Study Of Synchronization Problems For Multi-axis Motion Control Over Real-time Ethernet

Along with the continuous improvement in modern industrial and commercial systems, thereare increasing demands for applying a shared data network in control systems. For instance, in ap-plications with a large number of sensors and actuators, such as computer numerical control (CNC)machining centers, humanoid robots, printing machines, and automobiles, real-time control net-works can be used to perform information exchanges between a main controller and distributedaxes. Compared with traditional centralized control systems with directly wiring devices together,these network-based systems provide several advantages such as scalability, reliability, and remote-control capability. In addition, they reduce the problems of wiring connection and transmit-lengthlimitation, and decrease installation, reconfiguration and maintenance time and costs. However, thenetwork factors, such as real-time communication and network synchronization, message schedul-ing and network-induced delays, will be the big challenges for the motion control over controlnetworks, especially for the high-accuracy multi-axis coordinated control.This dissertation focuses on the synchronization problems for multi-axis motions in net-worked motion control systems, which come from the applications of real-time Ethernet for ad-vanced CNC systems and shaftless-based printing machines. Considering the field device networksused for information exchanges among sensors, actuators and motion controllers in the manufac-turing field, the synchronization problems of the networked motion control systems can be dividedinto two aspects: time synchronization and motion synchronization. Time synchronization is usedto make all network nodes share a common sense of a time, while the motion synchronization isachieved by synchronizing the motion of each axis with those of others. And we focus on address-ing the following three problems: design and build open simulation and experimental platformsof the networked motion control systems for the algorithm verification; model the entire time syn-chronization process and design time synchronization algorithms to achieve the synchronizationaccuracy on the order of tens of nanoseconds for cascaded real-time Ethernet-based systems andreduce the growth rate of synchronization error, and thus, the maximum number of networkednodes can be correspondingly increased; design decentralized multi-axis synchronous controllerswith considering the message scheduling and the network-induced delays of real-time Ethernet,and prove the asymptotic stability of the global system. The main research work and the novel contributions are listed as follows:First, we design and develop open simulation and experimental platforms to study the prob-lems of the decentralized control in the networked multi-axis motion control system. These plat-forms are designed as research tools, thus simplified the verification of new ideas about the timesynchronization and the decentralized coordinated control. Two case studies are then given toverify the effectiveness of the developed open platforms.Second, in terms of the time synchronization, which is the essential technology for a net-worked motion control system to achieve multi-axis motion synchronization, we propose an IEEE1588-based synchronization strategy. The presented synchronization method adopts the frequencycompensation and peer-to-peer transparent clock mechanisms, and utilizes a real-time synchroniza-tion protocol to embed the time synchronization into the real-time communication cycle. An op-timal proportional-integral (PI) controller is then designed for the frequency compensation mech-anism based on the model and analysis of the time synchronization process. Besides, in real-timenetwork-based systems with long linear paths, the growth rate of time synchronization error is themajor barrier to the scalability of systems even if a transparent clock mechanism is used. In orderto reduce the growth rate of synchronization error due to the quantization error in timestamping, aKalman filter is designed based on a state-variable model, which is built for the PI controller-tunedslave clock. In addition, the quantization effect is analyzed and the variance of quantization error isquantitatively estimated for each slave node. Experiments are performed on the open experimen-tal platform to validate its effectiveness and demonstrate that the peak-to-peak jitter is measuredto be only59.37ns after four hops, which is comparable with the current state of the art of theresearch. And the growth rate of synchronization error can also be significantly reduced by thepresented synchronization method. This indicates that the maximum number of networked nodescan be correspondingly increased. It should be also noted that the experimental results are attrac-tive as the low-cost crystal oscillators (XOs) are used instead of temperature-compensated XOs oroven-controlled XOs.Third, in real-time Ethernet-based systems, together with the different inertias and distur-bances in different axes, the message scheduling and the network-induced delays of real-time net-works are the main factors which cause synchronous problem. We thus design two decentralizedsynchronous controllers for such systems: one is the decentralized position synchronization con-troller, the other is the decentralized trajectory tracking controller. In the proposed controllers,position synchronization error is defined as a subset of all possible pairs of preceding axes due tothe limitations of message scheduling and network bandwidth. And, a motion message estimator isadopted in the synchronous controllers to reduce the effect of network-induced delays. It is proventhat the proposed controllers with the delay compensation can asymptotically stabilize both posi-tion and synchronization errors to zero. Simulations and experiments are performed to validate its effectiveness and demonstrate that it can achieve good position synchronization performance orgood contouring performance for the multi-axis motion over the real-time Ethernet.The high performance time synchronization and motion synchronization algorithms proposedin this dissertation will provide theoretical and technical supports to achieve multi-axis synchro-nization control in real-time Ethernet-based systems.