The Swarmathon: An Autonomous Swarm Robotics Competition

Sarah M. Ackerman∗1, G. Matthew Fricke1, Joshua P. Hecker1, Kastro M. Hamed2, Samantha R. Fowler2,Antonio D. Griego1, Jarett C. Jones1, J. Jake Nichol1, Kurt W. Leucht3, and Melanie E. Moses1,4,5

Abstract— The Swarmathon is a swarm robotics program-ming challenge that engages college students from minority-serving institutions in NASA’s Journey to Mars. Teams competeby programming a group of robots to search for, pick up,and drop off resources in a collection zone. The Swarmathonproduces prototypes for robot swarms that would collectresources on the surface of Mars. Robots operate completelyautonomously with no global map, and each team’s algorithmmust be sufficiently flexible to effectively find resources froma variety of unknown distributions. The Swarmathon includesPhysical and Virtual Competitions. Physical competitors testtheir algorithms on robots they build at their schools; theythen upload their code to run autonomously on identicalrobots during the three day competition in an outdoor arenaat Kennedy Space Center. Virtual competitors complete anidentical challenge in simulation. Participants mentor localteams to compete in a separate High School Division. Inthe first 2 years, over 1,100 students participated. 63% ofstudents were from underrepresented ethnic and racial groups.Participants had significant gains in both interest and corerobotic competencies that were equivalent across gender andracial groups, suggesting that the Swarmathon is effectivelyeducating a diverse population of future roboticists.

I. INTRODUCTION

A. The Journey to Mars

The National Aeronautics and Space Administration(NASA) Journey to Mars program has set the ambitious goalof sending manned missions to Mars by the 2030s.1 Thesesurface missions may last months or years, and becausetransporting sufficient materials from Earth is not practical,astronauts will need to utilize the resources already availableon Mars. For example, small pockets of ice can be melted forwater and converted to oxygen and hydrogen for fuel. Thisapproach is known as in-situ resource utilization (ISRU).

B. Meeting the Challenge with Robot Swarms

Autonomous robot swarms could locate, collect, and storeresources in advance of human arrival. A swarm of small,

1Computer Science Department, The University of New Mexico, Albu-querque, NM, USA, 2Education and Interdisciplinary Studies, FloridaInstitute of Technology, Melbourne, FL, USA, 3NASA Kennedy SpaceCenter, FL, USA, 4Department of Biology, The University of New Mexico,5Santa Fe Institute, Santa Fe, NM, USA, This work was supported byNASA Minority University and Research Education Program (MUREP)#NNX15AM14A and a generous donation from Google. We thank TheresaMartinez, MUREP STEM Engagement Manager; Paul Secor and MaryBaker of Secor Strategies, LLC; Kate Cunningham and Beatriz PalaciosAbad of UNM ADVANCE; Elizabeth Esterly and the students of theMoses Biological Computation Lab, Vanessa Svihla, Aeron Haynie, NASAvolunteers, UNM support staff, KSC Visitor Complex, and the Swarmathonstudents and faculty mentors.∗Correspondence: sherbet@unm.edu1https://www.nasa.gov/content/nasas-journey-to-mars

inexpensive robots acting autonomously provides severaladvantages over a few large, expensive, manually controlledrovers [1], [2] because they are robust, scalable and able toexplore unmapped environments [3].

ISRU is closely related to central-place foraging (CPF),in which resources are transported to a central collectionzone. Inspired by the success of social insects gatheringresources for their colonies, an early study of robot foragingcomputationally modeled a swarm collecting rock sampleson a distant planet [4]. CPF is a key robot swarm application[5], [6], and research continues to improve algorithms andengineering approaches for foraging swarms [7], [8].

Our team has designed and built a swarm of foragingrobots called Swarmies. Swarmies are rugged enough tooperate outdoors for hours, and feature sensors and grippersthat enable them to complete CPF tasks. We also developedcustom software packages and a graphical user interface(GUI), enabling rapid development and testing of CPF al-gorithms in simulation and physical hardware.

Swarm foraging provides an ideal educational challengebecause successful strategies require foundational roboticsskills such as grasping, localization, navigation, exploration,and decentralized communication and coordination [9].

C. The Swarmathon Competition

In the Swarmathon annual competition, teams of collegestudents develop algorithms for autonomous swarm foraging.Teams in the Physical Competition program groups of 3 - 6Swarmies to collect the most of 128−256 resources placed in225− 484m2 outdoor arenas. The Virtual Competition runsin a Gazebo simulation of the same arenas and robots. Robotsmust search for, collect, and return resources to a centralcollection zone completely autonomously. The competitionruns in a series of rounds, and in each round resources areplaced in different distributions (i.e., at uniform random, orin various sizes and shapes of clusters), and the same codeis run in each round. Score is determined by the numberof resources in the collection zone at the end of the round(resources pushed out of the collection zone by robots do notcount toward the score). Robots have limited view and noglobal map, making it difficult to find resources. In a thirdHigh School Competition, younger students are mentored bySwarmathon college students to complete a similar challengein a simplified simulation. Winners of the Virtual competitionadvance to the Physical competition the following year.

D. Diversifying the STEM/Robotics Pipeline

Tackling NASA’s real-world ISRU problem provides anopportunity to engage students in robotics and computerscience (CS), helping meet the growing need for science,technology, engineering, and mathematics (STEM) talentin the workforce. While women and minorities accountfor a growing proportion of US college graduates and USworkers, they are underrepresented among STEM collegegraduates and professionals, especially in CS and engineer-ing [10]–[13]. Their inclusion in STEM helps to addressan anticipated shortfall in these fields over the next fewdecades [12], [13]. Additionally “workers from a varietyof backgrounds enhance the quality of science insofar asthey are likely to bring a variety of new perspectives. . . interms of both research and application” [11]. For thesereasons, the Swarmathon recruits competitors from 2- and4-year colleges and universities that are minority-servinginstitutions (MSIs) including Historically Black Colleges andUniversities (HBCUs), Hispanic-Serving Institutions (HSIs),and Tribal Colleges and Universities (TCUs). This providesan educational opportunity that may not otherwise exist forparticipating students:

“The Swarmathon is an opportunity for studentsto work on a real-life engineering problem that’sinterdisciplinary and hard. They know that I don’tknow the answer which makes it fun for all of us.I teach at a small community college and it simplywouldn’t be possible for us do this level of workfrom scratch.” -Swarmathon Faculty Mentor

The Swarmathon has engaged a diverse student populationwith 81% of students from minority groups (including Asianstudents) and 63% of the 818 college participants identifyingas belonging to an underrepresented racial or ethnic group(Black/African American, American Indian/Alaskan Native,or Hispanic/Latino), substantially more diverse than USundergraduate CS majors, of which only 15.8% are fromthese underrepresented groups [14].

College students mentored K-12 students in their commu-nities, including 300 high school students who compteed ina simplified High School Swarmathon in Netlogo. Addition-ally, the Swarmathon supported 60 students to attend work-shops and the 2016 and 2017 Robotics Science and Systems(RSS) conferences and 17 students participated in researchinternships with mentors as other Swarmathon schools. Allof these activities further developed pathways into STEM. Asan example, 65% of RSS workshop participants indicated adesire to pursue a Ph.D. in robotics.

II. RELATED WORK

A. Foraging Robot Swarms

Though swarm robot foraging has been studied analyti-cally for decades [4], hardware implementations are rare [9].For example, studies of CPF often only simulate the collec-tion and transport of objects [15], and recent advances inhardware implementations, such as collaborative warehouse

robots, require permanent infrastructure such as buried guide-wires or visual markers to operate [16].

Many attempts have been made to develop swarm algo-rithms and robot systems that address various aspects ofthe swarm foraging problem including scalability, energyefficiency, task allocation, and collection speed [17]–[22].However, autonomous swarm foraging remains a challengingopen problem with no generally recognized best solution.

Projects using physical robot swarms include the Rob-otarium, a swarm robotics testbed providing remotely ac-cessible robots and an arena [23]. This innovative projectallows virtual experimentation with physical robot swarms,but it uses an overhead camera for localization and therobots themselves have no on-board sensors. Kilobots aresimple, small robots intended to be integrated into largeswarms. They have sensors and operate autonomously andcollaboratively to push items through a maze or into specificconfigurations, but they have relatively limited mobility andoperate in controlled laboratory environments [24]. TheSwarmanoid project demonstrated successful implementationof a heterogeneous swarm whose robots collaborate in orderto solve tasks like object retrieval in a highly specialized en-vironment [25]. The challenges of operating physical swarmsresult in robots that are usually semi-autonomous in prac-tice; they require frequent human support [26]. The realitygap between performance in simulation and performance inphysical hardware [27] makes implementation in real robotsparticularly difficult. The hardware and software platformsfor the Swarmies address these challenges so that CPF canbe developed and subsequently tested in a system that iscompletely autonomous, operates without global knowledgeor control, and can function outdoors.

B. Educational Robotics Environments and Competitions

Educational robotics is a growing field that is particularlyeffective at improving student performance in and attitudestowards STEM disciplines, and reducing gender and ethnicachievement gaps [12], [28]–[31]. Examples for youngerstudents include the distributed Robot Garden [32] and low-cost AERobots [33].

The NASA Robotic Mining Competition (RMC) is anannual engineering competition to design and build robotscapable of mining the Martian surface as part of the ISRUmission. Students focus on hardware design rather thanautonomous control. Only 16% of the 900 RMC participantsin 2017 identified as members of underrepresented groups2.The Swarmathon was designed to emulate the success andpopularity of RMC while emphasizing robot autonomy andengagement of students from MSIs.

C. Hardware

Teams selected for the Physical Competition are sentparts and instructions3 for building 3 Swarmies. Each robothas a front mounted gripper with actuated wrist and finger

2https://www.nasa.gov/offices/education/centers/kennedy/technology/nasarmc.html

3https://github.com/BCLab-UNM/Swarmathon-Robot

Fig. 1: Swarmie Architecture. Software packages (shaded)interface with hardware through a USB connection, or withthe Gazebo simulation.

joints for grasping resources. Robots sense objects in theenvironment using three ultrasound range finders and a webcamera. The ultrasounds have a 3m range. The camera hasa narrow field of view with a 1 rad arc and range of 1m.

Orientation and positional (pose) data are provided by aglobal positioning system (GPS) unit and inertial measure-ment unit (IMU) with a magnetometer, along with wheelencoder odometry. Computation is provided by a small on-board computer (an Intel NUC). An Arduino Leonardo mi-crocontroller provides the hardware interface between the on-board computer and both the wheel and gripper actuators andthe encoders that measure wheel velocity. The robot body isbuilt from laser cut and 3D printed components. The batterypack allows robots to run for 8 hours between charges, whichwith a default speed of speed of 0.2m/s results in a rangeof 5.75 km. This allows a team of 6 Swarmies to searcha linear distance of 34.5 km on a single charge, nearly thedistance of a marathon that the remote-controlled Mars RoverOpportunity took 11 years to complete.

The resources that Swarmies locate and collect are cubesmarked on all sides with AprilTags [34]. The perimeter ofthe central collection zone is also marked with AprilTags.

D. Software

The basic robot software, based on the robot operatingsystem (ROS) [35], is publicly available4 and provided toall teams. So teams may focus on the ISRU challenge, weimplement the low-level code that interfaces with physicaland simulated actuators and sensors, performs necessarycalibration and pre-configuration, and defines the simulatedworld for competition, including Swarmies, resources, andthe search environment.

The software base is designed so that students can imple-ment the same algorithms to control simulated or physicalrobots (Figure 1). The GUI and ROS master either connectto the physical robots through a wireless network, or runin a Gazebo simulation. The diagnostic package monitorshardware components and alerts the user to problems. The

4https://github.com/BCLab-UNM/Swarmathon-ROS

robot pose is estimated using two extended Kalman filters(EKFs) [36] that fuse encoder, IMU, and GPS data.

This base code includes: Gazebo simulation of the robots,target cubes, and competition arena (Figure 2); a GUI (Figure2b) showing output of each hardware sensor, a map of therobots’ estimated positions (Figure 2c); customizable ROSpackages to control behaviors (for search, communication,obstacle avoidance, and target pick up and drop off) anddiagnostics (for sensors, actuators, wireless quality and themicrocontroller); a ROS package that interfaces with theArduino to communicate with actuators and and sensors;open source packages including AprilTag and EKF packages.

The simple base code algorithm (modeled after the sub-sumption architecture [37]) is designed to be extended byteams to improve foraging. It implements simple collisionavoidance, a random walk for search, and when a targetappears in the camera view, the robot picks up the targetand returns to the central collection zone to drop it off.

III. RESULTS

In the first 2 years, 29 of the 31 Physical teams and 20of the 27 Virtual Teams successfully uploaded code to com-pete. The skills these students required for the competitionincluded programming proficiency in ROS, swarm algorithmdesign, and experimental design and testing.

Several teams developed effective new search algorithmsthat significantly improved the base code. The 2017 Phys-ical Competition winner, Southwestern Indian PolytechnicInstitute (SIPI), experimented with different search strategiesinspired by snail shells, lawnmower paths, and spirograph-style geometric designs, ultimately deciding on a strategyinspired by the spokes of a bicycle wheel. SIPI’s algorithmdisplayed the greatest flexibility, performing at a high levelacross a variety of target distributions.

Figure 2c shows the Montgomery College search algo-rithm in an environment with one large cluster of targetcubes. The zig-zag pattern shows the robots’ search paththrough the arena. Once the large cluster of resourceswas located, some robots continued searching while othersfocused on collecting from the cluster. This strategy suc-cessfully balanced the explore versus exploit trade-off anddramatically outperformed the other teams on nearly everydistribution of resources to win the 2017 Virtual Competition.Engineering innovations (i.e., improving localization andresource pickup and drop-off) improved scores more thannew search algorithms.

A. Assessed Student Learning Outcomes

263 students completed a comprehensive self-assessment(29%). 44 Faculty Mentors (76%) completed the survey withthe same questions to evaluate their students’ performance.

Student growth was rated on 9 Accreditation Board for En-gineering and Technology (ABET) Criteria for AccreditingComputing Programs, and 2 ABET Criteria for AccreditingEngineering Programs: “An ability to design and conductexperiments, as well as to analyze and interpret data,” and,

(a) Gazebo Simulation (b) User Interface (c) Example Search Pattern

Fig. 2: Gazebo and GUI. (a) Two robots bring target cubes marked with AprilTags to the collection zone. (b) Gazebointerface showing robot status, location tracing, and camera view. (c) Trace of the Montgomery College winning swarm of6 robots; the dense activity shows robots traveling between a large target cluster and the collection zone.

Fig. 3: Physical Competition. Students compare strategiesand cheer on Swarmies from the University of the Districtof Columbia as they drop cubes in the collection zone.

“An ability to function on multidisciplinary teams.”5 Studentsand faculty rated student degree of competency before andafter the Swarmathon on a 4-point scale: none (1), low degree(2), moderate degree (3), and high degree (4).

Data for each year was analyzed with a dependent-samplest-test. Differences between the 2016 and 2017 cohorts wereanalyzed with an independent samples t-test. Table I showsthe significant gains in self-assessed performance each year.2017 gains were greater than 2016 gains in all categories andall results are statistically significant with p < 0.05.

There were no statistically significant differences in ourassessment or survey data based on race, ethnicity, or gender,indicating that the Swarmathon provided a level playingfield for competition, eliminating the achievement gap notedin other studies [30], [31]. Faculty mentor surveys closelymirrored student self-assessments on all but two ABETstandards. Mentors ratings were more than 0.50 points lowerthan student ratings on (1) An understanding of professional,

5www.abet.org/accreditation/accreditation-criteria/

ethical, legal, security and social issues and responsibilities,and (2) An ability to analyze the local and global impact ofcomputing on individuals, organizations, and society.

TABLE I: Changes in Student Abilities and Attitudes

2016 2017

ABET Outcomes +0.35 +0.59Higher Degree Interest +0.21 +0.53

Career Motivation +0.19 +0.38

Across both competition years, students also reportedsignificant gains in their “desire or motivation to pursuea higher degree beyond the one I am currently pursuing,”and their “desire or motivation to pursue a career in CS, orRobotics, or other technical discipline in STEM” as seen inTable I. Gains were significant even among students whoalready possessed a moderate to high degree of interest inpursuing a more advanced degree.

The Swarmathon is effectively inspiring the next gener-ation of engineers to tackle major scientific problems likethe NASA ISRU mission and to bring their diversity ofbackground and experience into the STEM workforce.

REFERENCES

[1] M. Brambilla, E. Ferrante, M. Birattari, and M. Dorigo, “Swarmrobotics: a review from the swarm engineering perspective,” SwarmIntelligence, vol. 7, no. 1, pp. 1–41, 2013.

[2] O. Zedadra, N. Jouandeau, H. Seridi, and G. Fortino, “Multi-AgentForaging: state-of-the-art and research challenges,” Complex AdaptiveSystems Modeling, vol. 5, no. 1, p. 3, 2017.

[3] J. P. Hecker, K. Letendre, K. Stolleis, D. Washington, and M. E.Moses, “Formica ex machina: ant swarm foraging from physical tovirtual and back again,” in Swarm Intelligence. Springer, 2012, pp.252–259.

[4] L. Steels, “Cooperation between distributed agents through self-organisation,” in Intelligent Robots and Systems’ 90.’Towards a NewFrontier of Applications’, Proceedings. IROS’90. IEEE InternationalWorkshop on. IEEE, 1990, pp. 8–14.

[5] A. Drogoul and J. Ferber, “From Tom Thumb to the Dockers: Someexperiments with foraging robots,” From Animals to Animats, vol. 2,p. 451, 1993.

[6] A. Campo and M. Dorigo, “Efficient multi-foraging in swarmrobotics,” in Advances in Artificial Life. Springer, 2007, pp. 696–705.

[7] A. F. T. Winfield, “Towards an engineering science of robot foraging,”Distributed autonomous robotic systems, vol. 8, pp. 185–192, 2009.

[8] M. Dorigo, M. Birattari, and M. Brambilla, “Swarm robotics,” Schol-arpedia, vol. 9, no. 1, p. 1463, 2014.

[9] A. F. T. Winfield, “Foraging Robots,” in Encyclopedia of Complexityand Systems Science. Springer, 2009, vol. 6, ch. Foraging R, pp.3682–3700.

[10] A. L. Griffith, “Persistence of women and minorities in STEM fieldmajors: Is it the school that matters?” Economics of Education Review,vol. 29, no. 6, pp. 911–922, 2010.

[11] C. B. Leggon, “Women in science: Racial and ethnic differencesand the differences they make,” The Journal of Technology Transfer,vol. 31, no. 3, pp. 325–333, 2006.

[12] S. D. Museus, R. T. Palmer, R. J. Davis, and D. Maramba, Racial andEthnic Minority Student Success in STEM Education: ASHE HigherEducation Report, Volume 36, Number 6. John Wiley & Sons, 2011.

[13] US Department of Education, “FACT SHEET: Spurring African-American STEM Degree Completion [Press release],” 2016.[Online]. Available: https://www.ed.gov/news/press- releases/fact-sheet-spurring-african-american-stem-degree-completion

[14] S. Zweben and B. Bizot, “2016 Taulbee Survey: Generation CSContinues to Produce Record Undergrad Enrollment; Graduate DegreeProduction Rises at both Masters and Doctoral Levels,” ComputingResearch News, vol. 29, no. 5, 2017.

[15] A. Brutschy, L. Garattoni, M. Brambilla, G. Francesca, G. Pini,M. Dorigo, and M. Birattari, “The TAM: abstracting complex tasks inswarm robotics research,” Swarm Intelligence, vol. 9, no. 1, pp. 1–22,2015.

[16] J. Enright and P. R. Wurman, “Optimization and Coordinated Auton-omy in Mobile Fulfillment Systems.” in Automated action planningfor autonomous mobile robots, 2011, pp. 33–38.

[17] W. Liu, A. Winfield, J. Sa, J. Chen, and L. Dou, “Strategies for energyoptimisation in a swarm of foraging robots,” in International Workshopon Swarm Robotics. Springer, 2006, pp. 14–26.

[18] N. R. Hoff, A. Sagoff, R. J. Wood, and R. Nagpal, “Two foraging algo-rithms for robot swarms using only local communication,” in Roboticsand Biomimetics (ROBIO), 2010 IEEE International Conference on.IEEE, 2010, pp. 123–130.

[19] J. P. Hecker and M. E. Moses, “Beyond pheromones: evolving error-tolerant, flexible, and scalable ant-inspired robot swarms,” SwarmIntelligence, vol. 9, no. 1, pp. 43–70, 2015.

[20] E. Ferrante, A. E. Turgut, E. Duenez-Guzman, M. Dorigo, andT. Wenseleers, “Evolution of self-organized task specialization in robotswarms,” PLoS computational biology, vol. 11, no. 8, p. e1004273,2015.

[21] Q. Lu, J. P. Hecker, and M. E. Moses, “The MPFA: A Multiple-Place Foraging Algorithm for Biologically-Inspired Robot Swarms,”IEEE/RSJ International Conference on Intelligent Robots and Systems,2016.

[22] G. Fricke, J. Hecker, A. Griego, L. Tran, and M. Moses, “A dis-tributed deterministic spiral search algorithm for swarms,” in IEEEInternational Conference on Intelligent Robots and Systems, 2016.

[23] D. Pickem, P. Glotfelter, L. Wang, M. Mote, A. Ames, E. Feron, andM. Egerstedt, “The Robotarium: A remotely accessible swarm roboticsresearch testbed,” in Robotics and Automation (ICRA), 2017 IEEEInternational Conference on. IEEE, 2017, pp. 1699–1706.

[24] M. Rubenstein, C. Ahler, N. Hoff, A. Cabrera, and R. Nagpal,“Kilobot: A low cost robot with scalable operations designed forcollective behaviors,” Robotics and Autonomous Systems, vol. 62,no. 7, pp. 966–975, 2014.

[25] M. Dorigo, D. Floreano, L. M. Gambardella, F. Mondada, S. Nolfi,T. Baaboura, M. Birattari, M. Bonani, M. Brambilla, and A. Brutschy,“Swarmanoid: a novel concept for the study of heterogeneous roboticswarms,” IEEE Robotics & Automation Magazine, vol. 20, no. 4, pp.60–71, 2013.

[26] A. Rosenfeld, N. Agmon, O. Maksimov, and S. Kraus, “Intelligentagent supporting human-multi-robot team collaboration,” ArtificialIntelligence, 2017.

[27] N. Jakobi, P. Husbands, and I. Harvey, “Noise and the reality gap: Theuse of simulation in evolutionary robotics,” in European Conferenceon Artificial Life. Springer, 1995, pp. 704–720.

[28] S. Freeman, S. L. Eddy, M. McDonough, M. K. Smith, N. Okoroafor,H. Jordt, and M. P. Wenderoth, “Active learning increases studentperformance in science, engineering, and mathematics,” Proceedingsof the National Academy of Sciences, vol. 111, no. 23, pp. 8410–8415,2014.

[29] A. G. Welch, “Using the TOSRA to Assess High School Students’Attitudes toward Science after Competing In the FIRST RoboticsCompetition: An Exploratory Study.” Eurasia Journal of Mathematics,Science & Technology Education, vol. 6, no. 3, 2010.

[30] A. S. Richmond, G. A. Boysen, and R. A. R. Gurung, An Evidence-based Guide to College and University Teaching: Developing theModel Teacher. Routledge, 2016.

[31] C. M. Steele, Whistling Vivaldi: How Stereotypes Affect Us and WhatWe Can Do. New York, NY, US: WW Norton & Co, 2010.

[32] L. Sanneman, D. Ajilo, J. DelPreto, A. Mehta, S. Miyashita, N. A.Poorheravi, C. Ramirez, S. Yim, S. Kim, and D. Rus, “A DistributedRobot Garden System,” in Robotics and Automation (ICRA), 2015IEEE International Conference on. IEEE, 2015, pp. 6120–6127.

[33] M. Rubenstein, B. Cimino, R. Nagpal, and J. Werfel, “AERobot: Anaffordable one-robot-per-student system for early robotics education,”in Robotics and Automation (ICRA), 2015 IEEE International Confer-ence on. IEEE, 2015, pp. 6107–6113.

[34] E. Olson, “AprilTag: A robust and flexible visual fiducial system,” inRobotics and Automation (ICRA), 2011 IEEE International Conferenceon. IEEE, 2011, pp. 3400–3407.

[35] M. Quigley, K. Conley, B. Gerkey, J. Faust, T. Foote, J. Leibs,R. Wheeler, and A. Y. Ng, “ROS: an open-source Robot OperatingSystem,” in ICRA workshop on open source software, vol. 3, no. 3.2.Kobe, 2009, p. 5.

[36] T. Moore and D. Stouch, “A Generalized Extended Kalman FilterImplementation for the Robot Operating System,” in Proceedings ofthe 13th International Conference on Intelligent Autonomous Systems(IAS-13). Springer, 2014, pp. 335–348.

[37] R. Brooks, “A robust layered control system for a mobile robot,” IEEEjournal on robotics and automation, vol. 2, no. 1, pp. 14–23, 1986.