International Journal of Control, Automation and Systems

In this paper, consensus problem for high-order multi-agent systems that are at most critically unstable is researched under directed graph. Under the assumption that the delay appears in the communication and being unknown, a distributed protocol with delayed relative state information is proposed to solve the problem, and the design of the constant control gain is not utilizing the precise information of delay. If the agent dynamics has nonzero poles on the imaginary axis, an allowable delay bound is provided to guarantee consensus by studying the joint effects of agent dynamics, network topology and communication delay; otherwise, consensus is tolerant for any large yet bounded communication delay. Especially, it is shown that the unknown delay in communication is allowed to be time-varying if the network topology is undirected, and in this case the delay bound can be enlarged by improving the synchronizability of the undirected graph. ^Finally, two numerical examples are presented to illustrate the effectiveness of the theoretical result.

A controllable networked system is steered from an initial state to the desired state with the help of a set of driver nodes. The set of driver nodes used to control a networked system may not be unique. There exist multiple set of driver nodes which may be utilized to control the networked system. This is imperative to characterize these sets of driver nodes for identification of an optimal set of driver nodes. This paper presents mathematical formulation and two algorithms using the Region of Attraction (ROA) to identify an optimal set of driver nodes. Optimal set of driver nodes is identified to maximize the stability region. For any practical networked systems, driver node has limited actuating capacity, this paper consider this limitation a priori. The proposed mathematical formulation and algorithms are verified with the help of numerical examples using the MATLAB simulation. The proposed algorithms have been validated by applying on a robotic network.

In this paper, the problem of global stabilization control via partial-state feedback is addressed for a class of uncertain nonholonomic systems with the disturbed virtual control directions. The investigated nonholonomic system has the input-to-state stable (ISS) dynamic uncertainties. The small gain theorem and the changing supply function technique are employed to derive a global stabilization controller. Additionally, we develop a switching control strategy in order to get around the smooth stabilization burden associated with nonholonomic systems. The simulation results illustrate the efficacy of the presented algorithm.

Data matrices are widely used in the automotive, aerospace and computer manufacturing industries. In industry, they are used to identify objects used in process control. In this paper, we focus on detecting data matrices where a camera is configured to see it in a perpendicular direction that is typical in machine vision applications. In this case, the image projection can be modeled as a similarity transform. Data matrices are attached or marked by laser on the surface of objects, and have L-shaped solid lines which act as references for decoding. Under a similarity transform, distances from the center of a data matrix to each side of the L-shape are equal. This symmetric property is used to detect a data matrix, and experimental results show the feasibility of the proposed algorithm.

This paper proposes a three-level framework to detect texts in a single image. First, a salient feature map of text is extracted using a Fully Convolutional Network (FCN) that achieves good performance in semantic segmentation. Label combination using both boxes of word and characters level is proposed to improve the detection of uneven boundaries of text regions. Second, in the feature map of FCN, the text region has a higher probability value than the background region, and the coordinates in the character area are very close to each other. We segment the text area and the background area by using the characteristics of text feature map with Hierarchical Cluster Analysis (HCA). Finally, we applied a Convolutional Neural Networks (CNN) to classify the candidate text area into text and non-text. In this paper, we used CNN which can classify 4 classes in total by separating the background area and three text classes (one character, two characters, three characters or more). The text detection framework proposed in this paper have shown good performance with ICDAR 2015, and high performance especially in Recall criterion, finding more texts than other algorithms.

This paper investigates the aperiodically intermittent adaptive event-triggered control strategy for general linear multi-agent systems. The aperiodically intermittent adaptive event-triggered control inherits the respective advantages of aperiodically intermittent control strategy, event-triggered control strategy and adaptive control strategy, which improves communication efficiency, reduces control update frequency and is closer to the practical situations. Firstly, to reach leader-following consensus and save more control resources, a distributed aperiodically intermittent event-triggered scheme is devised, in which the transmission channels among agents only open if the local event-trigger condition is satisfied in predefined time intervals. Then, in order to get rid of continuous inter-agent communication for monitoring the triggering condition, a more general triggering mechanism is presented, in which discrete-time combinational measurement is adopted instead of using continuous-time tracking error directly. Next, to overcome the unexpected large feedback gains in real applications and appropriately tune the feedback gains, the aperiodically intermittent adaptive event-triggered controller is further devised. With aid of the matrix theory, stability theory of switching systems and Lyapunov function, some sufficient criteria are deduced. Moreover, the analyses of excluding the Zeno behavior are included, and explicit positive lower bounds between any two consecutive time intervals are rigorously guaranteed. Finally, the effectiveness for the designed control strategies is validated by simulations.

H ∞ control is an effective approach to handle model uncertainties. However, when modeling mismatch is large, it tends to be challenging to meet the desired requirements of both stability and performance by only using a single H ∞ controller. This study presents a switching method to enhance the robust stability and performance of H ∞ control by dividing the range of dynamics into multiple uncertain models. The candidate robust controllers are designed by solving a set of linear matrix inequalities for each uncertain model. A structural scheduling logic that selects the most proper controller into closed-loop is proposed. The selected controller can ensure bounded exponentially weighted H ∞ norm of the closed-loop switching systems. This work analyses their robust stability and disturbance attenuation performance via a linear fractional transformation by using the small gain theorem. The effectiveness of this method is validated with a fist-order inertial system with pure time delay.

This paper develops a concurrent learning-based approximate dynamic programming (ADP) algorithm for solving the two-player zero-sum (ZS) game arising in H ∞ control of continuous-time (CT) systems with unknown nonlinear dynamics. First, the H ∞ control is formulated as a ZS game and then, an online algorithm is developed that learns the solution to the Hamilton-Jacobi-Isaacs (HJI) equation without using any knowledge on the system dynamics. This is achieved by using a neural network (NN) identifier to approximate the uncertain system dynamics. The algorithm is implemented on actor-critic-disturbance NN structure along with the NN identifier to approximate the optimal value function and the corresponding Nash solution of the game. All NNs are tuned at the same time. By using the idea of concurrent learning the need to check for the persistency of excitation condition is relaxed to simplified condition. The stability of the overall system is guaranteed and the convergence to the Nash solution of the game is shown. Simulation results show the effectiveness of the algorithm.

A self-learning optimal control algorithm for episodic fixed-horizon manufacturing processes with time-discrete control actions is proposed and evaluated on a simulated deep drawing process. The control model is built during consecutive process executions under optimal control via reinforcement learning, using the measured product quality as a reward after each process execution. Prior model formulation, which is required by algorithms from model predictive control and approximate dynamic programming, is therefore obsolete. This avoids several difficulties namely in system identification, accurate modeling, and runtime complexity, that arise when dealing with processes subject to nonlinear dynamics and stochastic influences. Instead of using the pre-created process and observation models, value-function-based reinforcement learning algorithms build functions of expected future reward, which are used to derive optimal process control decisions. The expectation functions are learned online, by interacting with the process. The proposed algorithm takes stochastic variations of the process conditions into account and is able to cope with partial observability. A Q-learning-based method for adaptive optimal control of partially observable episodic fixed-horizon manufacturing processes is developed and studied. The resulting algorithm is instantiated and evaluated by applying it to a simulated stochastic optimal control problem in metal sheet deep drawing.