Intelligent Optimization and Scheduling of Networked Control Systems Using Neural Network
Essay Preview: Intelligent Optimization and Scheduling of Networked Control Systems Using Neural Network
Report this essay
[1] Intelligent Optimization and Scheduling of Networked Control Systems Using Neural NetworkFatin I. Telchy, [email protected], Hisham M. Al-Bakry, [email protected] Abstract—This paper presents the use of Neural Networks (NN) for transmission time scheduling for the Networked Control System (NCS) where a network is widely used to connect sensors and actuators to the control systems. The need to respect typical timing constraints of the applications supported in these systems requires suitable scheduling strategies in order to devise an appropriate sequence for transmission of the information produced by the processes using the communication system.The proposed model for NCS scheduling assesses its computational complexity, pointing out the drastic reduction in the time needed to generate a schedule as compared with the algorithmic scheduling solutions. The applied approach allows real-time NCS scheduling and makes it possible for the scheduling table to adapt the changes in process control features. Finally an on-line scheduling strategy is developed based on the neural model which can achieve real-time adaptation of the scheduling table changes in the manufacturing environment. Index Terms— Scheduling, Networked Control System, Neural Network, Rate Monotonic AlgorithmINTRODUCTIONMAJOR advancements over the last decades in wired and wireless communication networks gave rise to the new paradigm of Networked Control Systems (NCS). Within this paradigm, sensing and actuation signals are exchanged among various parts of a single system or among many subsystems via communication networks [1]. With the development of NCS, more and more researchers focus on the scheduling of network to realize the cooperation between network bandwidth requirement and control performance and can improve the Quality of Service (QoS) of network and reduce the chance of collision and congestion in network, then it can reduce the network induced time delay and the rate of data packet loss, so scheduling has great signification on improving the performances of NCS [2]. The most important part of network scheduling issue is how often a plant should be scheduled to transmit the data and with what priority the packet should be sent out regardless how the packet gets to the destination from the source efficiently, and what to do if the route is congested, these problems are up to the routing algorithms and congestion control algorithms [3].The use of the communication network in the feedback control systems (wherein the control loops are closed through a real-time network) makes the analysis and design of NCS complex. Scheduling of the network tasks has to be involved when a set of NCSs are connected to the network which competes for network bandwidth [4].The problem of network scheduling of NCS is finding an optimal/feasible schedule that can minimize a given performance measure. Network scheduling in NCSs is comparable to CPU scheduling in hard real time computing systems, where a set of concurrent CPU tasks are executed on a single CPU with timing constraints. Both cases involve allocating a shared resource to a set of a concurrent tasks; both involve frequent invocations of concurrent tasks, and both tasks have real time constraints and have deadlines to be met. However, in the case of network scheduling in NCS, the shared resource becomes the network instead of the CPU processor, and the execution of a real time task has been replaced by the transmission of a data packet [5].
Many contributions have been accomplished in this field; in Zhang (2001) [6] considered the scheduling of a set of controls system when their feedback control loops are closed through a communication network using Rate Monotonic Scheduling (RMS) algorithm. The optimal scheduling with RMS schedulability constraints with NCS stability constraints had been considered, Branicky et. al. (2002) [4] applied RMS algorithm for optimal scheduling of set of NCSs. They worked on scheduling when a set of NCSs are connected to the network and arbitrating for network bandwidth. They formulated the optimal scheduling problem under both RMS schedulability constraints and NCS-stability constraints using Sequential Quadratic Programming (SQP) optimization algorithm, Lin et. al (2009) [7] worked on co-design of scheduling and control of NCSs. The sampling periods are scheduled for multiple-control loops of NCSs depending on TrueTime toolbox and non-preemptive RMS algorithm. It was found that NCS scheduling enhances the performance of control systems, but also improves the network efficiency, and Jie and Wei-dong (2011) [2] worked on control and scheduling co-design of NCSs by approximate response-time analysis under fixed-priority scheduling to improve the control performance of NCS and enhance utilization rate of network resource. The Proposed IntelligentThe proposed Neural Feedback Scheduler (NFS) technique consist of two intelligent stages: The first stage produces optimal sampling periods by using Feedforward Neural Network (FFNN) named as Neural Network Optimizer (NNO) which replaces traditional optimization algorithm. The second stage of the NFS, schedules the NCS tasks using another Feedforward Neural Network (FFNN) named as Neural Network Scheduler (NNS), which works online and replaces the traditional offline RMS algorithm. This leads to improvement in the overhead (optimize the required time for a task to be completed) and computational complexity.The developed framework of the intelligent NFS is shown in Fig. 1. The highlighted block illustrates the proposed technique which effectively provides high efficiency and low overhead with respect to the convenient applied methods as can be seen in [4, 7, 2].[pic 1] Fig. 1. Neural Feedback Scheduler “NFS”Optimized Sampling PeriodsLiu and Layland [8], showed that RMS is optimal among all fixed priority assignments in the sense that no other fixed priority algorithm can schedule a task set that cannot be scheduled by RMS. Accordingly, RMS has been chosen as scheduling method for NCS, and to be developed to overcome the issues of finding an optimized sampling periods and overhead issue, i.e. develop the system performance by employing an intelligent technique.The performance measure function of the NCS is associated with the control cost function Ji(hi), as function of transmission period (hi), the selection of the performance measure function is crucial in the optimization problem. It directly relates the control cost to the NCS transmission period hi [6].The formulation of the optimization problem is [6]:[pic 2] (1)Subjected to:RMS Algorithm schedulability constraints: