Abstract: This paper introduces the basic principles of Autonomous Distributed Systems (ADS) and analyzes the shortcomings of traditional C/S-based integrated subway monitoring systems. Through the design and analysis of a station-level integrated monitoring system based on ADS technology, it demonstrates that ADS technology possesses excellent real-time performance, reliability, fault tolerance, online scalability, and online maintenance characteristics in subway integrated monitoring systems, providing guidance for the future development of subway integrated monitoring systems. Keywords: C/S, ADS, integrated monitoring system Application Research on Subway Supervisory Control System with ADS Abstract: This paper introduces the basic principle of Autonomous Decentralized System and analyzes the defects of traditional subway supervisory control system based on Client/Server technology. Utilizing the characteristics of ADS technology, a new station control system is designed and analyzed, which demonstrates good real-time performance, reliability, compatibility, expansion, and maintenance online. It provides guidance for the development of future subway monitoring and control systems. Keywords: C/S, ADS, supervisory control system 0 Introduction With the development of the social economy, the market demands increasingly higher requirements for subway integrated monitoring systems, specifically in terms of larger amounts of transmitted information, faster data processing speeds, and stronger real-time performance. The metro monitoring system is a main control system interconnected with various subsystems of the metro system[1]: integrating substation automation system (PSCADA), fire alarm system (FAS), electromechanical equipment monitoring system (EMCS), platform screen door system (PSD), floodproof door system (FG), and interconnecting with broadcast system (PA), closed-circuit television system (CCTV), vehicle information system (TIS), station information system (SIS), automatic control system (ACS), automatic fare collection system (AFC), signal system (SIG) and clock system (CLK) to realize resource information sharing, centralized management and maintenance of equipment and monitoring of subsystem faults. However, the current communication uses a master/slave structure, the main control machine has a heavy workload and cannot meet the requirements of dynamic changes and expansion of the system. Moreover, the application layer protocols between C/S developed by different manufacturers are not the same, and it is difficult to integrate software and hardware. In addition, the C/S structure has the problem that the system load is concentrated on the server. As the system scale expands and the amount of information increases, the server access will inevitably surge, increasing the server burden and affecting the monitoring process. 1 Introduction to ADS 1.1 Concept. Autonomous Decentralized System (ADS) [2] is a new system concept that has gradually developed in recent years. It breaks through the original centralized distributed C/S model and establishes a brand-new system model. Autonomous Decentralized System assumes that faults in the system are normal phenomena and that the system is composed of subsystems. In this system, all units (subsystems) are independent and equal, and there is no subordinate relationship between them. Each unit can complete its own task independently without interference from other units. At the same time, each unit can also coordinate its work to realize the operation of the whole system. The network monitoring diagram based on ADS is shown below: Figure 1 The network system structure based on ADS can be replaced by a ring network in addition to the bus network. Autonomous Decentralized System has two major characteristics: (1) Self-controllability. That is, if any subsystem in the system fails, is under maintenance or has just been added, it will not affect the self-management and operation of other subsystems. (2) Self-coordination. That is, if any subsystem in the system fails, is under maintenance or has just been added, other subsystems can coordinate their respective tasks and operate in a cooperative manner to realize their respective functions. These two characteristics ensure the system's online expansion, online maintenance, and fault tolerance. Therefore, each subsystem is required to be able to "intelligently" manage itself without interfering with the affairs of other subsystems and without being interfered with by other subsystems, but it can also coordinate with other subsystems. The key concept for realizing this system model is the communication method of data domain and broadcast. Each unit in the system actively broadcasts its internal processing information to the data domain and receives information from the data domain according to its own needs. Each subsystem only interacts with the data domain, and there is no direct coupling relationship between them. This better ensures online expansion, online maintenance, and fault tolerance. 1.2 Characteristics. The realization of self-regulation and self-regulation and coordination requires each subsystem to meet the following: (1) Equality. Each subsystem can manage itself and cannot be managed by other systems. There is no master-slave relationship between subsystems. This is consistent with the characteristics of the subway integrated monitoring subsystem. The various subsystems of the subway are relatively independent, such as the broadcast system, closed-circuit television system, and vehicle information system. (2) Locality. Each subsystem can manage itself and coordinate with other subsystems by relying only on local information. The metro integrated monitoring system itself does not interfere with the internal affairs of each subsystem. Each subsystem independently completes its own information transmission and processing functions (such as automation system, fire alarm system, electromechanical equipment monitoring system, etc.), and then transmits the monitoring information to the central server. (3) Self-sufficiency. The functions of each subsystem in managing itself and coordinating others are self-sufficient. That is, each system can independently handle its own affairs and coordinate with other subsystems to achieve resource sharing and information integration. The above three characteristics indicate that even if other subsystems fail or terminate communication with a certain subsystem, the subsystem can still work. The metro integrated monitoring system has the characteristics of equality, locality, and self-sufficiency of ADS, thus making ADS adaptable to the metro integrated monitoring system. 1.3 Model of the self-disciplined decentralized system The most basic ADS system consists of atomic nodes (Atom) and data fields (Data Field). The physical entities that atomic nodes can correspond to are computers, intelligent devices, or other hardware. The data field is the space for information transmission in ADS. From a physical perspective, it is equivalent to a network or memory. Information emitted by each atom circulates within the data domain, while the atom also retrieves information from the data domain. There are no direct connections between atoms; they are only interested in the content of the information in the data domain and do not need to know where the information comes from. Therefore, information flowing in the data domain contains a content code to identify its attributes, and each atom determines whether it needs this information by recognizing the content code. Figure 2 shows the ADS framework structure 1.3.1 Data Domain. In an autonomous distributed system, all atomic nodes are autonomous units, and their connections are only realized through the data domain. All data is broadcast to the data domain and circulates within it. The data domain is equivalent to a communication network or memory; nodes actively send information to the data domain and simultaneously retrieve information from the data domain as needed to complete the functions of their internal modules. The portion of the data domain extending into the atom is called the Atom Data Field, where the data required by the atom's system or application modules flows. All information carries a content code that specifies its attributes. In network communication, content codes are equivalent to message identifiers (such as sequence numbers, flow identifiers, etc.). Computer terminals filter and receive information based on the content code table according to their needs. 1.3.2 Atomic Nodes. Each atomic node is an autonomous subsystem that can select the required data information according to a built-in content code table. The selected data information then flows in the internal data field (ADF). All system and application modules in an atomic node use the same mechanism: once all the data required by a module has arrived, the system or application software will automatically start execution; this is called a data-driven mechanism. No module can control other modules or instruct other modules to receive and process data. There is only a loose coupling relationship between the modules; there is no control or being controlled relationship, and each module can independently determine and manage its own behavior. For a specific network communication entity, an atomic node is equivalent to a computer processing terminal. Each terminal can handle its own internal affairs. After receiving a message or data frame, the terminal processes it accordingly according to the driving mechanism. As the structure of the system changes with its expansion, shrinkage or partial failure, in order to ensure that the operation of the subsystem is not disturbed, each subsystem sends information with CC in a broadcast manner. Since it does not have a destination address, the receiver can only select information based on CC and does not know the sender. This communication method based on content code ensures the autonomous information transmission and autonomous information reception of each subsystem. That is, each subsystem does not need to know the relationship between the information source and destination, thus realizing the locality of each subsystem. 2 Composition of the subway monitoring system The subway integrated monitoring system [3] connects the central monitoring system, station-level monitoring system and depot monitoring system into an organic whole through the communication backbone network. The communication backbone network is a wide area network, which is the backbone connecting the station-level monitoring network and the central monitoring network. The station-level system and the depot system are located in the station depot and the depot, respectively. The architecture of the subway monitoring network is shown in Figure 3: Figure 3 Subway monitoring network architecture The central monitoring system is located in the central monitoring center (Operating Control Center, OCC). It is based on a 64-bit UNIX redundant real-time server. This layer of the system includes redundant real-time data servers, historical servers, operator workstations, front-end processors (FEPs), peripheral devices, and a central monitoring network. The metro integrated monitoring system establishes a central monitoring network at the central monitoring center, the core of which is a redundantly configured Ethernet switch. Station-level monitoring systems and depot monitoring systems are located at the stations and depots, respectively. Based on a 64-bit UNIX redundant real-time server, it includes front-end processors, station/depot operator workstations and peripheral devices, and station/depot TCP/PI LANs. 3. Traditional C/S Communication Method Traditional metro monitoring networks are based on a C/S communication method. All integrated and interconnected system data is uniformly accessed through the front-end processor of the integrated monitoring system. The front-end processor is responsible for periodically accessing and converting protocols with connected systems, converting real-time data of different formats into the internal data object format of the metro integrated monitoring system, and submitting it to the system's depot, station-level, and central real-time servers. This can easily create a communication bottleneck for the front-end processor. As the system expands, the real-time performance of information transmission will be affected. The following analysis uses a subway central monitoring network example (Figure 4): This central monitoring system implements central monitoring functions such as the Automatic Fire Alarm System (FAS), Supervisory Control and Data Acquisition (SCADA), Electromechanical Equipment Monitoring System (EMCS), Platform Screen Door System (PSD), and Floodproof Door System (FG). It also enables interconnection and control of the onboard information display system, station information display system, signaling system, broadcasting system, communication system, and wired telephone system. It adopts an industrial-grade 100M redundant ring-type switched Ethernet and TCP/IP protocol, using a client/server (C/S) access method. The central monitoring system is equipped with redundant central real-time and historical servers to support comprehensive system operation; it includes a general dispatch workstation, power dispatch workstation, train dispatch workstation, engineer maintenance and system management workstation, environmental control dispatch workstation, central ticketing center, management center, maintenance workstation, and website management workstation. It also includes dual-screen workstations as operator monitoring workstations, serving as backups for each other. Due to the use of a C/S access method, it has the following disadvantages: 1) Weak scalability. With the continuous development and improvement of the subway system, there is an increasing need for a large-scale wide-area monitoring network to achieve system integration and interconnection. C/S remote access requires specialized technology, and the system needs specialized design to handle distributed data. Changes in business operations require redesign and redevelopment, increasing the difficulty of maintenance and management, and limiting system scalability. 2) Clients need to install dedicated client software. Firstly, there is the workload of installation; secondly, any computer malfunction, such as a virus or hardware failure, requires installation or maintenance. Furthermore, system software upgrades require reinstallation on each client machine, resulting in very high maintenance and upgrade costs. Figure 4 shows the configuration of a central-level subway monitoring network. 3) Operating system limitations. There are generally limitations on the client's operating system. It may be compatible with Win98, but not with Win2000 or Windows XP. Or it may not be compatible with newer Microsoft operating systems, not to mention Linux, Unix, etc. 4) Real-time performance. As the monitoring network expands, transmission latency will inevitably increase, and congestion may occur when numerous pieces of information pass through the central router, affecting the transmission of fault diagnosis information (such as PSCADA) with strict real-time requirements, and in severe cases, causing data packet loss. 5) Load Concentration. The load of a C/S-based service system is overly concentrated on the server. As the system scales up, the amount of information increases, inevitably increasing the server load (as shown in the figure, FEP, central server, etc.). In severe cases, a surge in access volume during a certain period can cause the server to be unable to respond, affecting system reliability. 6) Fault Tolerance. The dual-machine redundancy technology shown in the figure is fundamentally a fault-prevention technology (such as dual FEP redundancy, dual router redundancy, dual printer redundancy, etc.). In practical applications, it suffers from high cost and low reliability. 4 ADS Communication Method To meet the needs of the subway network monitoring system, it is necessary to overcome the drawbacks of the C/S model's centralized server processing. Based on the characteristics of ADS technology itself, a peer-to-peer architecture can be adopted for the subway monitoring system. Simultaneously, a publish/subscribe communication model is implemented. Figure 5 shows a schematic diagram of a station-level ADS monitoring system. The station-level ADS monitoring system interconnects monitoring systems such as AFC, PSD, ACS, CLK, PA, SIG, CCTV, SCADA, TIS, EMCS, SCI, and FAS, as well as other auxiliary management and maintenance equipment. In this design, each subsystem is equivalent to an atomic node. Each node can manage itself but cannot be managed by other nodes, and there is no master-slave relationship between atomic nodes. 4.1 Reliability. In traditional systems, if the control center fails, the entire system will be paralyzed. The station server, as a special atomic node, retains the control center function but does not employ redundancy. If it fails, the system will automatically use an election competition method to elect a new control center (such as a SCADA monitoring system, TIS monitoring system, etc.) among the normal atomic nodes until the original control center recovers and transfers control. This greatly improves the flexibility and reliability of traditional systems. The same design concept is still adopted within each subsystem and the backbone network, which will not be elaborated further. 4.2 Real-time Performance. In the station-level ADS monitoring system, information producers (such as EMCS atomic nodes, FAS atomic nodes, etc.) publish information instantly without specifying a particular recipient. Nodes that need this information (such as the station server) can selectively receive data, thereby better ensuring the real-time performance of the system. The server does not need to periodically access each atomic node; it only needs to retrieve the data of interest from the data domain according to a certain algorithm, alleviating access pressure. For data that no atomic node receives, the system will automatically discard it. 4.3 Fault Tolerance. The ADS monitoring system architecture and its data-driven mechanism allow software modules to execute asynchronously. The number of copies of each software is determined by its importance in the application. Therefore, multiple replicated software modules are placed in different subsystems of the monitoring system, and all these identical software modules run independently and receive and send information independently. Some faulty modules may send incorrect information, while other normal modules send correct information. The software using this information can select the correct information for its own use through content codes ("event sequence number mechanism") and the number of events that occurred ("voting mechanism"). This approach effectively solves the problem of using backup fault tolerance technology in the past, namely, the inability to use backup when the switching device itself fails. When the ADS system handles fault tolerance issues, faulty software can run simultaneously with normal software, and the decision of whose information to use depends entirely on the receiver. In this sense, it achieves true fault tolerance. 4.4 Online Expansion. Online system expansion includes expansion at both the system and node levels. For system-level expansion, there are two types. Homogeneous system expansion simply requires merging its data domains, while heterogeneous system expansion requires a gateway. This example does not involve this, so it will be briefly introduced. For the expansion of internal software modules of a node, it only needs to register the new content code in the internal content code table of the system without reporting to other nodes. For example, online modifications to an access control system will not affect the operation of other nodes, nor will they affect the execution of other software modules within the node. 4.5 Online Maintenance. The autonomous distributed system can perform tests on some subsystems or their internal software modules while running. The information sent by the running nodes is divided into two categories: online information and test information. Correspondingly, the running nodes are also divided into online and test states. It is stipulated that online nodes only receive online information, while test nodes receive both types of information. Therefore, nodes in the test state will not interfere with the normal operation of online nodes, and at the same time, they can receive online information for individual testing or receive test information for joint testing. Taking advantage of this, subsystems can detect each other's online status (i.e., fault) and then report the fault status to the server. The above five advantages make the monitoring system built based on the ADS concept very suitable for the architecture of subway monitoring systems to meet the needs of dynamic system expansion. The construction of the subway monitoring system can be completed step by step as each department is involved. The corresponding control subsystems can operate autonomously and in a coordinated manner, depending on the number of departments involved. As the system scales up, subsystems can be expanded online without interfering with existing operational processes. Once a newly added subsystem is successfully debugged, it can be seamlessly integrated with the existing system to form a larger system, jointly achieving all tasks of the entire subway system. 6. Conclusion The continuous expansion of modern industrial production and the rapid development of computer technology have placed new demands on subway monitoring systems. Traditional subway monitoring systems can no longer meet certain user requirements. The autonomous distributed system, with its unique data domain structure and data-driven mechanism, achieves self-regulation, controllability, and coordination of its subsystems, giving the entire system real-time performance, reliability, fault tolerance, online expansion, and online maintenance capabilities. These characteristics meet the ever-evolving requirements of monitoring systems. References: [1] LIU De-qiang. The research and coming true of subway main control system[J]. Electric Power Automatization Equipment, 2004, 24(9): 45-48. [2] Mori K. Autonomous decentralized systems: Concept, Data field Architecture and Future trends. In: Proceedings of Autonomous Decentralized Systems, ISADS'93. Los Alamitos (CA): IEEE Computer Society, 1993: 28-34. [3] WU Chao. Research of reliability evaluation method of subway main control system[D]. Southwest Jiaotong University excellence Master Degree Thesis, 2007: 6-8. About the author: Yang Jing (1983-), male, from Yichang, Hubei Province, holds a master's degree and his research focuses on the reliability of integrated monitoring systems for rail transit.