Introduction to Parallel Blockchain

The concept of parallel blockchain was first proposed at the first International Symposium on Blockchain and Knowledge Automation held at the University of Denver in the United States from April 2 to 3, 2017. ISBKA 2017 was jointly sponsored or co-sponsored by 12 domestic and foreign academic organizations, including the Association for the Advancement of Intelligent Industries, Association for Intelligent Science and Technologies, the Chinese Automation Society, the Chinese Artificial Intelligence Society, and the Qingdao Intelligent Industry Technology Research Institute, and hosted by the University of Denver. At ISBKA 2017, more than 50 experts from China, the United States, Canada, Australia and other countries and regions discussed hot issues and technological progress in the field of blockchain and knowledge automation. Professor Wang Feiyue from the Institute of Automation, Chinese Academy of Sciences, gave a keynote speech on "Parallel Blockchain: Concept, Techniques and Applications", and for the first time proposed and interpreted the concept and technology of parallel blockchain and its preliminary application practice in finance, transportation, health and agriculture. In October 2017, Associate Researcher Yuan Yong and Professor Wang Feiyue from the Institute of Automation, Chinese Academy of Sciences, published the first academic paper in the field of parallel blockchain in China, "Parallel Blockchain: Concept, Method and Connotation Analysis".

Definition: Parallel blockchain technology is an organic combination of parallel intelligence theory and blockchain technology. It is committed to adding computational experiments and parallel decision-making functions to the current blockchain technology through the parallel interaction and co-evolution of actual blockchain systems and artificial blockchain systems, and realizing blockchain system management and decision-making that combines description, prediction, and guidance.

Parallel blockchain is a theoretical method that effectively solves problems related to blockchain modeling, experimentation, and decision-making. It is a deep combination of parallel intelligence, an original research paradigm proposed at the beginning of this century, and emerging blockchain technology.

Specifically, parallel blockchain is based on parallel intelligence theory and ACP method (Artificial systems + Computational experiments + Parallel execution). Its basic idea is to construct an artificial blockchain system by formally describing the static characteristics and dynamic behaviors of the core elements of the blockchain ecosystem (such as computing nodes, communication networks, consensus algorithms, incentive mechanisms, etc.), use computational experiments to conduct trial and error experiments and optimization for specific blockchain application scenarios, and realize decision optimization and parallel tuning through virtual-real interaction and closed-loop feedback between artificial blockchain systems and actual blockchain systems. In essence, the parallel blockchain system uses the artificial blockchain system as a "computing laboratory" and uses the offline trial and error experiment and rational deliberation of "responding to changes with changes" in the artificial blockchain system under normal conditions to achieve real-time management and decision-making of the real blockchain system under abnormal conditions.

Background introduction: Blockchain technology originated from "Bitcoin" in late 2008. A new digital cryptocurrency designed by a scholar with the pseudonym "Satoshi Nakamoto" is the underlying support technology of Bitcoin.

Blockchain technology has many advantages that other technologies cannot match: first, the fundamental feature of the blockchain system is decentralization; second, the blockchain system forms a consensus between nodes through mathematical algorithms. New data must be verified by all or most nodes before it can be written into the blockchain ledger maintained by all nodes, so it is extremely difficult to tamper with and forge; third, the blockchain system adopts an open and transparent data reading method based on privacy protection. The blockchain ledger data is publicly queried to all nodes at zero cost, thereby reducing the trust cost of nodes and system uncertainty. These significant advantages have important and broad application prospects in modern social systems. However, as an emerging technology, blockchain-related theoretical research and industrial practice are still in the initial stage. Core technical elements at the micro level, such as consensus algorithms, network structures, smart contracts, and incentive mechanisms, are still in the state of exploration, experimentation, and continuous optimization. The macro-level blockchain industry ecology and its impact on social economy also urgently need experiments, analysis, evaluation, and necessary supervision.

From the perspective of academic research, the existing blockchain technology is essentially still a new type of chain data structure and distributed computing architecture, which can effectively realize the descriptive modeling and calculation of complex social, economic, and financial systems, but lacks the ability to conduct computational experiments and predictive analysis of blockchain systems under different configuration conditions and various application scenarios. At the same time, it also lacks the ability to guide and make decisions by combining virtuality and reality, guiding reality with virtuality, and guiding reality with artificiality. This is the fundamental reason why blockchain technology can only rely on incremental trial and error experiments on the "chain" of real systems, or empirical decision-making methods such as sandbox supervision to achieve the development of blockchain technology and optimization of industrial ecology. To solve this problem, there is an urgent need to develop a set of new theories and methods for blockchain modeling, experiments and decision-making, aiming to provide blockchain technology and related industries with a set of computable, feasible and comparable descriptive modeling, predictive analysis and decision-making guidance methods.

At present, the theory of parallel intelligence has achieved significant practical benefits and preliminary theoretical results in more than ten typical application fields, such as national defense security, parallel transportation, parallel economy, parallel control, parallel vision, parallel image and parallel data. Parallel intelligence research is mainly aimed at social-physical information or human-machine-object ternary systems (Cyber-physical-social systems, CPSS) with "people in the loop" and high social complexity and engineering complexity. By studying data-driven descriptive intelligence, experimental-driven predictive intelligence, and interactive feedback-guided intelligence, it provides agile, focused and convergent solutions for uncertain, diverse and complex problems. At the first International Symposium on Blockchain and Knowledge Automation held by the University of Denver in April 2017, Professor Wang Feiyue of the Institute of Automation, Chinese Academy of Sciences, delivered a keynote report on "Parallel Blockchain: Concept, Techniques and Applications", proposing and interpreting the concept and technology of parallel blockchain for the first time and its preliminary application practice in finance, transportation, health and agriculture.

Conceptual framework: Generally speaking, blockchain can be narrowly defined as a decentralized shared ledger that combines data blocks into a specific data structure in a chain in chronological order and cryptographically guarantees that it cannot be tampered with or forged. This ledger can safely store simple, sequential, and verifiable data within the system. Correspondingly, the broad blockchain is a new technology framework composed of data links, communication networks, consensus algorithms, incentive mechanisms, smart contracts, and application scenarios, as well as the emerging industries and ecosystems derived from them. This new technology framework can use encrypted chain block structures to verify and store data, use distributed node consensus algorithms to generate and update data, and use automated script codes (smart contracts) to program and operate data. It is a new decentralized infrastructure and distributed computing paradigm.

It can be seen that the block "chain" itself is only an objective carrier and manifestation of data access. What is more essential and complex is the complex ecosystem behind the blockchain, which is composed of deep coupling and mutual feedback between nodes and elements. Therefore, parallel blockchain is not a simple superposition and interaction of multiple independent blockchains, but a new research framework that realizes blockchain system modeling, prediction and guidance in a "human-machine integration, virtual-real integration" way, through the co-evolution and parallel feedback of artificial blockchain system and actual blockchain system.

The conceptual framework of parallel blockchain is shown in Figure 1. Its core idea is to realize the modeling, experimentation and decision-making of blockchain system based on ACP method, that is: using artificial system (A) method to establish an artificial blockchain system corresponding to the actual blockchain system and capable of reflecting the state and evolution law of the actual system; using computational experiment (C) method, experiment, analyze and evaluate the actual blockchain system in the artificial system, so as to master the evolution law of the actual blockchain system in various possible scenarios; using parallel execution (P) method, through the parallel execution and co-evolution of artificial system and actual system, the actual blockchain system can be managed and controlled.

Specifically, parallel blockchain technology comprehensively considers various complex factors of physical, network and social ternary space, and adopts the method of organic combination of theoretical modeling, empirical modeling and data modeling to establish one or more artificial blockchain systems "associated" with the actual blockchain system. The problems caused by the lack of effective modeling, experimentation and evaluation methods in the actual blockchain system can be solved by building corresponding experimental scenarios in the artificial blockchain system. By accurately modeling the characteristics and behaviors of individual blockchain systems (such as miner nodes or transaction nodes), a large number of computational experiments can be implemented in a bottom-up emergent manner to simulate and "live broadcast" the various states and development characteristics of the blockchain system, thereby assisting in reasoning and predicting the evolution laws and interaction relationships of the core elements of the actual blockchain system under normal and abnormal conditions; the actual blockchain system co-evolves with the artificial blockchain system throughout its life cycle, and the two are connected to each other through specific parallel interaction mechanisms and protocols. On the basis of real-time synchronization of data, models, scenarios and decision-making elements, through the "What-if" scenario deduction and experimental exploration in the artificial system, the "estimation" of their respective future states and their mutual "reference" can be achieved, and their respective control and management methods can be adjusted accordingly.

Advantages: The core advantage of parallel blockchains is that they can effectively realize the learning and training, experimentation and evaluation, and management and control of blockchain systems.

(1) Learning and training: The emerging blockchain technology has generated huge market cultivation and technical training needs. Generally speaking, as learners become more familiar with and master blockchain technology, they will inevitably experience an evolutionary process from offline to online, and from off-chain to on-chain. On the one hand, online operations on the chain may bring security risks to the real blockchain system, and on the other hand, they may also generate actual costs due to the execution of specific operations (such as executing on-chain code). Parallel blockchains can realize scenario-based or even gamified learning and training processes in a safe, flexible and low-cost manner. Based on the real blockchain system, parallel blockchains can instantiate one or more artificial blockchain systems according to specific learning objectives. Through appropriate connection and combination of artificial and actual systems, learners can quickly master the various operations of the blockchain system and their possible effects in the artificial system, and quantify the actual effects of learning and training.

(2) Experimentation and evaluation: The real blockchain system is usually unable to conduct certain important destructive experiments and innovative experiments due to cost, security and legal reasons. Parallel blockchains can implement these experiments in the form of computational experiments, thereby providing decision-making basis for quantitative evaluation of blockchain system performance and realization of blockchain element innovation. For example, by simultaneously implementing various "stress" experiments, "limit" experiments, and "attack" experiments in an artificial blockchain that simulates a real system and multiple artificial blockchains with different configurations, it is possible to test and evaluate the security performance of the real blockchain while searching for the blockchain optimization configuration that can effectively resist such destructive attacks; in addition, parallel blockchains can effectively support open source experiments and innovations similar to the "Trading agent competition" model. The basic idea is to define appropriate experimental scenarios and goals according to specific needs, build a corresponding artificial blockchain system environment, fix certain blockchain control variables, and open several experimental variables (such as consensus mechanism, incentive mechanism, etc.) to the public or scientific researchers. Through competition or crowdsourcing, the collective wisdom is gathered to achieve the evaluation and optimization of specific experimental variables, thereby promoting the innovation and development of blockchain technology.

(3) Management and control: Parallel blockchains can serve as "parallel sandboxes" for government agencies and industry organizations to implement macro-regulation and trend forecasting, and realize the management and control of the blockchain ecosystem in a virtual and real way. On the one hand, the new technologies, new models and new formats emerging in the blockchain field can first be experimented, evaluated and improved in one or more artificial blockchain systems that are as close to the actual state as possible, and can be applied to the actual blockchain system after meeting specific regulatory goals and performance requirements, thereby realizing the "incubation" function of the parallel sandbox in a way of "artificial approach to reality"; on the other hand, new problems, new demands and new trends found in the actual blockchain system can also be imported into the artificial blockchain system in real time, and the most optimized new solutions can be obtained through a large number of computational experiments and search optimization in the artificial system, and the development and evolution of the actual blockchain system can be guided accordingly, thereby realizing the "innovation" function of the parallel sandbox in a way of "real approach to artificial".

Research framework: The core feature of the blockchain is a multi-agent system based on distributed consensus and chain data structure. On the one hand, blockchain consensus is a large-scale group coordination and collaboration process in a multi-agent social network. Affected by economic incentives and other factors, there are highly uncertain psychological and behavioral factors (such as selfish mining, malicious dust attacks, etc.), highly diverse consensus mechanisms and strategies, and highly complex intelligent competition and cooperation games in the consensus process. This is the social complexity brought to the blockchain by the participation of "people"; on the other hand, the chain data structure of the blockchain integrates a variety of special technical processing to achieve security, trustworthiness and non-tamperability, such as timestamps, hash operations, cryptographic algorithms and decentralized P2P networks. This is the engineering complexity of the blockchain at the technical level. It can be seen that the blockchain system is a typical "human in the loop" complex system with both social complexity and engineering complexity.

Uncertainty, diversity and complexity (UDC) characteristics make it difficult to directly apply traditional theories and methods based on mechanism analysis to blockchain system research, and must be solved through experimental methods. However, due to the introduction of complex factors such as human psychology, behavior, and strategic interactive games, the essential difficulty in studying and optimizing blockchain systems is how to quantitatively and in real time model, analyze, and evaluate the behaviors, mechanisms, strategies, structures, and other elements within the blockchain system when it is difficult or even impossible to conduct experiments. In essence, this is to deal with the contradiction between "modeling the unmodelable", "predicting the unpredictable", and "decision-making the undecidable". Parallel intelligence is an effective theory and method to resolve this essential contradiction.

Basic theory: Parallel intelligence theory is the logical extension and innovation of complex adaptive system theory and complexity science in the complex environment of CPSS in the new era. It is a new research paradigm that combines reality with artificiality, holistic with reduction, and qualitative with quantitative. Research on parallel blockchain based on parallel intelligence theory mainly solves the following three key problems:

Research question 1. Organic integration and unification of holistic modeling and reduction modeling of complex blockchain ecosystems. The holistic modeling and reduction modeling of complex systems are two contradictory and unified research methods. The former emphasizes the high-level emergence and evolution laws at the macro system level, while the latter focuses on the characterization and behavior interaction at the micro individual level. Parallel blockchain theory must organically combine the two. Generally speaking, blockchain (especially public chain) systems usually contain a large number of individual participants, such as mining nodes, transaction nodes, mining pools, etc. These participants are connected to each other through the blockchain network, and follow specific interaction protocols and consensus algorithms to jointly maintain and update the data chain. Therefore, it is necessary to first perform micro-level reduction modeling for a large number of individual participant nodes. Comprehensively and accurately characterize the static characteristics, dynamic behaviors and interaction mechanisms of participants. The more accurate and finer the restoration modeling, the higher the complexity of the subsequent overall modeling, but the high-level emergence and evolution laws obtained are more accurate and reliable. Therefore, blockchain system modeling must organically integrate the two research methods, taking into account the restoration modeling granularity and overall modeling complexity, and seeking the optimal balance between the two.

Research Question 2: Blockchain computational experiments and predictive analysis of "people in the loop" Due to factors such as economic costs, technical conditions, and laws and regulations, new ideas and technologies in the blockchain field are difficult to directly apply to actual blockchain systems. This is also the main reason why many Bitcoin improvement proposals (BIPs) are still in the state of proposal and draft and cannot be truly activated and implemented. Using computational experimental methods to test their feasibility and evaluate their efficiency and effectiveness is an effective way to solve this problem. The key research problem is that the blockchain system is not a "Newton system" composed of simple engineering technologies that are controllable and predictable, but a "Merton system" with "people in the loop" and human and social factors that deeply affect the system's behavior laws and have self-implementation characteristics. Therefore, blockchain experiments cannot be limited to experimental scenarios and environments such as the Bitcoin test network, which are intended to be as "simulated" as possible. Instead, they should be based on various actual or virtual computing experimental scenarios, using adaptive evolution algorithms, parallel learning and other algorithms to drive experiments, so as to observe and quantitatively evaluate the performance of various parameter configurations, new technology solutions and system architectures in different experimental scenarios, and predict their evolution laws.

Research Question 3: Bidirectional Guidance and Co-evolution of Actual and Artificial Blockchain Systems The main goal of parallel blockchains is not to guide artificial blockchain systems to approach real blockchain systems in a narrow sense, but to make actual and artificial blockchain systems co-evolve, closed-loop feedback and bidirectional guidance in a more general sense, so as to achieve the optimization of actual blockchain systems and promote the convergence of the entire parallel blockchain system to the set or emerging goals. In this process, the interaction protocol and synchronization mechanism of virtual and real blockchain systems are key issues worthy of in-depth study. Under normal circumstances, artificial blockchains explore various new models, scenarios, mechanisms and strategies through a large number of virtual computational experimental scenarios, forming a complete "situation-response" knowledge base; under abnormal circumstances (such as the TheDAO hard fork event), parallel blockchains should have the ability to adaptively switch to the best response plan for the current situation, and gradually guide the actual blockchain to approach the optimal state of artificial blockchain through real-time synchronization of data, mechanisms, strategies and algorithms between virtual and real blockchain systems. These three research questions correspond to the data-driven descriptive intelligence, experiment-driven predictive intelligence and interactive feedback-guided intelligence in the basic theory of parallel intelligence. Parallel blockchain uses the "trinity" parallel intelligence theory to add predictive analysis and parallel guidance functions to the current "descriptive" blockchain technology through "on-chain" parallel interaction and co-evolution of actual systems and artificial systems, so as to better serve the modeling, experimentation and decision-making needs of complex social and economic systems in the future.

Research method: The blockchain system is a typical distributed multi-agent system. Therefore, parallel blockchain follows the bottom-up research method in complexity science, and realizes the "description + prediction + guidance" parallel intelligence of blockchain system through blockchain modeling based on artificial system, prediction analysis based on computational experiment and guided decision-making based on parallel execution in ACP method.

Idea: Evolution of BDI morphology of blockchain driven by ACP

In the parallel blockchain research paradigm, the blockchain system can be regarded as a virtual blockchain "agent" composed of large-scale intelligent agent nodes connected by social network. The agent can be described by its BDI (Belief, desire, intention) model, where: belief is the objective cognition of the current world state by the blockchain system, and is a descriptive record of the data generated within the system and the state parameters of the external environment; desire is the consensus of each node within the blockchain system on the desired state, and is the goal of overall optimization of the blockchain; intention is the optimal value selected by the blockchain system from multiple possible plans and action sets to achieve the desire (goal), and is the action to be implemented at the next time node of the system. Correspondingly, in the parallel blockchain system, each actual blockchain system will build three types of artificial blockchain systems that co-evolve with it, namely

(1) "Record" blockchain (corresponding to the belief model): including an artificial blockchain built by "simulation" means that is consistent with the current actual blockchain system, and one or more artificial blockchains that are virtually built based on configuration conditions or experimental parameters that have appeared in history or may appear in the future.

(2) "Experimental" blockchain (corresponding to the intention model): for a set of experimental scenarios selected arbitrarily under normal or abnormal conditions, all "record" blockchains applicable to it are selected, so that they evolve together in a clock-synchronized manner under the same parameter configuration and experimental scenario settings, and the performance indicators of each "experimental" blockchain system (such as fitness, security, consensus speed, etc.) are evaluated in real time.

(3) "Ideal" blockchain (corresponding to the desire model): that is, for each possible optimization goal (such as performance priority, security priority or efficiency optimization, etc.), a set of optimally configured artificial blockchains are obtained through trial and error and search and optimization.

As a hybrid intelligent system of the above three forms of blockchain, parallel blockchain drives the self-evolution between blockchain forms based on the ACP method, and through a large number of computational experimental scenarios, forms a knowledge base of the optimal ideal state that the actual record state of the blockchain can achieve in the virtual experimental scenario, and then forms a "situation-response" rule library, thereby transforming the generation process of emergency response rules under abnormal conditions into a rational deliberation process based on computational experiments under normal conditions, and forming plans and actions to guide the actual blockchain system to actively approach the optimal ideal state of the artificial blockchain.

Method: Blockchain knowledge automation based on parallel learning

As shown in Figure 2, parallel blockchain uses ACP-based parallel learning method to realize knowledge automation of blockchain system, that is, to realize the closed-loop processing process of seven steps, including open source data acquisition for blockchain system, artificial blockchain system modeling, computational experimental scenario deduction, experimental analysis and prediction, management and control decision optimization and implementation, parallel feedback of virtual and real systems, and real-time evaluation of implementation effect.

Parallel learning is an emerging machine learning theoretical framework in the past two years. This framework combines the advantages of various existing machine learning theoretical frameworks. Its novelty lies in the three characteristic methods derived from the ACP method, namely, using software-defined artificial systems for big data preprocessing, data learning including predictive learning and integrated learning, and data-action-guided instruction learning based on Merton's law. The essential difference between parallel learning and traditional machine learning methods is that most machine learning methods are based on actual historical data, while parallel learning is based on hybrid parallel data of actual "small data" + artificial "big data", in which artificial data is virtual data that may appear in future scenarios. Therefore, if traditional machine learning is a "historical machine learning" method, then parallel learning is a "future-oriented machine learning" method, which is a lifelong learning method in which virtual and real systems coexist and evolve together.

The artificial intelligence Go program AlphaGo that has emerged in recent years is the most successful application case of the parallel learning method: AlphaGo first collects and analyzes the "small data" of real historical chess games of human chess players, and then generates a large number of virtual chess games through "left-right fighting" computational experiments to achieve self-game, self-adaptation and self-evolution, generating Go "big data" in virtual space; then through algorithmic reduction to judge the situation and determine the value network and strategy network and other "small knowledge", and through continuous games with human chess players to achieve parallel evolution.

Correspondingly, the parallel blockchain system first collects the node status data, in-chain transaction data and system operation data generated by the actual blockchain system in real time based on open source intelligence and big data analysis methods. The "small data" of these actual systems can be used as seed data to assist in building models of actual blockchains, and by flexibly changing the core elements of blockchains (such as consensus mechanisms, network structures or data structures, etc.), algorithms (such as difficulty adjustment algorithms, token production algorithms, etc.) and parameters (such as handling fees, number of nodes, etc.) to expand the blockchain model, and then generate a large number of "record" forms of software-defined artificial blockchain systems through instantiation. Secondly, according to the system optimization goals, several computing experiment scenarios are generated, and the actual and artificial blockchain systems are run simultaneously in each experimental scenario. Through the self-game, adaptive and self-evolution process driven by algorithms similar to AlphaGo, a larger-scale artificial blockchain "big data" is generated, and combined with the "small data" of the actual system to form parallel data of the blockchain system. At this point, traditional machine learning methods can be used to learn and predict the evolution laws and trends of blockchain systems based on virtual and real parallel data, and reduce "big data" to "small knowledge" that is applied to certain specific scenarios or tasks, suitable for solving specific problems or achieving specific optimization goals. Finally, the acquired small knowledge is applied to the actual blockchain system, and the quantitative evaluation of the implementation effect of these knowledge is achieved through the parallel closed-loop feedback and co-evolution of the virtual and real systems. Based on the "small knowledge" library thus formed, when a specific scenario or problem occurs in the actual blockchain system, the knowledge base can be quickly queried to obtain the matching accurate knowledge and implement it in the actual system.

Platform architecture: The basic elements of the parallel blockchain platform are shown in Figure 2 As shown in the platform layer in , it consists of the underlying element library and the upper-layer application components. It should be noted that what is proposed here is the simplest reference implementation of the parallel blockchain platform. Various components can be flexibly added according to needs during the actual platform construction process.

The element library includes eight categories: model library, ontology library, mechanism library, strategy library, scenario library, algorithm library, contract library and knowledge library. It can form a complete parallel blockchain system through the instantiation and reasonable assembly of various elements. Among them, the model library stores various explicit models of the blockchain, such as intelligent agent model, blockchain data structure model (Merkle tree, Patricia tree, etc.), network structure model (P2P network, MeshNet network, etc.); the ontology library stores the domain ontology of potential application fields (such as agricultural ontology, financial ontology, etc.) to enhance the semantic interoperability of the interaction between various intelligent agents within the platform; the mechanism library stores the interaction protocols and various consensus mechanisms of the intelligent agents; the strategy library stores the typical strategies and behavior patterns presented by the intelligent agents in the process of mining and trading; the scenario library stores the predefined and configurable experimental scenarios and parameters of the platform; the algorithm library stores the endogenous algorithms of the blockchain system (such as the difficulty adjustment algorithm) and external algorithms (such as the co-evolutionary algorithm and deep learning algorithm that drive the experiment); the contract library stores various smart contracts of the blockchain; the knowledge base stores the management decisions and situation-response rules obtained after the system is optimized.

It is worth mentioning that the parallel blockchain platform can learn from the intelligent agent platform similar to Trading agent competition, and conveniently configure various elements by designing specific standards and specifications, and open the element library to the public. Through multi-agent competition, it attracts research and engineering personnel to design and evaluate various novel models, mechanisms, strategies and other elements, so as to continuously enrich and improve the platform with collective wisdom and promote innovation and development in the blockchain field.

The upper-level application components include multi-agent platform components, scenario generators, consensus-driven engines, algorithm analysis tools and large-scale visualization tools. The multi-agent platform component provides platform users with the modeling capabilities, communication protocols and interaction mechanisms of blockchain nodes, and is one of the most important components in the bottom-up modeling method. The multi-agent platform component usually follows the FIPA (Foundation for intelligentphysical agents) specification, and is composed of an intelligent agent management system, a directory server and an intelligent agent component (Agentware), and can unify the syntax and semantics of internal message transmission and content language. The scenario generator can dynamically extract and configure real or virtual computational experiment scenarios from the scenario library, and select appropriate mechanisms, strategies or algorithms to instantiate and form one or more artificial blockchain systems. Furthermore, the consensus-driven engine can complete the computational experiment of the blockchain consensus process based on the artificial blockchain system, and update each element library according to the computational experiment results; the consensus-driven engine can be implemented based on a variety of algorithms. For example, discrete event simulation technology can dynamically simulate the interaction, communication and consensus between intelligent agents (i.e., blockchain nodes) and the external environment by advancing the simulation clock and processing discrete events. The algorithm analysis tool achieves its optimization goal by real-time collection and analysis of the data generated during the blockchain computational experiment, prompting the blockchain system to evolve from an "experimental" form to an "ideal" form. Finally, the visualization tool presents the process of computational experiments and blockchain consensus control in a variety of forms through a dynamic and real-time human-computer interaction interface.

Connotation analysis: Parallel blockchain is a new blockchain system research paradigm, which is characterized by providing description, prediction and decision-making guidance services for blockchain systems through the parallel execution and co-evolution of actual blockchain systems and artificial blockchain systems. It should be noted that the concept of "parallel chain" or "Parallel blockchain" has been used by blockchain technology and industry practitioners at home and abroad in different contexts, but its connotation is essentially different from the parallel blockchain proposed in this article. Generally speaking, the connotation of parallel blockchain can be summarized into three modes: cross-chain parallel, O2O parallel and virtual-real parallel. The similarities and differences are explained below.

First, with the development and popularization of blockchain technology in recent years, the number of various blockchains (especially alliance chains and private chains) has grown rapidly, and cross-chain communication and interoperability have become an inevitable trend in the future development of blockchain. Therefore, the "cross-chain parallel" mode refers to multiple actual blockchains that are "parallel" in this form. The original intention of this mode is to connect multiple independently running blockchains through specific technical means to form a "rope network structure" of blockchain groups, so as to enhance the interoperability between blockchains and the compatibility and interoperability of assets on the chain, and avoid the emergence of blockchain "data islands". The cross-chain interoperability technologies currently proposed include centralized or multi-sig notary schemes, sidechains or relays, and hash-locking. For example, the parallel chain technology proposed by Polkadot uses a relay mode to connect public chains (such as Ethereum) with alliance chains and private chains. "Cross-chain parallelism" is a blockchain parallel mode limited to cyberspace. It is worth mentioning that researchers have also proposed blockchain technology with parallel execution as the goal, aiming to improve the operating performance and efficiency of blockchain by designing new blockchain models to facilitate parallel processing of transactions, data or smart contracts within the blockchain. Here, parallel means "parallel" rather than "parallel". The difference between the two is that the former solves the problem through the divide-and-conquer method of "big and small", while the latter expands a real system to N artificial systems in virtual space by "small expansion", solving the problem through computational experiments and parallel evolution.

Secondly, the "O2O parallel" model uses the actual blockchain system as a bridge to connect the virtual cyberspace and the real physical space, forming an O2O (Online to offline or Offline to online, online and offline connection) parallel society. In recent years, modern society, industrial organization and enterprise form have increasingly shown a trend of parallel existence of virtual cyberspace and the real physical world. For example, the Cyber-physical systems (CPS) proposed by researchers, the digital factory proposed by industrial companies such as Siemens, the digital twin plan proposed by General Motors, and the software-defined enterprise proposed by SAP are all examples of the future O2O parallel trend. Blockchain technology can be used as a secure, reliable, decentralized distributed ledger to communicate between the virtual and real societies: on the one hand, online network big data can be naturally integrated into the blockchain; on the other hand, blockchain technology can also be combined with the Internet of Things technology to form the currently rapidly developing Blockchain of things technology, so that the equipment and facilities, physical assets, etc. in the offline physical space can be digitized and integrated into the blockchain. For example, smart IoT devices will be a typical application scenario of blockchain, which can monitor the entire life cycle of equipment production in a safe and reliable way, realize data transmission and negotiation transactions between devices, and use smart contracts to realize the automatic operation of equipment.

Finally, the parallel blockchain of the "virtual-real parallel" model is essentially different from the above two models. It is a CPSS parallel system formed by the extension of the blockchain system existing in the physical and cyberspace to the third social space. This model incorporates the complex social and human factors contained in the distributed blockchain system into the research scope, and uses computational experiments and parallel optimization of actual and artificial blockchain systems to give blockchain technology the trinity of parallel intelligence that describes, predicts and guides. Compared with the first two models, parallel blockchain is more of a new research paradigm rather than a specific technology or method.

Application field: Parallel blockchain is particularly suitable for application scenarios with "people in the loop" and complex social and human factors, such as smart transportation, agricultural production, financial technology, medical health, and education and training. In 2016, the Institute of Automation of the Chinese Academy of Sciences and the Qingdao Institute of Intelligent Industry Technology proposed the "Tianlian Project" plan, which aims to use technologies such as blockchain, big data and knowledge automation to help build a decentralized, secure, reliable and agile programmable intelligent industrial ecosystem, and has carried out some preliminary exploratory work in the fields of intelligent transportation, smart agriculture, smart health and organizational management. Taking the field of intelligent transportation as an example, parallel blockchains, parallel traffic control systems and "five-in-one" (i.e. urban transportation, public transportation, static transportation, logistics transportation and social transportation) demonstration applications together constitute the core modules of the parallel transportation system. The parallel blockchain provides a flexible and configurable underlying blockchain environment, laying a secure and reliable data and trust foundation for the upper-level parallel traffic control decision-making and demonstration applications. At present, the exploratory work of the parallel transportation blockchain mainly revolves around the storage and authentication of important traffic data, the decentralized transportation finance ecosystem, and the monitoring and traceability of traffic equipment based on the Blockchain of Things (BoT).