Reinforcement learning-based optimization of locomotion controller using multiple coupled CPG oscillators for elongated undulating fin propulsion

This article proposes a locomotion controller inspired by black Knifefish for undulating elongated fin robot. The proposed controller is built by a modified CPG network using sixteen coupled Hopf oscillators with the feedback of the angle of each fin-ray. The convergence rate of the modified CPG network is optimized by a reinforcement learning algorithm. By employing the proposed controller, the undulating elongated fin robot can realize swimming pattern transformations naturally. Additionally, the proposed controller enables the configuration of the swimming pattern parameters known as the amplitude envelope, the oscillatory frequency to perform various swimming patterns. The implementation processing of the reinforcement learning-based optimization is discussed. The simulation and experimental results show the capability and effectiveness of the proposed controller through the performance of several swimming patterns in the varying oscillatory frequency and the amplitude envelope of each fin-ray.

Autonomous speed adaptation by a muscle-driven hind leg robot modeled on a cat without intervention from brain

This study aims to design a nervous system model to drive the realistic muscle-driven legs for the locomotion of a quadruped robot. We evaluate our proposed nervous system model with a hind leg simulated model and robot. We apply a two-level central pattern generator for each leg, which generates locomotion rhythms and reproduces cat-like leg trajectories by driving different sets of the muscles at any timing during one cycle of moving the leg. The central pattern generator receives sensory feedback from leg loading. A cat simulated model and a robot with two hind legs, each with three joints driven by six muscle models, are controlled by our nervous system model. Even though their hind legs are forced backward at a wide range of speeds, they can adapt to the speed variation by autonomously adjusting its stride and cyclic duration without changing any parameters or receiving any descending inputs. In addition to the autonomous speed adaptation, the cat hind leg robot switched from a trot-like gait to a gallop-like gait while speeding up. These features can be observed in existing animal locomotion tests. These results demonstrate that our nervous system is useful as a valid and practical legged locomotion controller.

Three-dimensional walking of a simulated muscle-driven quadruped robot with neuromorphic two-level central pattern generators

We aim to design a neuromorphic controller for the locomotion of a quadruped robot with muscle-driven leg mechanisms. To this end, we use a simulated cat model; each leg of the model is equipped with three joints driven by six muscle models incorporating two-joint muscles. For each leg, we use a two-level central pattern generator consisting of a rhythm generation part to produce basic rhythms and a pattern formation part to synergistically activate a different set of muscles in each of the four sequential phases (swing, touchdown, stance, and liftoff). Conventionally, it was difficult for a quadruped model with such realistic neural systems and muscle-driven leg mechanisms to walk even on flat terrain, but because of our improved neural and mechanical components, our quadruped model succeeds in reproducing motoneuron activations and leg trajectories similar to those in cats and achieves stable three-dimensional locomotion at a variety of speeds. Moreover, the quadruped is capable of walking upslope and over irregular terrains and adapting to perturbations, even without adjusting the parameters.