Article Preview
TopIntroduction
Manipulators are widely used in industry, agriculture, and daily life for their ability to mimic certain movements of human arms, which allows them to grab things, carry objects, and operate tools (Ji & Huang, 2015). As a typical multi-input and multi-output non-linear system, a manipulator has time-varying and strong coupling characteristics. A manipulator requires high precision and fast tracking. Proportional-Integral-Derivative (PID) control (Liu, 2008) is a typical traditional control algorithm featuring a simple principle and easy parameter adjustment. PID control, however, is prone to large cumulative errors when trajectory tracking is carried out continuously for each joint of the manipulator. Trajectory tracking has evolved from the traditional control method to the current modern control algorithm (Galicki, 2015; Subramanian et al., 2017; Verrelli et al., 2015).
Neural network control, fuzzy control, and other advanced methods have been intensively investigated in recent years for manipulators. In terms of the trajectory tracking of the multi-joint manipulator, the manipulator’s dynamic model is divided into nominal model and modeling error (Ma & Jiang, 2016). The function of the sliding mode is defined as a tracking position error and velocity error aiming at the nominal model and modeling error, design control law for the nominal model, and the modeling error. The robust sliding mode PID is designed based on the manipulator’s nominal model. An improved adaptive PID algorithm was proposed (Zhan & Zhou, 2018) based on fuzzy control, which can enhance smoothness and accuracy and ensure smooth operations without impact damage in the process of grasping parts.
They also design and make a controller based on STM32 that demonstrates the feasibility and reasonability of the algorithm by comparing it with the performance curves of the ordinary PID controller in the position control test.
Li et al. (2019) combined the BP neural network and the ant colony algorithm to realize the manipulator trajectory control. Experiments showed that the angular displacement of the proposed algorithm was more fitting with the expected angular displacement, and the displacement error in the spatial three-dimensional coordinate system was smaller. In order to improve the response speed and control the accuracy of the underwater robot manipulator system, Deng (2019) proposed a sliding mode variable structure control method based on exponential reaching law. Simulation results show that the control system had fast response speed and a small control error.
Gang et al. (2021) designed a fuzzy neural network controller for the manipulator system. The parameters of the fuzzy neural network controller (FNNC) were optimized by combining the particle swarm optimization algorithm and Back Propagation (BP) algorithm. Simulation results showed that this scheme was very effective in solving the problem of the manipulator.
Yin et al. (2021), designed an adaptive fuzzy sliding mode controller for series manipulators, which used the fuzzy logic system to approach high frequency uncertainties, and a parameter adaptive method was used to update the low frequency uncertainty in real time. Simulation and experiments showed the model excelled in tracking performance and had stronger robustness against large disturbances.
Li et al. (2021), proposed a sliding mode control for a discrete-time robot system based on gain switching to solve the jitter problem in disturbed discrete systems with various unknown uncertainties. Local genetic algorithm and global genetic algorithm are used to solve the possible singular points of the manipulator moving on a given trajectory (Reboucas et al., 2019). The simulation r shows that this method can guarantee the optimal performance of the manipulator with the least trajectory of errors, singular points, and calculation. Zhong et al. (2021) proposed a fuzzy adaptive PID terminal sliding mode controller, which utilized the advantages of PID and a terminal sliding mode control to improve the convergence speed of the control algorithm and reduce steady-state errors.