The state of the art in biomimetics This article has been downloaded from IOPscience. Please scroll down to see the full text article. 2013 Bioinspir. Biomim. 8 013001 (http://iopscience.iop.org/1748-3190/8/1/013001) Download details: IP Address: 193.145.56.241 The article was downloaded on 10/01/2013 at 15:39 Please note that terms and conditions apply. View the table of contents for this issue, or go to the journal homepage for more Home Search Collections Journals About Contact us My IOPscience
Transcript
The state of the art in biomimetics
This article has been downloaded from IOPscience. Please scroll down to see the full text article.
2013 Bioinspir. Biomim. 8 013001
(http://iopscience.iop.org/1748-3190/8/1/013001)
Download details:
IP Address: 193.145.56.241
The article was downloaded on 10/01/2013 at 15:39
Please note that terms and conditions apply.
View the table of contents for this issue, or go to the journal homepage for more
Home Search Collections Journals About Contact us My IOPscience
The state of the art in biomimeticsNathan F Lepora1, Paul Verschure2 and Tony J Prescott1
1 Sheffield Centre for Robotics, University of Sheffield, UK2 Synthetic Perceptive, Emotive and Cognitive Systems (SPECS), Institucio Catalana de Recerca iEstudis Avancats, Universitat Pompeu Fabra, Barcelona, Spain
Received 17 August 2012Accepted for publication 3 December 2012Published 9 January 2013Online at stacks.iop.org/BB/8/013001
AbstractBiomimetics is a research field that is achieving particular prominence through an explosion ofnew discoveries in biology and engineering. The field concerns novel technologies developedthrough the transfer of function from biological systems. To analyze the impact of this fieldwithin engineering and related sciences, we compiled an extensive database of publications forstudy with network-based information analysis techniques. Criteria included publications byyear and journal or conference, and subject areas judged by popular and common terms intitles. Our results reveal that this research area has expanded rapidly from less than 100 papersper year in the 1990s to several thousand papers per year in the first decade of this century.Moreover, this research is having impact across a variety of research themes, spanningrobotics, computer science and bioengineering. In consequence, biomimetics is becoming aleading paradigm for the development of new technologies that will potentially lead tosignificant scientific, societal and economic impact in the near future.
(Some figures may appear in colour only in the online journal)
1. Introduction
Biomimetics is the development of novel technologies throughthe distillation of principles from the study of biologicalsystems. Biomimetic technologies arise from a flow of ideasfrom the biological sciences into engineering, benefitingfrom the millions of years of design effort performed bynatural selection in living systems (Bar-Cohen 2006a, Allen2010). This transfer of function from the natural world toartificial devices has driven novel research agenda across manydisparate disciplines, from materials science and architectureto computer science and robotics. More recently, advancesin robotics have facilitated the development of biomimeticrobotics inspired by the different design plans found inthe animal kingdom (examples shown in figure 1). Suchbiomimetic artifacts can provide excellent models of theirbiological counterparts, allowing us to ask and answerquestions about the biological system that cannot be addressedthrough experiments alone. We emphasize that it is thetransfer of function from biology to the machine that allowsbiomimetics to test hypotheses from the biological sciences;
otherwise, there is a danger of merely blind copying ormimicry of design principles with no further insight intothe living system. In this sense of transferring biologicalfunction, biomimetic systems can thus provide a test bedfor theoretical ideas in biology and a means for generatingbiological solutions to challenges in science and technology,and we may consider them an implementation of ‘livingmachines’.
Over the last decade, there has been an explosionof important discoveries within the many research topicscomprising biomimetics. The societal and economic impactsexpected to emerge from these advances will have futurebenefits for our health and quality-of-life, due to advancesin information and computation technologies, robotics,brain–machine interfacing and nanotechnology applied tolife sciences. Given this potential of biomimetics, manyinternational funding initiatives are underway to drive thefield forward. However, for the field to fully realize itspotential, it is necessary to have information gathering andcoordination initiatives that can inform policy makers about
Figure 1. Examples of biomimetic robots. (A) Stickybot: Gecko-inspired climbing-robot developed at Stanford University (Kim et al 2008,photo credit: Mark Cutkosky). (B) Robobee: Insect-like micro-air vehicle from the Harvard Microrobotics Lab (Sreetharan et al 2012, photocredit: The Harvard Microrobotics Lab). (C) Shrewbot: rodent-like, whiskered robot developed by Bristol Robotics Laboratory (Prescottet al 2009, Sullivan et al 2012, photo credit: Ben Mitchinson). (D) Robotic octopus tentacle: developed at the University of Pisa, Italy(Mazzolai et al 2012, photo credit: Cecilia Laschi). (E) iCub: a child-like humanoid robot developed at the Italian Institute of Technology(Metta et al 2008, photo credit: Lorenzo Natale). (F) Robot lobster: developed at North-Eastern University (Ayers et al 2011, photo credit:Daniel Blustein).
appropriate strategic decisions. A recent initiative towardthis goal in the European research area is the ConvergentScience Network (CSN) of biomimetic and biohybridsystems (http://csnetwork.eu), which facilitates surveying androad-mapping exercises combined with other coordinationactions for bringing researchers in the field together. Activitiesinclude organizing Living Machines: The First InternationalConference on Biomimetic and Biohybrid Systems (Prescottet al 2012), from which selected papers will be published inBioinspiration and Biomimetics.
This work is a survey of the state of the art of biomimeticsbased on an information analysis of a comprehensive databaseof publications on biomimetics in engineering and relatedsciences. While there are already many excellent reviews ofbiomimetics (Bar-Cohen 2006b, Vincent et al 2006, Barthelat2007, Teeri et al 2007, Pfeifer et al 2007, Fratzl 2007, Bhushan2009, Bongard 2009, Helms et al 2009, Gebeshuber et al2009, Johnson et al 2009, Wilson et al 2010, Nagel and Stone2011, Nosonovsky and Rohatgi 2012, Rawlings et al 2012),such accounts usually emphasize subjects that are part of theauthors’ expertise and research priorities. Instead, our aim is toprovide a complementary survey to these existing reviews fromthe viewpoint of a more objective statistical survey across thefield of biomimetics in engineering and related sciences. Inparticular, we focus on the following four main questions.Where is biomimetic research published? How rapidly isthe subject of biomimetics expanding? What subjects doesbiomimetics encompass? And are there research communitieswithin biomimetics? By focusing on these questions, our
intention is that the answers will clarify the current state ofthe fields comprising biomimetic research, how these fieldsevolved to their present state, and where they appear to beheading in the future.
2. Background to biomimetics
Biomimetics can, in principle, extend to all fields ofbiological research from physiology and molecular biologyto ecology and from zoology to botany. Promisingresearch areas include system design and structure, self-organization and co-operativity, new biologically activematerials, self-assembly and self-repair, learning, memory,control architectures and self-regulation, movement andlocomotion, sensory systems, perception, and communication.Biomimetic research, particularly at the nano-scale, shouldalso lead to important advances in component miniaturization,self-configuration and energy efficiency. Another key focusis on complete behaving systems in the form of biomimeticrobots that can operate on different substrates in a sea,on the land, or in the air. A further central theme is thephysiological basis for intelligent behavior as explored throughneuromimetics—the modeling of neural systems. Excitingemerging topics within this field include the embodiment ofneuromimetic controllers in hardware, termed neuromorphics,and within the control architectures of robots, sometimestermed neurorobotics.
Historically, the term ‘biomimetics’ was first usedby Otto Schmitt during the 1950s, when he made a
distinction between an engineering/physics approach to thebiological sciences, which was termed ‘biophysics’, anda biological approach to engineering, which he termedbiomimetics. Schmitt is also credited with establishing thefield of biomedical engineering, which now encompasses theimportant discipline of biomaterials that retains its strongconnections to biomimetics. A related term used in engineeringis ‘bionics’, which was introduced by Jack Steel of the USAir force to mean copying and taking ideas from nature(popularized in Daniel Halacy’s book Bionics: The Scienceof Living Machines (Halacy 1965)). Another term that isnow commonly used is bioinspired or biologically inspired,such as in the modern discipline of biologically inspiredcomputing. Bioinspired computing tends to focus on bottom-up, decentralized approaches to computation, such as geneticalgorithms, in contrast with the more traditional top-downapproach of artificial intelligence. Terms such as neuro-based and brain-based are now also being used, where theemphasis is more specifically based on the central nervoussystem of animals.
3. An information analysis methodology
To strive to be as objective as possible with this work, weadopted a methodology in which the state of the art wassurveyed using techniques from information analysis. Animportant aspect of this type of analysis of large datasets ishow the results are visualized. Friedman (2008) stated that the‘main goal of data visualization is to communicate informationclearly and effectively through graphical means.... To conveyideas effectively, both aesthetic form and functionality needto go hand in hand...’. Accordingly, we used a variety oftraditional and modern visualization techniques, including piecharts, word clouds and connected graphs, to give a range ofperspectives on the implications of the analysis.
Our general strategy was to construct a comprehensivedatabase of publications on biomimetic research inengineering and related sciences, using general web-basedresources for journals and conferences, such as IEEEXplore, Elsevier’s Scopus and the Thomson Reuters Web ofKnowledge. A range of synonyms for biomimetics were usedas search terms, including biomimetics, biomimetic, bionics,bionic, biomimicry, bioinspired and bioinspiration. The focusof this study was specifically on biomimetics in engineeringand related disciplines, and hence we narrowed the searchto include only papers in engineering, physics, mathematics,robotics, computer science and related disciplines. Theresulting database was then analyzed to infer the generalbreakdown of the field, e.g., by year, journal or conference,and subject areas are judged by common words in titles.
The implementational details of this methodology aregiven in two appendices: database construction in appendix Aand database analysis in appendix B. In particular, we describehow the information was extracted from the online databasesin a form that could then be analyzed from a standard desktoppersonal computer. In total, we extracted approximately 18 000publications on biomimetic research covering the years 1995–2011. Our analyses consisted of a series of tests, including
the analysis of year and journal or conference in which theywere published; a survey of biomimetic publications by topicbased on the most frequent terms in the titles of papers; andthen an analysis of community structure within the connectednetwork of papers linked by pairs of words in the titles that arecommon.
We emphasize from the outset that there are two mainlimitations of this information analysis approach. First, papersnot clearly labeled as biomimetic research, by their title orotherwise, were not included in our database, even though theircontent might be considered as clearly biomimetic. Second,papers may be incorrectly labeled as being biomimetic eventhough their subject matter is not. These are unavoidablelimitations of the information analysis tools and databaseinformation that are currently available. As such, the approachadopted here identifies those papers that have been labeled asbiomimetic by their authors, and we assume that the resultsobtained on this dataset are the representative of biomimeticsas a research field.
4. Where is biomimetic research published?
The first question is: In which journals and conferencesbiomimetic research is published? We then consideredsecondary queries, such as the impact of the journals and theproportion that biomimetics comprises of the total publishedcontent.
From a total of about 18 000 biomimetic publications inthe database, close to 10 300 (57%) were published in journalsand 7700 (43%) in conference proceedings. Altogether, 1925distinct journals and 1543 distinct conferences had at leastone publication on biomimetics. The 57% to 43% overallsplit for journals to conferences in comparison with anapproximate split of 11% to 89% for papers from just theIEEE Xplore database, 63% to 38% for papers from Scopusand 89% to 11% for papers from the Web of Knowledge.Based on these statistics, biomimetics overall has a higherproportion of research published in journals than conferences.However, there are considerable differences between databasesdue to differences in coverage. In particular, the IEEEdatabase favors engineering and robotics and thus containsmore biomimetics papers from conferences, consistent with aculture of dissemination through conference publications onthose subjects.
The 12 leading journals in biomimetics ordered by thetotal number of publications are visually displayed with apie chart (figure 2(A)), with only journals having more than100 publications in biomimetics shown in the figure. Furtherdetails of the leading journals are given in table 1, includingthe journal abbreviation (to interpret the labeling in figure 2),publisher, 2011 impact factor and biomimetic publication in2011 compared to the total journal output. The top five journalsin order of publication number are: Biomaterials (Elsevier)with a total of 484 papers, followed by Bioinspiration andBiomimetics (IOP) with 343 papers, the Journal of BiomedicalMaterials Research A (Wiley) with 228 papers, Langmuir(ACS) with 164 papers and Acta Biomateriala (Elsevier) with159 papers. The content of these journals are split over general
3
Bioinspir. Biomim. 8 (2013) 013001 Perspective
Figure 2. Journals and conferences publishing research on biomimetics. The pie charts represent the proportion of papers on biomimeticspublished in the leading journals and conferences ranked by the total number of publications. A total of close to 10 300 biomimeticpublications in journals and 7700 in conferences were considered. The legend gives abbreviations for these journals and conferences, with akey for the complete names given in tables 1 and 2.
Table 1. Leading journals publishing research on biomimetics. Journals are taken from the chart on left in figure 2 and represent the leadingjournals ranked by the total number of research papers on biomimetics. The journals listed in this table were selected by having more than100 publications in biomimetics. The abbreviations used in the key in figure 2 are given with the full journal name. The publisher, 2011impact factor, number of papers in biomimetics in 2011 and total number for 2011 are also given.
BIOMATERIALS Biomaterials Elsevier 7.4 55/1007BIOINSPIR BIOMIM Bioinspiration and Biomimetics IOP 2.0 89/89J BIOMED MATER RES A Journal of Biomedical Materials Research A Wiley 2.6 18/270LANGMUIR Langmuir ACS 4.1 48/1936ACTA BIOMATER Acta Biomaterialia Elsevier 4.9 27/454J MATER SCI Journal of Materials Science: Materials in Medicine Elsevier 2.3 18/285J BIONIC ENG Journal of Bionic Engineering Elsevier 1.0 25/52ADVANCED MATERIALS Advanced Materials Wiley 13.9 27/789NATURE Nature NPG 36.0 20/841BIOPHYS J Biophysical Journal Cell 3.7 6/696BIOMED MATER Biomedical Materials IOP 2.1 27/170IEEE T SYST MAN CY IEEE Transactions on Systems, Man and Cybernetics IEEE 2.1 7/109
biomimetics and materials science in medicine. At presentBioinspiration and Biomimetics is publishing the maximumnumber of papers on biomimetics per year, followed byBiomaterials and then Langmuir.
The six leading conferences that cover biomimetics aredisplayed in a pie chart (figure 2(B)), which were selectedaccording to those with more than 100 publications inbiomimetics. Further details are given in table 2, includingthe journal abbreviation and the number of biomimeticpublications in 2011 compared with the total numberof publications that year. The top three conferences inbiomimetics are: Robotics and Biomimetics (ROBIO) with atotal of 2796 papers, the International Society for Photonicsand Optics (SPIE) with 415 papers and the InternationalConference on Automation and Logistics (ICAL) with 201papers. Clearly, ROBIO dominates the research output onbiomimetics published in conferences, giving around one-third of all conference papers on the robotic aspects ofbiomimetics. In addition, ICRA and IROS also focus on
robotics, particularly automation and intelligent systems, andare regarded as leading conferences on robotics with highlycompetitive publication requirements. Five of the six leadingconferences are sponsored by IEEE, which indicates theemphasis of these applications of biomimetics in relation toengineering and technology.
5. How rapidly is the subject of biomimeticsexpanding?
Our next question concerns: How rapidly biomimetics isexpanding as a subject area? From counting the number ofpublications each year in our database, we see that in the firstdecade of this century there has been an explosive growthin biomimetic research, with the number of published paperseach annum doubling every 2–3 years (figures 3 and 4). Froma relatively small field in the mid-1990s of less than 100 papersper year, biomimetics has exponentially expanded thereafterto reach critical mass of several hundred papers per year by
4
Bioinspir. Biomim. 8 (2013) 013001 Perspective
Figure 3. Growth of biomimetic journals and conferences. The plots shows the number of papers published each year in biomimetics for theleading journals (left panel) and conferences (right panel) starting from 1995. The legend gives abbreviations for these journals andconferences, with a key for the complete names given in tables 1 and 2.
Table 2. Leading conferences in which research on biomimetics ispublished. Conferences are taken from the right-hand side of thechart in figure 2 and are ranked according to the total number ofpublications on biomimetics. The conferences listed in this tablewere selected by having more than 100 publications in biomimetics.The abbreviations used in the key in figure 2 are given with the fullconference name. The number of papers or abstracts in biomimeticspublished in 2011 is also given with the total for the conference.
ROBIO Robotics and Biomimetics 536/536SPIE International Society for 62/2800
Photonics and OpticsICAL International Conference 42/107
on Automation and LogisticsEMBC Engineering in Medicine 23/2100
and Biology ConferenceICRA International Conference on 19/1032
Robotics and AutomationIROS Intelligent Robots and 8/719
Systems
2002, a mature field with more than 1000 publications peryear by 2005, and currently has close to 3000 publication peryear. Over the last 15 years, this growth has far outpaced thatacross science in general, which averages close to 6% peryear (doubling every 13 years) (Larsen and von Ins 2010).Furthermore, the rapid growth in biomimetics has not yetsaturated, so this expansion compared to science as a whole isexpected to continue in the near future.
Based on this analysis, there is a boom in bioinspiredresearch. Leading discoveries in biomimetics have laid thefoundations for large areas of present and future researchworks. This changing research landscape should lead tochanges in the composition of academic departments. Weexpect that a greater number of researchers, research groups
Figure 4. Growth of biomimetic research. The bar chart plots thenumber of papers published each year in biomimetics starting from1995. The black bars indicate the proportion of journal papers andthe white bars indicate the proportion in books and conferences.
and departments in leading universities will be explicitlyfocused around biomimetic research. This is consistent withthe range of biomimetic research groups belonging to thesubscribers of the Convergent Science Network of biomimeticsand biohybrid systems (http://csnetwork.eu/members_list).
This expansion in biomimetic research and technologicaldevelopment is also reflected in the increased numbersof publications within individual journals and conferences(figure 3). In some cases, this has been from the publication
Figure 5. Popular topics in biomimetics. The word cloud shows the popularity of terms occurring in the titles of papers on biomimeticresearch. The word size is proportional to the frequency of word occurrence.
of new journals and conferences in biomimetics that havesince flourished, in particular the journal Biomimetics andBioinspiration (2006) and the conference proceedings forROBIO (2004). In other cases, this expansion has related toa change of focus of journals and conferences that have beenestablished for many years, such as the journal Biomaterials(1980), and the conferences ICRA (1984) and IROS (1988).The rapid expansion of ROBIO as a conference is clearlyevident in figure 3, compared with a more modest but stillappreciable expansion of the other conferences.
As these developments in robotics and engineeringare translated into technology, they have the potential forsignificant societal and economic impacts. The growth oftechnological transfer is seen from a corresponding risein patents granted in biomimetics (Bonser 2006, Bonserand Vincent 2007). By a similar procedure to the presentanalysis of academic publications, Bonser searched the USPatent and Trademark Office (USPTO) from 1985 to 2005for the keywords ‘biomimetic’, ‘bionic’, or ‘biologicallyinspired’. He then found that the cumulative number ofpatents approximated the leading half of a sigmoid distribution(Bonser 2006, figure 1). That being said, we would commentthat even though his data clearly show a remarkable growth inpatents on biomimetics (of about 100 per year in 2005), thereis little evidence for the claimed saturation from the presenteddata. He then continued his arguments with the claim that sincepatents secure the legal right to exploit an invention but areexpensive to prepare, the very act of patenting an inventionindicates that the inventor has some confidence that a producthas fair chance to be brought to market. Hence, Bonser’s patentstudy indicates that rapid development of new technologiesderived from biological models is taking place.
6. What subjects does biomimetics encompass?
The next question for analysis is: What are the individual topicsthat make up biomimetic research, and how do they vary inpopularity? To answer this question, we gave a simple pictorialrepresentation of these subject areas in a word cloud (figure 5),accompanied by the statistics for the 100 most common topics(summarized in table 3).
Word clouds, and data clouds more generally, are a visualdepiction of the frequency of words within a larger set obtainedby scaling the font size of each word within the cloud by itsfrequency of occurrence. The clouds used in this work placethe words randomly, with the overall layout determined mainlyby aesthetics and readability. Technical details of how the wordclouds were constructed are described in appendix B for thedatabase analysis, although we note here that common butuninformative terms such as parts of speech were excludedfrom this analysis. Note that since the area of word scales withthe square of its font size, word clouds tend to emphasize themost common words while filtering out those words that areless frequent.
Given that the database was constructed from papersconcerned with biomimetics, it is expected that many ofthe leading concepts in the word cloud (figure 5) relatedirectly to the overall subject area of biomimetics and itssynonyms such as bionic and bioinspired. Furthermore, wordsindicating the biomimetic research process are also popular,such as ‘based’, ‘model’ and ‘design’. Overall, the second mostpopular concept for this database is ‘robot’. This indicates thatmuch of contemporary biomimetic research published in theEngineering journals is focused on applications in robotics,as reinforced by ‘control’ being another popular term. Ourinterpretation of this robot–control pairing of concepts is thatthe control of robots is a problem as important as designingand building the hardware itself. Indeed, in many ways, therehave been huge advances in building robots due to progress inmicrocomputer-based technology. However, the utilization ofthese sophisticated devices lags behind the effortlessly smoothcontrol displayed by apparently primitive insects or infantanimals. The fact that control is a key concept in biomimeticsindicates that a main research topic is to take inspiration fromhow animals control their bodies and sensory systems.
Other, less common, topics from the word cloud indicatea wide variety of research taking place in biomimetics.A broad variety of research topics is directly inspiredby nature, including polymers, composites, fish, muscle,collagen and vision. Some terms are taken directly frombiomedical research, such as bone, tissue and cell, reflectingthe large impact of biomimetics upon this research area. This
6
Bioinspir. Biomim. 8 (2013) 013001 Perspective
Table 3. Top 100 most common topics in biomimetics. Topics are taken from the titles in a database of around 18 000 publications onbiomimetics.
Word Frequency(%) Word Frequency(%) Word Frequency(%) Word Frequency(%)
emphasis on biomedical research is consistent with the journalBiomaterials and Biomedical Materials Research havingleading publications for biomimetic research. In addition,concepts from control engineering and artificial intelligenceare also represented, including model, network, algorithm,simulation, learning, adaptive and optimization. These subjectareas are consistent with biomimetics being published inrobotics and engineering journals and conferences, such as thejournal IEEE Transactions on Systems, Man and Cyberneticsand the conferences ROBIO, IROS and ICRA.
7. Are there research communities withinbiomimetics?
Our final questions are concerned with the topography andinter-connectedness of biomimetics as a research field. Aspecific point of interest is whether the field of biomimeticsfunctions as a coherent whole or fractures into distinct fieldswith little connection between disparate areas. We commentthat there have been several academic networks concernedwith biomimetics, many of which have had a specific interestarea. However, it is not clear that these endeavors have causedbiomimetics to fracture into separate disciplines, in that theyindividually title their work based on a biomimetic sub-discipline to which they belong. To determine if there aresub-disciplines of biomimetics and what these are, we insteadconsider a network analysis of common nomenclature in thetitles of papers.
To address these questions, we use techniques fromnetwork theory to analyze the connectedness of biomimeticresearch. We begin with the most frequent topics inbiomimetics that were discussed in the previous section and
displayed in figure 5. We then quantify the connectednessbetween pairs of topics by their co-occurrence withinpublication titles, defined from a statistical measure of theirtendency to pair together relative to their chance level ofrandom co-occurrence (appendix B). Interpreting the frequentbiomimetic terms as nodes on a graph, this co-occurrencemeasure defines the strength of the connections between thesenodes. The results of such an analysis are displayed in figure 6,with the thickness of the connections proportional to the co-occurrence measure between topic pairs and the size of thetopic words proportional to their frequency (as in figure 5).The positioning and color of the connections and nodes arerelated to further analysis that we describe below.
For ease of interpretation, the top 50 word pairs are alsoshown in table 4. Some of these word pairs are from phrasessuch as ‘three dimensional’, which are not informative abouthow research areas are related but do serve to bind such termstogether for interconnection with other terms. Other wordpairs are more informative about how areas of research arerelated. For example, ‘flapping’ with ‘wing’ indicates thatinspiration from biology is the key theme for flying robots,while ‘underwater’ with ‘vehicle’ indicates an application ofbiomimetic robotics.
Related terms can then be positioned together by theuse of network analysis, for which we used the Gephivisualization software and toolkit (Bastian et al 2009). First,we applied the Force Atlas algorithm, which pulls togetherstrongly connected nodes while repelling all other nodes(further details in appendix B). In consequence, terms thattend to occur together are positioned closely on the graphwhile words that infrequently co-occur are positioned furtherapart. This positioning is shown in figure 6. The next stage
7
Bioinspir. Biomim. 8 (2013) 013001 Perspective
Figure 6. Connectedness of popular terms in biomimetics. Two words in the word cloud in figure 5 are considered connected if theyco-occur within the same titles, with the co-occurrence frequency giving the connection strength. A Force Atlas algorithm was applied tothese node words and connection strengths, which pulls together the connected terms. The graph is colored according to a modularityanalysis, which finds communities within the connected network, where a community is defined to be a group of nodes that have denserintra-connections but sparser connections with other communities.
of the investigation used a modularity analysis to looks forcommunities that tend to group together within the network.This modularity analysis results in a modularity score of 0.4,which is greater than the threshold of 0.3 usually taken toindicate community structure. This result implies that thereare distinct communities, or groups of terms that are moredensely connected together within their community than tothe other communities.
Five communities are evident from the modularityanalysis, which are denoted with the distinct colors infigure 6. Our interpretation of these communities are that theyare related to the following. (1) Robotics (blue)—coveringtraditional robotics with an emphasis on having control andintelligent, autonomous operation that is based on biology;applications include computer vision, walking robots andmanipulators; methods include pattern recognition, neuralnetworks, and other areas of machine learning. (2) Ethology-based robotics (green)—with the emphasis on constructingrobot hardware based on animals; examples include flyingrobots based on insects and birds, and underwater robots basedon fish. (3) Biomimetic actuators (yellow)—in particular,artificial muscle and its underlying technologies in material
science. (4) Biomaterials science (red)—with an emphasis onbiological materials such as bone, tissue and collagen, andtheir assembly and fabrication. (5) Structural bioengineering(black)—with the emphasis more concerned with the micro-structure of the biological materials.
Therefore, overall biomimetics as subject area is fairlywell inter-connected, indicating that it may be considered as asingle discipline. Within this connectivity, the field does formdistinct communities, as was confirmed by observing that eachcommunity has a recognizable theme.
8. Summary of main results
By applying an information analysis to a comprehensivedatabase of publications on biomimetics over the last 15 years,we could answer a set of strategic questions about the past,present and future of the research field.
First, where biomimetic research is published? Froma total of nearly 18 000 biomimetic publications, about57% were published in journals and 43% in conferenceproceedings (figure 2). Across the databases, there was alarge variation in these proportions, with the IEEE database
8
Bioinspir. Biomim. 8 (2013) 013001 Perspective
Table 4. Top 50 most common word pairs in biomimetics. Words are taken from the titles in a database of around 18 000 publications onbiomimetics, and the pairings measured by the frequency of co-occurrence relative to chance.
Word A Word B Co-occurrence measure Word A Word B Co-occurrence measure
Three Dimensional 72 Fin Fish 17Real Time 68 Investigation Experimental 16Flapping Wing 64 Insect Wing 16Phosphate Calcium 55 Autonomous Underwater 16Fluid Body 36 Assembly Self 16Stem Cells 33 Muscle Artificial 16Assembled Self 32 Mechanical Properties 16Underwater Vehicle 29 Preparation Characterization 15Lipid Membranes 29 Image Processing 15Acid Poly 29 Bio Inspired 15Pattern Recognition 29 Robust Controller 15Power Low 29 Engineered Tissue 15Transfer Energy 26 Between Interaction 15Insect Flapping 25 Titanium Phosphate 15Coating Titanium 25 Metal Polymer 15Controller Fuzzy 24 Wing Vehicle 15Parallel Manipulator 23 Machine Interface 15Neural Network 20 Fin Underwater 14Flapping Vehicle 20 Mimetic Bio 14Coatings Phosphate 19 Navigation Mobile 14Flexible Fin 19 Localization Mobile 14Tissue Engineering 19 Porous Scaffolds 13Autonomous Vehicle 18 Stem Cell 13Metal Composite 18 Flapping Micro 13Information Processing 17 Mimetic Peptide 13
finding a much greater proportion of conference papersthan the other databases, presumably related to differentpublishing strategies in engineering and computer sciencecompared with other disciplines. Overall, the top five journalswere (table 1): Biomaterials (Elsevier), Bioinspiration andBiomimetics (IOP), Journal of Biomedical Materials ResearchA (Wiley), Langmuir (ACS) and Acta Biomateriala (Elsevier).The top six conferences were (table 2): ROBIO, SPIE, ICAL,EMBC, ICRA and IROS. Of these, ROBIO dominated thepublication numbers, comprising more than one-third of theconference output.
Second, how rapidly is biomimetics expanding as asubject? From a relatively small field of tens of papersin the mid-1990s, biomimetics has exponentially expandedthereafter to now reach nearly 3000 papers per year (figure 4).The subject area has doubled in size every 2–3 years, faroutstripping the modest expansion of about 6% per year forscience in general (Larsen and von Ins 2010). Based on thisfinding, there is a boom in bioinspired research, with leadingdiscoveries in biomimetics laying the foundations for largeareas of present and future research. As these developmentsin robotics and engineering are translated into technology,they have the potential for significant societal and economicimpacts. This growth of technological transfer is also seen inthe rapid rise of patents granted in biomimetics (Bonser 2006,Bonser and Vincent 2007), indicating that a rapid developmentof new technologies derived from biological models is takingplace.
Third, what subjects does biomimetics encompass? Theresults of this analysis are displayed in a word cloud of frequentterms in biomimetic research (figure 5). As expected, theword biomimetic is the most popular word. Then, perhaps
more revealingly, other leading terms are ‘robot’ and ‘control’,which suggests that a main thrust of biomimetic research is totake inspiration from how animals control their bodies andsensory systems for application to robotics. Other conceptsfrom the word cloud indicate a wide variety of researchin biomimetics, including the taxonomy and abilities ofbiological organisms, terms from biomedical research andbioengineering, and concepts from computer science andartificial intelligence.
Finally, are there distinct research communities withinbiomimetics? This question was addressed with techniquesfrom network theory applied to a graph of frequent biomimetictopics linked given by common pairings within the titles ofpapers. Terms that are strongly connected can then be pulledtogether on the graph, while disparate topics are pushed apart(figure 6). Applying a modularity analysis to this networkshowed that the field of biomimetics was well connectedand may thus be considered a single discipline. Underlyingthis inter-connectivity was a community structure into fiveidentifiable research themes: robotics and control, ethology-based robotics, biomimetic actuators, biomaterials science andstructural bioengineering.
9. Conclusions
Biomimetics is a research field that is achieving particularprominence through a wide variety of new discoveries inbiology and engineering. By applying an information analysisto a comprehensive database of publications on biomimeticsover the last 15 years, we answered a set of strategic questionsabout the past, present and future of the research field. Themost notable result was that there has been a rapid expansion
9
Bioinspir. Biomim. 8 (2013) 013001 Perspective
of publications on biomimetics from the mid-1990s to presentday, doubling every 2–3 years to now reach a mature fieldof nearly 3000 papers per year. Furthermore, the field isstill expanding, and so more growth can be expected. Thesecond main result was that there are a number of distinctthemes into which biomimetics can be partitioned, which weidentified as robotics and control, ethology-based robotics,biomimetic actuators, and biomaterials science and structuralbioengineering. Taken together, these findings indicate thatbiomimetics is becoming a dominant paradigm for robotics,materials science and other technological disciplines, withthe potential for significant scientific, societal and economicimpact over this decade and into the future.
Acknowledgments
The authors thank the anonymous referees and are grateful fordiscussions with G Indiveri, S Vassanelli, A Mura, M Evansand other researchers from the Active Touch Laboratory at theUniversity of Sheffield and the Synthetic Perceptive, Emotiveand Cognitive Systems Laboratory at the Universitat PompeiFabra. This work was supported by the FP7 coordination actionConvergent Science Network (ICT-248986).
Appendix A. Database construction
The IEEE Xplore database (http://ieeexplore.ieee.org) wasused as a source of papers on biomimetics published inIEEE journals and conferences. The Thomson Reuters Web ofKnowledge database (http://apps.webofknowledge.com) andElsevier’s Scopus (http://scopus.com) were used for otherjournals and conferences. The search within the Web ofknowledge was restricted to engineering, physics, biophysics,robotics, computer science, mathematics and mathematicalcomputational biology, while the search within Scopuswas restricted to engineering, physics, computer science,mathematics, neurosciences and the decision sciences. Theseterms restricted the type of biomimetics that we consider tothat published in engineering and related sciences. We didconsider using other search terms within areas of biology,but our experience was that this resulted in many papers onjust pure biology without any clear relation to engineering ortechnology. A range of synonyms for biomimetics were usedas search terms, including biomimetics, biomimetic, bionics,bionic, biomimicry, bioinspired and bioinspiration. The searchengines in the three databases matched the terms with storedmetadata for each document, including the title, publicationtitle, abstract and author-defined keywords for each publishedpaper.
Since the information analysis was carried out inMATLAB (Mathworks, MA), it was necessary to first convertthe results of the search to a readable document in a non-propriety format. This was achieved in two steps. First,the search results were exported to Endnote referencingsoftware (Thomson Reuters, NY). By default, Endnote storesa bibliographic database in a propriety format. However, itdoes give an option to save this database in extensible markuplanguage (XML), a human readable text-based format that is
closely related to hypertext markup language (HTML) usedfor web pages.
A separate XML database was saved for each of thebiomimetic search terms given above. These were thenaccessible from MATLAB by loading the text file into a singlealphanumeric string. Within XML, the content of the metadatapertaining to each publication is enclosed within standardizedtabs. Hence, we could use standard text-matching commandswithin MATLAB to extract the metadata of relevance to theinformation analysis. Finally, before applying the informationanalysis, this database of metadata was pre-processed toeliminate repeated entries from overlapping search terms.
Appendix B. Database analysis
The metadata from the database of research articles inbiomimetics (appendix A) were then analyzed using a varietyof methods.
An initial analysis considered the biomimetic publicationsby year and journal or conference in which they werepublished. The results were plotted in figures 2–4 anddisplayed in tables 1 and 2, as described in the main textof this work.
The next analysis was to survey biomimetic publicationsby topic, which was estimated from counting the most frequentterms occurring in titles of papers. One complicating factoris that the most common terms are just words that occurgenerally in titles, such as parts of speech including ‘of’,‘for’ and ‘a’. After experimenting with various methods, wefound the most reliable way to remove these was to constructa list of these non-informative terms, and then remove themfrom the database entries to leave concepts that are informativeabout the field of biomimetics. These were presented in a wordcloud by exporting the list of words and their frequencies toan appropriate web-based tool (wordel; http://wordel.com),which was then manipulated into an appropriate graphicalformat.
Our final analysis was to split these topics into overallthemes, judged from research terms that occur commonlytogether. A suitable co-occurrence measure for a pair of wordsis the frequency that the words occur together in the titlepopulation (number of co-occurrences per number of titles)normalized by the product of the frequencies of the individualwords in the pair:
c(wordA, wordB) = f (wordA, wordB)
f (wordA) f (wordB). (B.1)
If the words were independent and randomly distributed, thenthe co-occurrence measure should equal 1; meanwhile, valuesof c less than or greater than 1 indicate words occurringtogether less than or greater than chance, respectively. This co-occurrence measure can then be used to find overall researchthemes by application of modularity analysis to determinecommunity structure. Co-occurrences between pairs of wordsdefine links between nodes (the words) on an undirectedgraph, with link strength given by the co-occurrence measure.The network-analysis software Gephi (http://gephi.org) wasthen used to display the links graphically with related words‘pulled’ together by the Force Atlas algorithm (figure 6).
We used standard Modularity and Force Atlas tools withinthe Gephi software, which are described more fully in thesoftware’s documentation (Bastian et al 2009).
References
Allen R (ed) 2010 Bulletproof Feathers: How Science Uses Nature’sSecrets to Design Cutting-Edge Technology (Chicago, IL:University of Chicago Press)
Ayers J, Westphal A and Blustein D 2011 A conserved neuralcircuit-based architecture for ambulatory and undulatorybiomimetic robots Marine Tech. Soc. J. 45 147–52
Bar-Cohen Y 2006b Biomimetics—using nature to inspire humaninnovation Bioinspir. Biomim. 1 1
Barthelat F 2007 Biomimetics for next generation materials Phil.Trans. R. Soc. A 365 2907–19
Bastian M, Heymann S and Jacomy M 2009 Gephi: an open sourcesoftware for exploring and manipulating networks Proc. Int.AAAI Conf. on Weblogs and Social Media pp 361–2
Bhushan B 2009 Biomimetics: lessons from nature—an overviewPhil. Trans. R. Soc. A 367 1445–86
Bonser R H C 2006 Patented biologically-inspired technologicalinnovations: a twenty year view J. Bionic Eng.3 39–41
Bonser R H C and Vincent J F V 2007 Technology trajectories,innovation, and the growth of biomimetics Proc. Instn Mech.Eng. C 221 1177–80
Fratzl P 2007 Biomimetic materials research: what can we reallylearn from nature’s structural materials? J. R. Soc. Interface4 637–42
Friedman V 2008 Data visualization and infographics SmashingMag.
Gebeshuber I C, Gruber P and Drack M 2009 A gaze into the crystalball: biomimetics in the year 2059 Proc. Instn Mech. Eng. C223 2899–918
Halacy D 1995 Bionics: The Science of Living Machines (NewYork: Holiday House)
Helms M, Vattam S S and Goel A K 2009 Biologically inspireddesign: process and products Des. Stud. 30 606–22
Johnson E A C, Bonser R H C and Jeronimidis G 2009 Recentadvances in biomimetic sensing technologies Phil. Trans. R.Soc. A 367 1559–69
Kim S, Spenko M, Trujillo S, Heyneman B, Santos Dand Cutkosky M R 2008 Smooth vertical surface climbing withdirectional adhesion IEEE Trans. Robot. 24 65–74
Larsen P O and von Ins M 2010 The rate of growth in scientificpublication and the decline in coverage provided by sciencecitation index Scientometrics 84 575–603
Mazzolai B, Margheri L, Cianchetti M, Dario P and Laschi C 2012Soft-robotic arm inspired by the octopus: II. From artificialrequirements to innovative technological solutions Bioinspir.Biomim. 7 025005
Metta G, Sandini G, Vernon D, Natale L and Nori F 2008 The iCubhumanoid robot: an open platform for research in embodiedcognition Proc. 8th Workshop on Performance Metrics forIntelligent Systems pp 50–6
Nagel J K S and Stone R B 2011 A computational conceptgeneration technique for biologically-inspired, engineeringdesign Des. Comput. Cogn. 10 721–40
Nosonovsky M and Rohatgi P K 2012 Biomimetics in MaterialsScience: Self-Healing, Self-Lubricating, and Self-CleaningMaterials (Heidelberg: Springer)
Pfeifer R, Lungarella M and Iida F 2007 Self-organization,embodiment and biologically inspired robotics Science318 1088–93
Prescott T J, Lepora N F, Mura A and Verschure P F M J (ed) 2012Living Machines: The First International Conference onBiomimetic and Biohybrid systems (London: Springer)
Prescott T J, Martin J P, Mitchinson B, Sullivan J C and Pipe A G2009 Whisking with robots Robotics and AutomationMagazine 3 42–50
Rawlings A E, Bramble J P and Staniland S S 2012 Innovationthrough imitation: biomimetic, bioinspired and bioklepticresearch Soft Matter 8 6675–9
Sreetharan P, Whitney J P, Strauss M and Wood R J 2012Monolithic fabrication of millimeter-scale machinesJ. Micromech. Microeng. 22 05502
Sullivan J C, Mitchinson B, Pearson M J, Evans M, Lepora N F,Fox C W, Melhuish C and Prescott T J 2012 Tactilediscrimination using active whisker sensors IEEE Sensors J.12 350–62
Teeri T T, Brumer H, Daniel G and Gatenholm P 2007 Biomimeticengineering of cellulose-based materials Trends Biotechnol.25 299–306
Vincent J F V, Bogatyreva O A, Bogatyrev N R, Bowyer Aand Pahl A K 2006 Biomimetics: its practice and theory J. R.Soc. Interface 3 471–82
Wilson J O, Rosen D, Nelson B A and Yen J 2010 The effects ofbiological examples in idea generation Des. Stud. 31 169–86
