Intelligent Scheduling And Optimization Of Networked Control Systems

A considerable research attention has recently been focused on a class of networked control systems (NCSs) which are typically spatially distributed systems wherein the communication between sensors, actuators, and controllers occurs through a shared resources-limited communication network. However, the insertion of communication network into traditional point-point control system makes the analysis and design of control applications complex.From resource scheduling perspective in this dissertation, networked control systems integrated control and scheduling codesign are reasonably modeled, and stability results and control laws design methods are also presented by some new techniques. An intelligent scheduling algorithm for the priority of network message is proposed. Some different solutions are also studied for optimization to the network resource and control performance. The main contributions are as follows:Firstly, by employing the notion of communication sequence and mixed logical dynamical framework, a networked control system subject to communication constraints is modeled to an integrated control and scheduling of discrete-time linear period system with resource constraints. The schedulability conditions and asymptotic stability results of this closed loop system are discussed in detail. The solution of control law is also given under a certain periodic scheduling of network message.Secondly, for an integrated control and scheduling of resource-constrained system with resource constraints, the optimal solution of discrete-time linear period system for period communication sequence and H-infinity control co-design is depicted by linear matrix inequalities (LMIs). A heuristic search method, namely increasing optimal sequence algorithm, can effectively seek the optimal solution of scheduling scheme and H-infinity controller for its asymptotical stability and r-exponential stability, respectively. The algorithm, in conjunction with the convex optimization of the LMIs, can successfully give a solution to the problem of scheduling and control co-design, and can greatly save the computational cost, simultaneously.Thirdly, considering the error and error difference of system response, a fuzzy feedback scheduler that shares communication network is designed with the bandwidth constraint. A scheduling algorithm of the scheduler, namely fuzzy maximum first (FMF), can schedule dynamically the priority of network message. Four different scheduling algorithms with different stochastic delay series are compared, respectively. The results of simulation highlight that the proposed scheduling algorithm can optimize the performance of control loop and has more flexible than other algorithms in uncertain running conditions.Fourthly, in order to optimize an unavoidable tradeoff between the control performance and the bandwidth consumption, an intelligent control approach to manage bandwidth dynamically, namely fuzzy bandwidth management (FBM), is proposed based on fuzzy logic control technique. The lower and upper bound of the assignable bandwidth, which can guarantee the system's stability, are evaluated in terms of LMIs and the resource constraints, respectively. In addition, the normalizable criterions of quality of control (QoC) and requirement of bandwidth (RoB) are also defined, which can estimate the performance of the whole networked control systems. The proposed approach, traditional fixed bandwidth allocation and optimal bandwidth allocation based on linear programming are also compared under these criterions.Fifthly, considering control performance and network resource, a multi-objective programming with a set of constraints combining expert knowledge expressed as rules is presented in order to maximize control performance and minimize bandwidth consumption. An optimal resource manager based on neural network (NN) as good and robust nonlinear function approximator is employed to provide the optimal solution. The optimal algorithm combining expert knowledge as connotative constraints expressed as rules is used as a teacher to label the data samples for the NN well-trained. Consequently, the NN is employed to allocate requirement of bandwidth for each control loop at runtime. The simulation results show that the proposed optimal strategy is a simple and effective approach in comparison with other three different resource allocation strategies.Finally, due to the variation of the workload and unpredictable open environment, a feedback scheduler based on least squares support vector machines (LSSVM) is designed in order to guarantee the stability of the system. The mechanism and its applications are discussed in detail. The feedback scheduler monitors the network resources, predicates the available utilization for the next period, and adopts interpolated method to calculate the next sampling period from predicative value. Consequently, the system's resources are dynamically allocated by this feedback scheduling mechanism. The results of simulation indicate that the feedback scheduling strategy can guarantee the stability of the system under flexible workload. These results also prove that the proposed strategy is an effective tradeoff method between quality of control and quality of service.