Implementing a Cooperative Framework Among

Bio-inspired Robots Based on Phonotaxis

Jos

e Camacho

, Rosa M. de Molina

, Eugenio Mart

ın

and Mart

ın Mellado

Instituto de Autom

atica e Inform

atica Industrial, Universidad Polit

ecnica de Valencia,

Camino de Vera, s/n. 46022 Valencia.

Dept. Biolog

ıa Animal y Ecolog

ıa, Facultad de Ciencias,

Universidad de Granada, 18071 Granada

Abstract. This paper proposes the development of a cooperative framework

among mobile robots, inspired in the phonotaxic behavior (tracking the source of

a speciﬁc sound) observed in cricket mating. By means of this behavior, in com-

bination with two other individual behaviors (for communication and obstacles

avoidance) a set of ﬁve cooperation primitives is reached. A simulation platform

has been used to test the design. Furthermore, two real robots, one acting as the

female and the other one as the male, have been developed and tested: male emits

a calling song (at a speciﬁc frequency) and female tracks or moves away from the

sound source.

1 Introduction

Cooperation allows robots to deal with large-scale problems obtaining two principal ad-

vantages over Single-Agent Systems: improved performance and fault tolerance [14].

Multi-Agent Systems are based on the thought that a group of robots, of little func-

tionality, can obtain better performance than a single highly qualiﬁed robot in complex

domains, such as planetary science tasks [6] or cleaning tasks [9]. However, coordinat-

ing that kind of systems is a complex problem.

Nowadays, Multi-Agent Systems have a good test framework called RoboCup Chal-

lenge [3] [10], the robotic soccer competition. Numerous solutions proposed there [17]

[18] use a hierarchical architecture for robots design, with a low level made up of reac-

tive behaviors and a high level or higher levels for more complex tasks.

Mataric [13] proposes the use of basis behaviors or cooperation primitives, building

blocks from which cooperative behaviors emerge. Although it is stated that the choice

of the basis behaviors depends on the domain and the goals, the next set was chosen:

Safe-Wandering, Dispersion, Aggregation, Homing and Following. Other behaviors,

such as group navigation (ﬂocking) or searching (foraging), can be obtained by combi-

nations of the basis set. Mataric proposes two types of combinations: complementary,

for concurrent output behaviors, and contradictory, for mutually exclusive behaviors.

To develop the basis behavior set mentioned, robots have to be able to distinguish

the members of its group from the environment obstacles and to know their position.

Camacho J., M. de Molina R., Martín E. and Mellado M. (2005).

Implementing a Cooperative Framework Among Bio-inspired Robots Based on Phonotaxis.

In Proceedings of the 1st International Workshop on Multi-Agent Robotic Systems, pages 31-38

DOI: 10.5220/0001192600310038

 SciTePress

Phonotaxis (tracking the source of a speciﬁc sound) provides robots with these capabil-

ities, as well as individuals and groups identiﬁcation. At this point, the research work

carried out by Barbara Webb et al. [19] [11] [12], where cricket phonotaxis is studied

using robotics as a tool, has been the main reference of this work. Behaviors studied by

biology are an inspiration source for cooperative robotics [1][2][5].

By using phonotaxis in several ways and by combining it with other individual

behaviors (communication and navigation) it will be shown that it is possible to generate

the ﬁve cooperation primitives, obtaining an alternative to other methods, such as those

used by Mataric [13]. The aim of this paper is to prove the usefulness of phonotaxis

in the area of cooperative robotics, as a complementary mechanism to other proposals

offered in the literature.

This paper is organized as follows. Section 2 details the materials and methods used.

Section 3 introduces individual behaviors. Section 4 describes the cooperation primi-

tives implementation. Section 5 presents the experimental result from both simulation

and real testing. Section 6 gives some conclusions, remarks and future research lines.

2 Materials and Methods

The proposed cooperative framework has been tested at two levels. As a ﬁrst step, the

cooperation primitives set was implemented for a homogeneous system of n (dynami-

cally set) robots in a simulation framework, Virtual Robot Simulator (VRS) [15]. Af-

terwards, a pair of real robots, one acting as the cricket female and the other one as the

male in the cooperative behaviors, were developed.

In both cases, robots show a behavior-based architecture [2] [4]. Each independent

behavior is in charge of a speciﬁc task, and gets inputs from sensors and its outputs

rule actuators. New behaviors can be added incrementally to the system with this archi-

tecture. Robots can achieve three individual behaviors: phonotaxis, navigation avoiding

obstacles and communication. The latter two have been chosen for two reasons: a) they

are important or even fundamental for autonomous operation and b) they are useful to

test phonotaxis in the cooperative robotics area. The cooperative primitives are imple-

mented at a higher level.

In simulation, each robot is modelled with a thread. Threads only communicate with

the environment (main thread of VRS) by the simulated sensors and actuators, as robots

do in the real world. For more detail, see [16].

Real robots made possible to test whether sound is a valid way of identifying and

locating mates in the real world. These robots have a small, exoskeleton type structure

mainly made of hard plastic. They use a differential drive platform with two wheels at

the back and a caster one at the front. This paper presents the results of the experiences

developed with a pair of real robots. A homogeneous group of more than two robots is

being developed and the real multi-robot framework will be studied in the future.

3 Individual Behaviors

The combination of the individual behaviors depends on the objectives. Thus, a speciﬁc

behavior management algorithm, which deﬁnes the (rigid) combination of individual

behaviors, is needed for each primitive. Although, in principle, the way of implement-

ing these individual behaviors should not affect the resultant framework, a brief com-

ment on this implementation is made below. Of course, the speciﬁc implementation is

determined by the features and capabilities of the robots used.

Phonotaxis provides robots with the capability to locate and identify individuals and

groups. Two sub-behaviors have been implemented, inspired in crickets mate.

The male behavior consists in generating the calling song, in order to attract or

repel ”conspeciﬁc” females. This song is made of: syllables, bursts of sound with a

species-speciﬁc carrier frequency; chirps, each one with a number of syllables; and

silent spaces between chirps [12]. Thus, the calling song can be parameterized by [7]:

carrier frequency, syllable rate, number of syllables per chirp, duty cycle and chirp rate.

In the cooperative framework proposed, the song of a robot codiﬁes its identity at

syllable level. The identiﬁcation of a robot contains two codes: group and individual

code. The individual code starts from 0 and grows until the number of robots of the

group. A sung syllable represents a logic 1 and a non-sung syllable is a logic 0. First

syllable in a chirp has to be sung, the rest can be used to store the identiﬁcation code.

The female behavior consists in tracking (or moving away from) the male. First

of all, females need the recognition capability of their species (group) speciﬁc calling

song. Carrier frequencies different to the speciﬁcs of its group are rejected by hardware

ﬁltering. The other song parameters can be detected by software.

An explicit communication model has been implemented. Thus, not only robots

know of mates existence, but a communication protocol is also established [5]. The

communication sensor uses infrared technology. The protocol has been implemented

over Philips RC5 code, so that a remote control can be used to send commands or

communicate with the robots.

To achieve navigation avoiding obstacles, two types of sensors have been added to

the robots: bumpers and infrared sensors. The readings of these sensors are the inputs

of a Fuzzy Rules Based Navigation System [8] with feedback variables. The feedback

variables take their value from previous decisions of the system, so that robots are able

to maintain continuous movement and avoid hesitation.

4 Cooperative Primitives

The ﬁve basis cooperative behaviors proposed by Mataric [13] have been developed to

make up the cooperative framework. Phonotaxis function has to be outlined because of

its locating and identifying capabilities. Moreover, as it will be shown later, phonotaxis

gets even more useful for primitives development because of its distance limitation,

combined with infrared communication distance limitation. The maximum distance for

sound reception has to be longer than the maximum distance for the infrared communi-

cation. Next, the basis cooperative behaviors are explained.

Safe wandering is deﬁned as the ability of a group of agents to move about while

avoiding collisions with obstacles and each other. To achieve this task, the main chal-

lenge for a robot is avoiding moving mates. In infrared communication, the emitter has

to be oriented towards the receiver. This feature makes this type of communication very

useful for safe wandering. If a robot receives a communication, it understands that the

emitter robot is moving towards it. To avoid collision, the receiver robot emits a calling

song which codiﬁes the emitter’s identiﬁcation (sent in the communication), so that this

one knows that it has to change its movement direction. Robots should communicate

their identiﬁcation in the same direction of movement of the wheels. Therefore, a pair

of communication emitters, one looking forward and one looking backward, are needed

to emit when moving forward and backward, respectively.

Dispersion is the ability of a group of agents to spread out in order to establish and

maintain some minimum inter-agent distance. This minimum distance, in our frame-

work, is the maximum distance for calling song detection. Each robot in turn sings and

listens. It sings so the rest can move away from it, and listens to move away from the

rest. Every robot adopts a movement trajectory in the opposite direction to the sound

source, made up of orientation (by using phonotaxis behavior) and translation periods

(by using navigation). The turns in orientation periods have a constant component and

a random component, so the turning angle changes. This prevents from inﬁnite loops.

Aggregation is deﬁned as the ability of a group of agents to gather in order to es-

tablish and maintain some maximum inter-agent distance. In this case, the maximum

distance is the maximum distance for communication detection. When a robot wants

the rest to approach, it starts singing. The other robots which detect the calling song try

to get to the ﬁrst one adopting a movement trajectory towards the sound source, made

up of orientation (by using phonotaxis behavior) and translation periods (by using nav-

igation). Before each navigation period, robots send a ”Hello Request” command to

establish communication with the calling robot. If any of them is close enough to this

robot, it receives the communication request, stops singing and establishes communi-

cation. Once the communication protocol is ﬁnished, the calling robot will start singing

again unless all its mates have arrived.

Following is deﬁned as the ability of an agent to move behind another retracing

its path and maintaining a line or queue. Probably, This is the most complex primitive

to perform. To establish a queue, one assumption has to be done: each robot knows

the identity of the immediate preceding and following robots or whether it is the ﬁrst

or the last one in the queue. The robots form a queue according to their identiﬁcation.

Thus, the previous assumption is satisﬁed if each robot knows its own identiﬁcation and

the number of robots in the group. Again, this information is needed to codify calling

songs, so the assumption is deﬁnitely true.

Every robot, except the ﬁrst one, waits until hearing its predecessor’s calling song

to start the algorithm, and it does not react to the calling song of other robots. When it

hears the proper song, it turns towards the sound source by using phonotaxis behavior.

In this only primitive, the turns are done by moving only a wheel, so that the robot grav-

ity center moves. These little translations make possible for the robot to recover the trail

when it has lost. During this process, the robot in turn rotates and sends its identiﬁca-

tion through the communication sensor. If the predecessor receives the identiﬁcation, it

understands that the next one is properly oriented. Thus, it stops singing and navigates

for a time interval. This algorithm is repeated for every pair of consecutive robots from

the front to the end of the queue. When the last but one robot stops singing, the last one

waits for a delay (time supposed for predecessor navigation) and then navigates.

Fig.1. Simulation tests: (a) Safe wandering, (b) Dispersion, (c) Aggregation, (d) Following, (e)

Homing.

Finally, we can deﬁne homing as the ability to ﬁnd a particular region or location.

This behavior is very similar to the aggregation. Actually, the only difference is the

calling element, a static base, which is used instead of the calling robot. The fact that

a static base is needed to locate a place can be seen as a disadvantage. If there is no

base located in the place the robots have to reach, one of the group of robots has to

perform another homing strategy, like greedy local pursuit, used in [13]. In this case,

one cooperative homing strategy can be the combination of an aggregation around that

robot and a following towards the desired location.

5 Experimental Results

5.1 VRS Simulation

In this section, the results obtained in the simulation of a group of four robots are shown.

The experiments were performed from a challenging initial formation. In the safe wan-

dering test (Figure 1(a)), the four robots were set out in a straight line, the ﬁrst two

looking towards the other two and vice versa. In the dispersion test (Figure 1(b)), the

robots show an initial formation where every one is facing the same direction and they

are close to each other. In the aggregation test, the calling robot stayes at the top of the

image (Figure 1(c)) and the other robots are facing the opposite direction to the center

of the image. In the following test (Figure 1(d)) the robots initially are arranged in a

queue, in order to pass through the columns. The formation of a queue starting from an

arbitrary formation is tested in the homing test (Figure 1(e)). Homing can be seen as a

particular aggregation. Actually, it is exactly the same if we use a robot of the group to

mark the desired location. Therefore, the test performed for aggregation can be valid for

homing, too. In section 4, it was proposed an aggregation phase followed by a following

Fig.2. Real robots (a) and tests: (b) Safe wandering, (c) Dispersion, (d) Aggregation, (e) Follow-

ing, (f) Homing.

phase to achieve homing without a marking base. The automatic formation of a queue

of robots is fundamental for this deﬁnition of homing and this is why it has been tested.

The outcomes show that the framework based on a phonotaxic behavior works prop-

erly. Robots are able to avoid themselves, to join or separate and to form and maintain

a queue while avoiding obstacles.

5.2 Real Environment

The ideal system consists of a group of n robots which dynamically adopt the male or

female role, depending on the cooperative behavior they are performing, their role in

that behavior and/or the part of the management algorithm they are executing. The ideal

system performance has been tested in the simulation, but no noise was implemented.

Now, sound as a feasible communication medium in real environments has to be proved.

Two robots have been developed (Figure 2(a)), one with the male capabilities and one

with the female capabilities. The experiences in the real world were performed in a big

(10 meters x 4 meters), closed room, with white furniture. The presence of furniture is

important to evaluate the effect of the bounces of the sound. The environment of the test

is not specially noisy, as those of some applications for industrial automation. Dealing

with noisy environments is part of the future work to be done. All the experiences shown

were ﬁlmed and then represented as drawing pictures to illustrate motions

In the tests done, only the female robot navigates because it is the only robot able

to react to sound. The safe wandering tests (Figure 2(b)) show how the female reacts to

sound (light triangles) and then navigates avoiding objects. In the ﬁrst experience, the

female robot even has to turn all the way round. Dispersion, aggregation and homing

tests are very similar. The female robot moves away from (Figure 2(c)) or towards to

(Figures 2(d) y 2(f)) the male adopting the typical zig-zag movement of the crickets [7].

Figure 2(d) show the two phases of the aggregation: ﬁrst the female gets closer to the

male and then, once it is close enough, the communication protocol is established. Note

that homing is very similar to the aggregation but with no second phase. For the other

deﬁnition of homing (aggregation + following), the queue formation of the two robots

Real videos can be downloaded from: http://ttt.gan.upv.es/ jcamacho/crickets/videos.

(not shown) is achieved just by steering the male robot after the aggregation. Finally, in

Figure 2(e), the female and male position and orientation for every time interval during

the following test are shown. Odd intervals are drawn in dark color while even intervals

are shown in light color.

As it can be seen, the primitives are performed correctly by the female robot, but

there are some erratic decisions in the phonotaxis. Several consecutive errors in the

phonotaxis sensor can cause non expected behaviors, as the turn towards the top in

Figure 2(c). The error percentage has been empirically calculated as 13.2% (without

echo), a high error value. This is, nonetheless, a good result which suggests that the

framework would get even better performance with a more detailed design of the sensor.

6 Conclusion

This paper is another contribution to demonstrate that bio-inspired behaviors are an

alternative methodology. There is a current of opinion which claims that complex plan-

ning methods can be replaced by biologically observed behaviors, which emerge from

the combination of simple behaviors. Probably, the best way to obtain intelligence is by

imitating it. Based on the results obtained, it can be concluded that:

1. Phonotaxis is a useful behavior for mobile robotics in environments where sound

is a feasible communication medium.

2. A cooperative framework, made of phonotaxis and other individual behaviors, has

been developed and tested in simulation and real world, in this latter case with two

robots. This framework is mainly based on the combination of two communication

methods: one directional (infrared) and one omnidirectional (sound).

From the second conclusion, it can be stated that the system keeps functional what-

ever two communication methods are used, as long as one is directional and the other

one is omnidirectional and the distance limitation of these methods is proper. The meth-

ods choice will depend on the environment characteristics. In this paper, sound was

chosen as the omnidirectional communication method. Sound can be an alternative to

visual techniques when the number of obstacles reduces the performance. Future work

is mainly aimed at:

a) A real, homogeneous, n robots system development.

b) Improving the sound sensor and testing the framework in noisy environments.

c) Comparing this framework with visual cooperative systems in environments with

many and/or big obstacles.

d) More complex cooperative tasks development using cooperative primitives.

Generalization of the real framework for n robot seems to be a very challenging

task. Two main problems have to be addressed: codiﬁcation of the identiﬁcation in the

sound (at syllable level) and interferences among songs. In simulation, the coopera-

tive system lose efﬁciency as more robots are included in the same environment. To

avoid interferences, a Frequency Division Multiplexing (FDM) is proposed. Moreover,

as these ”cricket” robots, by deﬁnition, only react to calling songs with speciﬁc para-

meters, different groups of robots can act over the same environment by using different

parameters songs.

Acknowledgements

This work has been partially funded by FEDER-CICYT project with reference DPI2002-

04434-C04-04 and by the FPU grants program, Secretar

ıa de Estado de Educaci

on y

Universidades (Ministry of Education and Science, Spain). Authors wish to thank An-

tonio Gonz

alez Mu

noz, Luis Javier Herrera Maldonado and Pedro Latorre Carmona.

Special thanks to Fernando Camacho P

aez, El Golico.

References

1. Arai, T., Pagello, E., Parker, L.E. (ed.): Advances in Multi-Robot Systems. IEEE Transactions

on Robotics and Automation, Vol. 18 (2002) 655–661

2. Arkin, R.C.: Behavior-Based Robotics. 3rd edn. Hardcorver, E.E.U.U. (2000)

3. Asada, M., Kuniyoshi, Y., Drogoul, A., Asama, H., Mataric, M., Duhaut, D., Stone, P., Kitano,

H.: The RoboCup Physical Agent Challenge: Phase-I. Applied Artiﬁcial Intelligence, Vol. 12

(1998) 251–263

4. Brooks, R.A.: A Robust Layered Control System for a Mobile Robot. IEEE Journal of Robot-

ics and Automation, Vol. 2 (1986) 14–23

5. Cao, Y.U., Fukunaga, A.S., Kahng A.B.: Cooperative Mobile Robotics: Antecedents and Di-

rections. Autonomous Robots, Vol. 4 (1997) 1–23

6. Estlin, T., Yen, J., Petras, R., Mutz, D., Castao, R., Rabideau, G., Steele, R., Jain, A., Chien,

S., Mjolsness, E., Gray, A., Mann, T., Hayati, S., Das, H.: An Integrated Architecture for

Cooperating Rovers. 5th International Symposium on Artiﬁcial Intelligence, Robotics and

Automation in Space (i-SAIRAS) (1999) 255–262

7. Huber, F., Thorson, J.: Cricket auditory communication. Scientiﬁc American, Vol. 253 (1985)

46–54

8. Vadiee, N., Jamashidi, M., Ross, T.J.: Fuzzy Logic and Control: Software and Hardware Ap-

plications. Prentice Hall, Englewood Cliffs, E.E.U.U. (1993).

9. Jung, D., Cheng, G., Zelinsky, A.: Experiments in Realising Cooperation between Au-

tonomous Mobile Robots. Proceedings International Symposium on Experimental Robotics

(ISER) (1997)

10. Kitano, H., Tambe, M., Stone, P., Veloso, M., Coradeschi, S., Osawa, E., Matsubara, H.,

Itsuki, N., Asada, M.: The RoboCup Synthetic Agent Challenge. Proc. of IJCAI (1997) 24–29

11. Lund, H.H., Webb, B., Hallam, J.: A Robot Attracted to the Cricket Species Gryllus bimac-

ulatus. Proc. of 4th European Conference on Artiﬁcial Life (1997)

12. Lund, H.H., Webb, B., Hallam, J.: Physical and Temporal Scaling Considerations in a Robot

Model of Cricket Calling Song Preference. Artiﬁcial Life, Vol. 4 (1998) 95–107

13. Mataric, M.J.: Designing and Understanding Adaptive Group Behavior. Adaptive Behavior,

Vol. 4 (1995) 51–80

14. Mataric, M.J., Sukhatme, G., Ostergaard, E.: Multi-Robot Task Allocation in Uncertain En-

vironments. Autonomous Robots, Vol. 14 (2003) 255–263

15. Mellado, M., Correcher, C., Catret, J.V., Puig, D.: VirtualRobot: An open general-purpose

simulation tool for Robotics. European Simulation and Modelling Conference, (2003)

16. Mellado, M., Correcher, C., de Molina, R.M., Camacho, J., Benet, G.: Simulation of mobile

robot applications with VirtualRobot. International Industrial Simulation Conference (2005)

17. Stone, P., Veloso, M.: A Layered Approach to Learning Client Behaviors in the RoboCup

Soccer Server. Applied Artiﬁcial Intelligence, Vol. 12 (1997) 165–188

18. Veloso, M., Stone, P., Han, K.: CMUnited-97: RoboCup-97 Small-Robot World Champion

Team. AI Magazine, Vol. 19 (1998) 61–69

19. Webb, B.: A robot cricket. Scientiﬁc American, Vol. 275 (1996) 94–99