Zeng-Guang Hou

Neural-Network-Based Adaptive Leader-Following Control for Multiagent Systems With Uncertainties

A neural-network-based adaptive approach is proposed for the leader-following control of multiagent systems. The neural network is used to approximate the agents uncertain dynamics, and the approximation error and external disturbances are counteracted by employing the robust signal. When there is no control input constraint, it can be proved that all the following agents can track the leaders time-varying state with the tracking error as small as desired. Compared with the related work in the literature, the uncertainty in the agents dynamics is taken into account; the leaders state could be time-varying; and the proposed algorithm for each following agent is only dependent on the information of its neighbor agents. Finally, the satisfactory performance of the proposed method is illustrated by simulation examples.

A Mean Square Consensus Protocol for Linear Multi-Agent Systems With Communication Noises and Fixed Topologies

The mean square consensus of linear multi-agent systems with communication noises is studied in this note. Each agent is modeled by a continuous-time linear time-invariant dynamics and the fixed communication topology is described by a digraph. The proposed consensus protocol is composed of two parts: the agents own state feedback and the relative states between agent and its neighbor agents. Due to the existence of communication noises, the relative states cannot be obtained accurately. To attenuate the noise effect, a time-varying gain vector a(t)K is applied to the inaccurate relative states. It is proved that: 1) if the communication topology has a spanning tree and every node has at least one parent node, then the proposed protocol can solve the mean square consensus problem if and only if a(t) satisfies ∫₀^∞a(s)ds = ∞ and ∫₀^∞ a²(s)ds <; ∞; and all roots of the polynomial whose coefficients are the elements of vector K are in the left half complex plane; 2) if the communication topology has a spanning tree and there exists one node without any parent node, then the condition ∫₀^∞ a²(s)ds <; ∞ is only sufficient but not necessary; and 3) if the communication topology has no spanning tree, then the proposed protocol cannot solve the mean square consensus problem.

Recurrent Neural Network for Non-Smooth Convex Optimization Problems With Application to the Identification of Genetic Regulatory Networks

A recurrent neural network is proposed for solving the non-smooth convex optimization problem with the convex inequality and linear equality constraints. Since the objective function and inequality constraints may not be smooth, the Clarkes generalized gradients of the objective function and inequality constraints are employed to describe the dynamics of the proposed neural network. It is proved that the equilibrium point set of the proposed neural network is equivalent to the optimal solution of the original optimization problem by using the Lagrangian saddle-point theorem. Under weak conditions, the proposed neural network is proved to be stable, and the state of the neural network is convergent to one of its equilibrium points. Compared with the existing neural network models for non-smooth optimization problems, the proposed neural network can deal with a larger class of constraints and is not based on the penalty method. Finally, the proposed neural network is used to solve the identification problem of genetic regulatory networks, which can be transformed into a non-smooth convex optimization problem. The simulation results show the satisfactory identification accuracy, which demonstrates the effectiveness and efficiency of the proposed approach.

Sampled-data based average consensus of second-order integral multi-agent systems: Switching topologies and communication noises

A distributed sampled-data based protocol is proposed for the average consensus of second-order integral multi-agent systems under switching topologies and communication noises. Under the proposed protocol, it is proved that sufficient conditions for ensuring mean square average consensus are: the consensus gain satisfies the stochastic approximation type condition and the communication topology graph at each sampling instant is a balanced graph with a spanning tree. Moreover, if the consensus gain takes some particular forms, the proposed protocol can solve the almost sure average consensus problem as well. Compared with the previous work, the distinguished features of this paper lie in that: (1) a sampled-data based stochastic approximation type protocol is proposed for the consensus of second-order integral multi-agent systems; (2) both communication noises and switching topologies are simultaneously considered; and (3) average consensus can be reached not only in the mean square sense but also in the almost sure sense. (c) 2013 Elsevier Ltd. All rights reserved.

Solving a modified consensus problem of linear multi-agent systems

A distributed protocol is proposed for a modified consensus problem of a network of agents that have the same continuous-time linear dynamics. Each agent estimates its own state using its output information and then sends the estimated state to its neighbor agents for the purpose of reaching a consensus. The modified consensus problem requires the group decision value to be a linear function of initial states and initial estimated states of all agents in the network, and the transformation matrix associated with this linear function not to be a zero matrix. It is proved that under the proposed control protocol, the modified consensus problem can be solved if and only if the system matrices of the agents dynamics are stabilizable and detectable, the input matrix is not a zero matrix, and the communication topology graph has a spanning tree. The proposed protocol can also be extended to multi-agent systems where agents are described by discrete-time linear dynamics. The corresponding necessary and sufficient conditions are provided as well.

A Neutral-Type Delayed Projection Neural Network for Solving Nonlinear Variational Inequalities

A neutral-type delayed projection neural network is proposed to deal with nonlinear variational inequalities. Compared with the existing delayed neural networks for linear variational inequalities, the proposed approach apparently has the larger application domain. By the theory of functional differential equation, a delay-dependent sufficient stability condition is derived. This stability condition is easily checked, and can guarantee that the proposed neural network is convergent to the solution of nonlinear variational inequality problem exponentially, which improves the existing stability criteria for the neutral-type delayed neural network. Moreover, many related problems, such as the projection equation and optimization problems, can also be dealt with by the proposed method. Finally, simulation examples are given to illustrate the satisfactory performance of the proposed method.

A Recurrent Neural Network for Hierarchical Control of Interconnected Dynamic Systems

A recurrent neural network for the optimal control of a group of interconnected dynamic systems is presented in this paper. On the basis of decomposition and coordination strategy for interconnected dynamic systems, the proposed neural network has a two-level hierarchical structure: several local optimization subnetworks at the lower level and one coordination subnetwork at the upper level. A goal-coordination method is used to coordinate the interactions between the subsystems. By nesting the dynamic equations of the subsystems into their corresponding local optimization subnetworks, the number of dimensions of the neural network can be reduced significantly. Furthermore, the subnetworks at both the lower and upper levels can work concurrently. Therefore, the computation efficiency, in comparison with the consecutive executions of numerical algorithms on digital computers, is increased dramatically. The proposed method is extended to the case where the control inputs of the subsystems are bounded. The stability analysis shows that the proposed neural network is asymptotically stable. Finally, an example is presented which demonstrates the satisfactory performance of the neural network

Seeking Consensus in Networks of Linear Agents: Communication Noises and Markovian Switching Topologies

The stochastic consensus problem of linear multi-input multi-output (MIMO) multi-agent systems (MASs) with communication noises and Markovian switching topologies is studied in this technical note. The agents full state is first estimated by the state observer, and then the estimated state is exchanged with neighbor agents through a noisy communication environment. The communication topology is randomly switching and the switching law is described by a continuous-time Markovian chain. Then a consensus protocol is proposed for this MAS, and some sufficient conditions are obtained for ensuring the mean square and almost sure consensus. In addition, if the communication topology is fixed, some necessary and sufficient conditions for the mean square consensus can be obtained according to whether or not each agent in the system has parents.

A behavior controller based on spiking neural networks for mobile robots

Spiking neural networks (SNNs), as the third generation of artificial neural networks, have unique advantages and are good candidates for robot controllers. A behavior controller based on a spiking neural network is designed for mobile robots to avoid obstacles using ultrasonic sensory signals. Detailed structure and implementation of the controller are discussed. In the controller the integrated-and-firing model is used and the SNN is trained by the Hebbian learning algorithm. Under the framework of SNNs, fewer neurons are employed in the controller than those of the classical neural networks (NNs). Experimental results show that the proposed controller is effective and is easy to implement.

Neural-Network-Based Nonlinear Model Predictive Control for Piezoelectric Actuators

Piezoelectric actuators (PEAs) have been widely used in nanotechnology due to their characteristics of fast response, large mass ratio, and high stiffness. However, hysteresis, which is an inherent nonlinear property of PEAs, greatly deteriorates the control performance of PEAs. In this paper, a nonlinear model predictive control (NMPC) approach is proposed for the displacement tracking problem of PEAs. First, a “nonlinear autoregressive-moving-average with exogenous inputs” (NARMAX) model of PEAs is implemented by multilayer neural networks; second, the tracking control problem is converted into an optimization problem by the principle of NMPC, and then, it is solved by the Levenberg-Marquardt algorithm. The most distinguished feature of the proposed approach is that the inversion model of hysteresis is no longer a necessity, which avoids the inversion imprecision problem encountered in the widely used inversion-based control algorithms. To verify the effectiveness of the proposed modeling and control methods, experiments are made on a commercial PEA product (P-753.1CD, Physik Instrumente), and comparisons with some existing controllers and a commercial proportional-integral-derivative controller are conducted. Experimental results show that the proposed scheme has satisfactory modeling and control performance.