This paper focuses on developing a robotic object grasping approach that possesses the ability of self-learning, is suitable for small-volume large variety production, and has a high success rate in object grasping/pick-and-place tasks. The proposed approach consists of a computer vision-based object detection algorithm and a deep reinforcement learning algorithm with self-learning capability. In particular, the You Only Look Once (YOLO) algorithm is employed to detect and classify all objects of interest within the field of view of a camera. Based on the detection/localization and classification results provided by YOLO, the Soft Actor-Critic deep reinforcement learning algorithm is employed to provide a desired grasp pose for the robot manipulator (i.e., learning agent) to perform object grasping. In order to speed up the training process and reduce the cost of training data collection, this paper employs the Sim-to-Real technique so as to reduce the likelihood of damaging the robot manipulator due to improper actions during the training process. The V-REP platform is used to construct a simulation environment for training the deep reinforcement learning neural network. Several experiments have been conducted and experimental results indicate that the 6-DOF industrial manipulator successfully performs object grasping with the proposed approach, even for the case of previously unseen objects.

To increase the cruising distance and braking reliability of various light electric vehicles (LEVs) actuated by brushless DC machines, this study proposes three cost‐effective braking commutation strategies based on a general full‐bridge DC/AC inverter without using position sensors. The braking commutation strategies investigated in this study are named according to the number of power switches in action over each commutation state; designated as single‐switch, two‐switch, and three‐switch, respectively. Theoretical analysis of the operation principle and important performance indices such as maximum voltage conversion ratio, braking torque, energy recovery ratio etc, are conducted and compared among each commutation strategy. Moreover, a cost‐effective, sensorless method for each different braking commutation strategy is used. The commutation signals are extracted directly from the average terminal voltages. Since additional power switches, passive components and costly position sensors are not essential, the proposed method is particularly suitable for various LEVs such as electric bicycles, electric scooters, electric wheelchairs and so on. Finally, several experiments are performed to justify the effectiveness of the proposed method.

By exploiting the concept of closed-loop input error (CLIE), this article proposes an identification approach for the dynamic parameters of industrial robot manipulator based on evolutionary algorithms (EAs). The proposed algorithm estimates the dynamic parameters by using joint torque residuals between the actual robot and a parallel estimated model. Both the actual robot and the estimated model use the same reference trajectories and the same control law structure tuned with the same gain. The state of the estimated model is generated by filtering the measured signal through a model state generator, so by adjusting the cutoff frequency of the filter, different estimated models can be generated. By using EA, one can search for an estimated model in the solution space corresponding to the model state generator so that the joint torque difference between the actual robot and the estimated model is minimized. In addition, EA will automatically adjust the filtering strength of the measured signals to provide a noise-free signal for the estimated model so that the identification result can be more accurate. The dynamic parameters of the estimated model obtained through the optimization process are the optimal identification results of the actual robot. Moreover, to reduce the computation cost of the estimated model, this article also provides an approximation algorithm for the observation matrix so as to enhance the search efficiency of EA. Several experiments have been conducted on a 6-DOF industrial robot manipulator to verify the effectiveness of the proposed approach.

Oftentimes, due to mechanical wear and ambient environment variations, friction-related parameters may vary with respect to the position of the iron-core linear servomotor. In addition, cogging force is also a function of the iron-core linear servomotor's position. In order to cope with the adverse effects caused by friction and cogging force, this paper proposes a disturbance compensation approach based on Non-Uniform Rational B-Spline (NURBS), in which the parameters of the proposed NURBS-based disturbance compensation approach are represented as a function of the linear servomotor's position. The parameters of the proposed NURBS-based friction model are updated using online adaptive laws designed based on the Lyapunov stability theory. Moreover, in order to attain excellent tracking performance for a motion stage actuated by iron-core linear servomotors, this paper combines the proposed NURBS-based friction model and cogging force model with Fast Nonsingular Terminal Sliding Mode Control (FNTSMC) to develop Adaptive Nonlinear Sliding Mode Control (ADNSMC). The proposed ADNSMC inherits the advantages of both FNTSMC and the proposed NURBS-based disturbance compensation approach. Experimental results indicate that the proposed ADNSMC exhibits satisfactory performance on tracking control and disturbance compensation.

This paper proposes a robotic assistance control scheme for intuitive teaching tasks by integrating the motion intention of human and real-time hand impedance compensation based upon a fuzzy RBF (Radial Basis Function) compensator. The motion intention of a human is estimated using a feedforward neural network. The parameters of the proposed fuzzy RBF hand impedance compensator are adjusted by considering hand impedance and robot dynamics. Three robotic assistance control schemes are compared: 1) robot without an impedance compensator; 2) robot with a constant-parameter impedance compensator; 3) robot with a fuzzy RBF hand impedance compensator. Several experiments have been conducted to verify the effectiveness of the proposed approach by comparing the contour error, exerted force and task time spent on teaching tasks. Experimental results indicate that the proposed fuzzy RBF hand impedance compensator has the best assistance results among the tested robotic assistance control schemes.