Ground substrates classification and adaptive walking through interaction dynamics for legged robots

Adaptive locomotion in different types of surfaces is of critical importance for legged robots.The knowledge of various ground substrates,especially some geological properties,plays an essential role in ensuring the legged robots'safety.In this paper,the interaction between the robots and the environments is investigated through interaction dynamics with the closed-loop system model,the compliant contact model,and the friction model,which unveil the influence of environment's geological characteristics for legged robots'locomotion.The proposed method to classify substrates is based on the interaction dynamics and the sensory-motor coordination.The foot contact forces,joint position errors,and joint motor currents,which reflect body dynamics,are measured as the sensing variables.We train and classify the features extracted from the raw data with a multilevel weighted k-Nearest Neighbor(kNN) algorithm.According to the interaction dynamics,the strategy of adaptive walking is developed by adjusting the touchdown angles and foot trajectories while lifting up and dropping down the foot.Experiments are conducted on five different substrates with quadruped robot FROG-I.The comparison with other classification methods and adaptive walking between different substrates demonstrate the effectiveness of our approach.

Multi-Layered CPG for Adaptive Walking of Quadruped Robots

This work concerns biped adaptive walking control on slope terrains with online trajectory generation. In terms of the lack of satis- factory performances of the traditional simplified single-layered Central Pattern Generator （CPG） model in engineering applications where robots face unknown environments and access feedback, this paper presents a Multi-Layered CPG （ML-CPG） model based on a half-center CPG model. The proposed ML-CPG model is used as the underlying low-level controller for a quadruped robot to generate adaptive walking patterns. Rhythm-generation and pattern formation interneurons are modeled to promptly generate motion rhythm and patterns for motion sequence control, while motoneurons are modeled to control the output strength of the joint in real time according to feedback. Referring to the motion control mechanisms of animals, a control structure is built for a quadruped robot. Multi-sensor models abstracted from the neural reflexes of animals are involved in all the layers of neurons through various feedback paths to achieve adaptability as well as the coordinated motion control of a robot＇s limbs. The simulation experiments verify the effectiveness of the pre- sented ML-CPG and multi-layered reflexes strategy.

Adaptive Walking Control of Biped Robots Using Online Trajectory Genera- tion Method Based on Neural Oscillators

This work concerns biped adaptive walking control on irregular terrains with online trajectory generation. A new trajectory generation method is proposed based on two neural networks. One oscillatory network is designed to generate foot trajectory, and another set of neural oscillators can generate the trajectory of Center of Mass （CoM） online. Using a motion engine, the characteristics of the workspace are mapped to the joint space. The entraining property of the neural oscillators is exploited for adaptive walking in the absence of a priori knowledge of walking conditions. Sensory feedback is applied to modify the gen- erated trajectories online to improve the walking quality. Furthermore, a staged evolutionary algorithm is developed to tune system parameters to improve walking performance. The developed control strategy is tested using a humanoid robot on ir- regular terrains. The experiments verify the success of the presented strategy. The biped robot can walk on irregular terrains with varying slopes, unknown bumps and stairs through autonomous adjustment of its walking patterns.