基于机器视觉的非特定物体的智能抓取系统的研究

针对传统机械臂局限于按既定流程对固定位姿的特定物体进行机械化抓取,设计了一种基于机器视觉的非特定物体的智能抓取系统;系统通过特定的卷积神经网络对深度相机采集到的图像进行目标定位,并在图像上预测出一个该目标的可靠抓取位置,系统进一步将抓取位置信息反馈给机械臂,机械臂根据该信息完成对目标物体的抓取操作;系统基于机器人操作系统,硬件之间通过机器人操作系统的话题机制传递必要信息;最终经多次实验结果表明,通过改进的快速搜索随机树运动规划算法,桌面型机械臂能够根据神经网络模型反馈的的标记位置对不同位姿的非特定物体进行实时有效的抓取,在一定程度上提高了机械臂的自主能力,弥补了传统机械臂的不足.     

基于多感知的空间机械手爪控制研究

对融合了视觉、滑觉、角位移等多种传感器的欠驱动空间机械手爪,研究其对不同形状、质地的物体实现自适应抓取控制。通过传感器反馈控制机械手运动、抓取力,提高机械手的自主能力。在抓取模式选择中,采用基于专家系统的抓取规划,根据物体不同的形状、尺寸选择不同的抓取模式;在抓取力控制中,通过由PVDF制作的滑觉传感器反馈,采用基于滑觉信号的模糊控制方法,对不同质地的物体选择不同的控制参数。通过实验研究验证基于多感知的控制方法对各种物体可以进行可靠的抓取。     

基于手势识别远程同步控制的智能机械臂研究

设计一种基于手势识别远程同步控制的智能机械臂系统,系统由手势识别器和智能机械臂组成;手势识别器穿戴在手中,能够感知和识别手势指令,并通过无线传输方式把手势指令传输给智能机械臂;智能机械臂收到手势指令后其自主决策系统迅速规划最优控制,实现同步控制机械臂伸展和机械手的抓取等动作;实验结果表明,手势识别器的手势指令简单易于操作,手势指令识别准确高效,智能机械臂动作规划协调,系统大大降低了机械臂的操控难度,完全满足作业任务实时控制的需要,具有较好的实用性和广泛的应用前景.     

绳驱动纱筒抓取仿生机械手设计

针对纱筒上下料对人力过度依赖的问题,在研究仿生学手指基础上,构建面向智能制造的纱筒抓取仿生机械手。首先,采取模块化设计思想,设计适合纱筒抓取的仿生机械手结构模型,并选择绳索传动作为驱动方式;其次,详细分析仿生机械手的组成及其抓取原理,运用D-H坐标法,实现机械手指坐标系和手指基座坐标系之间变换,推导机械手末端位置方程,得到最优抓取姿态;最后,利用有限元软件,建立三维欠驱动仿生机械手模型并对其进行虚拟装配与运动仿真分析,以验证机械手抓取纱筒的可行性和稳定性,形成机器人智能抓取仿生机械手的关键技术。     

作业型飞行机器人动态滑翔抓取的鲁棒自适应控制

作业型飞行机器人是指将多自由度机械臂固连在飞行机器人上的一类新型机器人系统,它能够对周围环境施加主动影响,同时也存在较为复杂的动力学性能.本文针对作业型飞行机器人滑翔抓取物体时所受到的摩擦力和接触力问题以及在飞行过程中产生的转动惯量变化问题,设计了一种整体式鲁棒自适应控制策略.首先在作业型飞行机器人系统动力学建模中引入...     

基于目标检测的机器人手眼标定方法

智能协作机器人依赖视觉系统感知未知环境中的动态工作空间定位目标,实现机械臂对目标对象的自主抓取回收作业。RGB-D相机可采集场景中的彩色图和深度图,获取视野内任意目标三维点云,辅助智能协作机器人感知周围环境。为获取抓取机器人与RGB-D相机坐标系之间的转换关系,提出基于yolov3目标检测神经网络的机器人手眼标定方法。将3D打印球作为标靶球夹持在机械手末端,使用改进的yolov3目标检测神经网络实时定位标定球的球心,计算机械手末端中心在相机坐标系下的3D位置,同时运用奇异值分解方法求解机器人与相机坐标系转换矩阵的最小二乘解。在6自由度UR5机械臂和Intel RealSense D415深度相机上的实验结果表明,该标定方法无需辅助设备,转换后的空间点位置误差在2 mm以内,能较好满足一般视觉伺服智能机器人的抓取作业要求。     

基于卷积神经网络的机械臂抓取控制系统设计

为保证机械臂的抓取精度,保证物体抓取的稳定性,本文设计基于卷积神经网络的机械臂抓取控制系统。在系统硬件部分,加设图像、位置和压力传感器,改装机械臂抓取控制器和运动驱动器,利用图像传感器设备,获取满足质量要求的机械臂抓取目标图像,为机械臂抓取控制功能提供硬件支持。软件部分利用卷积神经网络算法提取图像特征,确定机械臂抓取目标位置。结合机械臂当前位置的检测结果,规划机械臂抓取路线,预估机械臂抓取角度与抓取力。最终通过机械臂抓取参数控制量的计算,在控制器的支持下实现系统的机械臂抓取控制功能。实验结果表明,所设计系统应用下位置控制误差和速度控制误差的平均值分别为0.192m和0.138m/s,同时物体抓取掉落概率明显降低。     

Motion and force control with a nonlinear force error filter for underwater vehicle-manipulator systems

This paper deals with a control scheme for autonomous underwater robots equipped with manipulators. Several motion and force controllers have been developed. Most of them were designed in disregard of the dynamics of marine thrusters to develop a controller with a simple structure. However, the robot body propelled by thrusters generally has a considerably slower time response than the manipulator driven by electrical motors. Therefore, it may be difficult to construct a high-gain feedback control system to achieve a good control performance, because the high gain may excite the slow thruster dynamics ignored in the controller design, and the excitation will degrade the control performance. In this paper, we develop a motion and force controller for mathematical models with the dynamics of thrusters. It includes a nonlinear force error filter which allows us to construct a stable motion and force control system. To investigate its control performance, we conducted numerical simulations for comparing the proposed control scheme with an existing control scheme designed in disregard of the thruster dynamics. Simulation results demonstrate the usefulness of the proposed controller.     

基于软体手的机器人遥操作人机交互系统研制

为了提升抓取软体物体、易碎物体的完整性,设计制作了一种基于软体手的机器人遥操作人机交互系统。研制了一种嵌入了弯曲传感器的软体主手,获取人手的弯曲电信号,以此判断人手的弯曲程度,实现对软体从手系统的遥操作。并在远端软体从手中嵌入压力传感器,获取软体从手抓取目标时的表面压力电信号,以此来判断软体手抓取目标时的力度,实现力反馈。软体手人机交互系统能够减小传统刚体机械手抓取目标的限制,可有效抓取软体或易碎目标,实验表明了其有效性。     

三维立体视觉机械臂智能抓取分类系统的开发

采用工业相机、工业投影机、普通摄像头、计算机和机械臂开发了一套具有三维立体视觉的机械臂智能抓取分类系统。该系统采用自编软件实现了对工业相机、工业投影机的自动控制和同步，通过前期研究提出的双波长条纹投影三维形貌测量法获取了物体的高度信息，结合opencv技术和普通摄像头获取的物体二维平行面信息，实现了物体的自动识别和分类；利用串口通信协议，将上述处理后的数据传送至机械臂，系统进行几何姿态解算，实现了智能抓取，并能根据抓手上压力反馈自动调节抓手张合程度，实现自适应抓取。经实验证明该系统能通过自带的快速三维形貌获取装置实现准确、快速的抓取工作范围内的任意形状的物体并实现智能分类。     

Soft manipulator inspired by octopi: object grasping in all anatomical planes using a tendon-driven continuum arm

Artificial Life and Robotics - In this paper, we focus on the development of a soft continuum arm manipulator inspired by the morphology of octopi and their intelligent behavior. The proposed arm...     

MM-UAV: Mobile Manipulating Unmanned Aerial Vehicle

Given significant mobility advantages, UAVs have access to many locations that would be impossible for an unmanned ground vehicle to reach, but UAV research has historically focused on avoiding interactions with the environment. Recent advances in UAV size to payload and manipulator weight to payload ratios suggest the possibility of integration in the near future, opening the door to UAVs that can interact with their environment by manipulating objects. Therefore, we seek to investigate and develop the tools that will be necessary to perform manipulation tasks when this becomes a reality. We present our progress and results toward a design and physical system to emulate mobile manipulation by an unmanned aerial vehicle with dexterous arms and end effectors. To emulate the UAV, we utilize a six degree-of-freedom miniature gantry crane that provides the complete range of motion of a rotorcraft as well as ground truth information without the risk associated with free flight. Two four degree-of-freedom manipulators attached to the gantry system perform grasping tasks. Computer vision techniques and force feedback servoing provide target object and manipulator position feedback to the control hardware. To test and simulate our system, we leverage the OpenRAVE virtual environment and ROS software architecture. Because rotorcraft are inherently unstable, introduce ground effects, and experience changing flight dynamics under external loads, we seek to address the difficult task of maintaining a stable UAV platform while interacting with objects using multiple, dexterous arms. As a first step toward that goal, this paper describes the design of a system to emulate a flying, dexterous mobile manipulator.     

基于卷积神经网络的软硬触觉感知方法研究

智能机器手的应用已经遍布医疗、军工、农业及装配行业等领域.软硬作为物体的重要物理属性之一,对机器手的抓取控制物体有重大影响.在深度学习框架下,基于卷积神经网络提出了用于触觉感知的软硬物体的识别方法.使用薄膜压力传感器采集手指按压软硬物体的数据,建立训练和测试数据集,在Caffe中训练网络,以模拟触觉识别软硬物体.实验结果显示:对软硬物体的识别准确率达94.52%,表明,卷积神经网络对于识别软硬物体有比较好的分类效果.     

Intelligent Control for Manipulators with Moving Bases

Much research has recently been dedicated to the control of manipulators (rigid, flexible or with flexible joints) on stationary bases. To enlarge the workspace of manipulators, to avoid obstacles or singularities, and to open new application areas, a new generation of robots has been created. Among them is the mobile manipulator, a manipulator mounted on a vehicle. The consequent base motion and coupling forces not only affect the manipulator's kinematics but also its dynamics. Since it is usually difficult, if not impossible, to model them, an intelligent control method is proposed. The main idea is to design a regular manipulator motion controller that considers the dynamic coupling effects as external disturbances. Robustness is assured not by computation but by estimating and compensating for the coupling effects, along with any other disturbances.     

Soft Robot Arm Inspired by the Octopus

The octopus is a marine animal whose body has no rigid structures. It has eight arms composed of a peculiar muscular structure, named a muscular hydrostat. The octopus arms provide it with both locomotion and grasping capabilities, thanks to the fact that their stiffness can change over a wide range and can be controlled through combined contractions of the muscles. The muscular hydrostat can better be seen as a modifiable skeleton. Furthermore, the morphology the arms and the mechanical characteristics of their tissues are such that the interaction with the environment (i.e., water) is exploited to simplify control. Thanks to this effective mechanism of embodied intelligence, the octopus can control a very high number of degrees of freedom, with relatively limited computing resources. From these considerations, the octopus emerges as a good model for embodied intelligence and for soft robotics. The prototype of a robot arm has been built based on an artificial muscular hydrostat inspired to the muscular hydrostat of the Octopus vulgaris. The prototype presents the morphology of the biological model and the broad arrangement of longitudinal and transverse muscles. Actuation is obtained with cables (longitudinally) and with shape memory alloy springs (transversally). The robot arm combines contractions and it can show the basic movements of the octopus arm, like elongation, shortening and bending, in water.     

Parameterizing robust manipulator controllers under approximate inverse dynamics: A double‐Youla approach

We consider the goal of ensuring robust stability when a given manipulator feedback control law is modified online, for example, to safely improve the performance by a learning module. To this end, the factorization approach is applied to both the plant and controller models to characterize robustly stabilizing controllers for rigid‐body manipulators under approximate inverse dynamics control. Outer‐loop controllers to stabilize the nonlinear uncertain loop that results from approximate inverse dynamics are often derived by lumping uncertainty in a single term and subsequent analysis of the error system. Here, by contrast, the well‐known norm bounds of these uncertain dynamics are first recast into a generalized plant configuration that preserves the characteristic uncertainty structure. Then, the overall loop uncertainty is expressed with respect to the nominal outer‐loop feedback controller by means of an uncertain dual‐Youla operator. Therefore, using the dual‐Youla parameterization, we provide a novel way to rigorously quantify permissible perturbations of robot manipulator feedforward/feedback controllers. The method proposed in this paper does not constitute another robust control law for rigid‐body manipulators, but rather a characterization of a set of robustly stabilizing controllers. The resulting double‐Youla parameterization for the control of robot manipulators is amenable to numerous advanced design methods. The result is thoroughly discussed by a planar elbow manipulator and exemplified with a six‐degree‐of‐freedom robot scenario with varying payload.     

A hybrid learning/adaptive partial state feedback controller for RLED robot manipulators

In this paper, we present an adaptive partial state-feedback repetitive learning control algorithm for a rigid-link electrically-driven (RLED) robot manipulator actuated by brushed DC (BDC) motors. The proposed controller is designed to compensate for repeatable mechanical uncertainty via a learning control term while an adaptive control loop is used to compensate for parametric uncertainty in the electrical dynamics. The proposed controller guarantees semi-global asymptotic link position tracking while only requiring measurements of link position and electrical winding current (e.g. measurements of link velocity are not required).     

基于点云采样权重估计的未知物体抓取位姿生成方法

针对机械臂在非结构环境中对未知物体抓取位姿生成困难及抓取稳定性差的问题,提出一种基于点云采样权重估计的抓取位姿生成方法.首先通过移动深度相机的方式拼接得到较完整的物体点云信息,并对物体的几何特性进行分析,有效避开物体不宜抓取的位置进行抓取位姿样本生成;然后结合几何约束条件实现抓取位姿搜索,并利用力封闭条件对样本稳定性进行评估;最后为了对实际的抓取位姿进行评价,根据其稳定性、夹取深度、夹取角度等设定抓取可行性指标,据此在工作空间输出最佳抓取位姿并完成指定的抓取任务.实验结果表明,采用所提方法能够高效生成大量且稳定的抓取位姿,并在仿真环境中有效实现机械臂对单个或多个随机摆放的未知物体的抓取任务.     

智能获取装箱管状工件抓取位置的研究

机械手智能抓取工件时,工件抓取位置的获取是基于机器视觉精确装配的重点。针 对环境较复杂的装箱管状工件的装配抓取环节,建立合理的机器视觉系统,利用正向照明系统对 工件表面产生的反光直光条特征,提出了一种Gaussian 拟合与Hough 变换相结合的拟合算法。首 先利用Gaussian 拟合提取各光条法向上的中心坐标,然后将获取的坐标点集运用Hough 变换进行 拟合,最后根据获取的各光条中心线计算各工件的抓取位置。实验结果表明该方法能同时实现多 条光条直线的拟合,又能抑制干扰点或噪声的影响,有效实现装箱工件抓取位置的智能获取。     

柔性机械臂速度控制研究

结合机械臂的结构特点,将神经网络与预测控制、滚动优化理论相结合,设计出用于柔性机械臂轨迹跟踪的神经网络动态优化调整的智能控制系统。从而能够以良好的控制柔性机械臂跟踪各状态量的期望值,并同时达到末时刻速度调节的目的。