development of small nervous systems in addition to improving the biological fidelity of artificial intelligence. Our approach adds a richness to simulated intelligence that is often a black box in our observations of organismal behavior.

Utilizing a developmental version of the Braitenberg Vehicle, we demonstrate the developmental process as being a combination of differentially timed internal processes and a process of emergent growth and form shaping the agent’s body. Each developmental Braitenberg Vehicle can possess a meta-brain, which is a layered set of neural representations mimicking the functional and structural lamination found in biological brains. Meta-brains control the acquisition and processing of information in a developmentally relevant manner. A laminar meta-brain features several distinct layers of representation, from a lower level shaped by direct perception that is representation-free, to higher layers more reliant on symbolic representation and its associated processing. The usefulness and power of this approach will be demonstrated by extending it to several naturalistic contexts.