Sensor architecture and task classification for agricultural vehicles and environments

Senso s 2010, 10, 11226-11247; doi:10.3390/s101211226 senso s ISSN 1424-8220 www.mdpi.com/jou nal/senso s A icle Senso A chi ec u e and Task Classi ica ion o Ag icul u al Vehicles and En i onmen s F ancisco Ro i a-Más Depa amen o de Ingenie ía Ru al y Ag oalimen a ia, Uni e sidad Poli écnica de Valencia, Camino de Ve a s/n, 46022 Valencia, Spain; E-Mail: [email p o ec ed]; Tel.: +34-963-877-291; Fax: +34-963-877-299 Recei ed: 20 Oc obe 2010; in e ised o m: 26 No embe 2010 / Accep ed: 1 Decembe 2010 / Published: 8 Decembe 2010 Abs ac : The long ime wish o endowing ag icul u al ehicles wi h an inc easing deg ee o au onomy is becoming a eali y hanks o wo c ucial ac s: he b oad di usion o global posi ioning sa elli e sys ems and he inexo able p og ess o compu e s and elec onics. Ag icul u al ehicles a e cu en ly he only sel -p opelled g ound machines commonly in eg a ing comme cial au oma ic na iga ion sys ems. Fa m equipmen manu ac u e s and sa elli e-based na iga ion sys em p o ide s, in a join e o , ha e pushed his echnology o unp eceden ed heigh s; ye he e a e many un esol ed issues and an unlimi ed po en ial s ill o unco e . The complexi y inhe en o in elligen ehicles is oo ed in he selec ion and coo dina ion o he op imum senso s, he compu e easoning echniques o p ocess he acqui ed da a, and he esul ing con ol s a egies o au oma ic ac ua o s. The ad an ageous design o he ne wo k o onboa d senso s is necessa y o he u u e deploymen o ad anced ag icul u al ehicles. This a icle analyzes a a ie y o ypical en i onmen s and si ua ions encoun e ed in ag icul u al ields, and p oposes a senso a chi ec u e especially adap ed o cope wi h hem. The s a egy p oposed g oups senso s in o ou speci ic subsys ems: global localiza ion, eedback con ol and ehicle pose, non- isual moni o ing, and local pe cep ion. The designed a chi ec u e esponds o i al ehicle asks classi ied wi hin h ee laye s de o ed o sa e y, ope a i e in o ma ion, and au oma ic ac ua ion. The success o his a chi ec u e, implemen ed and es ed in a ious ag icul u al ehicles o e he las decade, es s on i s capaci y o in eg a e edundancy and inco po a e new echnologies in a p ac ical way. OPEN ACCESS Senso s 2010, 10 11227 Keywo ds: senso a chi ec u e; in elligen ehicles; o - oad au onomous ehicles; obo ics; p ecision ag icul u e 1. In oduc ion A Uni ed Na ions panel held in 2008 concluded ha ― he way he wo ld g ows i s ood will ha e o change adically … o cope wi h g owing popula ion and clima e change‖ [1]. The G een Re olu ion o he las i y yea s, including e icien i iga ion sys ems, p oduc i e a ie ies, and he high impac o mechaniza ion o massi ely apply sophis ica ed pes icides and syn he ic e ilize s has mo e han doubled adi ional yields. Howe e , hese echnological ad ances ha e come wi h ecological cos s esul ing in aqui e deple ion, salinized soils, chemical poisoning om o e use o chemicals, and pollu ed s eams. The ac ha we ha e been consuming mo e ood han a me s ha e been p oducing o mos o he pas decade led Bou ne [1] o conclude ha in o de o mee ising ood demand we need ano he g een e olu ion, and we need i in hal he ime. Di e en al e na i es ha e been p oposed o achie e his g eene e olu ion, such as bio echnology and sus ainable a ming, bu he mo e immedia e solu ions b ough by he so-called new echnologies, i.e., p ecision a ming and ag icul u al obo ics, seem o be e ma ch he e olu ion sough . The inco po a ion o hese echnologies in o ag icul u al p oduc ion no only bene i s p oduc i i y and en i onmen al condi ions, bu i also imp o es he wo king condi ions o a m manage s, labo e s, and ehicle ope a o s. In a s udy epo ed by Wexle [2], p o essionals whose wo k equi es long hou s alone in mono onous senso y en i onmen s ha e desc ibed al e a ions in mood, pe cep ion, and cogni ion esul ing om ex ended pe iods o dec eased senso y s imula ion. Howe e , many o he de ici s associa ed wi h senso y dep i a ion can be p e en ed i he subjec s exe cise du ing he dep i a ion pe iod. This is he si ua ion encoun e ed when ac o d i e s spend mos o hei wo king ime plan ing o sp aying in mono onous ields, and e en mo e se e ely when ha es e ope a o s employ day and nigh o ha e he yield collec ed on ime when he mois u e le el is op imum, he p ice a o able, and wea he condi ions accep able. The men al a igue incu ed by hese d i e s is no only caused by he mono ony o he wo k, bu also due o he s ess c ea ed by he need o s ee ing accu a ely wi hin igh ows and lanes wi hou causing any damage o he ege a ion while main aining a sui able wo king pace. The elie o he ope a o om d i ing con inuously, e y much like ai line pilo s, allows a mul iplici y o ac ions ha help o imp o e he ealiza ion o asks while educing he exhaus ion o he d i e . Fu he mo e, ha ing mo e oppo uni ies o in e ac wi h he ex e nal wo ld, he addi ional in o ma ion supplied by obo ized ehicles le s ope a o s make wise da a-based decisions. The p ope in eg a ion o he se o senso s and au oma isms ha will make u u e ehicles mo e in elligen is key o he success ul consumma ion o a new ag icul u al e olu ion. The singula condi ions ound in ag icul u al p oduc ion, especially he d udge y and epe i i e na u e o usual asks, has mo i a ed he wish o a m machine y au oma ion since as ea ly as 1924, when Will od [3] designed a s ee ing a achmen capable o ollowing u ows o guide a machine au oma ically ac oss he ield. Un il he appea ance o elec onics and compu e s, he sensing de ices used o au oma e ope a ions we e pu ely mechanical. In ac , he majo i y o senso s used in ag icul u al ehicles ha e been ela ed o au onomous na iga ion. Fo his pu pose, he de ices used Senso s 2010, 10 11228 bo h in No h Ame ica [3] and in Eu ope [4] ha e been mechanical eele s, compu e ision came as, global posi ioning sys ems, geomagne ic di ec ion senso s, lase scanne s, and ul asonic ange inde s. Howe e , he e a e many o he senso s o equen use in p ecision ag icul u e such as yield moni o ing es ima o s, soil p ope ies p obes, mois u e con en analyze s, and many o he s being de eloped a p esen . The usage o senso s in ag icul u al ehicles has e ol ed h ough ime. In a s udy o pa en s de o ed o in- ield au oma ic na iga ion, Ro i a-Más [5] ound ha beacons, pseudoli e localiza ion de ices, and op ical senso s excluding came as we e popula du ing he pe iod 1985–2000, bu ine ial measu emen uni s, GPS-based applica ions, and imaging de ices became p edominan in he 2001–2008 pe iod. The pa icula case o GPS can be jus i ied by he cancella ion o selec i e a ailabili y in May 2000, which pe mi ed he use o mo e accu a e posi ioning da a o ci ilian applica ions. The a ie y and use o senso s in ag icul u al ehicles has been inc easing in he las wo decades. Some o hem, such as GPS ecei e s, ha e gained uni e sal accep ance. O he senso s, on he con a y, coexis o he same pu pose. Le us ake, o ins ance, he case o eedback senso s o au o-s ee ing con ol loops. La ge au o-guided ac o s ypically inse low me e s in he s ee ing mechanism, whe eas smalle ehicles es ima e he angle u ned by he s ee ing wheel wi h linea po en iome e s o op ical encode s a ixed o he s ee ing cylinde [6]. Mos o he in o ma ion managed wi hin he amewo k o p ecision a ming is exp essed in e ms o globally- e e enced wo-dimensional maps, as o example yield, soil, he mal, o a iable a e p esc ip ion maps. This p ocedu e equi es a ehicle equipped wi h a global posi ioning ecei e , no mally a GPS, coupled wi h he speci ic senso s measu ing he ield p ope ies ep esen ed in he maps. As echnology ad ances, iche and mo e accu a e in o ma ion is being inco po a ed in he in elligen sys ems o ehicles, mo ing om wo-dimensional maps o eal ime h ee-dimensional ep esen a ions o ag icul u al scenes [7]. Fa om being an exhaus ed opic, he selec ion, ins alla ion, combina ion, and ac ua ion o senso s unde a pa icula a chi ec u e onboa d ag icul u al ehicles is in g owing in e es , no only o esea che s, bu o manu ac u e s and end-use s alike. When we hink abou ehicle a chi ec u e, we can ocus on embedded beha io s and he so wa e capable o implemen ing hem, we can cen e ou a en ion on he ha dwa e needs o sense and ac ua e acco ding o hose beha io s, o we can ollow a holis ic iew and conside bo h. This a icle is p ima ily conce ned wi h he second app oach, ha is, wi h he senso a chi ec u e necessa y o adap con en ional ag icul u al ehicles—no longe con en ional— o a ming asks and ield en i onmen s. The place and way o in eg a e beha io al and ac ua ion modes will be discussed along he ex . The se o beha io s ha a e suscep ible o o m he in elligen sys em o a ehicle is boundless, al hough he e exis some o hem which ha e gained b oad accep ance o he obo ics communi y. Blackmo e e al. [8], o ins ance, p o ide a lis o beha io s o an au onomous ac o , whe e simple p ocesses as wa ching and wai ing mingle wi h complex asks such as ou e planning and na iga ion. The implemen a ion o hese beha io s on he ac o ollows he wo k o A kin by which he design o ehicle a chi ec u e is iden i ied wi h so wa e a chi ec u e a he han ha dwa e, and consequen ly an objec -o ien ed sys em a chi ec u e, e y much like ha used by so wa e de elope s, is p oposed o con ol he ehicle. While so wa e ailu es ha e been iden i ied as one o he majo sou ces o p oblems o au oma ing ag icul u al ehicles, ha dwa e mal unc ion and disad an ageous senso assemblages on ehicles ha e esul ed in eliabili y issues as se e e as, o e en mo e han, so wa e p oblems. Fo ha eason, senso a chi ec u e is assumed (in his pape ) o be he i s s ep in he design o he mo e gene al and Senso s 2010, 10 11229 comple e ehicle a chi ec u e encompassing ha dwa e, so wa e, en i onmen s, a ge asks, and ehicle dynamics. Ye , he design o he senso ne wo k is hea ily in luenced by he es , and he e o e canno be unde aken in isola ion. The main objec i e o his a icle is o p opose a senso a chi ec u e ha be e adap s o he equi emen s and needs which make ag icul u al en i onmen s and asks so special, and signi ican ly dis inc om o he obo ic applica ions. I is impo an , hough, o always keep in mind ha he deg ee o au onomy pu sued hi he o wi h ag icul u al ehicles is semi-au onomy, which implies ha an ope a o always emains in he cabin o secu i y easons. Full au onomy, howe e , will come in a u he u u e i i e e comes. 2. Sensing Needs in Ag icul u al En i onmen s The e a e h ee basic p ope ies o ag icul u al sys ems ha make hese sys ems ideal o obo ic applica ions: i s , asks a e ha d and e y o en exposed o ha sh en i onmen s caused by ex eme empe a u es, s ong sunligh , dus , o chemical sp ay; second, usual du ies a e epe i i e, a iguing, and ime consuming; and hi d, he p esence o c op ows, ee lanes, and suppo ing s uc u es (Figu e 1) o e s a semi-s uc u ed se ing om which in elligen ehicles can ex ac c i ical in o ma ion. The as majo i y o a ming scenes aking place in open ields—i.e., ou doo s—can be classi ied unde one o he ollowing en i onmen s: 1. Ba en ields eady o be ploughed, disin es ed, o plan ed. 2. Swa hs o cu c ops p epa ed o be picked o baled. 3. Lanes delimi ed by ows o ui ees whose heigh impedes no mal ehicles o a el o e hem, as he ci us o cha d o Figu e 2(a). 4. Rows o bulk c ops o sho plan s easily a e sable in all o mos o hei g owing ime, as he soybean ield shown in Figu e 2(b). 5. Lanes o g ape ines which can be a e sed by all ehicles bu no by con en ional ac o s as illus a ed in Figu e 2(c). Figu e 1. (a) S uc u e o suppo hops. (b) T ellis in apple ees. (a) (b) Senso s 2010, 10 11230 Figu e 2. (a) En i onmen 3: ci us o cha d. (b) En i onmen 4: soybean ield. (c) En i onmen 5: Vineya d lanes. (a) (b) (c) All he en i onmen s desc ibed abo e, excluding he ba en ield, p o ide he pe cep ion sys em o in elligen ehicles wi h a numbe o de ec able ea u es ha can be help ul o gene a e in o ma i e maps o na iga e unde au oma ic modes. Howe e , plan s, ees and suppo ing s uc u es also limi he ee space ha ehicles need o ca y ou hei egula asks. As a ma e o ac , he igh spacing le be ween c op ows o ee lanes usually poses complex challenges o he au oma ion o ag icul u al asks. The h ee examples o Figu e 2 illus a e his ac ; he allowable de ia ion a ailable be ween ow and ac o on wheels in Figu e 2(b) is below decime e le els, and he e is no much oom o s ee ing co ec ions in he o cha d o Figu e 2(a) o he ineya d o Figu e 2(c). The posi ioning accu acy demanded in such si ua ions esul s in he highes le el o p ecision equi ed by GPS ecei e s, whe e sub-inch accu acies eached by Real-Time-Kinema ic (RTK) GPS a e becoming mo e and mo e popula among p oduce s. No o he ci ilian applica ion, no e en ae onau ics, is cu en ly demanding such le els o accu acy. Ne e heless, no all he challenges come om high accu acy posi ioning needs; being awa e o a ehicle’s su ounding is a undamen al capaci y o an in elligen ehicle. A wide a ie y o unexpec ed obs acles may in e e e wi h an ope a ing ehicle: o he a m ehicles (manually o au oma ically d i en), o go en ools, li es ock, and e en ield labo e s o hi ed pe sonnel no always amilia wi h ag icul u al equipmen . A sa elli e-based localiza ion sys em will ne e accoun o mobile o unexpec ed obs acles, he e o e he e is a eal need o an onboa d pe cep ion sys em. Apa om de ec ing po en ial obs acles, local pe cep ion can p o ide small adjus men s ha help o place he ehicle in he op imum cou se. The pe cep ion sys em o an in elligen ehicle will be as complex as i s sensing needs. I , o ins ance, people a e being acked o wa e s ess in plan s wan s o be es ima ed, an in a ed he mocame a will ha e o be ins alled. When ni ogen con en o lea disease is moni o ed, a hype spec al came a will be added. A obo ic shephe d migh iden i y ca le o sheep wi h a monocula came a, bu i ee densi y o h ee-dimensional mapping is pu sued he choice will be made by selec ing be ween a s e eo came a and a lase ange inde . In any case, ega dless o he le el o complexi y accomplished, he ope a ing mode o obo ic o - oad ehicles will be semi-au onomy, which always places an ope a o in he cabin o sa e y and eliabili y ailu es. The design o he senso a chi ec u e has o conside his assump ion, a leas in he nea and middle u u e. The ac ha obs acles need o be de ec ed by ehicles in ce ain Senso s 2010, 10 11231 applica ions does no imply ha obs acles ha e o be unambiguously iden i ied, which is a much mo e di icul ask. Fo hese si ua ions, he sensing needs o he ehicle could be, o ins ance, de e mining whe he an obs acle in e cep ing a p og ammed pa h is a e sable o no , which can be g aphically analyzed wi h a e sabili y maps. 3. Special Vehicles o Speci ic Tasks The pa icula con igu a ion o ag icul u al ehicles, in compa ison o o he land ehicles, is he esul o mo e han one hund ed yea s o e olu ion since he dawn o mechaniza ion. Du ing his pe iod o ime, a m equipmen has been adap ing o he speci ic needs and equi emen s o ield asks. Gi en ha mos o hese asks ha e no changed p o oundly, obo ized ehicles will ha e o ea u e new capabili ies wi hou de imen o hei p e ious ones. So, o ins ance, an au oma ed ha es e will ha e o keep i s e iciency a es cu ing c ops and sepa a ing g ain, and an au onomous sp aye will need o a e se o - oad e ains a he s ipula ed eloci ies. The implica ions o his assump ion a e essen ial o unde s and and design in elligen ag icul u al ehicles, as hey a e o ally di e en om small ligh weigh labo a o y p o o ypes. Robo ic pla o ms ul illing a m du ies need o wi hs and ha sh a mosphe ic condi ions, a e se ough e ains, possess wo king-day au onomy, deli e a minimum mechanical powe , ca y an ope a o (a leas in he nea u u e), keep up wi h cu en e iciency a es, and mos impo an ly, assu e high deg ees o sa e y o all he ope a ions in ol ed. Typical ag icul u al ehicles weigh be ween 2 and 20 ons, inco po a e diesel engines wi h a a ed powe be ween 20 kW and 500 kW, and can each e ail p ices o e $300,000. These igu es b ing he ollowing ad an ages o au oma ion and obo iza ion o ag icul u al ehicles. Fi s , he p oblem o hea y payloads o small obo s; adding senso s and ba e ies o se e al- on ehicles will ha e no e ec on he ehicle’s s abili y and pe o mance. Second, he cos e ec : some elec onics ha e a signi ican p ice which can ep esen an impo an pe cen age o he o al cos o usual obo s, bu pe cen ages quickly diminish when obo izing $70,000 ac o s o $250,000 combines. And hi d, he ypical p oblem o ba e ies and au onomy. Small obo s can only ca y a limi ed numbe o ba e ies, and in consequence canno un o ex ended pe iods o ime. Sola ene gy has been enough o powe ligh plane a y o e s bu is insu icien o deal wi h egula powe -demanding ield du ies. Howe e , o - oad ehicles powe ed by po en diesel engines can easily d aw some hund eds o wa s om he p opulsion engine o ene gize all he elec onics in eg a ed in he ehicle. In addi ion, he la ges pa o p esen day ag icul u al ehicles ea u e a well condi ioned cabin ha is ideal o hos elec onics and delica e componen s in a p o ec ed en i onmen . The se o easons adduced abo e a e all a o able o ehicle au oma ion. Howe e , he possibili y o an o e sized, o e weigh , and ex emely powe ul machine ge ing ou o con ol poses se ious challenges o he comme cial deploymen o in elligen ag icul u al ehicles. In ac , his is he p incipal cause o delaying he gene al p oduc ion o au oma ed mobile equipmen , as majo manu ac u e s canno assume his so o isk. Mo eo e , a se ious acciden p o oked by an au onomous machine would p obably suspend esea ch in he ield o a long pe iod o ime. This issue akes us o he pa adox o ehicle size: eliabili y e sus p oduc i i y. A ecu en ques ion in many obo ics and au oma ion con e ences is whe he u u e equipmen will inc ease o dec ease i s size. La ge sizes make he con ol o au oma ed machines c i ical; ye , minimum a es o p oduc i i y ha e Senso s 2010, 10 11232 o be g an ed. A ade-o will likely o e he bes answe . In gene al, o e sized ehicles migh ha e hei mass educed bu a minimum dimension, pe haps ha o medium size ac o s, may lead o a wo kable igu e. Solu ions, howe e , can ollow wo ends; he obo iza ion o con en ional equipmen , and he c ea ion o inno a i e ehicles endowed wi h speci ic capabili ies. The o me g oup is ep esen ed by usual ac o s, ha es e s, sp aye s and so o h; and he la e includes au oma ed u ili y ehicles, scou ing obo s, o si e-speci ic lying applica o s. Addi ionally, he e has ecen ly been a g owing in e es in he au oma ion o gol ehicles and domes ic ga dening p oduc s such as lawnmowe s, g een mowe s, and u ehicles. Figu e 3. Th ee-laye ask classi ica ion o in elligen ag icul u al ehicles. The de elopmen o ag icul u al ehicles has been s ongly in luenced by he di e si y o asks and compe ences equi ed in he ield; he e o e, i is con enien o conside he binomial ehicle + ask when inco po a ing new capabili ies, wha means ha ehicles no only ha e o adap o he ag icul u al en i onmen s p e iously desc ibed in Sec ion 2 bu also o he du ies commanded. The a ie y o asks ha can be au oma ed is coun less, om ha es e spou adap i e posi ioning o g ain uck con oy ollowing. The addi ion o capabili ies, and he e o e o senso s, ends o be g adual acco ding o p oduce s’ needs and he ma u i y s a e o echnology. The o ganiza ion o he in elligen sys em embedded in he ehicle can be dis ibu ed in o he h ee laye s ep esen ed in Figu e 3: Sa e y, In o ma ion, and Machine Ac ua ion. The Sa e y Laye is he laye holding he highes p io i y, and is esponsible o g an ing ehicle s a ic and dynamic s abili y, pa h clea ance and sa egua ding, so wa e and ha dwa e eliabili y, engine p ope beha io , and he sa es use o au oma ic modes, p e en ing, o example, ha he ope a o abandons he cabin while in au oma ic s ee ing. In ac , some Senso s 2010, 10 11233 comme cial au o-s ee ing sys ems include a load cell unde he d i e sea , hal ing he ehicle as soon as he weigh senso de ec s he absence o he d i e . Cali o nia law, howe e , allows ac o s in u ows a eling less han 3 km/h o a el wi hou a d i e , p o ided he h o le, clu ch, and b akes can be con olled emo ely, which implies a non- i ial s a e o au oma ion. In e es ingly, his p ac ice is widesp ead wi h ha es c ews [9]. Ne e heless, in spi e o all he ac ions—bo h legal and echnical—unde aken o educe he numbe o acciden s, ag icul u al machines a e in ol ed wi h many isky si ua ions e en be o e he implemen a ion o au oma ed sys ems. The Uni ed S a es Bu eau o Labo S a is ics indica es, o he yea 2007, ha he ag icul u al sec o has he highes a e o occupa ional a ali ies among high- a e sec o s [10]. The e olu ion o ag icul u al equipmen h ough he g adual inco po a ion o new echnology o e s a unique oppo uni y o imp o e he sa e y o ield ope a ions as well as he p oduce s’ li e s anda d. The In o ma ion Laye ecei es he inpu s coming om he se o senso s onboa d, and p o ides key in o ma ion o decision making algo i hms, s o ing his o ical da a and ab ica ing p esc ip ion maps commonly used in p ecision a ming applica ions. As a esul , his laye is inhe en ly passi e as i ge s in o ma ion om he en i onmen and dis ibu es i h oughou he o he laye s bu i does no esul in he ac i a ion o au oma ic ac ua o s. Some imes his in o ma ion is p ocessed in eal ime, o example when i akes pa in sa e y wa nings o s ee ing assis ance. On o he occasions, he In o ma ion Laye supplies he p oduce wi h impo an knowledge o he managemen o he a m, such as yield spa ial dis ibu ion, soil e ili y, o wa e s ess. Finally, he Machine Ac ua ion Laye is in cha ge o execu ing o de s and commands om easoning algo i hms and decision-making ou ines. These algo i hms con e he ehicle wi h i s in elligen beha io and con inuously need o be ed by he onboa d senso ne wo k. This is, he e o e, an ac i e laye hos ing om basic con ol loops o sophis ica ed beha io al-based a chi ec u es o he ype ou lined in [8]. Typical ac ions handled by his laye a e a iable a e plan ing, s ee ing, implemen posi ioning, sma sp aying, o eme gency b aking. Al hough hese h ee p incipal laye s ha e been enuncia ed independen ly, i is ob ious ha hey a e deeply in e ela ed; sa e y equi es bo h eliable in o ma ion and accu a e ac ua ion, and e e y ac ion exe ed au oma ically needs p ecise eedback in o ma ion om he onboa d senso s o close con ol loops and g an s abili y. The In o ma ion Laye , on he o he hand, may some imes un independen ly om he o he wo laye s when i basically p o ides manage ial da a o be s o ed and no p ocessed in eal ime, as o example o he elabo a ion o yield moni o ing maps o his o ical e olu ion g aphs o soil e ili y. 4. Resul s and Discussion: Senso A chi ec u e The senso a chi ec u e p oposed o mee he equi emen s o ag icul u al en i onmen s, ehicles, and asks desc ibed in p e ious sec ions is in line wi h he gene al ehicle a chi ec u e p esen ed in [11], and g aphically ep esen ed by Figu e 4, in which he en i e in elligen sys em is a icula ed a ound ou s uc u al subsys ems: local pe cep ion, global localiza ion, ac ua ion and con ol, and da a p ocessing. The ou h subsys em, da a p ocessing, comp ises he se o compu e s, p ocessing uni s, DSPs (digi al signal p ocesso s), and embedded con olle s hos ing decision making algo i hms, ecei ing senso da a, and sending ac ua ion commands acco ding o a gi en so wa e a chi ec u e. The o he h ee subsys ems inco po a e a mul iplici y o senso s ha ha e been g ouped in he subsec ions 4.1 o 4.4 de eloped below. The e will be, addi ionally, o he ancilla y componen s as ex a ba e ies, Senso s 2010, 10 11234 powe con e e s, signal condi ione s, o eme gency s op bu ons, which assis he en i e ehicle and he e o e do no belong o any o he ou s uc u al subsys ems illus a ed in Figu e 4. The p ac ical necessi y o edundancy and senso usion o enhance eliabili y o ces he luen coope a ion among he ou key subsys ems, which is a o ed by he modula designed p oposed. A popula example is gi en by he combina ion o local in o ma ion cap u ed by lida s o machine ision (pe cep ion subsys em) wi h global localiza ion p o ided by a sa elli e na iga ion sys em. Two c ucial aspec s need o be conside ed be o e using he in o ma ion coming om di e en senso s: equency and sys em o coo dina es. I is e iden ha global localiza ion and (local) ehicle- ixed posi ioning will na u ally ha e di e en coo dina e sys ems, and he e o e he app op ia e coo dina e ans o ma ion will ha e o be ca ied ou be o e me ging he da a. Each senso will gene a e eadings a a pa icula a e, howe e he e will be a main loop equency a which he o de s a e execu ed by he p ocessing subsys em. Main loop equencies in he o de o 10 Hz ha e been p o ed success ul o ope a ing o - oad ehicles au onomously [12]. Imaging senso s can acqui e images a 30 ames pe second ( ps), and ine ial senso s can easily each highe equencies, bu GPS ecei e s a e ypically se o send da a s ings a ei he 1 Hz o 5 Hz, which a e below he ecommended execu ion a es. Consequen ly, a sui able a chi ec u al design needs o be implemen ed in o de o achie e he bes pe o mance o he selec ed onboa d senso s. Figu e 4. Fou -co e subsys em a chi ec u e o in elligen ag icul u al ehicles. 4.1. Senso s o Local Pe cep ion and Vicini y Moni o ing The in o ma ion ha an in elligen ehicle can ex ac om i s su oundings is much iche han ha ga he ed by humans. Lase ange inde s can accu a ely ace p o iles a dis ances whe e he human eye can only dis inguish ague o ms, ul asonic de ices make use o equencies unno iceable o human hea ing, elepho os can see de ail a away while mac o lenses can pe cei e he de ail o he mic oscopic. Nea in a ed, ul a iole , o he mal in a ed a e bands o he ligh spec um ha a e un eachable o us. High speed shu e s can eeze apid mo emen s and ib a ions. Human pe cep ion is g ea ly impai ed a nigh bu he e a e se e al senso s ha wo k unde sca ce na u al ligh . Ye , he human b ain is unique and he e is no compu e compa able o i . Fo ha eason, in elligen ehicles Senso s 2010, 10 11241 in many pa h planning algo i hms. I p o ides he o ien a ion o he ehicle wi h espec o he no h, and can be es ima ed wi h an ine ial measu emen uni as he yaw angle is de e mined by in eg a ing he yaw a e a ound axes pe pendicula o he local angen plane. An op ional senso o es ima e headings is he luxga e compass, bu he amoun o elec onic de ices inside he cabin may c ea e magne ic ields and a ec he pe o mance o he compass. The ac ha onboa d GPS ecei e s can p o ide posi ion and ime o ehicles a a equency o 5 Hz makes hese senso s suscep ible o es ima e eloci y and heading. This ea u e may o may no be accep able acco ding o he applica ion pu sued. To begin wi h, GPS heading canno be known ins an aneously unless a se ies o poin s ha e been p ope ly eco ded. Bu e en in hese ci cums ances, accu acy and signal s abili y ha e o be qui e high; o he wise, luc ua ions a ound a e age alues will esul in unaccep able headings, as ins an aneous heading alues will oscilla e un ealis ically. Addi ionally, he ime needed o ge s able se ies o da a may be excessi ely long and no heading o speed will be a ailable un il he ehicle has a eled a signi ican po ion o he planned cou se. The ollowing wo igu es illus a e his issue when h ee-dimensional ins an aneous mapping es s we e conduc ed in a wine y ineya d in he summe o 2010. Figu e 10(a) shows he ajec o y ollowed by he mapping ehicle [ ac o in Figu e 2 (c)] ep esen ed in local angen plane coo dina es; a comple e ow wes -eas , and hal neighbo ing ow in he e u n di ec ion eas -wes . The ac o ollowed he s aigh lanes o he ineya d wi h app oxima e heading angles o +80º and -100º espec i ely, easily deducible om he GPS-based ajec o y o Figu e 10(a). The ins an aneous heading o each poin o he ajec o y, es ima ed wi h an algo i hm ha conside s 32-poin se ies, is ep esen ed in Figu e 10(b). A i s sigh , he plo seems s able and co ec excep o he wo ou lie s no iceable in Figu e 10, bu when heading angles we e in oduced in he mapping algo i hm designed o me ge mul iple 3D poin clouds in o a unique map, se e al inaccu acies showed up. Figu e 10. GPS-based heading es ima ion: (a) Vehicle ajec o y. (b) Ins an heading. (a) Senso s 2010, 10 11242 Figu e 10. Con . (b) Figu e 11(a, op) p o ides one o he 67 RGB images o he ineya d lane used o cons uc he 3D map, and Figu e 11 (a, bo om) displays i s 3D iew. The i ual ep esen a ion o he ines u ilizes ue colo o dis inguish ege a ion (da k g een) om soil (ligh b own). A op iew o he wo i ual ows is gi en in Figu e 11(b). Figu e 11. 3D e ain mapping: (a) Sample image ( op) wi h i s 3D iew (bo om). (b) Heading e o s in op iew. (a) (b) The indi idual maps wi h o igins loca ed a he coo dina es gi en by he wo ou lie s ound in Figu e 10 we e au oma ically elimina ed by he mapping algo i hm and he e o e do no appea in Senso s 2010, 10 11243 Figu e 11(b), bu hey we e he cause o la ge e o s in he alignmen o he images. No ice ha many o hese 3D images we e co ec ly displayed, bu when hey we e used o comple e he global map o he wo ows, he lack o accu acy in he es ima ion o he heading esul ed in equen misalignmen s and de ec i e o ien a ion o a ious po ions o he lane. The comple e map is ende ed in he bi d-eye iew o Figu e 11(b, bo om), and wo augmen ed po ions a e displayed in he op image. Fo his applica ion, a mo e eliable sou ce o headings is he e o e necessa y. Fu he mo e, he eliabili y o GPS was no sa is ac o y ei he , as wo impo an ou lie s appea ed, e en hough he e we e always be ween six and nine sa elli es in solu ion. 4.4. Non- isual Senso s o Moni o ing P oduc ion Pa ame e s Apa om all he in o ma ion acqui ed h ough he ision senso s men ioned in Sec ion 4.1, he e a e o he pa ame e s ha a e impo an o p oduce s and canno be de e mined emo ely. The eal ime es ima ion o he ha es ed c op is no mally acked by a yield senso moun ed inside he combine ha es e . A e age alues o yield a e globally e e enced wi h a GPS ecei e so ha yield maps can be gene a ed a he end o he season. Yield moni o ing is popula o g ain p oduc ion, mainly co n and soybeans, as well as o wine p oduc ion. O he in e es ing maps like hose ep esen ing ain all o soil p ope ies canno be buil ―on he ly‖, gene ally speaking, as pene ome e s, PH-me e s, conduc i i y p obes, and o he senso s ha e no been inco po a ed o ehicles wi h no mali y ye . The comple e au oma ion o an ag icul u al ehicle in ol es many mo e unc ions han au oma ic s ee ing. Na iga ion, o example, may equi e gea shi ing, b ake ac i a ion, h o le con ol, o di e en ial locking. All hese ac ions, when execu ed au oma ically, need o ack he posi ion o le e s and pedals wi h po en iome e s and encode s. An in elligen implemen , o ins ance, needs o sense i s posi ion (up o oad a eling and headlands; down o a ming) as well as he d ag o ce incu ed by he pulling ehicle (axle load cells). 4.5. Onboa d In eg a ion o he Comple e Senso Ne wo k All he senso s ha comp ise he a chi ec u e p oposed need o be op imally in eg a ed in he in elligen ag icul u al ehicle o i s use o be easy, com o able, and sa e. The physical loca ion o he senso s is decisi e and needs o be ca e ully planned. The main p ocesso (s) o he ehicle, as well as moni o s and sc eens, will p e e ably be ins alled inside he cabin, whe e ib a ion, dus , and mois u e will ha e a minimum impac . Consequen ly, p o isions should be made o se ing a nea amewo k o mul iple cables en e ing and exi ing he cabin. A second ba e y, independen om he ehicle’s own ba e y, is always e y help ul o p ese e he desi ed au onomy o he diesel engine. Code debugging and he simul anei y o mul iple senso s can easily exhaus he main ba e y when he engine is no unning bu compu e s and senso s a e on. In addi ion, s a ing he engine esul s in empo a y ol age d ops ha u n he onboa d senso s o , in alida ing p e ious ini ializa ion ou ines. GPS ecei e s, o example, need se e al minu es o lock he p ope numbe o sa elli es, and e e y ime he ol age is cu o , he cons ella ion sea ch needs o s a o e again. Fo his eason, an au oma ed double-ba e y cha ge sys em is e y con enien . This sys em o powe ing he added elec onic de ices was success ully implemen ed in he ac o o Figu e 2(a,c), and i cha ges bo h ba e ies wi h he engine al e na o when he engine is unning, bu powe s all he elec onics onboa d jus wi h he seconda y Senso s 2010, 10 11244 ba e y. Only in he unlikely e en o unning he seconda y ba e y ou while he engine is o , he p incipal ba e y would powe senso s and compu e s. The senso s which do no equi e a special posi ion in he ehicle, such as compasses o ine ial measu emen uni s, a e be e kep in he cabin; hey a e well p o ec ed and connec ions—powe and signal—a e kep sho . Fo many senso s, howe e , he e is an ad an ageous, o e en unique, loca ion in he ehicle. Op ical encode s, o ins ance, need o be moun ed on he ( on -axle) wheel kingpins, and he e o e no o he placemen makes sense o di ec ly ack s ee ing angles. The GNSS an enna ecei es be e da a when loca ed high, as mul ipa h e lec ions om he g ound and om low ege a ion can be a oided; hus, a cen e ed posi ion on he cabin oo is usually he p e e ed op ion. Lida s and came as may be moun ed ei he a he on o he ehicle o on he cabin, depending on he so o scenes being sensed. The complexi y o o ches a ing all he senso s, ac ua o s, and compu e s, while assu ing he igh ol age powe and he synch oniza ion o da a acqui ed a a ious equencies, calls o a well designed senso and sys em a chi ec u e. Figu e 12 shows a pic o ial ep esen a ion o a gene ic senso ne wo k o an in elligen ag icul u al ehicle. Figu e 12. Gene ic senso ne wo k o an in elligen ag icul u al ehicle. A e he ne wo k o senso s onboa d has been p ope ly designed, choosing he op imum senso s o each subsys em, selec ing hei mos a o able loca ion wi hin he ehicle, and linking hem eliably wi h he main p ocesso , i is ime o e isi he h ee-laye ask classi ica ion and discuss on he in elligen capabili ies o he a chi ec u e p oposed. The e is no physical embodimen o he ask laye s because hey a e concep ually concei ed as con aine s o such i ual elemen s as in o ma ion, isk p edic ion, o expe sys ems. The Machine Ac ua ion Laye is he laye ha holds he se o algo i hms con e ing in elligence o he sys em, which may be physically loca ed in he main compu e , in se e al DSPs, o e en in a mul iplici y o senso -based agen s as he p ocessing boa d o a sma came a. The Senso s 2010, 10 11245 design and in e ela ion o all hese algo i hms is wha some au ho s conside o be he sys em a chi ec u e, al hough in eali y i is he so wa e a chi ec u e. The model en isioned in his a icle conside s he sys em a chi ec u e o be he en elope ha comp ises bo h ha dwa e and so wa e a chi ec u es. The o me e e s o he senso and complemen a y ha dwa e desc ibed along his a icle. A de ailed exposi ion o he la e would equi e ano he pape in he line ollowed by [8], al hough a gene ic iew may be ou lined he e. Taken as a whole, he ac ua ion plan o he ehicle can ollow he biology-based eac i e app oach o he subsump ion a chi ec u e de eloped by B ooks [15], o on he con a y i may include a cogni i e engine inside he Ac ua ion Laye . While bo h ha e been p o ed o pe o m success ully o a numbe o obo s in pa icula si ua ions, ag icul u al ehicles usually bene i om bo h app oaches, and consequen ly he bes esul s a e o en achie ed wi h a hyb id model implemen ing ideas aken om bo h. Se e al so wa e a chi ec u es o ag icul u al o - oad ehicles a e desc ibed in he s udy cases p esen ed in [14]. 5. Conclusions The A al Sea in Asia and Lake Chad in A ica ha e su e ed a educ ion o hei su ace down o 10% o hei o iginal size in ba ely 30 yea s, as a consequence o unwise decisions made by he ag icul u al sec o [16]. The impac o he imp uden exploi a ion o na u al esou ces a ec s la ge a eas o he globe, o en in ol ing se e al na ions. Because he way he wo ld g ows i s ood needs o change in o de o igh amine while assu ing sus ainabili y, a echnological e olu ion is called o . The applica ion o new echnologies o ag icul u e, h ough he disciplines o p ecision a ming and ag icul u al obo ics, can b ing p ac ical solu ions o make ood p oduc ion mo e a ional and e icien . Ag icul u al ehicles a e p i ileged agen s in which new echnologies a e cu en ly being implemen ed. New ways o ca ying ou adi ional asks, such as au oma ic ha es ing, a iable a e applica ions, o wa e s ess si e-speci ic de ec ion can be key in he u u e o assu e no el p oduc ion sys ems compa ible wi h popula ion g ow h and en i onmen p ese a ion. These echnologies, howe e , equi e he op imum implemen a ion o senso s, ac ua o s, and compu e s in he so-called in elligen ehicles. This a icle p oposes a senso a chi ec u e o endow ag icul u al ehicles wi h he necessa y capabili ies o pe o m asks wi hin he amewo k o p ecision ag icul u e and ield obo ics. This senso a chi ec u e, in conjunc ion wi h he so wa e a chi ec u e, cons i u es he ehicle sys em a chi ec u e. The ha dwa e a chi ec u e de eloped de ines ou key g oups, o amilies, o senso s: local pe cep ion and icini y moni o ing, global posi ioning, a i ude and con ol, and non- isual acking o p oduc ion pa ame e s. These ou senso amilies a e no mally p esen in mos in elligen ehicles, al hough he pa icula senso s ac ually included in each g oup, depend on each speci ic applica ion. The a angemen o senso s acco ding o his a chi ec u e has a o ed edundancy and he p ac ical implemen a ion o new echnologies in ag icul u al o - oad ehicles, complying wi h especial equi emen s o asks and en i onmen s. Table 1 p o ides a c oss- able ela ing some ad anced ehicle asks wi h he senso s needed o accomplish hem. Fu u e needs will likely esul in new asks, no el senso s, and as a esul an augmen a ion o he o iginal a chi ec u e p oposed; bu he adop ion o new echnologies by he ag icul u al sec o is a ma e o ime. Senso s 2010, 10 11246 Table 1. Senso - ask c oss able. Acknowledgemen s The esea ch ac i i ies de o ed o he s udy o senso and sys em a chi ec u es o ag icul u al in elligen ehicles ca ied ou du ing 2010 ha e been suppo ed by he Spanish Minis y o Science and Inno a ion h ough P ojec AGL2009-11731. Re e ences and No es 1. Bou ne, J.K. The global ood c isis—The end o plen y. Na ional Geog aphe 2009, June, 26-59. 2. Wexle , B.E. B ain and Cul u e, 1s ed.; MIT P ess: Camb idge, MA, USA, 2006; pp. 75-83. 3. Reid, J.F.; Zhang, Q.; Noguchi, N.; Dickson, M. Ag icul u al au oma ic guidance esea ch in No h Ame ica. Compu . Elec on. Ag ic. 2000, 25, 155-167. 4. Keiche , R.; Seu e , H. Au oma ic guidance o ag icul u al ehicles in Eu ope. Compu . Elec on. Ag ic. 2000, 25, 169-194. 5. Ro i a-Más, F. Recen inno a ions in o - oad in elligen ehicles: in- ield au oma ic na iga ion. Recen Pa en s Mechan. Eng. 2009, 2, 169-178. 6. Ro i a-Más, F.; Zhang, Q.; Hansen, A.C. Dynamic beha io o an elec ohyd aulic al e: Typology o cha ac e is ic cu es. Mecha onics 2007, 17, 551-561. 7. Ro i a-Más, F.; Zhang, Q.; Reid, J.F. S e eo ision h ee-dimensional e ain maps o p ecision ag icul u e. Compu . Elec on. Ag ic.2008, 60, 133-143. 8. Blackmo e, S.; Foun as, S.; Ha e, H. P oposed sys em a chi ec u e o enable beha io al con ol o an au onomous ac o . In P oceedings o Au oma ion Technology o O - oad Equipmen Con e ence, Chicago, IL, USA, July 2002; pp. 13-23. Senso s 2010, 10 11247 9. Shi ley, R. Ven u a Coun y’s changing ag icul u e. Resou ce 2008, Feb ua y, 5-11. 10. Mye s, M.L. The ag icul u al sa e y and heal h challenge. Resou ce 2009, Oc obe /No embe , 7-9. 11. Ro i a-Más, F. Gene al a chi ec u e o in elligen ag icul u al ehicles. In P oceedings o Robo ics, Cle mon -Fe and, F ance, Sep embe 2010. 12. Bebel, J.C.; Raskob, B.L.; Pa ke , A.C.; Bebel, D.J. Managing complexi y in an au onomous ehicle. In P oceedings o Posi ion, Localiza ion, and Na iga ion Symposium, San Diego, CA, USA, May 2006; pp. 356-365. 13. G ewal, M.S.; Weill, L.R.; And ews, A.P. Global Posi ioning Sys ems, Ine ial Na iga ion, and In eg a ion; John Wiley & Sons: New Yo k, NY, USA, 2001. 14. Ro i a-Más, F.; Zhang, Q.; Hansen, A.C. Mecha onics and In elligen Sys ems o O -Road Vehicles. 1s ed.; Sp inge -Ve lag: London, UK, 2010. 15. B ooks, R. Camb ian In elligence: The Ea ly His o y o The New AI, MIT P ess: Camb idge, MA, USA, 1999. 16. Micklin, P.; Aladin, N.V. Reclaiming he A al Sea. Sci. Ame . 2008, Ap il, 44-51. © 2010 by he au ho s; licensee MDPI, Basel, Swi ze land. This a icle is an open access a icle dis ibu ed unde he e ms and condi ions o he C ea i e Commons A ibu ion license (h p://c ea i ecommons.o g/licenses/by/3.0/).