基于自学习中枢模式发生器的仿人机器人适应性行走控制

为了克服传统中枢模式发生器(Central pattern generator, CPG)关节空间控制方法的复杂性和局限性, 本文基于自学习中枢模式发生器模型, 提出了一套在线调制和融合多传感器信息的仿人机器人环境自适应行走控制方法.算法难点在于如何在机器人的工作空间将自学习CPG用于工作空间轨迹生成, 并使CPG参数直接和步态模式相关联.本文提出了利用自学习CPG来学习和实时生成机器人质心轨迹和脚掌轨迹的方法, 在线调节机器人步长、抬腿高度和步行速度等关键参数.参考生物反射行为, 利用传感反馈信息激发CPG以产生具有环境适应性的工作空间轨迹, 提升行走质量. 控制系统的参数通过优化算法来进一步改善行走性能.相比于传统的CPG关节空间法, 本文所采用的自学习CPG工作空间法不仅极大简化了CPG网络结构而且提高了仿人机器人行走的适应性.最后, 通过仿人机器人坡面适应性行走的仿真和实验, 验证了所提出控制策略的可行性和有效性.     

A Fast Dynamically Equilibrated Walking Trajectory Generation Method of Humanoid Robot

This paper describes a fast dynamically equilibrated trajectory generation method for a humanoid robot. From a given input motion and the desired ZMP trajectory, the algorithm generates a dynamically equilibrated trajectory using the relationship between the robot's center of gravity and the ZMP. Three key issues are denoted: 1) an enhanced ZMP constraint which enables the calculation of robot stability even if several limbs are contacting the environment, 2) a simplified robot model is introduced that represents the relationship between its center of gravity and ZMP, 3) a convergence method is adopted to eliminate approximation errors arising from the simplified model. Combining these three key issues together with online ZMP compensation method, humanoid robot H5 have succeeded to walk, step down and so on. Experimental results using humanoid robot H5 are described.     

基于神经网络的仿人跑步机器人步态规划

为了快速生成仿人机器人跑步运动轨迹,研究了一种用于仿人机器人跑步步态生成的步态规划器。采用三维弹簧倒立摆模型描述跑步过程中仿人机器人质心运动规律,奔跑时机器人质心轨迹及落脚点位置可以由四个步态参数来确定,从而将步态规划问题转化成步态参数优化问题,求解了500余种不同运动状态下的步态参数。建立了基于三层BP神经网络的步态规划器,将优化结果作为训练样本训练神经网络。用上述规划器实现了仿人机器人跑步步态规划并对规划结果进行了仿真验证。研究结果表明,基于BP神经网络的步态规划器可以实现步态参数的快速计算,生成的跑步步态逼真;提出的跑步运动步态规划方法可行,为仿人机器人实时轨迹生成提供了一种解决方法。     

复杂环境下仿人机器人有效支撑区域的感知

当主流的仿人机器人都采ZMP(zero moment point)理论作为稳定行走的判据.实时ZMP点落在支撑足与地面接触形成的多边形支撑区域内是仿人机器人实现稳定步行的必要条件.因此实现仿人机器人在复杂现实环境中稳定行走,必须要求机器人足部感知系统提供足够丰富的地面环境信息,从而可以准确获取支撑区域的形状以实现基于实时ZMP点的稳定控制.文中将柔性阵列力传感器应用于仿人机器人足部感知系统,提出了获取仿人机器人支撑区域形状的方法,而且通过实验验证了其可行性.     

Online Balance Controllers for a Hopping and Running Humanoid Robot

This paper describes online balance controllers for running in a humanoid robot and verifies the validity of the proposed controllers via experiments. To realize running in the humanoid robot, the overall control structure is composed of an offline controller and an online controller. The main purpose of the online controller is to maintain dynamic stability while the humanoid robot hops or runs. The online controller is composed of the posture balance control in the sagittal plane, the transient balance control in the frontal plane and the swing ankle pitch compensator in the sagittal plane. The posture balance controller makes the robot maintain balance using an inertial measurement unit sensor in the sagittal plane. The transient balance controller makes the robot keep its balance in the frontal plane using gyros attached to each upper leg. The swing ankle pitch compensator prevents the swing foot from hitting the ground at unexpected times while the robot runs forward. HUBO2 was used for the running experiment. It was designed for the running experiment, and is lighter and more powerful than the previous walking robot platform, HUBO. With the proposed controllers, HUBO2 ran forward stably at a maximum speed of 3.24 km/h and this result verified the effectiveness of the proposed algorithm. In addition, in order to show the contribution of the stability, the running performance according to the existence of each controller was described by experiment.     

Humanoid odometric localization integrating kinematic,inertial and visual information

We present a method for odometric localization of humanoid robots using standard sensing equipment, i.e., a monocular camera, an inertial measurement unit (IMU), joint encoders and foot pressure sensors. Data from all these sources are integrated using the prediction-correction paradigm of the Extended Kalman Filter. Position and orientation of the torso, defined as the representative body of the robot, are predicted through kinematic computations based on joint encoder readings; an asynchronous mechanism triggered by the pressure sensors is used to update the placement of the support foot. The correction step of the filter uses as measurements the torso orientation, provided by the IMU, and the head pose, reconstructed by a VSLAM algorithm. The proposed method is validated on the humanoid NAO through two sets of experiments: open-loop motions aimed at assessing the accuracy of localization with respect to a ground truth, and closed-loop motions where the humanoid pose estimates are used in real-time as feedback signals for trajectory control.     

Real-time continuous ZMP pattern generation of a humanoid robot using an analytic method based on capture point

This paper introduces an analytic method to generate a continuous ZMP pattern based on a capture point (CP). When a target CP is decided in real-time, the pattern generator makes the CP, ZMP, which it is within convex hull of the supporting feet area, and CoM patterns without discontinuity by closed-form solutions for a single step. Therefore, the proposed pattern generation method does not need a ZMP pattern modification, numerical iterations, and future ZMPs. The method is employed to treat applications such as step length change while walking and push recovery during walking in place. Furthermore, since compliant characteristics such as body oscillation appear in the humanoid robot, we introduce a system model, a double inverted pendulum model with flexible joints for the model-based control. Finally, the real-time walking pattern generation method and the walking control scheme are verified by experiments with the humanoid robot HUBO2.     

Momentum control with hierarchical inverse dynamics on a torque-controlled humanoid

Hierarchical inverse dynamics based on cascades of quadratic programs have been proposed for the control of legged robots. They have important benefits but to the best of our knowledge have never been implemented on a torque controlled humanoid where model inaccuracies, sensor noise and real-time computation requirements can be problematic. Using a reformulation of existing algorithms, we propose a simplification of the problem that allows to achieve real-time control. Momentum-based control is integrated in the task hierarchy and a LQR design approach is used to compute the desired associated closed-loop behavior and improve performance. Extensive experiments on various balancing and tracking tasks show very robust performance in the face of unknown disturbances, even when the humanoid is standing on one foot. Our results demonstrate that hierarchical inverse dynamics together with momentum control can be efficiently used for feedback control under real robot conditions.     

Adaptive hierarchical positioning error compensation for long-term service of industrial robots based on incremental learning with fixed-length memory window and incremental model reconstruction

Industrial robots have been extensively used in industry, however, geometric errors mainly caused by connecting rod parameter error and non-geometric errors caused by deflection and friction, etc., limit its application in high-accuracy machining. Aiming at addressing these two types of errors, parametric methods for error compensation based on the kinematic model and non-parametric methods of directly establishing the mapping relationship between the actual and target poses of the robot end-effector are investigated and proposed. Currently both types of methods are mainly offline and will be no longer applicable when the pose of the end-effector in the workspace changes dramatically or the working performance of the robot degrades. Thus, to compensate the positioning error of an industrial robot during long-term operation, this research proposes an adaptive hierarchical compensation method based on fixed-length memory window incremental learning and incremental model reconstruction. Firstly, the correlation between positioning errors and robot poses is studied, a calibration sample library is created, and thus the actively evaluating mechanism of the pose mapping model is established to overcome the problem of the robot’ workspace having a differential distribution of error levels. Then, an incremental learning algorithm with fixed-length memory window and an incremental model reconstruction algorithm are designed to optimize the pose mapping model in terms of its parameters and architecture and overcome the problem that the performance degradation of the robot exacerbates the positioning error and affects the applicability of the pose mapping model, ensuring that the pose mapping model runs stably above the target accuracy level. Finally, the proposed method is applied to the long-term compensation case of a Stäubli industrial robot and a UR robot, and compared to state-of-art methods. Verification results show the proposed method reduces the position error of the Stäubli robot from 0.85mm to 0.13mm and orientation error from 0.68° to 0.07°, as well as reduces the position error of the UR robot from 2.11mm to 0.17mm, demonstrating that the proposed method works in real world scenarios and outperforms similar methods.     

基于最优控制的仿人机器人行走振动抑制

针对仿人机器人行走过程中由腿部非刚性特性引起的振动,提出了一种基于最优控制的行走振动抑制方法．首先对振动进行建模,并将这一模型加入到原有机器人动力学模型中去．然后基于拓展的动力学模型,使用预观控制方法求取一条符合质心加速度约束的控制轨迹,作为求解最优控制问题的初始解．进而由此初始解出发,迭代求解此带约束的最优控制问题,利用得到的最优控制轨迹即能以前馈方式抑制行走过程中的振动．最后,使用预观控制方法和本文方法在仿人机器人“空”上进行行走对比实验．实验结果表明提出的方法显著减小了机器人行走过程中零力矩点（ZMP）的振荡和躯干的晃动．该方法对行走振动抑制的有效性得到了验证．     

Position-based impedance control of a biped humanoid robot

This paper describes position-based impedance control for biped humanoid robot locomotion. The impedance parameters of the biped leg are adjusted in real-time according to the gait phase. In order to reduce the impact/contact forces generated between the contacting foot and the ground, the damping coefficient of the impedance of the landing foot is increased largely during the first half double support phase. In the last half double support phase, the walking pattern of the leg changed by the impedance control is returned to the desired walking pattern by using a polynomial. Also, the large stiffness of the landing leg is given to increase the momentum reduced by the viscosity of the landing leg in the first half single support phase. For the stability of the biped humanoid robot, a balance control that compensates for moments generated by the biped locomotion is employed during a whole walking cycle. For the confirmation of the impedance and balance control, we have developed a life-sized humanoid robot, WABIAN-RIII, which has 43 mechanical d.o.f. Through dynamic walking experiments, the validity of the proposed controls is verified.     

Planning and executing navigation among movable obstacles

This paper explores autonomous locomotion, reaching, grasping and manipulation for the domain of navigation among movable obstacles (NAMO). The robot perceives and constructs a model of an environment filled with various fixed and movable obstacles, and automatically plans a navigation strategy to reach a desired goal location. The planned strategy consists of a sequence of walking and compliant manipulation operations. It is executed by the robot with online feedback. We give an overview of our NAMO system, as well as provide details of the autonomous planning, online grasping and compliant hand positioning during dynamically stable walking. Finally, we present results of a successful implementation running on the humanoid robot HRP-2.     

ADXL203型双轴加速计在机器人足部感知系统中的应用

仿人机器人要实现在复杂环境下稳定行走,仅仅依靠地面反力信息远不能满足应用要求,此时足部的倾角信息显得更为重要。脚面倾角可以反映地面倾斜状态,是仿人机器人稳定控制的一个重要依据。利用ADXL203双轴加速度传感器与DSP(TMS3202811)实现对倾角信息的实时高速采集与处理,并通过实验证明了倾角传感器在机器人足部感知系统中是可行的。     

Realization of Dynamic Stair Climbing for Biped Humanoid Robot Using Force/Torque Sensors

This paper proposes a control algorithm for the dynamic stair climbing of a human-sized biped humanoid robot. Dynamic stair climbing can cause more instability than dynamic biped walking on the ground because stair climbing requires an additional vertical motions and a large step length. We assume that stair configuration is already known and only use force/torque sensors at ankle joint to achieve a control algorithm for a stable dynamic stair climbing. We describe a stair climbing pattern generation and stair climbing stages, and then propose a real-time balance control algorithm which is composed of several online controllers. Each online controller is addressed in detail with experimental results. Finally, the effectiveness and performance of the proposed control algorithm are verified through a dynamic stair climbing experiment of KHR-2.     

A walking pattern generation method of humanoid robot MAHRU-R

This paper proposes an omni-directional walking pattern generation method for a humanoid robot MAHRU-R. To walk stably without falling down, a humanoid robot needs the walking pattern. Our previous walking pattern method generated the walking pattern with linear polynomials of the zero moment point (ZMP). It implemented the simple walking like forward/backward walking, side step walking and turning. However, this method was not sufficient to satisfy the various walking which is combined by forward/backward walking, side step walking and turning. We needed to upgrade the walking pattern generation method to implement an omni-directional walking. We use the linear inverted pendulum model consisted of ZMP and center of mass in order to simplify the computation of walking pattern. The proposed method assumes that the state of the following stride is same to the state of the current stride. Using this assumption of walking pattern, the proposed method generates the stable walking pattern for various walking. And the proposed scheme generates the ZMP trajectory with the quartic polynomials in order to reduce the fluctuation of ZMP trajectory by various walking. To implement the efficient walking pattern, this method proposes three walking modules: periodic step module, transient step module and steady step module. Each step module utilizes weighted least square method with future ZMP position information. The effectiveness of the proposed method is verified by simulations of various walking. And the proposed method is confirmed by the experiment of real humanoid robot MAHRU-R.     

A Biologically Inspired Biped Locomotion Strategy for Humanoid Robots: Modulation of Sinusoidal Patterns by a Coupled Oscillator Model

Biological systems seem to have a simpler but more robust locomotion strategy than that of the existing biped walking controllers for humanoid robots. We show that a humanoid robot can step and walk using simple sinusoidal desired joint trajectories with their phase adjusted by a coupled oscillator model. We use the center-of-pressure location and velocity to detect the phase of the lateral robot dynamics. This phase information is used to modulate the desired joint trajectories. We do not explicitly use dynamical parameters of the humanoid robot. We hypothesize that a similar mechanism may exist in biological systems. We applied the proposed biologically inspired control strategy to our newly developed human-sized humanoid robot computational brain (CB) and a small size humanoid robot, enabling them to generate successful stepping and walking patterns.     

基于ADAMS的双足机器人拟人行走动态仿真

在双足机器人HEUBR_1的设计中,下肢采用了一种新的串并混联的仿人结构,并在足部增加了足趾关节.为验证该仿人结构设计的合理性及拟人步态规划的可行性,在ADAMS虚拟环境中建立了双足机器人HEUSR_1的仿真模型.通过拟人步态规划生成了运动仿真数据,在ADAMS虚拟环境中实现了具有足趾运动的拟人稳定行走,经仿真分析,获得了双足机器人HEUBR_1拟人行走步态下的运动学和动力学特性,仿真结果表明:双足机器人HEUBR_1的串并混联的仿人结构设计能够满足行走要求,且拟人步态规划方法可行,有足趾运动的拟人行走具有运动平稳、能耗低、足底冲击力小的特点.稳定行走的仿真步态数据可为下一步双足机器人HEUBR_1样机行走实验提供参考数据.     

Adaptive Robot Biped Locomotion with Dynamic Motion Primitives and Coupled Phase Oscillators

In order to properly function in real-world environments, the gait of a humanoid robot must be able to adapt to new situations as well as to deal with unexpected perturbations. A promising research direction is the modular generation of movements that results from the combination of a set of basic primitives. In this paper, we present a robot control framework that provides adaptive biped locomotion by combining the modulation of dynamic movement primitives (DMPs) with rhythm and phase coordination. The first objective is to explore the use of rhythmic movement primitives for generating biped locomotion from human demonstrations. The second objective is to evaluate how the proposed framework can be used to generalize and adapt the human demonstrations by adjusting a few open control parameters of the learned model. This paper contributes with a particular view into the problem of adaptive locomotion by addressing three aspects that, in the specific context of biped robots, have not received much attention. First, the demonstrations examples are extracted from human gaits in which the human stance foot will be constrained to remain in flat contact with the ground, forcing the “bent-knee” at all times in contrast with the typical straight-legged style. Second, this paper addresses the important concept of generalization from a single demonstration. Third, a clear departure is assumed from the classical control that forces the robot’s motion to follow a predefined fixed timing into a more event-based controller. The applicability of the proposed control architecture is demonstrated by numerical simulations, focusing on the adaptation of the robot’s gait pattern to irregularities on the ground surface, stepping over obstacles and, at the same time, on the tolerance to external disturbances.     

仿人机器人视觉导航中的实时性运动模糊探测器设计

针对仿人机器人视觉导航系统的鲁棒性受到运动模糊制约的问题,提出一种基于运动模糊特征的实时性异常探测方法. 首先定量地分析运动模糊对视觉导航系统的负面影响,然后研究仿人机器人上图像的运动模糊规律,在此基础上对图像的运动模糊特征进行无参考的度量,随后采用无监督的异常探测技术,在探测框架下对时间序列上发生的图像运动模糊特征进行聚类分析,实时地召回数据流中的模糊异常,以增强机器人视觉导航系统对运动模糊的鲁棒性. 仿真实验和仿人机器人实验表明:针对国际公开的标准数据集和仿人机器人NAO数据集,方法具有良好的实时性（一次探测时间0.1s）和有效性（召回率98.5%,精确率90.7%）. 方法的探测框架对地面移动机器人亦具有较好的普适性和集成性,可方便地与视觉导航系统协同工作.     

工业机器人定位误差规律分析及基于ELM算法的精度补偿研究

针对工业机器人应用于飞机零部件自动化钻孔时绝对定位精度较差的问题,提出利用极限学习机（ELM）算法建立机器人法兰中心点理论位置与实际位置之间的误差模型,并优化补偿机器人定位精度的方法．首先基于空间网格采样方法,获得了机器人绝对定位误差沿机器人基坐标系不同方向的误差变化规律,分析了建模补偿的可行性;其次建立基于ELM算法的误差补偿模型,并针对误差模型训练中隐含层神经元个数取值问题进行了分析优化．实验结果表明,机器人绝对定位误差值沿其坐标系不同方向存在不同的变化规律,补偿前绝对定位误差分布范围为0.29 mm~0.58 mm,平均误差为0.41 mm;补偿后定位误差分布范围降低到0.04 mm~0.32 mm,平均误差为0.18 mm;采用ELM算法建模的补偿速度快,泛化性能好．