Fig. 7
From: Bio-inspired neural networks with central pattern generators for learning multi-skill locomotion
The robot is commanded to follow various velocity commands in real-world grasslands. (a) Snapshots of the robot performing fall recovery, lateral locomotion, steering, and forward locomotion in real world. (b) Measured velocity (\({v}_{x}\), \({v}_{y}\), \({\omega }_{yaw}\)) and target velocity (\(\hat{v}_{x}\), \(\hat{v}_{y}\), \(\hat{\omega }_{yaw}\)). Note that the measured velocity is estimated via Kalman filter. (c) Roll and pitch angles. (d) CPG oscillation offset \(\chi\) of front left leg joints. (e) CPG output signal \(\psi\) of front left leg joints. (f) Target joint angles of front left leg joints. (see Supplementary Video S2 for the corresponding video. Video URL: https://youtu.be/TCY2-pvkbjg).
