Efficient learning of robust quadruped bounding using pretrained neural networks

Bounding is one of the important gaits in quadrupedal locomotion for negotiating obstacles. The authors proposed an effective approach that can learn robust bounding more efficiently despite its large variation dynamic body movements. first pretrained neural network (NN) based on data from a robot operated by conventional model-based controllers, and then further optimised NN via deep reinforcement learning (DRL). In particular, designed reward function considering contact points phases to enforce gait symmetry periodicity, which improved performance. NN-based feedback controller was learned simulation directly deployed real quadruped Jueying Mini successfully. A variety environments are presented both indoors outdoors with authors’ approach. shows efficient computing good results over uneven terrain.