Solution of forward kinematics in Stewart platform using six rotary sensors on joints of three legs

A novel method is proposed for real-time solution of direct kinematics problem of Stewart platform (SP) using six measurements on three legs’ joints consisting of the rotations of three legs in two directions. After the application of the method on a laboratory sample SP, it is observed that the method is preferable to the conventional method that uses the length measurements of all six legs, in the grounds of industrial applicability. It is due to simpler implementation, less expense, easier maintenance, and stress-free assembly. The algorithms of both forward and inverse kinematics are fully derived based on geometric relationships between the platform states and the measurement data. The sensitivity to the measurement errors is analyzed theoretically and is applied through a computer simulation to several configurations of the sample SP which are uniformly distributed in the workspace. The variances of measurement errors for those configurations are compared between the conventional and proposed methods and it is observed that: the proposed method operates more accurate in position measurement especially in lateral movements. Additionally, the proposed method is not too sensitive to direction of movement and geometry of the SP.     

Quadruped walking with parallel link legs

We are proposing an underwater robot for the work. In this study, we designed the robot, which has body of rectangular plane and 4 legs at each corner. The leg is consisted with parallel mechanism of 2or3 cylinders, and the end of each cylinder is attached on the robot body with free rotational joint and the end of both piston rods are connected with pin joint. 2 cylinder leg’s motion is restricted in forward or backward direction but 3 cylinder leg can move any direction. We are studying the control scheme of walking for this robot, which is putting mind especially on smooth and steady movement without rolling, pitching, yawing or heaving motion and keeping the body horizontally. We confirmed the validity of control scheme with simulation and experiments.     

基于逆运动学和动力学的虚拟人行走仿真

在计算机上模拟真实人行走是计算机仿真的一个基本问题。人体行走是一种伴随着碰撞、摩擦和滑动的复杂的系统运动,为了实现模拟的逼真性,需要着重在运动控制上进行研究。首先对人体行走进行分析并建立了简单的人体模型,然后详细给出行走过程中关节点位置的数学描述,最后采用逆运动学求解雅可比矩阵的方法并结合动力学知识,运用VC＋＋．NET和OpenGL为编程工具以骨架模型实现了虚拟人行走。     

基于独立成分分析和核向量机的虹膜识别方法

针对虹膜识别过程中的特征提取及识别问题,提出了用独立成分分析提取虹膜特征,用核向量机进行识别的方法.从采集到的人眼图像中定位虹膜,并对其进行归一化处理和图像增强处理.用独立成分分析提取统计独立的特征,通过选择合适的特征个数可以达到较高的识别准确率.在得到虹膜特征编码后,用核向量机进行分类判决,核向量机是一种适合大规模数据集的快速支持向量机训练算法,并将结果与支持向量机的分类结果进行了对比.实验结果表明了该方法的可行性和有效性.     

伸缩腿双足机器人半被动行走控制研究

研究半被动伸缩腿双足机器人行走控制和周期解的全局稳定性问题.使用杆长可变的倒立摆机器人模型,以支撑腿的伸缩作为行走动力源,采用庞加莱映射方法分析了双足机器人行走的不动点及其稳定性.当脚与地面冲击时,假设两腿间的夹角保持为常数,设计了腿伸缩长度的支撑腿角度反馈控制率.证明了伸缩腿双足机器人行走过程不动点的全局稳定性.仿真结果表明,本文提出的腿伸缩长度反馈控制可以实现伸缩腿双足机器人在水平面上的稳定行走,并且周期步态对执行器干扰和支撑腿初始角速度干扰具有鲁棒性.     

Hybrid learning concepts based on self-organizing neural networks for adaptive control of walking machines

Online learning of complex control behaviour of autonomous mobile robots like walking machines is one of the current research topics. In this article, a hybrid learning architecture based on reinforcement learning (RL) and self-organizing neural networks for online adaptivity is presented. The hybrid concept integrates different learning methods and task-oriented representations as well as available domain knowledge. The proposed concept is used for RL of control strategies on different control levels on a walking machine.     

Path tracking with quadruped walking machines using discontinuous gaits

Discontinuous gaits for walking machines offer great advantages over wave gaits, and they seem more adequate for following a path over irregular terrain. This paper, focused on quadruped walking robots, addresses the problem of following an arbitrary path using both discontinuous crab and turning gaits. First, the paper presents the algorithms to generate these gaits as local motions and highlights their advantages in comparison with continuous gaits. This comparison considers stability, velocity, power consumption and terrain adaptability. The algorithms for tracking an arbitrary trajectory using these gaits are then introduced, and some simulations and experimental results are reported.     

基于DE-EDA多目标优化的受电弓模糊控制

针对传统控制器参数整定算法不能兼顾多个性能指标的问题,就受电弓-接触网的接触力控制问题设计了基于差分进化-分布估计算法（DE-EDA）的模糊控制器。以接触力的方差、误差的积分和控制器输出能量为优化目标,以模糊控制器量化因子和比例因子为优化变量,建立多目标优化模型。与LQR受电弓最优控制进行对比,仿真结果表明了所提出的基于DE-EDA多目标优化的模糊控制器的有效性与优越性。     

On energetically efficient motion algorithms of walking machines with cyclic drives

Walking machines with cyclic drives are considered, the program motion of the machine and its drives that provide minimum of power expenditures is synthesized.     

Walk control of segmented multi-legged robot based on integrative control of legs and 2-DoF active intersegment joints

This paper proposes a novel walk control method for multi-legged robots which have segmented bodies like a millipede. The segment of the robot which is addressed in this paper has a pair of legs which has 3 degrees of freedom (DoF) and 2-DoF active joint which connects the adjacent segmented body. The legs are controlled based on a decentralised control method, follow-the-contact point gait control, in which the legs contact on the contact-points where the forward legs have done. The joints between segments are controlled so as to make the bodies follow a desired trajectory. The legs and the intersegment joints are controlled with integrative and consistent manner thanks to local coordinate systems defined on each contact point. Finally, the proposed control method is verified via a robot simulator with physics engine.     

Force distribution for the legs of a quadruped walking vehicle

For a walking vehicle with active force control an important problem is how to distribute the forces of the feet. To prevent leg slippage, the active force control of a walking vehicle requires an efficient approach to optimize the foot force distribution. This article presents a new optimization method for a quadruped walking vehicle. We combine the force/moment equilibrium equations with some optimal relations and obtain a set of solvable linear equations. The solution of the equations is the optimal force distribution to prevent leg slippage. To solve the discontinuity problem of optimal solutions, we have used the principle of the convex combination that is an interpolation. Thus the optimal solution found with this algorithm is continuous during a complete locomotion cycle. To solve the force distribution problem for four feet on the ground the maximum computing time is 26 ms on an IBM-compatible DX 80386+803787 only. This optimization method is quite efficient for the controller in real time. © 1997 John Wiley & Sons, Inc.     

Impact of railroad ballast type on frontal plane ankle kinematics during walking

Five healthy male subjects walked on a control surface (level concrete), and two sloped rock surfaces (walking ballast-rock about 1.9 cm across; main line ballast-rock about 3.8 cm across) while their rearfoot motion (defined throughout as ankle inversion/eversion as seen from the frontal plane) was measured to determine if the different walking surfaces caused different ankle kinematics. The ballast was placed in 5m long trays that were tilted 7 degrees in the transverse plane. Rearfoot motion was measured while the subjects walked the length of the respective surfaces wearing work boots. A repeated measures ANOVA and a subsequent multiple comparison test revealed that the rearfoot range of motion was significantly greater walking on the main line ballast than walking on either the walking ballast or the level concrete. Meanwhile, the mean range of rearfoot motion for walking ballast was not significantly different from that resulting from walking on concrete. Variability was more than twice as great walking on main line ballast than walking on level concrete. Rearfoot angular velocities walking on level concrete and walking ballast were not significantly different, but both were significantly less than walking on main line ballast. Results suggested that rearfoot motion could be reduced if railroads placed walking ballast where trainmen have to walk as part of their jobs.     

变长度弹性伸缩腿双足机器人半被动起步行走仿人控制

传统双足机器人行走使用轨迹跟踪控制,而人类行走大部分时间处于被动状态.针对半被动变长度弹性伸缩腿双足机器人从静止状态开始起步行走的问题,提出了一种起步行走仿人控制方法.首先,使用串联弹性驱动双足弹簧负载倒立摆(B-SLIP)模型;然后,利用拉格朗日方法建立行走动力学方程,并利用模型的自稳定性在双支撑阶段采用能量误差比例...     

Neural virtual sensors for terrain adaptation of walking machines

When walking in realistic conditions, accurate, reliable sensorial information is critical to ensure the safe operation of legged robots. That means a large number of sensors, cabling, and electronic systems must be used, complicating the robot. On the other hand, the great complexity of the hardware of walking machines is one of the main obstacles preventing the introduction of this kind of vehicle in real applications; consequently this hardware should be simplified. These antagonistic requirements can be reconciled by the use of what are called virtual sensors. This paper addresses the design of virtual sensors for terrain adaptation developed with the aims of simplifying the hardware of the walking machine or increasing the reliability of the sensorial information available. These virtual sensors are based on neural networks and can estimate the forces exerted by the feet from data extracted from joint‐position sensors, which are mandatory in all robotic systems. The force estimates are used to detect foot/ground contact. Some experiments carried out with the SILO4 walking robot are reported to prove the efficacy of this method. © 2005 Wiley Periodicals, Inc.     

Forward kinematics of a class of parallel (Stewart) platforms with closed-form solutions

This article studies the geometrical condition for closed-form solutions of forward kinematics of parallel platforms. It is shown that closed-form solutions are available if 1 rotational degree of freedom (dof) of the moving platform is decoupled from the other 5 dof. Geometrically, this condition is satisfied when five end-points at the moving platform (or at the base) are colinear. A general case that these five points do not coincide with each other is studied first and is shown to have 16 possible closed-form solutions. The variations of parallel platforms that satisfy the above-mentioned geometrical condition are then discussed. Some of them have the additional feature that the three rotational dof are fully decoupled from the 3 translational dof and their closed-form solutions are further simplified. One particular case has extremely simple forward kinematics and could be used as an alternative to the Stewart platform.     

结合独立成分分析和核向量机进行人脸识别

在人脸识别过程中,首先利用独立成分分析得到独立的人脸基影像,所提取的特征就是人脸图像在基影像上的投影系数,通过选择合适的特征个数可以达到较高的识别准确率。然后采用支持向量机和核向量机分别对待识别图像在基影像上的投影系数进行分类判决,结果显示二者都能达到较高的识别准确率,但随着特征个数的增加,核向量机的准确率更高,训练时间更短,支持向量更少。实验表明方法可行有效的。     

Face recognition with lattice independent component analysis and extreme learning machines

We focus on two aspects of the face recognition, feature extraction and classification. We propose a two component system, introducing Lattice Independent Component Analysis (LICA) for feature extraction and Extreme Learning Machines (ELM) for classification. In previous works we have proposed LICA for a variety of image processing tasks. The first step of LICA is to identify strong lattice independent components from the data. In the second step, the set of strong lattice independent vector are used for linear unmixing of the data, obtaining a vector of abundance coefficients. The resulting abundance values are used as features for classification, specifically for face recognition. Extreme Learning Machines are accurate and fast-learning innovative classification methods based on the random generation of the input-to-hidden-units weights followed by the resolution of the linear equations to obtain the hidden-to-output weights. The LICA-ELM system has been tested against state-of-the-art feature extraction methods and classifiers, outperforming them when performing cross-validation on four large unbalanced face databases.     

Forward tracking of complex trajectories with non-Standard N-Trailers of non-minimum-phase kinematics avoiding a jackknife effect

It is a common conviction that forward motion control of tractor-trailer vehicles is a substantially simpler problem relative to reversing with trailers. This opinion may be misleading when considering the N-trailer vehicles moving forward with positive hitching offsets when a guidance point is located on a trailer. Due to the non-minimum-phase nature of vehicle kinematics, closing a feedback from a trailer posture can lead to the jackknife effect in this case. So far, there has been no solution to this problem for the N-trailers admitting trajectories of a varying curvature. To fill this gap, we propose a scalable and modular control strategy applicable to the N-trailer vehicles equipped solely with off-axle interconnections. The concept relies on a transformation of the control problem posed for the non-minimum-phase kinematics into a corresponding problem formulated for a virtual vehicle of minimum-phase kinematics, which can be solved by using the recently proposed cascade-like controller.     

Neural networks IC controlled multi-legged walking MEMS robot with independent leg mechanism

In this paper, we will compare the walking behavior of quadruped and hexapod walking MEMS robots. These robots are fabricated by connecting same modules, which are composed of a couple of independent leg mechanisms. Independent leg mechanisms can actuate the single leg by a single artificial muscle wire. The neural networks IC that mimics real living organisms controls the mechanical systems. The length and weight of the quadruped MEMS robot were 7.2 mm and 95.8 mg, respectively. The quadruped robot showed the walking speed of 24.6 mm/min. The robot tended to lose its balance and the weight balance is quite important for the moving quadruped. On the other hand, the length and weight of the hexapod MEMS robot were 9.0 mm and 162 mg, respectively. The hexapod robot showed stable walking. The speed was 27.0 mm/min.