Capturability-based Fuzzy Footstep Planner for a Biped Robot with Centroidal Compliance

Compliance motion and footstep adjustment are active balance control strategies from learning human subconscious behaviors.The force estimation without direct end-actuator force measurement and the optimal footsteps based on complex analytical calculation are still challenging tasks for elementary and kid-size position-controlled robots.In this paper,an online compliant controller with Gravity Projection Observer(GPO),which can express the external force condition of perturbations by the estimated Projection of Gravity(PoG)with estimation covariance,is proposed for the realization of disturbance absorption,with which the robustness of the humanoid contact with environments can be maintained.The fuzzy footstep planner based on capturability analysis is proposed,and the Model Predictive Control(MPC)is applied to generate the desired steps.The fuzzification rules are well-designed and give the corresponding control output responding to complex and changeable external disturbances.To validate the presented methods,a series of experiments on a real humanoid robot are conducted.The results verify the effectiveness of the proposed balance control framework.

移动机器人导航技术综述

作为移动机器人的核心技术之一,导航技术涉及感知、决策、控制等多层架构,包含建图、定位、规划等主要模块,兼容经典控制与机器学习等多样算法,其相关研究涉及面广,时间跨度大,联系较为不明显.为梳理不同工作间关联,帮助快速了解移动机器人导航技术发展与挑战,本文从“导航”的词语释义与数学描述出发,以“导”与“航”两个主要目标为线索贯穿全文.通过分析不同环境下的导航应用局限,探究环境的空间复杂性与时间不确定性对导航技术产生的影响,阐明复杂场景应用下的研究现状与关键挑战.全文借助对近几十年来相关文献的归类梳理,寻找不同研究间的相关联系、研究初衷以及可能方案,拟在帮助读者加深对导航技术理解的同时,为后续研究提供思路参考.

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.

Adaptive walking control of quadruped robots based on central pattern generator (CPG) and reflex

This paper presents a central pattern generator （CPG） and vestibular reflex combined control strategy for a quadruped robot. An oscillator network and a knee-to-hip mapping function are presented to realize the rhythmic motion for the quadruped robot. A two-phase parameter tuning method is designed to adjust the parameters of oscillator network. First, based on the numerical simulation, the influences of the parameters on the output signals are analyzed, then the genetic algorithm （GA） is used to evolve the phase relationships of the oscillators to realize the basic animal-like walking pattern. Moreover, the animal＇s vestibular reflex mechanism is mimicked to realize the adaptive walking of the quadruped robot on a slope terrain. Coupled with the sensory feedback information, the robot can walk up and down the slope smoothly. The presented bio-inspired control method is validated through simulations and experiments with AIBO. Under the control of the presented CPG and vestibular reflex combined control method, AIBO can cope with slipping, falling down and walk on a slope successfully, which demonstrates the effectiveness of the proposed walking control method.

基于弹性分析与函数回归模型的示教轨迹学习与泛化方法

示教学习是解决仿人机器人动态轨迹规划问题的高效方法.示教轨迹作为一种函数型数据,天然地具有耦合的时空特征,需要在函数型数据分析的理论框架下进行研究.传统的示教学习成果以“预处理”的形式解决上述耦合问题,不能充分利用示教轨迹的函数型数据本质.为了充分利用这一本质,本文采用弹性分析方法,将示教轨迹视作函数型数据,解决它们的时–空特征解耦问题.基于解耦后的示教轨迹,本文建立函数型回归模型并将其应用于动态轨迹的生成,实现示教轨迹在不同任务参数下的泛化.在实验中,JACO-2机械臂学习形态各异的技能并在不同的任务参数下完成任务,验证了提出方法的有效性.本文在统一的数学框架下,基于函数型数据分析的数学方法,解决一类仿人机器人的示教轨迹学习与泛化问题,具有较好的理论统一性与可实践性.

Active Balance Control of Humanoid Locomotion Based on Foot Position Compensation

A foot positioning compensator is developed in this paper for a full-body humanoid to retrieve its balance during continuous walking.An online Foot Position Compensator(FPC)is designed to improve the robustness of biped walking,which can modify predefined step position and step duration online with sensory feedback.Foot placement parameters are learned by the FPC based on the Policy Gradient Reinforcement Learning(PGRL)method.Moreover,the FPC assists the humanoid robot in rejecting external disturbances and recovering the walking position by re-planning the trajectories of walking pattern and the Center of Mass(CoM).An upper body pose control strategy is also presented to further enhance the performance of humanoid robots to overcome strong external disturbances.The advantages of this proposed method are that it neither requires prior information about the walking terrain conditions,nor relies on range sensor information for surface topology measurement.The effectiveness of the proposed method is verified via Webots simulation and real experiments on a full-body humanoid NAO robot.