Deep learning has achieved great success with big data and powerful computing, but its performance is poor under sample constraint, mainly due to the construction of function space (clusters) and the design of algorithms under dataset constraint. Accordingly, a categorical review of deep learning under restricted samples is presented. In addition, according to the current research on the brain, the cognitive process of humankind is categorized in the brain with different regions, and the cognitive functions of each region are also different. Therefore, the training function of each region should also be different. At this point, an idea of deep learning method using functional evolution is proposed, trying to create a network structure composed of multiple functional modules, and the training procedure of the functional module used in this method is studied, aiming to explore the new area of "humanoid learning".

With the continuous development of manipulator technology, the traditional rigid manipulator, such as space manipulator, surgical manipulator and man-machine interactive manipulator, etc. gradually has difficulties in meeting the needs of some new manipulator systems in terms of lightweight, motion flexibility, large space range, etc. More and more researchers pay attention to the research of flexible manipulator and its high-precision motion control. At present, some effective motion control methods have been proposed for flexible manipulator. However, when an actuator of the manipulator fails, it is difficult for conventional control methods to ensure the original control performances and may even lead to system instability. The research on the flexible manipulator motion control under the effect of actuator faults has important theoretical significance and application prospect. Firstly, the existing motion control methods of flexible manipulator are summarized. Then, according to the types of actuator faults, the effect of actuator performance fault, actuator completely damaged fault and sudden actuator fault on the system are analyzed, and the state of art methods that are used to overcome these actuator faults are summarized. Finally, the key problems to be further solved in the motion control of flexible manipulator under the effect of actuator faults are discussed, and three prospect directions summarized, which has reference value for the further research of flexible manipulator motion control.

Adaptive Dynamic Programming (ADP) algorithm, as a research focus in the field of optimal control, has been widely applied in the field of Energy Management System (EMS). ADP is an effective tool for solving optimal control problems in complex nonlinear systems, which can adaptively adjust the control strategy through the input and output data of the system. The research progress of ADP algorithm and its application in EMS are introduced. Then the research status and algorithm implementation of ADP algorithm in discrete EMS and continuous EMS are analyzed. And at last the Real-time Adaptive Dynamic Programming (RT-ADP) algorithm and its feasibility are introduced.

Through purposely employing beneficial nonlinear stiffness and damping with a constructed bioinspired reference model, a novel saturated PD with sliding mode control method, (referred to as saturated PD-SMC method), is designed for active suspension systems under saturated control input constraints. The designed control method has several distinct benefits, including the simple structure of PD control method, strong robustness of SMC method with respect to model uncertainties and external disturbances, without requirements of exact system parameters associated with traditional SMC method, as well as taking input constraints into consideration. In the designed control method, the PD part is employed to stabilize the controlled active suspension system, the SMC part is applied to provide strong robustness, and the saturation functions are introduced to prevent the violation of control input constraints. The corresponding stability analysis is provided by Lyapunov techniques. Several experimental results show that, the designed control method dramatically improves transient performance in comparison with some existing methods, including decreasing control energy by 30.65% and improving ride comfort, which can achieve a satisfactory trade-off between costs and results for suspension system control.

As an important problem in high-speed railway systems, train operation control plays a significant role in minimizing energy consumption of trains and improving the operational efficiency of the railway system. A train operation control scheme based on the symmetric alternating direction method of multipliers is proposed for a single train operating on a railway line with several stations. The optimal train operation control problem is formulated by taking the passenger comfort, energy consumption and train punctuality as the objective, and the train operation dynamics, departure time from stations, velocity and traction force limits as constraints. Within the framework of the symmetric alternating direction multiplier method, the optimization problem is recast as two independent subproblems, and an alternate solving mechanism is introduced to obtain the optimal solution of the original problem. Numerical simulations show that the symmetric alternating direction multiplier method can obtain the optimal train control sequence in a smaller number of iterative steps compared to the alternating direction multiplier method, thus illustrate the effectiveness of the algorithm.

The asynchronous control problem based on time-varying emission probability approach is studied for a class of mode-constrained linear jump systems. Firstly, considering the parameter and structure change of the actual system affected by external environmental factor, the semi-Markov jump system model is introduced to characterize the state change of this class of systems. Considering the mode constraints of the system, the hidden semi-Markov transition rates model and time-varying emission probability model are used to describe the mode changes of the system and the mode relationship between the system and the controller, respectively. Furthermore, an asynchronous static output feedback controller is constructed to ensure the stable operation of the system. In addition, based on Lyapunov stability theory and linear matrix inequality method, the stochastic stability conditions and the existence conditions of the controller for the closed-loop linear jump system are given. Compared with the traditional asynchronous control method based on time-invariant emission probability, the proposed asynchronous static output feedback control strategy can greatly reduce the conservatism. Finally, a numerical example is given to verify the effectiveness and correctness of the proposed asynchronous static output feedback controller.

Recently, unmanned autonomous vehicles (UAVs) have attracted a lot of attention in both military and civilian fields, where the trajectory tracking mission has been a popular research topic. In this paper, a robust Sliding Mode Control (SMC) is proposed for controlling a quadrotor UAV for 3D trajectory tracking in the presence of perturbations and parameter uncertainties. The nonlinear dynamics of a quadrotor with 6-DOF is first established. Then, a sliding mode controller with mass, inertia, and stiffness uncertainties is designed. The 3D tracking effectiveness of the controller is verified by modeling simulations in Matlab Simulink and Universal Mechanism software systems. Finally, further physical verification is done using a Pelican quadrotor platform with perturbations applied to the horizontal and vertical axes to verify its robustness. Both the simulation results and the practical implementation results show that the tracking effect and the robustness of the quadrotor UAV for a given trajectory are satisfactory, confirming the correctness and effectiveness of the proposed SMC control algorithm.

To address the green capacitated vehicle routing problem (GCVRP), a mixed integer programming (MIP) model is established to minimize the total freight cost, and an improved Lagrange relaxation algorithm (ILRA) is proposed to solve the problem. Firstly, the dual problem of the original problem is obtained by Lagrange relaxation technique, and the lower bound of the original problem is obtained by solving the dual problem by subgradient method; Secondly a repair algorithm and a neighborhood search algorithm are designed to obtain the upper bound of the original problem, and then update the multiplier iterative solution; Finally, a simulation experiment is carried out. The experimental results show that 10 tests are carried out on 19 cases of different scales under the same experimental environment. The average gap between the upper and lower bounds of MIP obtained by ILRA is 7.61%, while the average gap obtained by Gurobi solver is 15.47%. Therefore, compared with Gurobi solver, ILRA can obtain high-quality solutions of GCVRP efficiently.

With the development of the industrial internet, the control system is developing from time-triggered to event-triggered for avoiding network congestion. For a class of multi-rate industrial processes, a multi-rate model predictive control method with event trigger mechanism is proposed by combining model predictive control (MPC) with lifting and event-triggered technologies. The proposed method firstly adopts the lifting technology to solve the multi-rate problem, and then employs MPC algorithm to design controller to achieve the setpoint tracking. Furthermore, an event-triggered mechanism is designed to make the controller update only under the condition of trigger mechanism violation, while ensuring the system stability. Experiments have been carried out on an industrial grinding process, showing the effectiveness of the proposed method. The proposed method can save the communication resource and computational load, thereby providing a new method for the design of process industrial controllers in industrial internet framework.

The distributed model-free adaptive control (DMFAC) problem for single-input single-output nonlinear multi-agent systems (MASs) with false data injection attacks is studied. A new distributed dynamic linearization method is proposed to obtain an equivalent distributed compact form dynamic linearization data model for nonlinear MASs. Unlike the existing DMFAC results of MASs that the network topology is used in the controller design, the distributed model-free adaptive controller designed in this research uses the input/output data and the controller parameter does not depend on the eigenvalues of the Laplacian matrix. Simulation examples verify that the proposed distributed model-free adaptive control algorithm can acheive bounded consensus control of MASs in the mean square sense. The algorithm can ensure that the multi-agent system achieve the consensus control objective when it is under attacks.

Aiming at the problem that the existing coordinated control methods of dual permanent magnet synchronous motor can not give consideration to disturbance suppression and coordinated control precision, the coordinated control of dual permanent magnet synchronous motor system with uncertainty and unmatched disturbance is studied. Firstly, the mathematical model of dual motor system is established based on system uncertainty and unmatched disturbance. Secondly, the coordinated control model is obtained by combining cross coupling control with traditional PI control. Thirdly, an integral sliding mode method based on disturbance observer is proposed to design a coordination controller to suppress the unmatched disturbances effectively. Finally, a semi-physical simulation is carried out on a multi-motor experimental platform based on dSPACE. The experimental results show that the proposed method can effectively improve the speed synchronization and torque synchronization performance of the system under startup and load mutation.

Aiming to control the formation tracking and obstacle avoidance system of multi-mobile robots under the changing topology, an obstacle avoidance method based on distance and speed between robots and an artificial potential field method to avoid obstacles are proposed to establish a consistent control formation control protocol. Firstly, the communication topology between robots is established to facilitate the information exchange between robots. At the level of formation control, a formation control law with collision avoidance is designed. Then, at the level of formation tracking, the formation error motion problem is transformed into a minimum optimization problem according to the cost function by using model predictive control method. In order to efficiently solve the optimization problem online, a generalized projection neural network optimization method is used, in which the optimal solution is used as the control input. Finally, the simulation of multi-mobile robot formation verifies the effectiveness of the proposed strategy.

The research on the non-fragile H_∞-consensus control problem for a class of multi-rate multi-agent system under the event-triggered (ET) mechanism is focused on. In order to be more in line with actual need, a multi-rate sampling strategy is adopted, which leads to a multi-rate sampling that can be converted into a single-rate sampling via lifting technique. Considering the transmission burden among the agents, an ET mechanism is introduced to reduce the numbers of transmission among the agents. In addition, in view of the possible inaccuracy of the controller implementation, a controller that can tolerate the changes/fluctuations during the implementation is designed to make the multi-agent system more robust. An observer-based ET non-fragile controller is designed to achieve the H_∞-consensus control of the multi-agent system, in which the controller can tolerate the variations/fluctuations during the implementation. By using the linear matrix inequality technique, the sufficient conditions are obtained that can ensure the H_∞-consensus control of the considered system, and then the controller parameters are designed. Finally, a numerical simulation example is given to prove the effectiveness of the ET control method.

Aiming at the problem that the traditional vector control strategy can not eliminate the system disturbance on the basis of maintaining the dynamic performance of the system, and at the same time, in order to further improve the control accuracy and anti-interference ability of the system, a continuous twisting sliding mode control (CTSMC) method based on extended state observer (ESO) is proposed. Firstly, based on the double closed-loop vector control structure, the continuous twisting control method is adopted in the speed loop. Compared with the traditional control method, this method can converge the error to a certain region, not only improving the robustness of load disturbance, but also reducing the chattering phenomenon. Secondly, based on this controller, an ESO is added to estimate the total disturbance of the system, and the signal is used for the feedforward compensation controller to further improve the dynamic performance of the system. Finally, the closed-loop stability analysis is carried out by using Lyapunov stability theory, and the effectiveness of this method is verified by simulation and experiment.

The development of injection molding machines has always piqued the industry's interest as an important production and manufacturing equipment for plastic parts in modern industry. How to achieve high-precision, high-efficiency, green, and energy-saving injection molding parts is an important development direction for injection molding machines as the aerospace, power electronics, automobile manufacturing, and other industries grow. A strategy that combines the deep neural network (DNN) is proposed to realize the predictive control of the injection molding process of the injection molding machine, in response to the problem that traditional model predictive control (MPC) finds it difficult to guarantee real-time tracking in the injection molding process. A model predictive controller with constraints is created based on the dynamic model of the injection molding machine injection process, and the controller's operating data is collected and used to train the deep neural network to realize the predictive control of the injection molding process based on the deep neural network control. The simulation results show that this strategy can effectively avoid the complex calculation caused by the model predictive control in the injection molding process and meet the real-time requirements and has a broad application prospect.

Based on arbitrary switching rules, an adaptive neural network control scheme with time-varying full state constraints is proposed for a class of nonlinear uncertain stochastic switching systems. In the process of control research, neural network is used to approximate the uncertain items in the system. In order to solve the constraint problem of the system, the coordinate transformation technology is used to ensure that all states of the system are within the constraint boundary, and the sufficient criteria for the stability and convergence of the closed-loop system are given. Finally, the simulation results show that the control strategy proposed in this research can achieve better control effect. The control strategy designed here can greatly improve the security of the system.

A decentralized adaptive event-triggered funnel control for a class of uncertain large-scale nonlinear systems is studied. Firstly, a new adaptive decentralized funnel controller was constructed by using a new funnel control method with barrier Lyapunov function to achieve output tracking for a given transient behavior. Secondly, an auxiliary nonlinear function was introduced to solve the interconnection problem in controller design. At the same time, the command filtering technique was applied to backstepping design to avoid ‘complexity explosion’ in backstepping process. In addition, an event-triggered mechanism was designed to reduce unnecessary transfers between controllers and actuators, thus improving resource efficiency. The results show that the proposed control scheme can ensure that all the signals of the closed-loop system are bounded and the tracking error always evolves in the funnel. Finally, the effectiveness of the control method is verified by a numerical system.