Figure 3. Dynamically stable bipod gaits are optimal for fast ground locomotion in the absence of leg adhesion.

Gaits were optimized for forward velocity while climbing upward (left), downward (centre left) or sideways (centre) on a vertical surface using leg adhesion, walking on the ground with leg adhesion (centre right) or walking on the ground without leg adhesion (right). (a) The average number of legs in stance phase over five walking cycles. A dashed black line indicates three legs in stance phase as expected for the classic tripod gait. N=15 for each condition. (b) The percentage of time that the model's COM lies within a polygon of support delineated by each leg in stance phase when the gait is tested during ground walking. The level of adhesion is kept as during optimization. Optimized gaits are colour coded by class. Data points are randomly scattered along the x-axis for clarity. N=15 for each condition, except for vertical downward (N=11) and vertical sideways (N=12) locomotion. In these conditions several gaits were unable to support ground walking.