1

Shinerbot: Bio-inspired CollectiveRobot Swarm Navigation Platform

Enyu Luo, Xin Hui Fang, Yuting Ng and Grace Xingxin GaoUniversity of Illinois at Urbana-Champaign

BIOGRAPHIES

Enyu Luo is a Master’s student in the Electrical andComputer Engineering Department at the University ofIllinois at Urbana-Champaign. He received his B.S. degree inelectrical engineering from the same university. His researchinterests include robotics and UAVs.

Xin Hui Fang is an undergraduate student in the Electricaland Computer Engineering Department at the University ofIllinois at Urbana-Champaign. Her research interests includerobotics and UAVs.

Yuting Ng is a graduate student in the AerospaceEngineering Department at the University of Illinois atUrbana-Champaign. She received her B.S. degree in electricalengineering, graduating with university honors, from thesame university. Her research interests are in advanced signaltracking, navigation, control, robotics, RADAR and UAVs.

Grace Xingxin Gao received her B.S. degree in mechanicalengineering and her M.S. degree in electrical engineeringfrom Tsinghua University, Beijing, China in 2001 and 2003.She received her PhD degree in electrical engineering fromStanford University in 2008. From 2008 to 2012, she wasa research associate at Stanford University. Since 2012, shehas been an assistant professor in the Aerospace EngineeringDepartment at University of Illinois at Urbana-Champaign.Her research interests are systems, signals, control, androbotics. She is a senior member of IEEE and a member ofION.

Abstract—We designed and built a collective robot swarmnavigation platform, called Shinerbot, inspired by the emer-gent navigation behavior of the Golden Shiner Fish. Unliketraditional navigation robots that require location information,path-planning and communication between networked elements,each Shinerbot performs only two navigation operations. EachShinerbot modulates its speed based on sensing of its environmentat its current location. In addition, each Shinerbot moves towardsneighboring Shinerbots. Both our Shinerbot speed and directionincorporate some randomness. As a swarm, our Shinerbotscollectively navigate using minimal sensing and control.

We designed our Shinerbot swarm navigation platform touse vibration motors for mobility, a photodiode for environment

Enyu Luo and Xin Hui Fang are with the Department of Electrical and Com-puter Engineering; Yuting Ng and Grace Xingxin Gao are with the Departmentof Aerospace Engineering, University of Illinois at Urbana-Champaign, Ur-bana, IL 61801, USA. E-mails: enyuluo1@illinois.edu, xfang14@illinois.edu,yng5@illinois.edu, gracegao@illinois.edu.

light intensity sensing, reflective ranges for neighbor proximitysensing and a microcontroller for processing. In addition, to makelarge-scale swarm operations tractable, we designed a swarmmessaging system and a swarm charging plate.

We built 30 Shinerbots, each of size 40.6mm and of costUSD $20. This process includes drawing the circuit schematics,producing the PCB layouts, sending the PCB for fabrication,soldering the electronic components, writing custom software,programming the Shinerbots, calibration and testing. Follow-ing that, we experimentally demonstrated successful Shinerbotswarm navigation.

I. INTRODUCTION

In our investigation of collective intelligence, we are inspiredby swarms in nature, such as a colony of ants, a flock of birdsand a school of fish. These animal swarms achieve complexnavigation performances despite each individual in the swarmsperforming only minimal sensing and control.

The Golden Shiner is a type of fish that schools and prefersdark areas [1]. Each Golden Shiner performs light intensitysensing at its current location and modulates its swimmingspeed: swimming fast in bright environments, swimming slowin dark environments. This allows the Golden Shiner to quicklyswim out of bright areas while prolonging its time in darkareas. In addition, each Golden Shiner senses neighbors andswims towards neighbors. This sensing and movement strategyenables a school of Golden Shiners to collectively navigate todark areas.

We designed and built a collective robot swarm navigationplatform, called Shinerbot, inspired by the emergent navigationbehaviour of the Golden Shiner Fish.

The rest of the paper is organized as follows. Section IIprovides an overview of related work on robot swarm navi-gation platforms. Section III describes our Shinerbot swarmnavigation algorithm. Section IV describes our Shinerbotswarm navigation platform. Section V shows and discusses ourexperiment results. Through our experiment, we demonstratedsuccessful Shinerbot swarm navigation. Finally, Section VIsummarizes the paper.

II. RELATED WORK ONROBOT SWARM NAVIGATION PLATFORMS

Existing robot swarm navigation platforms includeMarXbot [3], eSwarbot [4], Pi-swarm [5], e-Puck [6],Khepera [7], Jasmine [8], Kilobot [9] and Droplet [10]. Theirsizes and costs are summarized in Table I.

2

TABLE ISIZE AND COST OF EXISTING ROBOT SWARM NAVIGATION PLATFORMS

Robot Size CostMarXbot [3] 17cm diametereSwarbot [4] 12.6cm diameterPi-swarm [5] 9cm diameter part cost of USD$100e-Puck [6] 7cm diameterKhepera [7] 7cm diameter retail price of USD$272Jasmine [8] 2.7cm cube part cost of USD$112Kilobot [9] 33mm diameter retail price of USD$113Droplet [10] part cost of USD$50 without the shell

The larger swarm navigation platforms cost most and usewheels and/or tracks for mobility. The smaller swarm nav-igation platforms, which include Kilobot and Droplet, costless and use vibration motors for mobility. Our objective is todesign our Shinerbot swarm navigation platform using a lowerpart cost, while achieving our collective navigation goals.

III. OUR SHINERBOT SWARM NAVIGATION ALGORITHM

We model the sensing and movement strategy of the GoldenShiner Fish as a random walk with an environment factorinfluencing speed and a social factor influencing direction.Thus, each Shinerbot performs the following dynamic update,as detailed in our prior work [2].

pi(t+ 1) = pi(t) + pi(t) (1)pi(t) = env(pi(t))soc(pi(t), {pj(t)}j ∈ Ni(t)) (2)

where pi(t) is the position of Shinerbot i at time t andpi(t) is the velocity update to be performed at time t. Thevelocity update is governed by an environment factor env(pi)and a social factor soc(pi, {pj}j ∈ Ni). Both env(pi) andsoc(pi, {pj}j ∈ Ni) incorporates some randomness.

The environment factor is a function of the light intensityat pi. It enables convergence to an optimal. The social factoris a function of neighbor proximity at pi. It enables collec-tive intelligence, expediting convergence. The random factorenables tolerance to noise and errors, allowing Shinerbots toescape from local optimals and converge to the global optimal.

A block diagram of our Shinerbot swarm navigation algo-rithm is shown in Fig. 1.

Environment factor on speedBright → Fast, Dark → Slow

Sense light intensity

Sense neighbors

Social factor on directionTowards Neighbors

Collective navigationto dark areas

Random factor on speed

Random factor on direction

speed

direction

Minimal sensing Minimal control Collective Navigation

Fig. 1. Block diagram of our Shinerbot swarm navigation algorithm.

IV. OUR SHINERBOT SWARM NAVIGATION PLATFORM

We designed and built each Shinerbot to have a compactform factor with a low part cost. The block diagram and

(a)

(b)

Fig. 2. Block diagram and schematics of our Shinerbot swarm navigationplatform. (a) Block diagram showing basic capabilities and related hardwarecomponents. (b) Schematics showing connections to the microcontroller,a small computer containing a processor, memory, and programmable in-put/output (I/O) pins.

schematics of our Shinerbot swarm navigation platform areshown in Fig. 2.

The rest of this section is organized into subsections onmobility, sensing, swarm messaging and swarm charging.

A. Mobility

Shinerbot mobility is achieved using vibration motors [11].By adding a H-bridge between the microcontroller and eachvibration motor, we enabled smooth forward, reverse androtational maneuvers.

B. Sensing

Shinerbot light intensity sensing is achieved through aphotodiode sensitive to the visible spectrum.

Shinerbot directional proximity sensing of neighboringShinerbots is achieved through one infrared (IR) transmitterand three IR receivers. The placement of the IR sensors andthe IR beam pattern are shown in Fig. 3.

An IR Received Signal Strength Indicator (RSSI) providesa distance measurement. Fig. 4 shows the IR sensor geome-try, RSSI measurements and distance estimation results. Thefollowing equations describe the relationship between RSSImeasurements and distances:

RSSI = (1− r

104) · kr· cos2(θ) (3)

= (1− r

104) · kr· 4h2

4h2 + r2

3

where r is the distance between two neighboring Shinerbots,h is the height of a Shinerbot and θ is the angle between thesignal path and the vertical. The coefficient k is a scaling factorselected for curve-fitting of RSSI measurements, as shownin Fig. 4 (b).

(a) (b)

Fig. 3. (a) Placement of 3 infrared (IR) receivers and 1 IR transmitter on theunderside of the Shinerbot; (b) IR beam pattern [12].

(a)

(b) (c)

Fig. 4. Distance estimation using IR Received Signal Strength Indicator(RSSI). (a) Path traveled by an IR signal between two neighboring Shinerbots;(b) RSSI measurements against distances; (c) distance estimation from RSSImeasurements.

Using three distance measurements to a single neighbor,Shinerbots perform triangulation to obtain directional proxim-ity to the neighbor. The equations for estimating the directionalproximity are given as follows:

x =r2A − r2B

2l3(4)

y =r2C − r2D2(l1 + l2)

+l1 − l2

2(5)

r2D =r2A + r2B −

l232

2(6)

where x is the distance to a neighboring Shinerbot in thex-direction, y is the distance to a neighboring Shinerbot in they-direction. The coordinate reference used is given in Fig. 5.rA, rB , rC are distances from the IR transmitter (tx) to theIR receivers (rxA, rxB , rxC). rD is the distance from tx to

the midpoint between rxA and rxB . l1, l2, l3 are geometricallengths on the Shinerbot, labeled in Fig. 5.

rA rBrD

rC

l3

l1

l2

X

Y

Shinerbot

Shinerbotneighbour

xy

Fig. 5. Direction estimation using three IR distance measurements.

4

C. Swarm Messaging

Large-scale swarm operations are made tractable throughsystem messages. Shinerbots can be put into different states(reset, sleep, wake, run) via system messages that spreadthrough neighbor-to-neighbor messaging, as shown in Fig. 6.

Fig. 6. Spreading of system messages from one Shinerbot, programmed usinga computer, to the entire swarm.

These messages are modulated on the IR transmissions usedfor neighbor directional proximity sensing, using a simpli-fied implementation of Carrier Sense Multiple Access withCollision Avoidance (CSMA-CA), following Infrared DataAssociation (IrDA) wireless protocols [13].

D. Swarm Charging

We designed a charging plate, with groove spacings calcu-lated such that a randomly placed Shinerbot will have at leastone leg on a positive voltage channel and one leg on a negativevoltage channel. This charging plate can simultaneously chargeup to 49 Shinerbots through their conducting legs.

V. EXPERIMENT RESULTS

We built each Shinerbot to have a compact form factor of40.64 mm in diameter with a low part cost of USD $20, asshown in Fig. 7. This process includes drawing the circuitschematics, producing the PCB layouts, sending the PCBfor fabrication, soldering the electronic components, writingcustom software, programming the Shinerbots, calibration andtesting. Following that, we performed an experiment using 30Shinerbots.

Fig. 7. Shinerbot: our bio-inspired collective robot swarm navigation platform.

We created the experiment environment using an overheadprojector, projecting light down onto a white horizontal sur-face, termed the experiment platform. We initialized our exper-iment with Shinerbots placed in somewhat random positionson the experiment platform. The experiment results are shownin Fig. 8, on the next page.

At time t = 0min, the Shinerbots were initialized in randomlocations. By time t = 3min, the Shinerbots have formed twoclusters. At time t = 11min, the Shinerbots have successfullynavigated to the dark area near the center of the experimentplatform and the dark edge of the experiment platform. Thelone Shinerbot in the middle ran out of power.

VI. CONCLUSION

In summary, we designed and built a collective robotswarm navigation platform, called Shinerbot, inspired by theemergent navigation behavior of the Golden Shiner Fish.Our Shinerbots perform minimal sensing and control. As aswarm, our Shinerbots are able to achieve synergistic overallnavigation performance with no data communication betweenswarm elements and no path planning. Using 30 Shinerbots,each of size 40.6mm and of cost USD $20, we experimentallydemonstrated successful Shinerbot swarm navigation.

ACKNOWLEDGMENTS

The authors would like to thank Kwok Bun (Gilbert) Chengfor his help in soldering electronic components onto the PCBsand for assembling the Shinerbots. The authors would also liketo thank Alexandra Bacula and Katie Caroll for their help incalibrating the Shinerbots.

5

t = 0 min t = 1 min t = 2 min t = 3 min

t = 4 min t = 5 min t = 6 min t = 7 min

t = 8 min t = 9 min t = 10 min t =11 min

Fig. 8. Snapshots from an experiment video demonstrating successful Shinerbot swarm navigation.

REFERENCES

[1] A. Berdahl, C. J. Torney, C. C. Ioannou, J. J. Faria, and I. D. Couzin,“Emergent sensing of complex environments by mobile animal groups,”Science, vol. 339, no. 6119, pp. 574–576, 2013.

[2] H. Liang and G. Gao, “Navigating robot swarms using collective intel-ligence learned from golden shiner fish,” in Proceedings of CollectiveIntelligence Conference (CI-2014), 2014.

[3] M. Bonani, V. Longchamp, S. Magnenat, P. Retornaz, D. Burnier,G. Roulet, F. Vaussard, H. Bleuler, and F. Mondada, “The MarXbot,a miniature mobile robot opening new perspectives for the collective-robotic research,” in Proceedings of 2010 IEEE/RSJ International Con-ference on Intelligent Robots and Systems (IROS-2010). IEEE, 2010,pp. 4187–4193.

[4] M. S. Couceiro, C. M. Figueiredo, J. M. A. Luz, N. M. Ferreira, andR. P. Rocha, “A low-cost educational platform for swarm robotics,”International Journal of Robots, Education and Art, vol. 2, no. 1, pp.1–15, 2011.

[5] J. Hilder, R. Naylor, A. Rizihs, D. Franks, and J. Timmis, The piswarm: A low-cost platform for swarm robotics research and education.Springer, 2014, pp. 151–162.

[6] S. Torrellas Socastro, “Cognitive stimulation through task-oriented teler-obotics,” Master’s Thesis, 2013.

[7] F. Mondada, E. Franzi, and A. Guignard, “The Development of Khep-era,” in Experiments with the Mini-Robot Khepera, Proceedings of theFirst International Khepera Workshop, ser. HNI-Verlagsschriftenreihe,Heinz Nixdorf Institut, 1999, pp. 7–14.

[8] S. Kernbach, D. Habe, O. Kernbach, R. Thenius, G. Radspieler,T. Kimura, and T. Schmickl, “Adaptive collective decision-makingin limited robot swarms without communication,” The InternationalJournal of Robotics Research, vol. 32, no. 1, pp. 35–55, 2013.

[9] M. Rubenstein, C. Ahler, and R. Nagpal, “Kilobot: A low cost scalablerobot system for collective behaviors,” in Proceedings of 2012 IEEEInternational Conference on Robotics and Automation (ICRA-2012).IEEE, 2012, pp. 3293–3298.

[10] N. Farrow, J. Klingner, D. Reishus, and N. Correll, “Miniature six-channel range and bearing system: algorithm, analysis and experimentalvalidation,” in Proceedings of 2014 IEEE International Conference onRobotics and Automation (ICRA-2014). IEEE, 2014, pp. 6180–6185.

[11] P. Vartholomeos and E. Papadopoulos, “Analysis, design and controlof a planar micro-robot driven by two centripetal-force actuators,” inProceedings of 2006 IEEE International Conference on Robotics andAutomation (ICRA- 2006). IEEE, 2006, pp. 649–654.

[12] V. Semiconductors. (2011, oct) TEMD7100X01 silicon PINphotodiode. [Online]. Available: http://www.mouser.com/ds/2/427/temd7100-244776.pdf

[13] I. . W. Group et al., “Wireless LAN medium access control (MAC) andphysical layer (PHY) specifications,” IEEE, 1997.

[14] P. Technology. (2016, jun) Powerstream Li-ion coin cell Lir2477 datasheet. [Online]. Available: http://www.powerstream.com/p/Lir2477.pdf