Bio-Inspired Protocols for Embodied Multi-Agent Systems

Vinicius Souza de Jesus

1 a

, Carlos Eduardo Pantoja

1 b

, Fabian Manoel

1 c

, Gleifer Vaz Alves

2 d

Jose Viterbo

3 e

and Eduardo Bezerra

1 f

Centro Federal de Educac¸

ao Tecnol

ogica Celso Suckow da Fonseca (CEFET/RJ), Rio de Janeiro, RJ, Brazil

Universidade Tecnol

ogica Federal do Paran

a (UTFPR), Ponta Grossa, PR, Brazil

Universidade Federal Fluminense (UFF), Niter

oi, RJ, Brazil

Keywords:

Bio-Inspired Protocol, Embodied MAS, Ecological Relations.

Abstract:

Bio-Inspired approaches and techniques are being used in different domains and applications in artiﬁcial in-

telligence, including the agent domain. Some agents are able to move from one system to another to establish

new relationships. In biology, ecological relations are concepts responsible for classifying the relationships

between living beings in an ecosystem, depending on the behavior and function that each one can assume.

The objective of this work is to propose bio-inspired protocols based on ecological relations: Predation, In-

quilinism, and Mutualism. These protocols aims to preserve agents’ knowledge as they can live as a tenant

in another physical body waiting for a similar hardware to predate, or acquire and transmit knowledge by

interacting with other agents while sharing the same physical body. To validate these protocols, a study case

and a scenario are implemented, tested, and evaluated in a real environment.

1 INTRODUCTION

The biology has inspired in so many ways Computer

Science and Artiﬁcial Intelligence ﬁelds contributing

to the development of new concepts, algorithms, and

techniques capable of improving learning capabili-

ties and social organization applied in different do-

mains of applications. Intelligent systems can be-

come increasingly efﬁcient by applying bio-inspired

approaches in their performance and group work for

achieving goals (Zedadra et al., 2016).

The agent approach also has some similarities

with biology concepts, such as reasoning, cogni-

tion, and interactions between agents. The agents

are independent and proactive entities with cogni-

tive ability to make decisions based on what they

can perceive in the environment and communicate

with other agents. Multi-Agent Systems (MAS) are

composed of multiples agents interacting with each

https://orcid.org/0000-0002-4534-6078

https://orcid.org/0000-0002-7099-4974

https://orcid.org/0000-0003-0614-0592

https://orcid.org/0000-0002-5937-8193

https://orcid.org/0000-0002-0339-6624

https://orcid.org/0000-0001-9177-5503

other, and they are capable of solving situations col-

lectively (Wooldridge, 2000).

Some MAS can be closed or opened depending

on the characteristics that agents can assume. An

open MAS allows agents to enter, interact, and leave

it in anytime. It provides that agents from a MAS

can share knowledge (beliefs, intentions, plans, and

goals), and help each other to achieve a common goal

(Huynh et al., 2006) in another MAS. It can reduce

the number of messages exchanged using an exter-

nal infrastructure in case of MAS in distant locations,

since an agent moves itself to a speciﬁc MAS instead

of sending messages using a network connection, for

example. Therefore, an open MAS intrinsic leads to

agents capable of moving from one MAS to another.

In biology, ecological relations deﬁne and explain

how interactions between living beings occur in an

environment. These relationships happen among liv-

ing beings of the same species or not, and it allows

members of a particular community of living beings

to relate to other ones from different communities, ex-

panding the relationships to an ecosystem level (Be-

gon et al., 2006). Similarly, an open MAS allows the

entrance and the exit of agents in its system, enabling

a dynamic integration of new agents with the existing

ones (Huynh et al., 2006).

312

Souza de Jesus, V., Pantoja, C., Manoel, F., Alves, G., Viterbo, J. and Bezerra, E.

Bio-Inspired Protocols for Embodied Multi-Agent Systems.

DOI: 10.5220/0010257803120320

In Proceedings of the 13th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2021) - Volume 1, pages 312-320

ISBN: 978-989-758-484-8

We assume the existence of an Embodied MAS,

which is a physical device composed of resources,

controlled by an embedded open MAS. It can be com-

pared to an Embodied agent (Rickel and Johnson,

2000), which is a single agent incorporated in a phys-

ical device. However, the agent is not part of an em-

bedded MAS. Then, one can deﬁne an ecosystem for

Embodied MAS as a group of Embodied MAS situ-

ated in a physical environment, and capable of com-

municating and interacting with each other.

There are related works inspired by biology con-

cepts in the agent domain to provide techniques and

algorithms to increase the level of interaction and

communication between MAS that act upon the same

environment (G

unay et al., 2015), allowing agents of

different systems to communicate and interact. A for-

aging situation (search of food resources based on

survival instincts) (Zedadra et al., 2016) is presented

where different open MAS act upon a simulated envi-

ronment. It allows agents from different MAS to in-

teract and perform complex tasks. These works do not

consider the risks of dealing with a real environment.

Since the real environment is unpredictable, they do

not offer mechanisms to preserve the integrity of the

agents’ knowledge in case of some incident.

This work proposes three protocols inspired by

ecological relations for transferring agents between

distinct Embodied MAS. The embedded open MAS

of the Embodied MAS has agents that can enter or

leave in a synchronized way depending on the proto-

col activated. The protocols allows Embodied MAS

to preserve their acquired knowledge during a situa-

tion where a menace exists or even if its physical parts

are damaged. Besides, it is possible to exchange their

knowledge by moving part of their systems to a host

for learning or sharing abilities. The main contribu-

tion is an architecture to create and assemble phys-

ical devices using embedded MAS capable of trans-

ferring agents using protocols inspired by ecological

relations.

The remainder of this paper is organized as fol-

lows. In section 2 the related works are discussed, in

section 3 is presented the the bio-inspired protocols

with each ecological relations implemented. Next, in

section 4 shows the formalization for the bio-inspired

protocols, in section 5 is presented the prototype used

in the experimental evaluation and the results of the

performed tests, and Section 6 shows the conclusion.

2 RELATED WORKS

Some works already use the bio-inspired techniques

and Multi-Agent Systems applied in real environ-

ment (Ferri et al., 2006; Chen et al., 2009). New tech-

nologies and methodologies have emerged (Zeghida

et al., 2018; G

unay et al., 2015) inspired by biology

and aiming to improve the usage of Open MAS.

A framework (G

unay et al., 2015) allows agents

to create a compromise protocol dynamically at run-

time, which allows the agents to change a compro-

mise assumed in the MAS’s design time or assume a

new one. Considering open MAS, this dynamic pro-

tocol allows agents transferred to another MAS to cre-

ate new compromises and update those in the original

MAS at runtime to these agents to better adapt to the

current MAS. Considering bio-inspired MAS, there

is a generic multi-paradigm model for bio-inspired

systems (Zeghida et al., 2018) to compare an agent

with a living being giving them some abilities that

the living being has. These abilities, such as evolu-

tion, self-correction, and decentralized control, im-

prove the agents’ capability to acquire new knowl-

edge and autonomy. Although, if the MAS is dam-

aged, these agents can not preserve their knowledge.

A bio-inspired foraging algorithm is presented in-

volving agents that can cooperate and perform com-

plex tasks by breaking them down into small and sim-

ple sub-tasks (Zedadra et al., 2016). It character-

izes the act of capturing and searching for food in

a particular storage location. Despite being a bio-

inspired application and allowing communication be-

tween agents from different MAS, this work does not

have a mechanism for preserving agents’ knowledge,

so if the MAS is damaged, all knowledge will be lost.

In order to employ MAS in a real environment,

it is necessary to have architectures that allow agents

to communicate with physical devices and a network

for providing a communication channel with others

MAS. ARGO is a customized architecture that inte-

grates microcontrollers and agents to automatically

perform actions and capture percepts in a real envi-

ronment. ARGO agents share the same architecture

of a Jason agent (Bordini et al., 2007), but they can

interface hardware components. This type of agent

cannot communicate with agents from other MAS,

since Jason framework only provide Closed MAS.

For this, Communicator agents uses an IoT architec-

ture and middleware that allow establishing commu-

nication from this speciﬁc type of agent with another

one of the same type hosted in another MAS. Con-

sidering these agents’ characteristics, they can be ex-

ploited to build an approach where an Open MAS

could communicate with other Open MAS, includ-

ing transferring agents based on some ecological rela-

tions implemented by the bio-inspired protocols pro-

posed. These protocols deal with issues that eventu-

ally arise in a physical environment to preserve and

Bio-Inspired Protocols for Embodied Multi-Agent Systems

313

share knowledge.

The open MAS approaches are gaining more im-

portance in the agents’ literature, allowing differ-

ent agents from different MAS to communicate, ex-

change knowledge, and transfer themselves to another

MAS (Golpayegani et al., 2019; Amaral and H

ubner,

2019). Therefore, the open MAS literature is con-

cerned with organizational structure after an agent

transference and the interaction between a transferred

agent and the destination MAS agent. Some questions

such as how will agents be transferred? Why trans-

fer agents? are not being answered, and these issues

are being explored in this work using bio-inspired

protocols for transferring agents aiming at preserving

agents’ knowledge.

3 BIO-INSPIRED PROTOCOLS

This section discusses the bio-inspired protocol pro-

posed based on ecological relations (inquilinism, pre-

dation, and mutualism) to be used along with Embod-

ied MAS. Besides, in section 3.1, we show the details

of engineering the protocols and how their execution

behaves.

In biology, ecological relations (Howe et al.,

1988) are the domain dedicated to explain how liv-

ing beings relate to each other within an environment.

For example, some species can live in collaboration

with other species in the same environment where

both of them get a win-win situation with this inter-

action, or they can compete for resources or life exis-

tence. These interactions occur in a determined place

where several communities of living beings constitute

a balanced and self-sufﬁcient system named ecosys-

tem (Begon et al., 2006). There are several ecological

relations, nevertheless, discussing all of them is out of

the scope of this work. However, three relations have

characteristics that we point to be useful for agents:

Predation, Inquilinism, and Mutualism.

Since MAS are composed of several agents inter-

acting for achieving commons objects or even com-

peting for the same resources in an environment, and

each of them may have roles in the system, then we

could consider that MAS can be interpreted as a com-

munity of agents (Rickel and Johnson, 2000). Fur-

thermore, once it is possible to consider a MAS as

a community of agents, we can extend this concept,

in this work, to be applied to Embodied MAS, which

contains a community of agents responsible for con-

trol physical resources and interacts with other Em-

bodied MAS in a real-world environment (Pantoja

and Viterbo, 2017). Once different Embodied MAS

can interact with each other, one must consider that

they are part of an ecosystem of Embodied agents.

We can deﬁne an Embodied MAS as a physical

device operated by an embedded open MAS com-

posed of multiples agents playing different roles.

Physical agents are responsible for interfacing phys-

ical resources for gathering sensing information and

acting in the real world. Besides, they have the ca-

pability of sharing the collected information to other

agents in the open MAS. The Communicator is a

unique agent responsible for all the external interac-

tions (including mobility). It is also responsible for

activating the bio-inspired protocols. Finally, there

are Traditional agents without any dedicated ability

other than internal communication and interaction be-

tween all types of agents. Based on this deﬁnition,

we assert that these Embodied MAS can make use or

be affected by some of the ecological relations men-

tioned before. In Table 1, it is shown a comparison

between the ecological relations from biology and our

proposed bio-inspired protocols.

The ecological relations are represented as a set

of transfer protocols activated by an Embodied MAS

to provide a mechanism for Communicator agents

to interact with Communicator agents from others

Embodied MAS for transferring agents and preserv-

ing knowledge. How they interact will establish the

ecological relation adopted. For example, in situa-

tions where an Embodied MAS has its physical re-

sources damaged but it contains valuable information

(its mental state), it can activate a protocol for trans-

ferring itself to an existing Embodied MAS with the

same physical resources, taking full control of it. If

the resources are not the same, it can co-exist with

the target Embodied MAS by interacting and learning

from its agents, or it can wait for a new Embodied

MAS appear in the ecosystem with the same kind of

resources to get control over it.

An Embodied MAS (Sender) activates a protocol

for transferring its agents to another Embodied MAS

(Receiver) in these situations. The Sender saves all

agents’ mental states that will be transferred and us-

ing the Communicator agent transfers them to the Re-

ceiver. This step is common for all protocols.

The ﬁnal step depends on which protocol has been

activated. In the predation protocol, all the agents

will be copied from the Sender Embodied MAS —

with their actual mental state — and sent to another

Embodied MAS. All agents of the Sender and Re-

ceiver Embodied MAS will be killed, and the agents

received from the Sender Embodied MAS will be ini-

tialized in the Receiver Embodied MAS. In the in-

quilinism protocol, all agents from the Sender Em-

bodied MAS is sent to the Receiver Embodied MAS,

where they co-exist with the existing ones. In this

ICAART 2021 - 13th International Conference on Agents and Artiﬁcial Intelligence

314

Table 1: Comparison of the ecological relations deﬁnition by biology and the ecological relations in the bio-inspired protocols.

Ecological

relations

For biology (Begon et al., 2006) For bio-inspired protocols

Predation Predation is an inharmonious relation-

ship, where a living being (the predator)

uses of its instincts to hunt and prey on

another living being (the prey).

In the predation relation, all agents are trans-

ferred to another Embodied MAS with the inten-

tion of predating and dominating the second one

eliminating its existing agents to get control of

its physical resources.

Inquilinism Inquilinism is a harmonious relationship,

where a living being (host) serves as

shelter for another one (tenant) without

being harmed. The tenant uses the host’s

body for protection permanently or tem-

porarily.

In the inquilinism relation, an Embodied MAS

sends all its agents to another one for protection

and shelter until a similar Embodied MAS be

identiﬁed. Besides, the transferred agents can-

not interfere in the other agents’ activities.

Mutualism Mutualism is a harmonious relationship,

where both living beings involved in the

relationship beneﬁt from it. This mu-

tual beneﬁt becomes indispensable for

the survival of the involved in the rela-

tionship.

In the mutualism relation, an Embodied MAS

can send an agent, a group of agents, or the en-

tire MAS to another Embodied MAS to learn

or transmit knowledge. The transferred agents

should co-exist during a speciﬁc time, and then

they return to their original Embodied MAS.

case, all agents from the Sender Embodied MAS are

killed. Finally, the mutualism protocol, an agent or a

group of agents will be sent to the Receiver Embod-

ied MAS to exchange knowledge in loco. After that,

these agents should get back to the Sender Embodied

MAS and no agents are killed. The process of activat-

ing the protocols is described in Figure 1.

Figure 1: Transport of agents using the bio-inspired proto-

col.

3.1 Implementing the Bio-Inspired

Protocols

This section describes the proposed bio-inspired pro-

tocol’s implementation, explaining each ecological

relation in the protocol (predation, inquilinism, and

mutualism), and how they are activated. For activat-

ing and executing a protocol, it is necessary to em-

ploy a reliable communication infrastructure where

agents from one Embodied MAS can exchange in-

formation or move to another Embodied MAS. The

ContextNet middleware provides a server-side com-

munication layer using the Scalable Data Distribu-

tion Layer (SDDL) middleware (Endler et al., 2011),

which extends the OMG’s standard Data Distribution

Service (DDS) addressing real-time applications and

embedded systems. It allows multiples devices con-

necting to a network at the same time, dealing with all

network issues such as reconnection and scalability.

Considering this, ContextNet allows multiple Em-

bodied MAS to connect in the same communication

infrastructure for interacting and exchanging informa-

tion. Besides, an Embodied MAS uses ContextNet to

communicate with other Embodied MAS, it can have

some agents capable of moving to another Embodied

MAS, and it also performs a bio-inspired protocol.

This work uses the Jason framework (Bordini

et al., 2007) for programming the embedded open

MAS because it is a well-explored platform and has

a customized agent architecture capable of control-

ling physical resources named ARGO (Pantoja et al.,

2016) to allow programming Physical agents. The

Jason has another customized agent architecture ca-

pable of communicating using the middleware Con-

textNet, an agent with this architecture is named Com-

municator (Pantoja et al., 2018). However, the Jason

does not have a mechanism to transfer agents.

Therefore, we extend the Communicator agent ar-

chitecture to create Open MAS to send and receive

agents from other MAS using the ContextNet. With

ARGO and Communicator agent extended architec-

ture, it is possible to have multiple Embodied MAS

coexisting in the same physical environment con-

Bio-Inspired Protocols for Embodied Multi-Agent Systems

315

nected to a communication infrastructure, character-

izing an Embodied MAS ecosystem. When a bio-

inspired protocol is activated, the Sender Embodied

MAS saves the current mental state (all beliefs, in-

tentions, plans, and goals) of each agent that will be

transferred. It creates a ﬁle to be transferred and used

to initialize these agents in the Receiver Embodied

MAS. Finally, the Receiver Embodied MAS identiﬁes

the protocol type that must be activated and execute.

In the inquilinism protocol, all agents are trans-

ferred, but they are only hosts in the Receiver MAS.

They cannot perform any action, and it is only possi-

ble to interact with existing agents. In the predation,

all Sender MAS’ agents are transferred as well; but it

dominates the Receiver MAS, and all its native agents

are killed, remaining only the agents transferred. In

both past protocols, the Sender Embodied MAS is

deleted for security and privacy purposes. In the mu-

tualism protocol, one or more agents are moved to

another Embodied MAS to learn new skills and then

return after the learning process is over. Afterward,

independently of the protocol chosen, all agents trans-

ferred are instantiated in the Receiver MAS.

To activate the bio inspired protocol, it was de-

veloped an internal action — it is an action that an

agent can perform that does not affect the environ-

ment in Jason framework (Bordini et al., 2007) —

named .moveout. To perform this internal action is

necessary to pass two or three parameters:

• Identiﬁer: it is the unique identiﬁer of a Commu-

nicator agent. This parameter identiﬁes the Em-

bodied MAS that will receive the agents;

• Ecological Relation: one of the three ecological

relations available that can be used;

• Agent Name: it identiﬁes the agents that will be

sent. It is an optional parameter that is only used

when the mutualism relation is activated. In other

relations, all agents are always transferred.

Once the .moveout internal action is called, two

algorithms are executed. The algorithm 1 is exe-

cuted by the Sender Embodied MAS receiving the

identiﬁer of the Communicator agent from the Re-

ceiver Embodied MAS, the ecological relation, and

the agents’ name in case of mutualism. After that,

all selected agents are prepared to be sent, preserv-

ing their actual mental state (beliefs, plans, intentions,

and goals). For this, it accesses the mental state of

all agents selected at runtime. Then, it sends these

selected agents, and wait for a acknowledgment mes-

sage. If they were instantiated in the Receiver Embod-

ied MAS correctly, these agents are killed, otherwise,

the process is aborted. It is important to remark that

kill all selected agents erases the embedded MAS.

Algorithm 1 : Algorithm of the MAS who active the bio-

inspired protocol (Sender Embodied MAS).

1: procedure trans f erAgents(idReceiverMAS,

ecologicalRelation, [agentName] )

2: selectedAgents ← null

3: if ecologicalRelation = ”Predation” or eco-

logicalRelation = ”Inquilinism” then

4: selectedAgents ← getAgents(all)

5: else

6: if ecologicalRelation = ”Mutualism” then

7: selectedAgents ← getAgents

(agentName)

8: end if

9: end if

10: sendAgents(idReceiverMAS,

ecologicalRelation, selectedAgents)

11: if trans f erence = OK then

12: kill(selectedAgents)

13: else

14: abort()

15: end if

16: end procedure

Considering the Receiver Embodied MAS, the Al-

gorithm 2 starts when the Communicator agent de-

tects the ecological relation to be performed, and the

agents received. Firstly, the algorithm creates and ini-

tializes all the received agents. A message is sent back

if all received agents are initialized correctly or not.

Then, if the ecological relation is the predation, all

existing agents at this moment — that were not trans-

ferred by the ecological relation — are killed.

Algorithm 2: Algorithm of the MAS who receive the agents

transferred with the bio-inspired protocol.

1: procedure receiveAgents(ecologicalRelation,

agentsTransfered)

2: createAndInitialize(agentsTrans f ered)

3: if ecologicalRelation = ”Predation” then

4: updateSerialPorts(agentsTrans f ered)

5: kill(myAgents − agentsTrans f ered)

6: end if

7: end procedure

The Embedded MAS uses ARGO for interfacing

and controlling physical resources and to interact with

a real environment using a physical body. It must be

composed of a microcomputer and one or more mi-

crocontroller; the ﬁrst is used for hosting the Embed-

ded MAS, and the microcontroller is used for con-

necting electronic components.

A Physical agent using the ARGO can con-

trol multiple microcontrollers; it has each microcon-

ICAART 2021 - 13th International Conference on Agents and Artiﬁcial Intelligence

316

troller’s serial port for accessing the sensors and ac-

tuators. As the serial port is not ﬁxed, it maintains a

belief for each serial port. Therefore, before all agents

are killed in the Receiver MAS, the transferred ARGO

agents update their serial ports using the ports set in

the Receiver ARGO agents. When developing MAS

in real environments, one must consider the unpre-

dictability and the risks that an Embodied MAS can

be exposed to. Agents are continuously learning and

it could become critical as time passes by. A mech-

anism to preserve agents’ knowledge becomes essen-

tial to MAS in a real environment. Our proposed bio-

inspired protocols aim to overcome this situation.

4 FORMALIZATION FOR THE

BIO-INSPIRED PROTOCOLS

In this section the ecological relations with its corre-

sponding use with the agents are formalized.

Deﬁnition 4.1 (Bio-inspired Protocol). A Bio-

Inspired Protocol (or Bio TP) is deﬁned by the fol-

lowing 3-tuple:

Bio

TP = hSender, Receiver, APi

where,

• Sender is a MAS comprised by the following sets

of Agents:

Sender = hPA

, TA

, CA

where,

is a set of n Physical Agents from the

Sender, where n ≥ 0.

is a set of n Traditional Agents from the

Sender, where n ≥ 0.

is a set with a single agent named Commu-

nicator from the Sender.

• Receiver is a MAS comprised by the following sets

of Agents:

Receiver = hPA

, TA

, CA

where,

is a set of n Physical Agents from the Re-

ceiver, where n ≥ 0.

is a set of n Traditional Agents from the

Receiver, where n ≥ 0.

is a set with a single agent named Commu-

nicator from the Receiver, responsible for es-

tablishing the communication between the Re-

ceiver and a given Sender.

• AP (or Activate Protocol): the third element from

the tuple deﬁnes which kind of bio-relation will be

executed by the protocol; There are three possible

kinds: Predation, Inquilinism, and Mutualism. As

it follows, we describe these three different proto-

cols executions using set theory formalisation.

1. Predation

= PA

− PA

; TA

= TA

− TA

= CA

−CA

0; TA

0; CA

NB-1: In the Predation relation the Sender

takes control over the Receiver. That is the

reason we use difference of sets in the ﬁrst line

above. In the third line all three sets are as-

signed as empty because the agents from the

Sender should be removed.

2. Inquilinism

= PA

∪ PA

; TA

= TA

∪ TA

= CA

∪CA

0; TA

0; CA

NB-2: In the Inquilinism all agents from the

Sender are merged with agents from the Re-

ceiver. That is why we have used union of sets

above. Similarly, in the third line (above) all

three sets are assigned as empty.

3. Mutualism

{PA

= PA

∪PA

;PA

= PA

∪PA

} | PA

{TA

= TA

∪ TA

;TA

= TA

∪ TA

} |

= TA

= CA

;

NB-3: In the Mutualism there is an option that

could be selected for the new set of PA

and

, i.e., it may occur a merge of sets (us-

ing union of sets) or the Sender’s agents may

remain in the Sender and they are not trans-

ferred to the Receiver. Besides all Sender’s

agents should be preserved, thus there is no

empty set operation. Notice that in case the

agents from the Sender have been sent to the

Receiver, then at some point, they should be

sent back to the Sender. This is represented by

the operations placed between brackets.

5 PROTOTYPE: DESIGNING AND

TESTING

This section presents the tests and results of the bio-

inspired protocols proposed. The tests use the concept

of autonomous underwater vehicles for each of the

Bio-Inspired Protocols for Embodied Multi-Agent Systems

317

proposed bio-inspired protocols. Autonomous Un-

derwater Vehicles (AUV) (Cruz, 2011) are crewless

vehicles with internal processing that autonomously

perform actions based on sensors’ information. AUV

are regularly used to explore the sea’s areas where hu-

mans cannot go. Considering this, our mechanism for

preserving the collected information could become

vital. The tests performed in this work show scenarios

where each bio-inspired protocol can be applied in an

AUV application as an Embodied MAS.

The ﬁrst scenario is characterized by two Embod-

ied MAS with exactly the same hardware exploring a

new sea area. One of these Embodied MAS (Leader)

has more important knowledge than the other (Sol-

dier). During the exploration, the Leader’s AUV is

damaged, and to protect the integrity of its knowl-

edge, it activates the bio-inspired protocol using the

predation relation and preys the Soldier. The sec-

ond scenario the Soldier’s AUV is damaged and has

more important knowledge than the Leader Embodied

MAS. Hence, the Soldier activates the bio-inspired

protocol with the inquilinism relation to transfer all its

agents to the Leader and remains tenant until another

AUV with the same hardware is available to the Sol-

dier control. Since it is the Soldier, and by hierarchy

purposes, it cannot prey its Leader. In the third sce-

nario, a new Embodied MAS (Student AUV) is sent

on an exploration mission with a more experienced

Embodied MAS (Teacher AUV), who knows the most

part of the area explored. In this case, the Student ac-

tivates the bio-inspired protocol with the mutualism

relation for sending a group of agents to the Teacher

to learn; afterwards, they return to the student MAS

with the knowledge acquired to perform a better ex-

ploration.

Two prototypes of autonomous vehicles were de-

veloped. These vehicles are equals considering the

hardware composition, and each one is composed of 1

Raspberry pi zero microcomputer and 1 Arduino Uno.

Besides, they employ 1 luminosity sensor (LDR), 1

temperature sensor (LM35), 2 white LEDs, 2 DC 3-

6V motors (1 for each wheel), 1 H bridge driver mod-

ule (L298N), and 2 power banks (one for the micro-

computer and the other for the microcontroller). All

these components are interconnected in a circuit com-

posed of resistors and capacitors connected to the Ar-

duino Uno. In ﬁgure 2 shows the prototypes.

5.1 Test Results

All developed MAS for the test scenarios had ver-

sions with different amount of agents — 10, 30, and

50 agents — to check if exists interference in the

transfer speed. Besides, we tested the effectiveness

Figure 2: Car prototypes for testing in a real world.

of performing the agents’ transferring and checked if

the agents’ knowledge was preserved. Considering

that all agents are always transferred in the predation

and inquilinism relations, the tests with mutualism as-

sumed the worst-case transferring all agents.

For each scenario, all MAS versions (10, 30, and

50 agents) were tested 10 times, resulting in 30 tests

per scenario and 90 tests combining all three sce-

narios. Moreover, an internet with 5 Mbps speed

was used for the ContextNet middleware to create the

communication infrastructure. The Table 2 shows the

transfer speed average for each scenario including the

ecological relation applied. Finally, the efﬁciency of

transport and knowledge preservation is shown.

Table 2: A comparative table considering the three scenar-

ios.

Time to complete

Eco. Rel. 10

Ag.

Eff.

Predation 0.43 s 0.74 s 1.46 s 100%

Inquilinism 0.39 s 0.70 s 1.03 s 100%

Mutualism 0.38 s 0.69 s 1.03 s 100%

When analyzing the results, we conclude that the

number of agents directly interferes with the proto-

cols’ execution. When the protocol sends multiples

agents to another MAS, it must access the entire base

of beliefs, desires, intentions, and plans of each agent

to transport. After that, when these agents arrive at

the Receiver MAS, they are instantiated individually.

Considering the scenarios and the results, we con-

clude that the predation is more appropriate when

the Embodied MAS hardware is damaged, and it still

needs to continue logically operating, not only saving

the agents’ knowledge because this relation is slower

than the others and requires another Embodied MAS

with exact the same hardware. The inquilinism suits

in cases where the Embodied MAS’s hardware is de-

teriorating, there is no other similar hardware, or there

is no need for it to remain operational. The main goal

is to preserve the agents’ knowledge. This relation is

one of the fastest and can be performed independently

of the MAS physical conﬁguration.

Finally, the mutualism relation is appropriate for

ICAART 2021 - 13th International Conference on Agents and Artiﬁcial Intelligence

318

cases where the Embodied MAS needs to learn new

knowledge, and when communicating does not apply

or is not secure. For example, when installing a new

sensor in the AUV scenario, the student MAS sends

a physical agent to the teacher MAS to learn with its

physical agents. Then, the transferred agent returns

with the knowledge to operate the sensor.

6 CONCLUSION

This work presented bio-inspired protocols for Em-

bodied MAS inspired by ecological relations. It

shows the importance of preserving the integrity of

the agents’ knowledge depending on the situation that

an Embodied MAS could be exposed to when it is

inserted in real environments where maintaining the

knowledge can be crucial to the mission’s success.

Three protocols based on ecological relations

were implemented: Predation, Mutualism, and In-

quilinism. In predation, all agents of a MAS is trans-

ferred to another MAS to dominate it. In this situa-

tion, the Sender MAS has crucial knowledge in a mis-

sion; nevertheless, it is physically damaged, then they

transfer their agents to a MAS with similar hardware

and dominates it. In mutualism, some MAS agents

can be transferred to another MAS to learn or ex-

change knowledge. After that, they can return to the

Sender MAS to share the new knowledge acquired

during the transference. In Inquilinism, all agents

from the Sender MAS are transferred to another MAS

to preserve their knowledge without dominating it.

As future work, we aim to implement an ecosys-

tem of Embodied MAS in the computer lab of our

university to broaden the testing environment and the

number of devices involved. Moreover, a new feature

called the dominance degree is being developed. The

degree of dominance maps all possible predation re-

lation between Embodied MAS. Thus, an Embodied

MAS with less dominance degree can not prey on the

other, but the bio-inspired protocol will automatically

change the relation to inquilinism if it tries.

REFERENCES

Amaral, C. J. and H

ubner, J. F. (2019). Goorg: Automated

organisational chart design for open multi-agent sys-

tems. In International Conference on Practical Ap-

plications of Agents and Multi-Agent Systems, pages

318–321. Springer.

Begon, M., Townsend, C. R., and Harper, J. L. (2006). Ecol-

ogy: from individuals to ecosystems. Blackwell Pub-

lishing.

Bordini, R. H., H

ubner, J. F., and Wooldridge, M. (2007).

Programming Multi-Agent Systems in AgentSpeak us-

ing Jason. John Wiley & Sons Ltd.

Chen, B., Cheng, H. H., and Palen, J. (2009). Integrating

mobile agent technology with multi-agent systems for

distributed trafﬁc detection and management systems.

Transportation Research Part C: Emerging Technolo-

gies, 17(1):1–10.

Cruz, N. (2011). Autonomous underwater vehicles. BoD–

Books on Demand.

Endler, M., Baptista, G., Silva, L., Vasconcelos, R.,

Malcher, M., Pantoja, V., Pinheiro, V., and Viterbo,

J. (2011). Contextnet: context reasoning and shar-

ing middleware for large-scale pervasive collaboration

and social networking. In Proceedings of the Work-

shop on Posters and Demos Track, page 2. ACM.

Ferri, G., Caselli, E., Mattoli, V., Mondini, A., Mazzolai,

B., and Dario, P. (2006). A biologically-inspired al-

gorithm implemented on a new highly ﬂexible multi-

agent platform for gas source localization. In Biomed-

ical Robotics and Biomechatronics, 2006. BioRob

2006. The First IEEE/RAS-EMBS International Con-

ference on, pages 573–578. IEEE.

Golpayegani, F., Dusparic, I., and Clarke, S. (2019). Using

social dependence to enable neighbourly behaviour in

open multi-agent systems. ACM Transactions on In-

telligent Systems and Technology (TIST), 10(3):1–31.

unay, A., Winikoff, M., and Yolum, P. (2015). Dynam-

ically generated commitment protocols in open sys-

tems. Autonomous Agents and Multi-Agent Systems,

29(2):192–229.

Howe, H. F., Westley, L. C., et al. (1988). Ecological re-

lationships of plants and animals. Oxford University

Press.

Huynh, T. D., Jennings, N. R., and Shadbolt, N. R. (2006).

An integrated trust and reputation model for open

multi-agent systems. Autonomous Agents and Multi-

Agent Systems, 13(2):119–154.

Pantoja, C. E., Soares, H. D., Viterbo, J., and El Fallah-

Seghrouchni, A. (2018). An architecture for the de-

velopment of ambient intelligence systems managed

by embedded agents. In SEKE, pages 215–214.

Pantoja, C. E., Stabile Jr, M. F., Lazarin, N. M., and Sich-

man, J. S. (2016). Argo: A customized jason ar-

chitecture for programming embedded robotic agents.

Fourth International Workshop on Engineering Multi-

Agent Systems (EMAS 2016).

Pantoja, C. E. and Viterbo, J. (2017). Prototyping ubiqui-

tous multi-agent systems: A generic domain approach

with jason. In International Conference on Practi-

cal Applications of Agents and Multi-Agent Systems,

pages 342–345. Springer.

Rickel, J. and Johnson, W. L. (2000). Task-oriented collab-

oration with embodied agents in virtual worlds. Em-

bodied conversational agents, pages 95–122.

Wooldridge, M. J. (2000). Reasoning about rational agents.

MIT press.

Zedadra, O., Seridi, H., Jouandeau, N., and Fortino, G.

(2016). A cooperative switching algorithm for multi-

Bio-Inspired Protocols for Embodied Multi-Agent Systems

319

agent foraging. Engineering Applications of Artiﬁcial

Intelligence, 50:302–319.

Zeghida, D., Meslati, D., and Bounour, N. (2018). Bio-ir-m:

A multi-paradigm modelling for bio-inspired multi-

agent systems. Informatica, 42(3).

ICAART 2021 - 13th International Conference on Agents and Artiﬁcial Intelligence

320