← Back

Visual evaluation

Extracted gait cycle

Displays a single locomotor cycle extracted automatically from the continuous trajectory, based on local minima in disk speed. This isolates one full oscillatory period for clearer visualization of inter-arm phase relationships.

Locomotion speed over time

Plots the disk’s translational speed throughout the evaluation episode.

Disk speed over time plot

Arm values

Direction alignment quantifies how well each arm aligns with the global direction of motion during a single gait cycle (1 signifying perfect alignment and −1 pointing in the opposite direction). Propulsive contribution measures the propulsive ground-contact force each arm generates in the movement direction during a single gait cycle, expressed relative to the most propulsive arm.

Arm value table

Arm midpoint angles over time

Shows the time evolution of each arm's midpoint angle, defined as the deviation of the arm from its straight, radially extended configuration.

Arm midpoint angles over time

Arm propulsive impulse over time

Depicts instantaneous propulsive impulses over time, quantifying ground reaction forces of each arm projected onto the movement direction.

Arm propulsive impulse over time

Arm propulsive impulse synchronisation matrix

Quantifies how synchronously pairs of arms generate propulsion. Each entry reflects the normalized correlation of propulsive impulse profiles between two arms (values near 1 indicate strong co-activation, 0 indicates independence).

Arm propulsive impulse synchronisation matrix

2D-embedding of ganglion states during simulation

Shows a dimensionality-reduced projection (UMATO) of ganglion states during the actual simulated behavior. Clustering and trajectories in this space reflect the underlying recurrent neural dynamics driving coordination.

UMATO embedding during simulation

2D-embedding of ganglion states during constant input

Displays a dimensionality-reduced projection (UMATO) of ganglion-state trajectories when sensory input is held constant, isolating the network’s intrinsic oscillatory dynamics.

UMATO embedding under constant input