Mapping the continuum of gait strategies and limb-coupling dynamics in mouse and fly locomotion

Animals move with remarkable agility and robustness, yet how they coordinate their limbs to produce stable, adaptable gaits remains incompletely understood. Research has shown that quadrupeds and hexapods employ compatively distinct strategies; however, prior work has relied on coarse metrics (e.g., footfall times, stance duration) lacking sub-cycle definition. The adoption of ML paradigms has enabled high-throughput treatment and pattern abstraction for analysis at unprecedented detail, which we specifically leverage in three ways: SLEAP for high-resolution trajectory tracking of mouse and fly body parts in over 10,000 spontaneous locomotion bouts; a periodic autoencoder for continuous phase extraction; and sparse Bayesian learning for neuromuscular coupling inference. From high-resolution coordinate and phase data, we map the full domain of exhibited gaits, identifying instantaneous gaits and transitions. We model limb dynamics across locomotion manifolds using a Kuramoto system of phase-coupled oscillators, estimating control parameters such as coupling strength and phase offsets from kinematics. We show that observed controller symmetries align with patterns established directly from known neural measurements, and coupling parameters vary with gait and speed. We compare locomotor strategies between the mouse and fly and discuss their implications for the development and conservation of kinematic and control paradigms.