This paper presents an artificial neural network (ANN)-based nonlinear model predictive visual servoing method for mobile robots. The ANN model is developed for state predictions to mitigate the unknown dynamics and parameter uncertainty issues of the physics-based (PB) model. To enhance both the model generalization and accuracy for control, a two-stage ANN training process is proposed. In a pretraining stage, highly diversified data accommodating broad operating ranges is generated by a PB kinematics model and used to train an ANN model first. In the second stage, the test data collected from the actual system, which is limited in both the diversity and the volume, are employed to further finetune the ANN weights. Path-following experiments are conducted to compare the effects of various ANN models on nonlinear model predictive control and visual servoing performance. The results confirm that the pretraining stage is necessary for improving model generalization. Without pretraining (i.e., model trained only with the test data), the robot fails to follow the entire track. Weight finetuning with the captured data further improves the tracking accuracy by 0.07–0.15 cm on average.

With the increasing demand for rehabilitation and the lack of professional rehabilitation personnel, robot-assisted rehabilitation technology plays an increasingly important role in neurological rehabilitation. In order to recover the exercise ability of patients with waist injury, a new type of wire-driven waist rehabilitation training parallel robot (WWRTPR) is designed. According to the motion trajectory planning of waist rehabilitation training, two coordinate systems are established: moving coordinate system and static coordinate system. The inverse kinematics modeling analysis is carried out, and the dynamic model of the robot is established by using Newton–Euler method. An intelligent control method of force/position hybrid control based on radial basis function neural network is proposed. The stability of the closed-loop system is analyzed, and the results show that WWRTPR tends to be stable. The simulation analysis of rehabilitation training on WWRTPR is carried out, and the simulation results show that the proposed intelligent control method can effectively control the robot system, which provides a reference for the development of a flexible intelligent rehabilitation training robot.

The work demonstrates a new approach to design of a level of intelligent control of robotic systems. The analytical model results in operational decisions. The structure of these decisions make them readily available to be implemented as an expert system. The approach is applied to a case study of mobile supervisory robots. The model presented here was motivated by manufacturing robotic systems and a type of autonomous robots that collect information at different sites for safety and other control purposes. The robot actions are directly affected by the obta~ned data. At each site the amount of available information (and thus the correctness of the robot decision) can be increased if the robot keeps collecting data at that site for a longer period of t~me. This means a delay in reacting and in reaching next site and accordingly, a decrease in the general amount of robot's information on the whole system.

The method of finding an economic amount of information collected by a robot at each site is based on the theory of controlled discrete event stochastic systems developed in our earlier works. This theory combines he basic concepts of discrete event control extended to stochastic systems with some aspects of information economics.

This paper is concerned with the design of intelligent subsystems that interface actuators and sensors to intelligent supervisors for robot work-cells. Benefits of our approach include removal of low level computation and stringent real-time constraints from the superviser, potential for combining and interpreting information from sensor combinations, and provision of a uniform information interface for disparate devices. Our approach to the design and organisation of these subsystems is based on the concept of virtual devices. We demonstrate the applicability of the concept by describing the design and implementation of an intelligent controller for a sensory gripper.