A Design of the ViaBots Model for Industrial Assembly Line

Application

Mara Pudane

, Arturs Ardavs

, Egons Lavendelis

, Leonards Elksnis

, Aiga Andrijanova

Peteris Ecis

, Arturs Oniscenko

and Agris Nikitenko

Department of Artificial Intelligence and Systems Engineering, Riga Technical University, Kalku 1, Riga, Latvia

{mara.pudane, arturs.ardavs, egons.lavendelis, leonards.elksnis, aiga.andrijanova, peteris.ecis,

Keywords: Adaptive Systems, Multi-agent Systems, Viable Systems Model, Heterogeneous Multi-robot Systems.

Abstract: Due to growing requirement of Industry 4.0 and general robotics infiltration into everyday life and industry

applications, the adaptive heterogeneous multi-robot systems have become highly significant topic. While the

adaptivity as a phenomena has not been researched for a long time in robotic systems, the organisational

theory has analysed the adaptivity in long term, or viability, for several decades. ViaBots is organisational

theory based framework for technical systems, that defines the functions that the system must fulfil to be

viable. The goal of this paper is to present a design of the ViaBots model in case of heterogeneous multi-robot

system, in particular, for an industrial assembly use-case.

1 INTRODUCTION

The number and functionality of robotic devices,

including autonomous robots, robot manipulators

etc., in the past years have grown rapidly. Such a rise

in the possibilities has also boosted new requirements

for robotic systems in general; they call for fully

adaptive multi-robot systems that can adapt to the

changes in the system itself, as well as to the

fluctuations of the external environment (Dario,

2017). Such adaptive systems would offer multiple

benefits, including fault-tolerance, enhanced

behaviours of the system, and minimal human

intervention. Only then it would be possible to talk

about true autonomy of robotic systems.

However, it is hard to achieve such general long-

term adaptivity that allows the system to adapt to

different tasks that are not foreseen in the original

design. There are some robotic systems, that manage

to achieve at short-to-middle-term autonomy in

https://orcid.org/0000-0002-9188-5478

https://orcid.org/0000-0002-4430-8037

https://orcid.org/0000-0001-9912-035X

https://orcid.org/0000-0001-8801-6071

https://orcid.org/0000-0002-2509-6235

https://orcid.org/0000-0001-6088-5873

https://orcid.org/0000-0002-2189-4053

https://orcid.org/0000-0002-5701-3094

specific cases, such as adaptivity to workload or

changes in system configuration (Ardavs et al., 2019).

However, there are little systems that are designed to

adapt to potential rapid changed.

Adaptivity and systems’ ability to persist over

long periods of time has been researched extensively

in the Organisational theory. The Viable Systems

Model (VSM) defines the functions and functional

dependencies that the system needs to be adaptive in

long term, i.e., viable (Beer, 1985). VSM has been

applied to technical systems as well and has showed

promising results.

The remainder of the paper is organized as

follows. In Section 2 we explain the existing

problems in the development of adaptive

heterogeneous multi-robot systems, in particular, the

need for theoretical model for multi-robot systems’

functional organisation, and the integration of

heterogeneous robots. In Section 3, the VSM is

described as well as its mappings for a technical

system. Section 4 details application of the model for

Pudane, M., Ardavs, A., Lavendelis, E., Elksnis, L., Andrijanova, A., Ecis, P., Oniscenko, A. and Nikitenko, A.

A Design of the ViaBots Model for Industrial Assembly Line Application.

DOI: 10.5220/0009173709330940

In Proceedings of the 12th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2020) - Volume 2, pages 933-940

ISBN: 978-989-758-395-7; ISSN: 2184-433X

933

the use-case design and finally, in Section 5,

conclusions and future work is given, as this is

research in progress.

2 RELATED WORKS

The related work section is divided into two parts:

first of all, the concept of adaptivity in the multi-robot

systems so far is reviewed, looking at existing

examples of the multi-robot systems frameworks, as

well as adaptivity as a concept in other areas.

Secondly, the methods for creating heterogenous

multi-robot systems are reviewed, with the focus on

collaboration among robots.

2.1 The Logical Structure of the

Heterogeneous Multi-robot Systems

In order to achieve a long-term operation, several

architectures have been proposed for adaptive multi-

robot systems. Some examples of such approaches

include NASA’s more universal Autonomous Nano

Technology Swarm (ANTS) concept (Vassev et al.,

2012), or narrower biology-inspired intelligent

control (Jafari & Xu, 2019). Unfortunately, while

most of the studies concentrate on specific missions or

applications of robot teams, only very few propose

formal frameworks or methods of system design. One

of such approaches uses Event-B and PRISM design

methods to derive technical design of the system

through iterative steps and assesses probability of goal

achievement thereby providing guidance for further

developments (Tarasyuk et al., 2013). Similarly, in

(Gerostathopoulos et al., 2016) it is proposed to design

self-adaptive system by using the invariant method.

The developed framework, IRM-SA, tackles the

design complexity of invariant methods thus enabling

the design of such systems.

Current requirement for the robotic systems has

created an emerging need for general formal

framework for design and development of resilient

and adaptive systems (Ardavs et al., 2019). This

framework would have to fulfil two main tasks –

ensure systems’ adaptivity to the changes in the

environment and within itself; and reduce the

complexity of the design process (Ardavs et al.,

2019).

In the organizational theory, a concept of the long

term adaptivity, or viability, has been researched for

several decades, resulting in the VSM (Beer, 1985).

Essentially, the VSM describes the functions and

functional dependencies that the systems need to be

viable. While the VSM initially was only used to

analyse and improve organisation related aspects,

such as information flows (Kirikova & Pudane,

2014), in the recent years due to the new requirements

of multi-robot systems, VSM has been adapted to

some technical systems. Examples of these include

the design of the smart distributed automation

systems (Bonci et al., 2019) as well as cyber security

management (Spyridopoulos et al., 2014). To

authors’ knowledge, the most extensive research on

VSM applicability in the technical systems was done

in (Ardavs et al., 2019) where the technical-systems-

appropriate VSM design was developed and adapted

to conveyor belt simulation in a multi-agent

environment. The results showed that using the VSM

as a logical-level framework enables agent self-

organisation and improves the output of the system.

This leads to conclusion that VSM-based model

would improve the performance of a real system as

well and is applicable to industrial assembly task

done by highly heterogeneous multi-robot system at

the conveyor belt. Another reason to believe this

conclusion, is that the fractal nature aside, the

functions of VSM are similar to MAPE-K loop

introduced by IBM (IBM Coorporation, 2005).

2.2 MAS as a Logical Abstraction

Layer for Multi-robot System

Another challenge regarding heterogeneous multi-

robot systems is enabling the control of the whole

system. As the complexity of robots is growing, the

control of each robot separately becomes increasingly

detailed. The control of multiple robots is an even

more challenging task. Additionally, different robots

of various manufacturers have different interfaces

and operating systems which leads to difficulties in

the integration.

To cope with the complexity in various domains,

a common approach is to develop a layered design

and decision-making architectures. In multi-agent

systems, for example, design is usually divided into

micro and macro level where the micro level concerns

the design of a single agent as opposed to macro level

which includes designing multi-agent system as a

whole (Wooldridge, 2009). Another related approach

to heterogeneous systems, is aggregate computing

which is alternative way for implementing such

systems (Beal et al., 2015). While (Bures et al., 2016)

criticizes MAS as an approach to implement complex

heterogeneous systems, we see VSM as a tool to

mitigate the critic, in particular, of lack of the

architecture in multi-agent systems.

Similarly, the decision making process must be

divided in the multi-robot systems. In this case, three

ICAART 2020 - 12th International Conference on Agents and Artiﬁcial Intelligence

934

abstraction layers must be used (Lavendelis &

Nikitenko, 2015). On the lowest level, the robots

control their mechanisms and perform discrete

actions. On the middle level, the decisions are made

that are needed to fulfil tasks (such as going to

particular location). At the top level, the multi-robot

management decisions are made. These decisions

include cooperation, task distribution and other high-

level tasks that are arguably the most challenging.

In several cases, the developers have used multi-

agent layer for the management of heterogeneous

systems, such as autonomous transport vehicles’

systems (Martin et al., 2019), or multi-robot systems

(Lavendelis & Nikitenko, 2015). In the latter case,

each agent represents a physical device (robot, sensor,

manipulator, etc.) within this abstraction layer. The

high-level decisions are negotiated among agents;

each agent manages the corresponding device. This

allows to abstract from low level details while

performing such tasks as allocation and reallocation,

collaborative mapping, coordination, etc.

Use of multi-agent paradigm also allows building

open systems since a new agent representing a new

device can be added to the system. Still, mechanisms

to adapt the situation to the changes in the

composition of agents are necessary. Since the VSM

deals with the management of the whole system, the

multi-agent system is the level where the VSM is

implemented.

3 ViaBots MODEL

The ViaBots model consists of the VSM functions

transformed into agent-related terms and concepts.

3.1 The Viable Systems Model

The viable systems model consists of five functional

blocks, called subsystems (Beer, 1985). These

functional blocks: S1, S2, S3, S4 and S5 respectively

represent five main functional groups: Operation,

Coordination, Control, Intelligence and Policy. The

instances of the S1 subsystem (Operation) are the

units that perform the actual work in the system, i.e.,

everyday tasks, while the rest of the subsystems

represent the Management functions. S2, or

Coordination performs the tasks of local

management. One S2 instance is attached to every S1

instance. S2 instances negotiate resources, and if no

consensus can be acquired, turn to S3. S3 (Control) is

global level management. The main function of S3 is

to oversee the execution of the global plan by

observing the work of S1 and S2 and, if needed,

redistributing resources. The plan is changed in close

coordination of S3 which is informed of the current

situation in the organisation, and S4 (Intelligence)

which observes external environment and holds the

model of changes in the external environment.

Finally, S5 (Policy) is the subsystem that generates

general policy for the system as well as fulfils the role

of representation.

There are different types of dependencies among

functions (Pudane, 2013). The type a link is between

S1 and S2, as well as between S1 units and external

environment. The goal of these links is to manage

complexity, i.e., reduce the number of states with

which system needs to deal. The b type link is

coordination and negotiation link that exists among

S2 instances. Type c link is directed link that

represents monitoring and is located between S3 and

S1, as well as between S4 and external environment.

Type d link is a cooperation or integration link

between S3 and S4. Finally, type k link is a directed

control link that exists between S5 and S3, S4; S3 and

S2, and S2 and S1.

3.2 The Adaptation of VSM to

Technical Systems

To adapt the VSM to multi-agent system, several

mappings were done (overview in Tables 1 and 2).

First, all the functions of VSM subsystems were

identified and mapped either to agent tasks, or

communications (e.g., negotiations), and the

subsystems themselves where mapped to agent roles.

The roles can then be comprised into agents

dynamically. Such an approach gives more flexibility

since there are no limitations on what physical units

must perform the functions. Then, the roles were

correspondingly mapped into behaviours and tasks

were implemented as procedures. Each agent that

carries out one or several behaviours, can now

represent either a software agent, or a robot. The

Table 1 contains identified functions, corresponding

tasks, task mappings into roles and then – agents.

Table 2 contains the interaction design with functions

that correspond to channels, and multi-agent system

interaction types. The mapping in detail is discussed

in (Ardavs et al., 2019).

4 USE-CASE DESIGN

The benefits of the ViaBots model is best seen in a

larger scale and complexity (Ardavs et. al, 2019), for

this reason use case is a physical assembly line.

Similarly to the simulation in (Ardavs et. al, 2019), it

A Design of the ViaBots Model for Industrial Assembly Line Application

935

Table 1: Task allocation to roles and role allocation to agents in ViaBots model.

VSM Function Task Role Agent

S1 Depends on the system

1..n tasks depending on the

variety of the agents

1.. n roles

depending on the

domain

Either physical units that carries

out roles (i.e., robots), or the ones

that will run on one computer or in

one domain.

Assigned to agents having variety

reduction and amplification

capabilities

Type A Attenuators and Amplifiers

Attenuate variety task and

amplify variety tasks between

external environment and syste

S1 element coordination

Negotiation Task

Solving of conflicts of S1

S1 resource relocation

Type A Attenuators

Attenuate variety task between

Operation and Managemen

S3 Resource distribution The resource distribution tas

Assigned to one agent irrelevant

from the already assigned roles

Interpretation of policy

decisions

Interpret high level instructions to

lower level tas

Development planning

according to environment and

system states

Opinion task-current situation

S4 Opinion task-future states

Assigned to one agent irrelevant

from the already assigned roles

Assigned to agent having

monitoring environment

capabilities

S4 Suggest. for safety polic

Safety and resilience tas

S4 Learning Learning tas

Management of external

contacts

Search task

S4 Environment monitoring Env. monitoring tas

S5 Representation HCI tas

Assigned to one agent irrelevant

from the already assigned roles

Investment in structure and

olicy formation

Policy task

Table 2: The interactions in ViaBots model.

In VSM

Function MAS interaction type

System Channel

Type B

S1 element coordination Coordination protocol

S2 Solving of conflicts of S1 Negotiation protocol

S2 S1 resource relocation Resource negotiation protocol

S2 Type

Control channel function Sent once, message-received response

Type

Control and

monitoring over S1 and S2

Sent once, message-received response

Type C

nform message

S3 and S4 Type D

Development planning according to

environment and system states

Negotiation protocol

S4 Type C Monitoring of system itself

nform message

Type C

Monitoring of cooperation of S3 and

nform message

Type

Sent once, message-received response

Channel Type A

Amplify variety between Operation

and Managemen

Broadcast

was decided to build a real-life conveyor belt with

multiple robots as workers.

4.1 The Conveyor Belt Use-case

The following robots are used with the conveyor belt:

in the robotics community well-known packaging

Rethink Robotics, 2015. Baxter™ SDK API

Documentation, retrieved from http://api.

rethinkrobotics.com/.

ABB Robotics, 2019. Robotics product range. Creating

the flexible, collaborative and connected Factory of the

Future. ABB Robotics.

robot Baxter with fixed base and two synchronized 5

DOF manipulators

, an industrial 5 DOF ABB

manipulator

and several custom-made 3 DOF

manipulators. Additionally, a humanoid robot with

human-like hands – Pepper

is used in case no other

robot is capable to do the tasks. Such a set of robots

is highly heterogeneous in terms of the manipulators

Softbank Robotics, 2019. Pepper, retrieved from:

https://www.softbankrobotics.com/us/pepper.

ICAART 2020 - 12th International Conference on Agents and Artiﬁcial Intelligence

936

used, time necessary to do the task and parts available

to each robot. At the same time, multiple robots can

do any or most of the tasks. This enables modelling

an assembly line with different robots and as a

consequence necessity to find appropriate task

allocation among the robots to optimize the

performance of the whole assembly line.

As tasks for the conveyor belt, boxes with round

holes were designed. Robots must insert cones in

these holes in a specific pattern (Figure 1). The

different types of parts are achieved by adding a

marker, to what type the box belongs (i.e., what is the

cone pattern for this particular box type). There are

two sizes of the cones, and not all the robots can insert

both types of the cones.

Figure 1: The tasks for the assembly line - different layouts.

4.2 The Implementation of the Tasks of

the ViaBots Model

The tasks of the ViaBots model were adapted to the

conveyor belt use-case (overview in the Table 3). The

S1 functionality in this case is part assembly that is

modelled as moving the cones from the dispenser to

the boxes. The complexity of the environment is

reduced by robot sensors; we added infrared sensors

to the conveyor belt to simplify the sensor data

processing. Negotiation is performed by S2 agents

based on the time needed for task completion (some

manipulators are slower than others, it takes more

time to pick up bigger cones) and availability of the

details. If manipulator runs out of cones, it is defected

and cannot continue working until additional cones

are added to dispenser.

The interaction among Operation and

Management is performed only when one of the

subsystems S1 cannot continue its work. Such a

situation can occur in the following cases: (a) there

are no more cones of a particular size; (b) the robot

that has the needed cones is working for another S1

unit. In such a case, S3 redistributes resources, i.e.,

makes the S1 units share the same robots.

S4 holds the environment model, i.e., it foresees

the upcoming assembly task sequence; together with

S3 they choose appropriate resource redistribution,

and S3 can ask for external resources (i.e., ask

appropriate resources to the human user). The

assembly task sequence can be changed as a change

in the external environment.

Due to complexity in real-life implementation, S5

was left to user. This concerns overall configuration

of the conveyor belt.

Table 3: The adaptation of ViaBots model tasks to conveyor belt use-case.

In VSM Task Implemented as

S1 1..n tasks depending on the variety of the agents

Robot tasks of picking up a part and placing it in the box in the

corresponding place.

Type A

Attenuate variety task and amplify variety tasks

between external environment and system

The environment was simplified by adding infrared sensors to

announce that the box had arrived. The robot must pick and

place the detail, based on fixed position. S1 is considered to be

the robot manipulator together with infrared sensor.

S2 Negotiation Task

Negotiating among S2 units based on time for part insertion

and available robots

Type A

Attenuate variety task between Operation and

Managemen

Intervention required when errors or resource lacking occurs

S3 The resource distribution task

Calculating the times for task execution and enforcing the new

tasks

Interpret high level instructions (policies) to

lower level tas

Implemented through human interaction with system

S3 Opinion task based on current situation Weights of the calculated

art sequence

S4 Opinion task based on future states

Weights of the foreseen

art sequence

S4 Safety and resilience tas

Calculating the number of lef

-over cones.

S4 Learning task

Part sequence which S4 learns; in this case the part sequences

will be fixed.

S4 Monitoring of environment tas

Due to simplicity of environment acquired through S1 and S2

S5 HCI tas

Performed by use

S5 Policy tas

Performed by user.

A Design of the ViaBots Model for Industrial Assembly Line Application

937

Figure 2: Deployment diagram of the assembly line use-case.

4.3 Overall Scheme of the Use-case

The deployment diagram of the use-case is depicted

in Figure 2. The multi-agent system of the use-case

was implemented in JADE (Bellifemine et al., 2007).

JADE container is populated with an agent for each

of the manipulators by the conveyor belt. Each of

these agents via simple ACL messages (FIPA, 2002)

manages the corresponding manipulators. The

subsystem roles, according to ViaBots model, are

implemented as JADE agent behaviours and are

assigned to agents dynamically. Depending on the

starting sequence, one of the agents runs not only S1

and S2, but also S3 and S4 functions. Physically, the

agents will run on the manipulators’ computers.

Additionally, on a separate computer, a GUI agent

captures the states of the conveyor belt and provides

the user interface to interact with conveyor belt

system. Finally, Conveyor manager is a specific agent

that would control the conveyor belt and move it,

when needed.

In Figure 3, a general sequence diagram of

interaction among the behaviours is depicted.

Conveyor agent serves as a sensor to all the S1

systems and message to S1 agents when the new box

has arrived. It also places incoming boxes on the belt.

All the S2 Behaviours will contain a task that

determines, if the incoming box requires insertion of

the given unit’s type. If the box refers to particular S1

unit, the message is sent to S2 of new task, otherwise,

the box is ignored. Then, the S2 requests to

corresponding S1 agents if they are able to perform

the task and in what time. If the answers are positive,

S2 picks one of the agents to perform the work.

Otherwise, the message is sent to S3 that there are no

resources available. S3 gathers the information from

all the S2 on current situation, as well as information

from the S4 on the future states. Finally, S3 generates

a new resource distribution. If resources cannot be

redistributed, the working cycle ends.

5 CONCLUSIONS AND FUTURE

WORK

The work presented in the paper is on-going research

and the use-case is about to be tested to acquire first

results. The theoretical results achieved earlier in

(Ardavs et al., 2019) based on a simulated

environment has proven the usefulness of the VSM in

the technical systems It was proven in the simulator

that overhead calculations are not significant,

especially with large number of robots. The purpose

of the implementation on a realistic assembly line

done within this paper is to collect data from close to

ICAART 2020 - 12th International Conference on Agents and Artiﬁcial Intelligence

938

Figure 3: Scenario sequence diagram of the assembly line use-case.

real environment and further analyse the efficiency of

the proposed model. The ultimate goal of the research

is to create a framework for long term adaptivity of

heterogeneous multi-robot systems. This would

enable such systems as robotized assemblers or other

production lines to adapt to unexpected events or

slight changes in the tasks done. At the moment, such

adaptation can be done only by human operator.

The future work is to do experiments in real

environment to compare the efficiency of the

ViaBots-model-based assembly line against the one

with fixed configuration to prove that the results

achieved in simulated environment are valid.

It is planned to introduce unpredictable events to

enable test long-term autonomy in these scenarios. It

must be considered that any system (including natural

systems) can adapt only to tasks that can be physically

done with available tools. For this reason. the

unpredictable events in conveyor case will be

different kinds of assembly tasks, such as a box with

different hole configuration (i.e., 9 holes).

Later on the model will be applied to different

mobile robot based scenarios to validate its

applicability to various tasks since the aim of the

research is to develop a general model that can be

applied to various multi-robot systems.

ACKNOWLEDGEMENTS

This work has been supported by the European

Council Seventh Framework Program FLAG-ERA

project “Rethinking Robotics for the Robot

Companion of the Future” (RoboCom++).

A Design of the ViaBots Model for Industrial Assembly Line Application

939

REFERENCES

Ardavs, A., Pudane, M., Lavendelis, E., Nikitenko, A.,

2019. Long-Term Adaptivity in Distributed Intelligent

Systems: Study of ViaBots in a Simulated

Environment. In Robotics, 8, 25.

Beal, J., Pianini, D., Viroli M., 2015. Aggregate

Programming for the Internet of Things. In Computer,

48, 9, pp. 22-30.

Beer, S., 1985. Diagnosing the Systems for Organizations,

John Wiley & Sons: Pennsylvania, USA.

Bellifemine, F.L., Caire, G., Greenwood D., 2007.

Developing Multi-Agent Systems with JADE (Wiley

Series in Agent Technology). John Wiley & Sons, Inc.:

Hoboken, NJ, USA.

Bonci, A., Pirani, M., Cucchiarelli, A., Carbonari, A.,

Naticchia, B., Longhi, S., 2018. A Review of Recursive

Holarchies for Viable Systems in CPSs. In Proceedings

of IEEE 16th International Conference on Industrial

Informatics (INDIN), Porto, Portugal, July 18-20, 2018,

pp. 37-42.

Bures, T. Plasil, P., Kit, M., Tuma, P., Hoch, N., 2016.

Software Abstractions for Component Interaction in the

Internet of Things. In Computer, 49, 12, pp. 50-59.

Dario, P., 2017. Fet-Flagship proof-of-concept Project:

Rethinking Robotics for the Robot Companion of the

future. Presentation at Ro-Man 2017: 26th IEEE

International Symposium on Robot and Human

Interactive Communication, August 28, 2017.

Foundation for intelligent physical agents (FIPA), 2002,

FIPA ACL Message Structure Specification.

Gerostathopoulos, I., Bures, T., Hnetynka, P., Keznikl, J.,

Kit, M., Plasil, F., Plouzeau, N., 2016. Self-adaptation

in software-intensive cyber–physical systems: From

system goals to architecture configurations. In Journal

of Systems and Software, 122, pp. 378-397.

IBM Coorporation, 2005, An architectural blueprint for

autonomic computing, Autonomic Computing, White

Paper, Third edition IBM: USA.

Jafari, M., Xu, H., 2019. A biologically-inspired distributed

fault tolerant flocking control for multi-agent system in

presence of uncertain dynamics and unknown

disturbance. In Engineering Applications of Artificial

Intelligence, 79, pp. 1-12.

Kirikova, M., Pudane, M., 2014. Viable Systems Model

Based Information Flows. In Proceedings of 17th East

European Conference on Advances in Databases and

Information Systems, New Trends in Databases and

Information Systems, Advances in Intelligent Systems

and Computing, 241(1), pp. 97-104.

Lavendelis, E., Ņikitenko, A., 2015. Software Abstraction

Layer Based Multi-Robot System Technology. In STO-

MP-AVT-241 - Technological and Operational

Problems Connected with UGV Application for Future

Military Operations, Poland, Rzeszow, 20-22 April,

2015. Rzeszow: NATO STO, pp.18.1-18.8.

Martin, J., Casquero, O., Fortes, B., Marcos, M., 2019. A

Generic Multi-Layer Architecture Based on ROS-

JADE Integration for Autonomous Transport Vehicles.

In Sensors, 19, 69.

Pudane, M., 2013. Knowledge Flow Analysis using Viable

Systems Model, Master Thesis, Riga Technical

University, Riga, Latvia (in Latvian).

Spyridopoulos, T., Maraslis, K., Tryfonas, T., Oikonomou,

G., Li, S., 2014. Managing cyber security risks in

industrial control systems with game theory and viable

system modelling. In Proceedings of 2014 9th

International Conference on System of Systems

Engineering (SOSE), Australia, June 9-13, pp. 266-271.

Tarasyuk, A., Pereverzeva, I., Troubitsyna, E., Laibinis, L.,

2013. Formal Development and Quantitative

Assessment of a Resilient Multi-robotic System. In

Proceedings of: SERENE 2013: International

Workshop on Software Engineering for Resilient

Systems, Kiev, Ukraine, October 3-4.

Vassev, E., Sterritt, R., Rouff, C., Hinchey, M., 2012.

Swarm Technology at NASA: Building Resilient

Systems. In IEEE IT Professional, 14, pp. 36-42.

Wooldridge, M., 2009. An Introduction to MultiAgent

Systems. Wiley Publishing, 2nd Edition.

ICAART 2020 - 12th International Conference on Agents and Artiﬁcial Intelligence

940