In order to ensure that the off-line arm of a two-arm-wheel combined inspection robot can reliably grasp the line in case of autonomous obstacle crossing,a control method is proposed for line grasping based on hand-ey...In order to ensure that the off-line arm of a two-arm-wheel combined inspection robot can reliably grasp the line in case of autonomous obstacle crossing,a control method is proposed for line grasping based on hand-eye visual servo.On the basis of the transmission line's geometrical characteristics and the camera's imaging principle,a line recognition and extraction method based on structure constraint is designed.The line's intercept and inclination are defined in an imaging space to represent the robot's change of pose and a law governing the pose decoupling servo control is developed.Under the integrated consideration of the influence of light intensity and background change,noise(from the camera itself and electromagnetic field)as well as the robot's kinetic inertia on the robot's imaging quality in the course of motion and the grasping control precision,a servo controller for grasping the line of the robot's off-line arm is designed with the method of fuzzy control.An experiment is conducted on a 1:1 simulation line using an inspection robot and the robot is put into on-line operation on a real overhead transmission line,where the robot can grasp the line within 18 s in the case of autonomous obstacle-crossing.The robot's autonomous line-grasping function is realized without manual intervention and the robot can grasp the line in a precise,reliable and efficient manner,thus the need of actual operation can be satisfied.展开更多
A grasping force control strategy is proposed in order to complete various free manipulations by using anthropomorphic prosthetic hand. The position-based impedance control and force-tracking impedance control are use...A grasping force control strategy is proposed in order to complete various free manipulations by using anthropomorphic prosthetic hand. The position-based impedance control and force-tracking impedance control are used in free and constraint spaces, respectively. The fuzzy observer is adopted in transition in order to switch control mode. Two control modes use one position-based impedance controller. In order to achieve grasping force track, reference force is added to the impedance controller in the constraint space. Trajectory tracking in free space and torque tracking in constrained space are realized, and reliability of mode switch and stability of system are achieved. An adaptive sliding mode friction compensation method is proposed. This method makes use of terminal sliding mode idea to design sliding mode function, which makes the tracking error converge to zero in finite time and avoids the problem of conventional sliding surface that tracking error cannot converge to zero. Based on the characteristic of the exponential form friction, the sliding mode control law including the estimation of friction parameter is obtained through terminal sliding mode idea, and the online parameter update laws are obtained based on Lyapunov stability theorem. The experiments on the HIT Prosthetic Hand IV are carried out to evaluate the grasping force control strategy, and the experiment results verify the effectiveness of this control strategy.展开更多
堆叠覆盖环境下的机械臂避障抓取是一个重要且有挑战性的任务。针对机械臂在堆叠环境下的避障抓取任务,本文提出了一种基于图像编码器和深度强化学习(deep reinforcement learning,DRL)的机械臂避障抓取方法Ec-DSAC(encoder and crop fo...堆叠覆盖环境下的机械臂避障抓取是一个重要且有挑战性的任务。针对机械臂在堆叠环境下的避障抓取任务,本文提出了一种基于图像编码器和深度强化学习(deep reinforcement learning,DRL)的机械臂避障抓取方法Ec-DSAC(encoder and crop for discrete SAC)。首先设计结合YOLO(you only look once)v5和对比学习网络编码的图像编码器,能够编码关键特征和全局特征,实现像素信息至向量信息的降维。其次结合图像编码器和离散软演员-评价家(soft actor-critic,SAC)算法,设计离散动作空间和密集奖励函数约束并引导策略输出的学习方向,同时使用随机图像裁剪增加强化学习的样本效率。最后,提出了一种应用于深度强化学习预训练的二次行为克隆方法,增强了强化学习网络的学习能力并提高了控制策略的成功率。仿真实验中Ec-DSAC的避障抓取成功率稳定高于80.0%,验证其具有比现有方法更好的避障抓取性能。现实实验中避障抓取成功率为73.3%,验证其在现实堆叠覆盖环境下避障抓取的有效性。展开更多
基金Project(2006AA04Z202)supported by the National High Technology Research and Development Program of ChinaProject(51105281)supported by the National Natural Science Foundation of China
文摘In order to ensure that the off-line arm of a two-arm-wheel combined inspection robot can reliably grasp the line in case of autonomous obstacle crossing,a control method is proposed for line grasping based on hand-eye visual servo.On the basis of the transmission line's geometrical characteristics and the camera's imaging principle,a line recognition and extraction method based on structure constraint is designed.The line's intercept and inclination are defined in an imaging space to represent the robot's change of pose and a law governing the pose decoupling servo control is developed.Under the integrated consideration of the influence of light intensity and background change,noise(from the camera itself and electromagnetic field)as well as the robot's kinetic inertia on the robot's imaging quality in the course of motion and the grasping control precision,a servo controller for grasping the line of the robot's off-line arm is designed with the method of fuzzy control.An experiment is conducted on a 1:1 simulation line using an inspection robot and the robot is put into on-line operation on a real overhead transmission line,where the robot can grasp the line within 18 s in the case of autonomous obstacle-crossing.The robot's autonomous line-grasping function is realized without manual intervention and the robot can grasp the line in a precise,reliable and efficient manner,thus the need of actual operation can be satisfied.
基金Project(2009AA043803) supported by the National High Technology Research and Development Program of China Project (SKLRS200901B) supported by Self-Planned Task of State Key Laboratory of Robotics and System (Harbin Institute of Technology),ChinaProject (NCET-09-0056) supported by Program for New Century Excellent Talents in Universities of China
文摘A grasping force control strategy is proposed in order to complete various free manipulations by using anthropomorphic prosthetic hand. The position-based impedance control and force-tracking impedance control are used in free and constraint spaces, respectively. The fuzzy observer is adopted in transition in order to switch control mode. Two control modes use one position-based impedance controller. In order to achieve grasping force track, reference force is added to the impedance controller in the constraint space. Trajectory tracking in free space and torque tracking in constrained space are realized, and reliability of mode switch and stability of system are achieved. An adaptive sliding mode friction compensation method is proposed. This method makes use of terminal sliding mode idea to design sliding mode function, which makes the tracking error converge to zero in finite time and avoids the problem of conventional sliding surface that tracking error cannot converge to zero. Based on the characteristic of the exponential form friction, the sliding mode control law including the estimation of friction parameter is obtained through terminal sliding mode idea, and the online parameter update laws are obtained based on Lyapunov stability theorem. The experiments on the HIT Prosthetic Hand IV are carried out to evaluate the grasping force control strategy, and the experiment results verify the effectiveness of this control strategy.
文摘堆叠覆盖环境下的机械臂避障抓取是一个重要且有挑战性的任务。针对机械臂在堆叠环境下的避障抓取任务,本文提出了一种基于图像编码器和深度强化学习(deep reinforcement learning,DRL)的机械臂避障抓取方法Ec-DSAC(encoder and crop for discrete SAC)。首先设计结合YOLO(you only look once)v5和对比学习网络编码的图像编码器,能够编码关键特征和全局特征,实现像素信息至向量信息的降维。其次结合图像编码器和离散软演员-评价家(soft actor-critic,SAC)算法,设计离散动作空间和密集奖励函数约束并引导策略输出的学习方向,同时使用随机图像裁剪增加强化学习的样本效率。最后,提出了一种应用于深度强化学习预训练的二次行为克隆方法,增强了强化学习网络的学习能力并提高了控制策略的成功率。仿真实验中Ec-DSAC的避障抓取成功率稳定高于80.0%,验证其具有比现有方法更好的避障抓取性能。现实实验中避障抓取成功率为73.3%,验证其在现实堆叠覆盖环境下避障抓取的有效性。
