Global Asymptotic Stability and Stabilization of Neural Networks With General Noise

Output Tracking Control of Random Nonlinear Time-Varying Systems

This paper is concerned with the output tracking control problem for random nonlinear systems with time-varying powers. A distinct feature of this paper is that we consider time-varying powers and the second-order moment process simultaneously, which is more practical in real applications than the existing results where only one factor is considered. We propose a new design scheme, which ensures that the fourth moment of the tracking error can be adjusted to be arbitrarily small and all the states of the closed-loop system are bounded in probability. Finally, a numerical simulation is given to demonstrate the feasibility of the control idea.

Stabilization and destabilization of nonlinear systems via aperiodically intermittent stochastic noises: Average techniques and scalar functions

In this paper, almost sure exponential stabilization and destabilization criteria for nonlinear systems are obtained via aperiodically intermittent stochastic noises based on average techniques and piecewise continuous scalar functions. Compared with existing results on almost sure exponential stability of stochastic systems, the requirement on the upper bound of the diffusion operator of a Lyapunov function is released. The upper bound is allowed to be a scalar function and even be unbounded. Simultaneously, by means of putting forward new concepts "average noise control rate" and "average noise control period," assumptions on infimum of control time and supremum of rest time in the previous references about aperiodically intermittent control can be removed without implementing in the upper limit of the uncontrolled rate, which reduces the conservativeness of stabilization criteria resulting from non-uniform distribution of control time and rest time. In addition, the main results are applied to coupled and uncoupled nonlinear spring-mass-damper oscillator systems, respectively, and corresponding numerical simulations are carried out to demonstrate the validity of the theoretical analysis.

Prioritizing Useful Experience Replay for Heuristic Dynamic Programming-Based Learning Systems

The adaptive dynamic programming controller usually needs a long training period because the data usage efficiency is relatively low by discarding the samples once used. Prioritized experience replay (ER) promotes important experiences and is more efficient in learning the control process. This paper proposes integrating an efficient learning capability of prioritized ER design into heuristic dynamic programming (HDP). First, a one time-step backward state-action pair is used to design the ER tuple and, thus, avoids the model network. Second, a systematic approach is proposed to integrate the ER into both critic and action networks of HDP controller design. The proposed approach is tested for two case studies: a cart-pole balancing task and a triple-link pendulum balancing task. For fair comparison, we set the same initial weight parameters and initial starting states for both traditional HDP and the proposed approach under the same simulation environment. The proposed approach improves the required average number of trials to succeed by 60.56% for cart-pole, and 56.89% for triple-link balancing tasks, in comparison with the traditional HDP approach. Also, we have added results of ER-based HDP for comparison. Moreover, theoretical convergence analysis is presented to guarantee the stability of the proposed control design.

Chaos versus noise as drivers of multistability in neural networks

The multistable behavior of neural networks is actively being studied as a landmark of ongoing cerebral activity, reported in both functional Magnetic Resonance Imaging (fMRI) and electro- or magnetoencephalography recordings. This consists of a continuous jumping between different partially synchronized states in the absence of external stimuli. It is thought to be an important mechanism for dealing with sensory novelty and to allow for efficient coding of information in an ever-changing surrounding environment. Many advances have been made to understand how network topology, connection delays, and noise can contribute to building this dynamic. Little or no attention, however, has been paid to the difference between local chaotic and stochastic influences on the switching between different network states. Using a conductance-based neural model that can have chaotic dynamics, we showed that a network can show multistable dynamics in a certain range of global connectivity strength and under deterministic conditions. In the present work, we characterize the multistable dynamics when the networks are, in addition to chaotic, subject to ion channel stochasticity in the form of multiplicative (channel) or additive (current) noise. We calculate the Functional Connectivity Dynamics matrix by comparing the Functional Connectivity (FC) matrices that describe the pair-wise phase synchronization in a moving window fashion and performing clustering of FCs. Moderate noise can enhance the multistable behavior that is evoked by chaos, resulting in more heterogeneous synchronization patterns, while more intense noise abolishes multistability. In networks composed of nonchaotic nodes, some noise can induce multistability in an otherwise synchronized, nonchaotic network. Finally, we found the same results regardless of the multiplicative or additive nature of noise.