Study On PPF-based Real-time Signal Network Transmission Technology And Its Applications

With the development of electronic communication technology and computer network technology, some research hotspots have appeared, such as networked control systems (NCSs) and sensor network. The signals that transmitted in NCSs or wireless sensor network are almost real-time signals sampled by sensors, so their time performance is of utmost importance to users. With the development of NCSs, the problem of real-time signal network transmission is becoming more and more important. Especially for control systems based on Internet and wireless network, network transmission uncertainties, such as delay, dropout and disorder of data frames, data incompleteness and asynchronism caused by single-packet or multi-packets transfer and network congestion, not only have a great impact on the real-time performance, reliability and stability of embedded network systems, but also make their analysis and design become fairly difficult. Transmission uncertainties are unavoidable. If the real-time performance is improved unilaterally, the reliability will be lowered and vice versa. The problem that must be solved for the development of NCSs is how to balance between real-time performance and reliability of signal network transmission.In the research field of real-time signal network transmission, most researchers have focused on the network transmission of audio/video signals, while studies on sensor signals or measurement signals are less. From the point of signal characteristics, sensor signals and audio/video signals are quite different, and their usage is also widely discrepant. The main contents in this dissertation can be summarized as follows:(1)Real-time signal network transmission technology based on polynomial predictive filtering (PPF)Combined with the study on characteristics of real-time sensor signal, a method of applying PPF theory in real-time signal network transmission is presented. Effects resulting from transmission uncertainties are alleviated by prediction, and the real-time performance and reliability of signal network transmission is improved. The basic ideas are presented as follows. Transmitters may only send samples which are necessary for reconstruction of the original signal, and bandwidth occupancy is reduced. At receivers, a polynomial predictive filter is adopted to restore data which are not sent, delayed and lost. The original signal is reconstructed by these restored data. Users can re-sample the reconstructed signal to gain estimation of signal at their expected time. During the whole transmission, in order to reduce those invalid transmissions caused by bit errors, channel coding is used to encode samples; in order to reduce the influence caused by dropouts, some redundancy data are sent. With signal reconstruction, remote signals are just like coming from virtual local sensors. The transmission and usage are separated, and errors resulting from transmission uncertainties and prediction are processed together as the output errors of virtual local sensor. From the user's perspective, the network is transparent. Users can make use of these signals without considering complex network characteristics directly and focus on output errors of virtual local sensor only.(2)PPF-based NCS modeling methodThe application of PPF theory is discussed in this dissertation. A new NCS modeling method which combines quering method with signal prediction is presented. First, aiming at smoothness of real-time signals in NCSs, PPF theory can be applied in NCSs, and real-time signals transmitted in NCSs can be handled from the perspective of signal processing. Effects on sensor signals and control signals resulting from transmission uncertainties are compensated. The basic ideas are presented as follows. Two buffer stacks are set up at the side of controller and actuator, which are used to store the predicted values of unreached samples and the control values or sensor samples received formerly. Based on the data in these stacks, controller or actuator can predict the future value of signal and use this value as input directly. Then, the necessary and sufficient gradual stability condition of NCS is given based on augmented state vector method. At last, compared with queuing method, some simulation instances are taken to validate the NCS model presented in this dissertation. The results have showed that the robust of NCS against network influence can be improved by using this method. By using the modeling method presented in this dissertation, the design methods for deterministic systems can be used in NCSs without suffering the disadvantages of queuing method.The main innovation points in this dissertation can be summarized as follows: (1) A new real-time signal network transmission model is presented, and data reliability can be improved in the premise of ensuring the real-time performance of transmitted signals.â‘ The real-time signal network transmission is studied from the perspective of signal processing, and a method is presented which combines PPF theory with channel coding. Real-time sensor signals are transmitted with'compression'and encoding, and samples are restored and used to reconstructe original signals. On one hand, effects resulting from transmission uncertainties are alleviated by prediction, and the reliability can be improved under the premise of meeting the real-time requirement; On the other hand, network load can be reduced, and bandwidth can be utilized more efficiently.â‘¡By constructing a virtual local sensor, transmission uncertainties are transformed into uncertainties of system model consequently, and the errors resulting from transmission uncertainties and prediction are processed together as the output errors of virtual local sensor.â‘¢The original signal is reconstructed, and users can choose different sampling period and re-sample the reconstructed signal to abtain predicted sample values at the desirable time.(2) Real-time signals transmitted in NCSs can be compensated by using PPF theory, and the dynamic performance of control system can be ensured to a certain extent. The method presented in this paper can use maximum information received through network, and transmission utilization and network capacity can be improved. PPF-based NCSs can utilize disordered samples or samples with long delay to predict the future value of signal. In former studies, these samples are discarded and considered to be invalid as dropouts, even though they are received later. PPF-based modeling method of NCSs is more universal, developers can focus on control systems themselves without considering the different network characteristics.(3) The multi-frequency sampling problem between controller and actuator can be solved by signal reconstruction and re-sampling.