Fig. 4
From: Bio-inspired neural networks with central pattern generators for learning multi-skill locomotion
(a) Target velocity and measured velocity of the robot in different simulated terrain conditions. (a1) Uneven terrain. (a2) Low friction. (a3) Actuation delay. (a4) External load. (a5) Slope. (b) Success rate of fall recovery maneuvers of the robot in different simulated terrain conditions. (b1) Uneven terrain. (b2) Slippery surface. (b3) Actuation delay.
