Towards Blockchain-based Multi-Agent Robotic Systems: Analysis, Classification and Applications
Ilya Afanasyev, Alexander Kolotov, Ruslan Rezin, Konstantin Danilov, Manuel Mazzara, Subham Chakraborty, Alexey Kashevnik, Andrey Chechulin, Aleksandr Kapitonov, Vladimir Jotsov, Andon Topalov, Nikola Shakev, Sevil Ahmed
TTowards Blockchain-based Multi-Agent RoboticSystems: Analysis, Classification and Applications
Ilya Afanasyev , Alexander Kolotov , Ruslan Rezin , Konstantin Danilov , Manuel Mazzara Subham Chakraborty , Alexey Kashevnik , Andrey Chechulin , Aleksandr Kapitonov Vladimir Jotsov , Andon Topalov , Nikola Shakev , Sevil Ahmed Innopolis University, Innopolis, Russia ITMO University, St.Petersburg, Russia University of Library Studies and Information, Sofia, Bulgaria Technical University of Sofia, Branch Plovdiv, Plovdiv, Bulgaria { i.afanasyev, a.kolotov, r.rezin, k.danilov, m.mazzara, s.chakraborty } @[email protected], [email protected], [email protected]@unibit.bg, { topalov, shakev, sevil.ahmed } @tu-plovdiv.bg Abstract —Decentralization, immutability and transparencymake of Blockchain one of the most innovative technology ofrecent years. This paper presents an overview of solutions basedon Blockchain technology for multi-agent robotic systems, andprovide an analysis and classification of this emerging field. Thereasons for implementing Blockchain in a multi-robot networkmay be to increase the interaction efficiency between agentsby providing more trusted information exchange, reaching aconsensus in trustless conditions, assessing robot productivityor detecting performance problems, identifying intruders, al-locating plans and tasks, deploying distributed solutions andjoint missions. Blockchain-based applications are discussed todemonstrate how distributed ledger can be used to extend thenumber of research platforms and libraries for multi-agentrobotic systems.
I. I
NTRODUCTION
The successes demonstrated in recent years in the inte-gration of robotic systems, wireless sensor networks (WSN),cloud computing, distributed planning and management, anddistributed ledgers provides and optimistic outlook towardsincreasingly popular technological solutions such as the Inter-net of Robotic Things (IoRT) [1], [2], [3], [4], [5] and theBlockchain-based Multi-Agent Robotic Systems (MARS) [6],[7], [8], [9]. It is known that one of the important problemsin developing multi-robot systems is the design of strategiesfor their coordination in such a way that the robots couldeffectively perform their operations and reasonably coordinatethe task allocation among themselves [10]. Real-world sce-narios usually require the use of heterogeneous robots and theperformance of tasks with various structures, constraints andcomplexity. The task distribution for decentralized solutions isappropriate, since the use of autonomous multi-robot systemsin complex scenarios becomes limited and inefficient, andcentralized solutions pose a danger of failure for the entiresystem. Since system agents have to share information, therequirements for the quality of communication in decentralizedsystems are increasing, including such important functions asmaintaining data integrity, resiliency and security in accessingdata. Therefore, implementation of blockchain technology for interaction and coordination of multi-agent robotic systemsbecome a reasonable solution for a dynamic and decentralizedtask distribution.Studies of distributed ledgers demonstrate that decentral-ization and immutable record technologies make blockchainone of the most powerful innovations, since in a decentralizednetwork, intercepting most network nodes using cyberattackslooks economically impractical [11]. However, many popularblockchain solutions suffer from such issues as scalability,delay and low throughput [11]. For example, Bitcoin canprocess less than 10 TPS (transactions per second), whileEthereum can process up to 40 TPS, which is clearly notenough compared to daily 2000 TPS VISA (which theoret-ically can increase up to 50,000 TPS). Since the blockchainnetwork grows with an increase in both the number of usersand transactions, a verification of transactions slows downthe transaction process and throughput. This is known asthe classical Blockchain Trilemma - when it comes to thechoice two of the three between decentralization, scalabilityand security [12]. One of the scaling methods that does notcompromise security or decentralization is called sharding,which involves fragmentation of the available dataset intosmaller datasets called shards [11], [12]. Although multi-agentrobotic systems (MARS) are not so critical to scalability andspeed as the financial and big data-based systems, they arenevertheless also very sensitive to delays and throughput ofthe information channels at data exchange between agents.The literature describes many types of developed algorithmsfor distributed consensus, each of which has distinctive fea-tures, advantages and disadvantages. The summary of the mostimportant distributed consensus algorithms, including Proofof Work (PoW), Proof of Stake (PoS), Proof of Activity(PoAc), Proof of Burn (PoB), Proof of Capacity (PoC), etc.are presented in the review [13]. The methodologies used toachieve consensus in blockchain networks largely determinekey performance characteristics, including scalability, trans-action speed, transaction completeness, security, and resourceconsumption. Each method requires a procedure to generate a r X i v : . [ c s . R O ] J u l nd then adopt a block. The block can be generated oroffered by a node in the network, and it encodes a numberof transactions (for example, in cryptocurrency, it can bemonetary transactions between accounts). Further, a key stepis the adoption of the proposed block/related transactionsby network participants, a process called consensus building.Once a block is accepted, it becomes part of the block chain,and the newly created blocks are cryptographically linked toit. After some time (depending on the consensus algorithmused), the block becomes a constant part of the blockchain,i.e. reaching a finality. However, the finality does not excludethe existence of a small statistical probability that a blockcan be changed (intentionally by design or due to an attack),although with each new block added, and for an establishedblockchain system, it becomes negligible.Let’s summarize the key points of the blockchain technol-ogy in relation to the work of multi-agent robotic systems [9]: • The blockchain-based database is a database for addingonly. Once the data is included in the database, theycannot be changed. The database forms the blockchainstate and is distributed among the nodes. • Each node is an agent on the blockchain network andstores a complete copy of the database. The node isresponsible for transferring all incoming data receivedfrom another node to all its neighboring nodes and couldgenerate records for changing the blockchain state. Allnodes are connected through peer-to-peer communicationchannels. Some nodes should play the role of a validator. • Validators check the correctness of records for changingthe state of the blockchain and approve them (for exam-ple, combining the records into blocks, linking the blockstogether and sending new blocks to the neighbors). Onlyverified entries are applied to all nodes to build the currentstate of the blockchain.Let’s emphasize the advantages of the blockchain technol-ogy for multi-agent systems [9]: • Data availability is achieved through multiply duplicationof data and communication. • Consistency of data is achieved through data validationand strict rules of changes appliance. • No way to remove or change the data stored in theblockchain. • Economic or reputational incentive forces nodes to notviolate the validation rules.The research relevance is based on the importance inthe development of distributed multi-agent robotic systemsthat could effectively perform different operations and inde-pendently coordinate the task allocation within the system.Information exchange during the interaction of robots has theparticular importance for reaching the goals of a multi-agentsystem in conditions of uncertainty, external interference,environmental changes or the presence of intruders when dataintegrity maintenance, resiliency and security have specialvalue. To this end, the blockchain technology for multi-agent robotic systems is designed to solve the problems of information exchange for a group of robots, provide a recordof the interaction history and validate the task execution,enhancing the efficiency of the whole system and extendingthe capabilities of MARS applications.This paper extends the study presented by the authorsin the paper [9] with new materials related to blockchain-based multi-agent systems, including aspects of implementingMARS via Wireless Sensor Network (WSN), ensuring theintegrity and security using distibuted ledgers, and deploymentprospects for these systems in Smart Buildings, Smart Citiesand Industry 4.0.The continuation of this paper is structured as follows.Section II introduces the present state of scientific and en-gineering development in the blockchain-based multi-agentsystems. Section III describes and classifies the most typicalcases, which we identified for blockchain-based robotics ap-plications. Section IV considers aspects of MARS realizationusing Wireless Sensor Network. Section V discusses multi-agent robotic systems related to Smart Buildings, Smart Citiesand Industry 4.0. Finally, we summarize the strong and weaksides of the blockchain-based MARS and discuss the barriersthat technology must overcome in order to prove its viabilityand become mass in the Section VI.II. R
ELATED P APERS
In this section, we analyze the state-of-the-art publicationsand applications where blockchain technology is used fordistributed multi-agent systems with the special focusing onrobotics. A recent research series focuses on the use ofBlockchain technology for the shared knowledge and repu-tation management system in studying the collective behaviorof robots [6], [7], [8], [14], [15], [16], [17].The study [14] presented a trust management model fordecentralized robotic networks that focuses on access con-trol and reputation management for each node. This modelprovides group access based on a robot-oriented trust that isselected and dynamically updated over time. The analysis ofthe system was carried out by compromising a robot usingattacks. The idea of using blockchain technology to solvesecurity problems in multi-robot systems were discussed in[7], [15]. The author [7] states that combining peer-to-peernetworks with cryptographic algorithms allows reaching anagreement by a group of agents (with the following recordingthis agreement in a verifiable manner) without the need fora controlling authority. He describes some blockchain-basedinnovations that could provide a breakthrough in MARSapplications: • New security models and methods to preserve data con-fidentiality and robot’s entity validation; • Design of distributed decision making and collaborativemissions using special transactions in the ledger thatallow robotic agents to vote and reach agreements; • Development of blockchain ledgers for using differentrobot’s parameters, corresponding to changing environ-ments without any changes in their control algorithm,llowing to increase the flexibility of robots withoutincreasing the complexity of MARS design; • Creation of infrastructure for MARS to follow certainlegal norms and safety rules adopted for human societythat could even result in building new business modelsfor MARS operation.In the paper [8], the theoretical concept of managing securityproblems in multi-robot systems using blockchain technologywas reinforced by the implementation and proof-of-conceptfor controlling Byzantine robots. The authors developed anapproach to using decentralized programs based on smartcontracts to create secure swarm coordination mechanisms,as well as for identifying and eliminating Byzantine swarmmembers through collective decision making.The study [6] is based on the organization of theblockchain protocol for multi-agent coordination and controlof unmanned aerial vehicles (UAVs). The paper [15] concernsthe consensus protocol of the blockchain, which uses an addi-tional procedure for verifying the liability execution to preventpayment transactions to questionable service providers. Forthis purpose, the liability execution for agent-based serviceproviders in the decentralized trading market is verified by aformal model checker. As the proof-of-concept, an applicationwas implemented, where a taxi was modeled with the subse-quent delivery check at the end of the completed mission.The article [18] proposes a modular architecture, combiningthe RobotChain [19] framework as a decentralized ledger forregistering events with robots, smart contract technology formanaging robots and Oracle for processing any data types.The modular architecture can be used in various contexts(manufacturing, network or robot management, etc.) since itis easy to integrate, adapt, maintain, and expand for newdomains. What is more, this architecture allows to refusefrom tokens to accelerate the validation process or replacethem by a reputation system for managing tasks and reachingconsensus, since the monetary value may not make sense forprivate blockchain-based networks. The examples of roboticapplications can be: • Task allocation between robot network; • Information support of robots in operations (for example,a robot cannot recognize objects, while others can); • Assessing the robot productivity or detecting performanceproblems; • Voting consensus for swarm robotics.To ensure the interaction of heterogeneous robots in thecyber-physical space, an ontology can be used that describesthe knowledge and competencies of the robots in the system,provides a quick exchange of information between coalitionmembers and smart contracts for the allocation of sensory,computational, control and service tasks between intelligentrobots, embedded devices and information resources [16]. Todo this, the study [16] presents a methodology for creatingcyber-physical smart space with instructions how to create andmanage coalitions of intelligent robots using knowledge pro-cessors and information stored in the blockchain. The similar way, the paper [17] discusses the cyber-physical-social system,which combines smart space technology and blockchain. Theinteraction between mobile robots and humans is related onontology-based publication/subscription mechanism, where alldata exchange is controlled and key information is stored inthe blockchain network.Intelligent cyber-physical systems can be implemented asmulti-agent systems with the ability to schedule tasks byagents [20]. In such multi-agent systems, the protocol of planexecution should lead to proper completion and optimizationof actions, inspite of their distributed execution. However, inunreliable scenarios there is a probability that agents will notfollow the protocol due to failures or malicious reasons thatresult in the plan failure. To prevent such situations, the plancan be executed by agents through smart contracts, ensuringthat the task is performed even in an untrusted environment[20]. Moreover, smart contracts can be automatically generatedfrom manufacturing plans, resulting in automation of theentire system with seamless integration of agents into onecyber-physical system [20]. In the similar way, self-sustainingcyber-physical environments are formed in which all criticalaspects at both the cyber and physical levels are effectivelystimulated, coordinated and supported using blockchain-basedmechanisms and protocols for data storage, communicationand coordination [21]. Thus, the advantages of a properlydeveloped blockchain framework for organizing multi-agentsystems (which can be extended to both robotic and cyber-physical systems) should meet the following requirements[21]: • Decentralization of data storage, which increases therigidness of the system; • Scalability and ease of joining new agents; • Participants reach a consensus in conditions without trust; • Ability to maintain trust among initially unknown agents; • Transparency and immutability; • Agents remain fully autonomous, they fully control theiridentity and private keys; • Blockchain data is complete, consistent and accessible.III. T HE C LASSIFICATION OF THE B LOCKCHAIN - BASED R OBOTICS A PPLICATIONS
In this section, we classify blockchain-based robotic multi-agent systems, which we revealed during the literature review.Let’s consider and discuss our vision of possible robotic multi-agent applications based on blockchain technology shown inthe Figure 1 (that is the extended version of the classificationpresented in [9]).
A. Agent Tasks Assigned in Executable Code from Blockchain
Let’s consider a scenario in which there are multiple agents(robots), without taking into account the hardware platform.Any agents have a pre-installed control program, but theyare configured to receive external executable code (bytecode),which is a set of commands in the package, to implementoperations, achieving the multi-agent system’s goal. The word”bytecode” means here the platform-independent executableig. 1: The classification of typical cases for using blockchaintechnology in multi-agent systems in robotics applicationscode that results in the same sequence of command executionfor any agents. Thus, at the moment of sending the bytecode,there is no need to know the exact hardware platform of theagent (robot), which must execute the commands specified inthe bytecode.The blockchain could be used here to distribute such byte-codes to the agents [9]:- The system which generates the tasks should not beconnected to the agent directly. The peer-to-peer network isused to deliver the message.- A message will be delivered even if an agent is turned off.- The agent is able to inform constantly about its statechanges (e.g. ”moved forward for 10 meters”, ”picked anobject”, etc.). The state is stored in the blockchain thereforeit could be recovered quickly in case of the agent was turnedoff.- The delayed bytecode execution could be scheduled.- Two or more command sequences could be automaticallyqueued for execution by the agent.
B. Distributed Decision Making by a Time-Limited Voting
Distributed decision making (DDM) is the challenging taskin multi-agent robotic systems. There was proposed a solutionto use the blockchain in SWARM systems [7], in which itwas used the idea of sending cryptocoins as the prizes to someaddresses that must be achieved. However, the development ofblockchain technology provides new opportunities that are alsoeffective for solving this problem, for example, using smartcontracts in Ethereum. For this purpose, smart contract(s)should be developed to create an infrastructure by conductingpolls with complex behaviors, such as time-limited voting orvote delegation.
C. Distributed Decision Making for Tasks Assigned in Byte-code
By combining the approaches (B) and (A), it is possibleto obtain another interesting solution, where the use of smart contracts for agents may contain some tasks formulated inbytecode. Other agents may vote for actions, resulting indefining co-generated script that can be obtained from a smartcontract.
D. Action Validation to Exclude Intruders or Faulty Agents
Agents can be used to check each others actions, locationsor states. Let’s look at the swarm where agents executesome actions to achieve the common goal. Periodically, agentssend telemetry information of sensor measurements and theirlocation based on odometry. Sometimes an agent may startworking incorrectly and send wrong data. Information ob-tained from other agents can be used to reach a consensusthat the agent working wrong and the recovery procedurecan begin. In this case, co-evolution scenarios can be appliedfurther. A consensus based on information obtained from otheragents can also be used to identify a robot that is behavingincorrectly, for example, it had been hacked or infected byan attacker. To solve the performance problem of validators,the Sharding approach can also be used here [11], [12]. Datafrom agents are combined depending on their location: thus,the separate shard is formed. Validators are coordinated forthe shard, therefore the information volume for processing issignificantly reduced.
E. Economic Incentive to Optimize Task Performance
The financial side of the blockchain can be used as abasis for stimulating a multi-robot system. Thus, researchersfrom Carnegie Mellon University performed multi-robot map-ping with a market approach, coordinating a robot teamand maximizing the information acquisition at minimal cost[22]. This approach demonstrated reliability and adaptabilityto a dynamic environment, even with the loss of colonymembers, in addition to its ability to withstand communicationlosses and disruptions. The researchers found that, the marketarchitecture in the negotiations of robots, improved the robotteam efficiency for environmental study in many times [22].Although the algorithm was designed to minimize the distancetraveled during the mission, minimizing the exploration timealso showed encouraging results for quick survey of the ter-rain. This approach also allows dynamical changing prioritiesto minimize the resource consumption for the robot team,fulfilling the multiple mission objectives for the robot group[22].
F. Automated Task Dispatching via Blockchain
The blockchain-based distributed consensus can be used todispatch, assign and execute tasks between competing agents.In this case, the dispatching code can be written in theform of a smart contract stored in the blockchain for theimplementation of the following possible scenario:1) The client sends a request for the task execution to thesmart contract dispatcher.2) The dispatcher notifies agents about the new request.3) The agents agree the task performance in the blockchainvia a peer-to-peer network.) Blockchain validators determine the order of agreementsin accordance with the commission that a particularagent pays for processing the agreement.5) The first agreement received by the dispatcher is con-firmed by the smart contract code, and the order detailsare provided to the appropriate agent.Thus, the market will regulate the choice of agents that willbe sent to complete the task. However, certain agents can usestrategies that allow them to pay more to be selected validatorsthat may lead to appearing the most efficient and stable serviceproviders. The case study on the creation of such a multi-agentsystem is presented in the paper [15], which implements anautomated dispatching taxi script with validation of obligationfulfillment by comparing the route traveled with a map by avalidator.
G. Authentication / Suitability Check
There are critical situations associated with the occurrenceof risks / threats that require authentication: • When agents do not trust each other, but use a commonphysical resource; • When agents may be attacked by third parties. Thus,hacker attacks can reveal confidential information and/orlocation, change the task and influence on the agent’sactivity outcome.In the paper [9], the example with a service station for elec-tric vehicles (EV) battery replacement was considered, wherea blockchain-based solution was proposed to provide batteryauthentication services. Since the smart contract code in theblockchain is not changeable, therefore battery amortizationstate is available to all participants in the deal. Therefore, theservice station and EV can connect to any blockchain nodesto check the battery, excluding data replacement at the man-in-the-middle attack.The similar system was described in article [23], whereblockchain was used to store information about the battery lifecycle and solve the problem of its replacement for a transactionwithout trust between the parties. Thus, the calculation of thebattery price and the exchange of digital currency between anEV owner and the station, as well as the key logic, were im-plemented using smart contracts to solve the problem of lackof trust. Similarly, the study [24] introduced an autonomouscharging architecture and a billing framework for EV chargingbased on IOTA technology that provides resistance to hackerattacks and preserving confidential user information.
H. Ensuring the Integrity and Security based on Blockchain
The one of the key features of the multi-agent roboticsystems is the increased requirements to the security inconditions of limited energy consumption of the devices,high requirements to productivity, physical characteristics andmobility of devices, their compatibility with each other. Thepresence of both software and hardware/software componentsinteracting with each other in multi-agent robotic systems and their environment as well as possible variability of cyber-physical environment determine the susceptibility of suchsystems to specific sets of attacks.To meet today’s challenges, it is necessary to develop anintegrated approach for security [25] of robotic systems. Thecomprehensiveness of the approach here means not only theunion of various security systems and it is also very importantto take into account the protection of the security system initself against the attacks. And here the blockchain can be veryeffective solution because of its reliability.There are three basic security concepts - confidentiality,integrity, and availability. Cryptographic protocols, encryptedstorage, etc. are usually used to ensure the confidentiality. Forthe availability support the intrusion detection and preventionsystems, backup communication links, etc. are usually used.To ensure the integrity, checksum or digital signature arewidely used. However, the use of blockchain technology cansignificantly improve the efficiency of ensuring the integrityof stored and transmitted data.Let’s consider the main areas of the distributed multi-agentrobotic systems where the blockchain can be used [7]. • Robot’s sensors [26]; • Robot’s storage [27]; • Robot’s architecture [28]; • Interaction of agent with neighboring agents by IoTprotocols [29]; • Interaction of agent with distant hosts or cloud throughInternet [30]; • Data aggregation, analysis and storage in the cloud [31].As an example, to enhance the security of the multi-agentrobotic systems interaction, the new possible architecture ofinternet - Named Data Networking (NDN) can be used as apart of the system. The NDN is formed with two basic things,i.e. Sending Request and Receiving the data packet. Regardingthe aspect of securities issues NDN can use approaches alongwith the blockchain to protect itself against the various threats.There are many types of attacks to be noted [32], e.g. Inter-est Flooding, Cache Misappropriation, Data Fishing, SelfishAttack etc. It is hard to avoid these security attacks by exist-ing security solutions due to the decentralized and dynamiccharacteristics of NDN. But the decentralized blockchain [33]approach can be applied to meet the security requirement ofNDN. It can use the hash of Interest or Data by smart contractand it will decrease the chance of user privacy leaking becauseboth data identifier and user identifier will be replaced withtemporary names. In addition, nobody will be able to changeor delete the fields due to the blockchain structure of hashedchain blockchain [34].It is already proven [35] that malicious actions can bedetected by NDN blockchain model and blockchain with smartcontract plays an efficient role in this scenario. The blockchainmaintains the trajectory of Interest and Data in a hashedmanner. The distributed blockchain for NDN contain seriesof blocks, and each block contains a hashed transition set.Each block has a head pointer (except the initiated block) thatlinked to a previous block. It consists of a timestamp record theime when the block is written, a bit that linked to a successorblock. In each block, it keeps a hash value that composed byseveral hashed Interest or Data transition records.IV. R
EALIZATION OF M ULTI -A GENT R OBOTIC S YSTEMVIA W IRELESS S ENSOR N ETWORK
Wireless sensor networks (WSN) are widely used for thepractical implementation of multi-agent systems, and the ad-dition of mobile robots to the WSN structure is a well-observed trend [36]. Robots that are active doers in a multi-agent robotic system (MARS) can provide flexibility wheninstalling network sensors and realizing active data acquisition,since they can perform various operations and interact withthe environment [36], [5]. Although these interactions can bepredetermined or based on real-time observations, however,the choice of a suitable communication protocol for robo-tized WSN can be a challenge, considering the complexityand multi-components of robots, as well as the type ofcommunication implemented in MARS: one-to-many. Onepossible solution is to use the HTTP protocol over a WiFiconnection, although it is not very suitable for bidirectionalcommunication due to such difficulties as specifying portsand sometimes IP addresses for each network component,large packet size, high power consumption, and transmissionproblems for control commands via the Internet connection[36]. The alternative solution is to use Cloud Computingand the Internet of Things (IoT) technologies for organizingcommunication between nodes and controlling the WSN robo-tized components, especially when using the Message QueueTelemetry Transport (MQTT) protocol. Due to small packets’size and ”publish/subscribe” concept, managing the connectionbetween network devices can be simpler and more feasible.Thus, at present 6LoWPAN networks with MQTT protocolare becoming a good solution for MARS applications dueto low power consumption, IP-driven nodes and support forlarge mesh networks. As an example, the studies [36],[37] proposed a robotized WSN suitable for solving variousproblems of environmental monitoring. The robotized WSNis an adaptive system in which intelligent agents are movingsensors for detecting and tracking areas, where the monitoredenvironment parameters are different from certain thresholdvalues. For the experiment, iRobot Create and KUKA youBotmobile platforms were used with additionally installed single-chip Gumstix Verdex pro TM XL6P computers and variousexpansion modules as robotic agents (mobile network nodes).The functionality of the proposed WSN robotic with dataexchange using 6LoWPAN is verified using the MQTTBoxplatform. It enables building MQTT clients for publishing orsubscribing topics, configuring MQTT virtual device networks,testing MQTT devices etc.However, security becomes a critical issue when applyingIoT concept to organize communication between the robotizedWSN nodes. A successful solution of the problem of interac-tion between the nodes, together with recording the interactionhistory and performing the verification task can be providedby the blockchain technology (see, the Section III H ). This can increase the efficiency of the robotized WSN and expandthe possibilities of their applications.The notion of a blockchain is associated with a publiclyavailable chronological database of transactions recorded by anetwork of agents (e.g. swarm of agents). It is obvious thatsoftware robots (holons) are suitable for this purpose and sucha research with robotic swarms had been executed [7]. Eachagent possesses private and public keys used to prove theorigin and/or encrypt messages. As stated in [7], the proofmay delay the task execution up to 10 minutes. In order toimprove this situation, it is reasonable to use other cryptocur-rencies instead of Bitcoin, for example, based on IOTA orEthereum. On the other hand, not all transactions should beincluded in blocks. In many cases, different preprocessingschemes should be applied to deeply model the environmentand classify the situation [38]. Such an example is shownin Figure 2, where six security-based ontologies have beendepicted and their combination defines dangerous processingnodes (in red), warning zones (in yellow), and safe (uncolored)zones. Blockchain-based solutions are required only in red andsometimes in yellow zones. The information preprocessingallows not only increasing the efficiency of the system, butalso reducing its vulnerability. Other applications for modelingand data processing, also suitable for blockchain agents, arediscussed in [38].Fig. 2: An ontology-related preprocessing scheme with sixsecurity-based ontologies and their combinations that identifydangerous processing nodes (red), warning zones (yellow), andsafe (uncolored) zonesV. S MART B UILDINGS , S
MART C ITIES AND I NDUSTRY
Internet as often as we do in the sameway we now simply connect devices in a stable manner verydifferently from what we used to do two decades ago, whenore conscious effort was necessary and the connection couldbe lost several times over a work session. The more the Internetwill enter our life, the less we will notice, the more pervasive itwill become in every aspect of personal and professional life.For example, the use of a multi-agent system with auditableblockchain voting helps to make the voting records transparentand unchangeable, allowing to overcome the main e-votingproblem how to increase the level of respondents’ trust in theelectronic voting system [39].This progressive integration is what today we call the
Inter-net of Things (IoT) [40], where every object is transparentlyconnected to the network and can communicate with otherobjects, systems or individuals. In this developing scenariorobots also play a role, and robotics as a discipline cannot beconsidered as a completely separate domain and independentlydeveloping. The
Internet of Robotics Things (IoRT) [1], [2],[3], [4], [5] is a recently defined concept, which aims atdescribing the integration of robotics technologies in IoTscenarios. Multi-Agent Robotic Systems, as described in thispaper, is a notable application scenario for which IoRT canconstitute the working infrastructure [41], both for domestic(domotics/Smart Buildings) and industrial (Industry 4.0) use.The papers [42], [43] examine the integration of theblockchain technology and the Internet of Things, which isexpected to transform human life and provide great economicbenefits. However, the main restrictions for such integrationare insufficient data security and a level of trust.Regarding Smart Buildings and Software-defined Buildings(SDB), blockchain is destined to represent a persistence in-frastructure of pervasive application in everything concerninghome resource, environment and processes [44], from energymanagement to billing, from environmental comfort to safety,surveillance and further. Developing the idea of Smart Build-ing system, researchers integrate subsystems, such as intelli-gent networks, services, buildings and household appliances,into models of Smart Spaces and even Smart Cities, usingblockchains to effectively exchange data when interactingsubsystems, connecting and remote control for reaching abetter life quality, sustainability, energy conservation and thedevelopment of socio-economic systems [45], [44]. The IoRTin the domotics context will be the enabling infrastructurefor blockchain-based Multi-Agent Robotic Systems. In thiscontext, and it also applies to Smart Cities as aggregationof Smart Buildings, a large number of sensors collect datawith high variability of accuracy, reliability and frequency.Therefore, the Blockchain technology could permit the man-agement of public immutable ledgers tracking all the activitiesand determining those that are more trustful and those thatare less, and act accordingly. This hardware and softwareinfrastructure will simplify the interaction of the differentagents in the different buildings allowing traceability of datacollection and enhancing trust, security and accuracy of thecooperation between multiple agents.Ecological and environmental monitoring is a sphere whereall advantages of blockchain are needed, and usage of IoT andIoRT is a good way to create a big independent sensor network. Transparency, immutability and security are highly demandedfor environmental monitoring. Peer-to-peer approach gives away of a cheap connection of the new sensor to the globalnetwork and start provide a data about environment publicly.Right now the concept of citizens’ observatories one of themost suitable for blockchain technology. Such projects asWeObserve [46] and WeSenseIt [47] demonstrate how citizenssensor networks helping to improve fullness of the ecologicalinformation, and with peering technologies it can be a reallyscalable solution.In the context of Industry 4.0 (Smart Factories) [48], andwith the increasing trend of automation, similar considerationson the efficacy of IoRT and Blockchain hold. Blockchaintechnology may eventually represent the pillar of a business ororganization thanks to better contract management, effectivequality control, better accountability, recognition and authen-tication of IoRT devices. In general, blockchain technologyin Industry 4.0 gives us the chance to innovate and refreshthe concept of cybersecurity, offering a mechanism by whichactivities can be immutably tracked and pseudonymized [49].The progress in various technologies and their cooperationfor robotics, automation, IoT, big data processing, cloud com-puting and blockchain lead to the fourth industrial revolution,when the interaction of the Smart Factory components withinthe company and external industrial IoT systems provide trustand reliable control over the resource distribution and products[50], [51]. For example, in logistics, where the supply chain isa multi-agent system in which each supplier has own behaviormodel and purposes, the blockchain can bring the necessarytransparency and trust, speeding up the supply processes andeliminating many shortcomings of current supply chains [42],[52]. VI. C
ONCLUSIONS AND D ISCUSSION
Currently, the approach consisting in the organization of animmutable distributed database storing all relevant informa-tion and providing access to agents of a multi-agent roboticsystem (therefore expanding the capabilities of the system asa whole) is of great relevance in the context of the FourthIndustrial Revolution. The key component of the approachis a distributed ledger technology (blockchain), which allowsagents to interact or allocate the tasks through responsiblesmart contracts. On the one hand, the reliability of the agentis mainly determined by the reputational model, allowing todetermine the trust level to the agent only after the fulfillmentof the agreed obligations. On the other hand, the automationof obligation fulfillment by an agent can provide a verificationprocedure that will allow to verify the liability execution evenamong initially unknown agents and reach a consensus inconditions without trust ([7], [15], [21]). In addition, the goalsfor the blockchain implementation in a multi-agent robotic sys-tem may be the increase of the interaction efficiency betweenagents by organizing more trusted information support, assess-ing the robot productivity or detecting performance problems,voting consensus for swarm robotics, plan scheduling andask allocation, deploying distributed decision making andcollaborative missions.The basic message of this paper is to bring together andprovide, as a guide, ideas on how frameworks, architecturesand structures supported by blockchain solutions can be usedto solve practical problems that face by multi-agent roboticsystems and cyber-physical systems. Relevant studies showthat the blockchain begins to play a large role in the develop-ment of systems and applications with many agents (robots),in which the development of strategies for their coordinationis conducted in such a way that agents can effectively performtheir operations and intelligently coordinate task allocation[10]. The analysis of the recent publications allowed to identifyand classify groups of tasks for multi-agent robotic systemsbased on blockchain technology. This classification is ourmain contribution by this paper. Blockchain technologies canbe used to expand the existing number of platforms andlibraries used by researchers, or to motivate them to usea common solution that is widely distributed and tested,rather than trying to develop their own software solutions tocover similar scenarios. Real-world scenarios may require theuse of disparate agents and the performance of tasks withdifferent structures, constraints and complexity. Therefore, therequirements for the quality of communication in decentralizedsystems are increasing, including such important functions asmaintaining resiliency, data integrity and security in accessingdata. Therefore, the introduction of blockchain technology forthe interaction and coordination of multi-agent robotic systemsbecomes a reasonable solution for many modern research andindustrial tasks.Based on modern investigations, the authors conclude thatat present one of the most promising tasks in the fieldof developing multi-agent systems is the development ofmethodologies, models, structures, architectures and meth-ods, aiming to integrate blockchain technology with highcomplexity systems, such as cyber-physical systems, robotswarm, the Internet of things, the Internet of Robotic Things,Smart Buildings, Smart Factories and Smart Cities. To fullyimplement such integration, it is necessary to automate andsolve subtasks of different difficulty levels, such as intellectualsupport for agent interaction, task and plan allocation, analysisof task performance and liability execution, evaluation of agentperformance, identification of improperly functioning agentsand intruders, security-based issues and many others.Many tasks remain to be done at the moment and requiresolutions, but some of them, according to the authors, are themost important and urgent [9]: • Development of a conceptual model of information sup-port for robot network during the task performance; • Design of a typical ontological model of a multi-robotnetwork; • Development of requirements and formal process model-ing for liability execution [53], [15]; • Design of a consensus protocol for a group interactionverification before launching a task based on the infor-mation from a distributed ledger; • Development of a validation methodology for task per-formance by the robotic system; • Design of multi-agent robotic system architecture; • Automatic process reconfiguration for multi-robot sys-tems based on real scenarios; • Analysis of cybersecurity [54]; • Improvement of existing frameworks allowing for multi-agent robotic networks to perform collaborative taskswith regard to scalability, decentralization and securityrequirements. R
EFERENCES[1] P. P. Ray, “Internet of robotic things: concept, technologies, and chal-lenges,”
IEEE Access , vol. 4, pp. 9489–9500, 2016.[2] P. Simoens, M. Dragone, and A. Saffiotti, “The internet of roboticthings: A review of concept, added value and applications,”
InternationalJournal of Advanced Robotic Systems , vol. 15, no. 1, 2018.[3] R. S. Batth, A. Nayyar, and A. Nagpal, “Internet of robotic things:Driving intelligent robotics of future-concept, architecture, applicationsand technologies,” in , pp. 151–160, IEEE, 2018.[4] C. Mahieu, F. Ongenae, F. De Backere, P. Bonte, F. De Turck, andP. Simoens, “Semantics-based platform for context-aware and person-alized robot interaction in the internet of robotic things,”
Journal ofSystems and Software , vol. 149, pp. 138–157, 2019.[5] I. Afanasyev, M. Mazzara, S. Chakraborty, N. Zhuchkov, A. Maksatbek,M. Kassab, and S. Distefano, “Towards the internet of robotic things:Analysis, architecture, components and challenges,” arXiv preprintarXiv:1907.03817 , 2019.[6] A. Kapitonov, S. Lonshakov, A. Krupenkin, and I. Berman, “Blockchain-based protocol of autonomous business activity for multi-agent systemsconsisting of uavs,” in
Workshop on Research, Education and Develop-ment of Unmanned Aerial Systems (RED-UAS) , pp. 84–89, IEEE, 2017.[7] E. C. Ferrer, “The blockchain: a new framework for robotic swarm sys-tems,” in
Proceedings of the Future Technologies Conference , pp. 1037–1058, Springer, 2018.[8] V. Strobel, E. Castell´o Ferrer, and M. Dorigo, “Managing byzantinerobots via blockchain technology in a swarm robotics collective decisionmaking scenario,” in
Proceedings of the 17th International Conferenceon Autonomous Agents and MultiAgent Systems , pp. 541–549, Inter-national Foundation for Autonomous Agents and Multiagent Systems,2018.[9] I. Afanasyev, A. Kolotov, R. Rezin, K. Danilov, A. Kashevnik, andV. Jotsov, “Blockchain solutions for multi-agent robotic systems: Relatedwork and open questions,” in
Proceedings of the 24th Conference ofOpen Innovations Association FRUCT , p. 76, FRUCT Oy, 2019.[10] T. L. Basegio, R. A. Michelin, A. F. Zorzo, and R. H. Bordini, “A decen-tralised approach to task allocation using blockchain,” in
InternationalWorkshop on Engineering Multi-Agent Systems
Renewableand Sustainable Energy Reviews , vol. 100, pp. 143–174, 2019.[14] I. Zikratov, O. Maslennikov, I. Lebedev, A. Ometov, and S. An-dreev, “Dynamic trust management framework for robotic multi-agentsystems,” in
Internet of Things, Smart Spaces, and Next GenerationNetworks and Systems , pp. 339–348, Springer, 2016.[15] K. Danilov, R. Rezin, A. Kolotov, and I. Afanasyev, “Towardsblockchain-based robonomics: autonomous agents behavior validation,”in
International Conference on Intelligent Systems , IEEE, 2018.[16] N. Teslya and A. Smirnov, “Blockchain-based framework for ontology-oriented robots coalition formation in cyberphysical systems,” in
MATECWeb of Conferences , vol. 161, p. 03018, EDP Sciences, 2018.[17] A. Kashevnik and N. Teslya, “Blockchain-Oriented Coalition Formationby CPS Resources: Ontological Approach and Case Study,”
Electronics ,vol. 7, p. 66, may 2018.18] V. Lopes, L. A. Alexandre, and N. Pereira, “Controlling robots usingartificial intelligence and a consortium blockchain,” arXiv preprintarXiv:1903.00660 , 2019.[19] E. C. Ferrer, O. Rudovic, T. Hardjono, and A. Pentland, “Robochain:A secure data-sharing framework for human-robot interaction,” arXivpreprint arXiv:1802.04480 , 2018.[20] A. Shukla, S. K. Mohalik, and R. Badrinath, “Smart contracts formultiagent plan execution in untrusted cyber-physical systems,” in , pp. 86–94, IEEE, 2018.[21] R. Skowro´nski, “The open blockchain-aided multi-agent symbiotic cy-berphysical systems,”
Future Generation Computer Systems , vol. 94,pp. 430–443, may 2019.[22] R. Zlot, A. Stentz, M. B. Dias, and S. Thayer, “Multi-robot explo-ration controlled by a market economy,” in
Robotics and Automation,2002. Proceedings. ICRA’02. IEEE International Conference on , vol. 3,pp. 3016–3023, IEEE, 2002.[23] S. Hua, E. Zhou, B. Pi, J. Sun, Y. Nomura, and H. Kurihara, “Applyblockchain technology to electric vehicle battery refueling,” in
Proceed-ings of the 51st International Conference on System Sciences , 2018.[24] D. Strugar, R. Hussain, M. Mazzara, V. Rivera, I. Afanasyev, and J. Lee,“An architecture for distributed ledger-based m2m auditing for electricautonomous vehicles,” in
Workshops of the International Conferenceon Advanced Information Networking and Applications , pp. 116–128,Springer, 2019.[25] V. Desnitsky, A. Chechulin, I. Kotenko, D. Levshun, and M. Kolomeec,“Application of a technique for secure embedded device design basedon combining security components for creation of a perimeter protectionsystem,” pp. 609–616, 2016.[26] A. Moinet, B. Darties, and J.-L. Baril, “Blockchain based trust &authentication for decentralized sensor networks,” 06 2017.[27] H. Shafagh, L. Burkhalter, A. Hithnawi, and S. Duquennoy, “Towardsblockchain-based auditable storage and sharing of iot data,” in
Proceed-ings of the 2017 on Cloud Computing Security Workshop , pp. 45–50,ACM, 2017.[28] V. Desnitsky, I. Kotenko, and A. Chechulin, “Configuration-based ap-proach to embedded device security,”
Lecture Notes in Computer Science(including subseries Lecture Notes in Artificial Intelligence and LectureNotes in Bioinformatics) , vol. 7531 LNCS, pp. 270–285, 2012.[29] E. E. Reilly, M. M. Maloney, M. Siegel, and G. Falco, “A smart city iotintegrity-first communication protocol via an ethereum blockchain lightclient,” in
Proceedings of the International Workshop on Software En-gineering Research and Practices for the Internet of Things (SERP4IoT2019), Marrakech, Morocco , 2019.[30] N. Fotiou, V. Siris, and G. Polyzos, “Interacting with the internet ofthings using smart contracts and blockchain technologies,” in
Security,Privacy, and Anonymity in Computation, Communication, and Storage ,pp. 443–452, Springer International Publishin, 2018.[31] Q. Xu, K. Aung, Y. Zhu, and K. Leong Yong,
A Blockchain-BasedStorage System for Data Analytics in the Internet of Things , pp. 119–138. 06 2018.[32] E. AbdAllah, H. Hassanein, and M. Zulkernine, “A survey of securityattacks in information-centric networking,”
IEEE Communications Sur-veys Tutorials , vol. 17, no. 3, pp. 1441–1454, 2015.[33] P. K. Sharma, S. Singh, Y.-S. Jeong, and J. H. Park, “Distblocknet: Adistributed blockchains-based secure sdn architecture for iot networks,”
IEEE Communications Magazine , vol. 55, no. 9, pp. 78–85, 2017.[34] N. Kshetri, “Can blockchain strengthen the internet of things?,”
ITProfessional , vol. 19, no. 4, pp. 68–72, 2017.[35] K. Zhu, Z. Chen, W. Yan, and L. Zhang, “Security attacks in nameddata networking of things and a blockchain solution,”
IEEE Internet ofThings Journal , 2018.[36] S. Ahmed, A. Topalov, and N. Shakev, “A robotized wireless sensornetwork based on mqtt cloud computing,” in , pp. 1–6, IEEE, 2017.[37] S. A. Ahmed, V. L. Popov, A. V. Topalov, and N. G. Shakev, “Envi-ronmental monitoring using a robotized wireless sensor network,”
AI &SOCIETY , vol. 33, no. 2, pp. 207–214, 2018.[38] V. S. Jotsov, “Proposals for knowledge driven and data driven applica-tions in security systems,” in
Innovative Issues in Intelligent Systems ,pp. 231–293, Springer, 2016.[39] M. Pawlak, A. Poniszewska-Mara´nda, and N. Kryvinska, “Towards the intelligent agents for blockchain e-voting system,”
Procedia ComputerScience , vol. 141, pp. 239–246, jan 2018.[40] N. Alam, P. Vats, and N. Kashyap, “Internet of things: A literaturereview,” in , pp. 192–197, IEEE, 2017.[41] A. Gautam and S. Mohan, “A review of research in multi-robot sys-tems,” in , pp. 1–5, IEEE, 2012.[42] F. Casino, T. K. Dasaklis, and C. Patsakis, “A systematic literature reviewof blockchain-based applications: current status, classification and openissues,”
Telematics and Informatics , 2018.[43] X. Wang, X. Zha, W. Ni, R. P. Liu, Y. J. Guo, X. Niu, and K. Zheng,“Survey on blockchain for Internet of Things,”
Computer Communica-tions , vol. 136, pp. 10–29, feb 2019.[44] M. Mazzara, I. Afanasyev, S. R. Sarangi, S. Distefano, and V. Kumar, “Areference architecture for smart and software-defined buildings,” arXivpreprint arXiv:1902.09464 , 2019.[45] C. Lazaroiu and M. Roscia, “Smart district through iot and blockchain,”in , pp. 454–461, IEEE, 2017.[46] A. Berti Suman and M. Van Geenhuizen, “Not just noise monitoring:rethinking citizen sensing for risk-related problem-solving,”
Journal ofEnvironmental Planning and Management , pp. 1–22, 2019.[47] V. Lanfranchi, S. N. Wrigley, N. Ireson, U. Wehn, and F. Ciravegna,“Citizens’ observatories for situation awareness in flooding,” in
ISCRAM2014 Conference Proceedings-11th International Conference on Infor-mation Systems for Crisis Response and Management
IEEE Access , vol. 7, pp. 45201–45218, 2019.[50] N. Teslya and I. Ryabchikov, “Blockchain-based platform architecturefor industrial iot,” in , pp. 321–329, IEEE, 2017.[51] A. Kapitonov, I. Berman, S. Lonshakov, and A. Krupenkin, “Blockchainbased protocol for economical communication in industry 4.0,” in , pp. 41–44, IEEE, 2018.[52] R. Casado-Vara, J. Prieto, F. De la Prieta, and J. M. Corchado, “Howblockchain improves the supply chain: Case study alimentary supplychain,”
Procedia computer science , vol. 134, pp. 393–398, 2018.[53] M. Mazzara, “Deriving specifications of dependable systems: toward amethod,”
CoRR , vol. abs/1009.3911, 2010.[54] N. Dragoni, A. Giaretta, and M. Mazzara, “The internet of hackablethings,” in