A simulated body approach to understanding motor control in a highly deformable biological body

Embodiment is vital; it impacts the central nervous system (CNS) and neural computation in animals, humans, and robots. The body's physicality, continuity, internal states, and homeostasis shape behaviors and neural networks. For accurate motor control, the neural output must consider the physical and muscular structure of the body. Deformable bodies, like certain animals and soft robots, have unique control challenges due to their vast movement possibilities. However, creatures like larvae and octopi skillfully manage this using local and global computation, suggesting an efficient neural representation.

Using the Drosophila larva as a model system we propose a simulated body approach to decipher the link between body, motor control and CNS. We introduce a numerical simulation framework, coded in taichy lang, modeling the larva body including its muscle structure extracted from a CT-scan. By combining material point method (MPM), muscle dynamics approximated by Hill model, muscle geometry as 2D ribbons wrapped around the inner body, a simplified inner body dynamics as an elastic rod and local friction as a mixed of coulomb and viscous friction, we demonstrate our ability to model larva dynamics. Finally, we address motion planning using a bayesian program synthesis approach.