Real-time extensive livestock monitoring using lpwan smart wearable and infrastructure

applied sciences A icle Real-Time Ex ensi e Li es ock Moni o ing Using LPWAN Sma Wea able and In as uc u e Robe o Casas 1,* , A u o He mosa 1,Ál a o Ma co 1,2 , Te esa Blanco 1,2 and F ancisco Ja ie Za azaga-So ia 1   Ci a ion: Casas, R.; He mosa, A.; Ma co, Á.; Blanco, T.; Za azaga-So ia, F.J. Real-Time Ex ensi e Li es ock Moni o ing Using LPWAN Sma Wea able and In as uc u e. Appl. Sci. 2021,11, 1240. h ps://doi.o g/ 10.3390/app11031240 Academic Edi o : Ak am Alomainy Recei ed: 31 Decembe 2020 Accep ed: 25 Janua y 2021 Published: 29 Janua y 2021 Publishe ’s No e: MDPI s ays neu al wi h ega d o ju isdic ional claims in published maps and ins i u ional a il- ia ions. Copy igh : © 2021 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 (CC BY) license (h ps:// c ea i ecommons.o g/licenses/by/ 4.0/). 1A agon Ins i u e o Enginee ing Resea ch, Uni e si y o Za agoza, 50018 Za agoza, Spain; a u ohache@uniza .es (A.H.); ama [email p o ec ed] (Á.M.); [email p o ec ed] (T.B.); ja y@uniza .es (F.J.Z.-S.) 2GeoSpa ium Lab S.L., Ca los Ma x 6, 50015 Za agoza, Spain *Co espondence: casas@uniza .es; Tel.: +34-976-762-856 Fea u ed Applica ion: The wea able and in as uc u e p esen ed in his wo k has p o en o be a aluable and lexible ool o eal- ime ex ensi e li es ock acking and moni o ing wi h egu- la o y compliance. Main ea u es a e online and o line moni o ing and da a logging; adap a i e dual adio (FSK and LoRaWAN) o bes ene gy/bandwid h/ ange balance; adap a i e mo emen da a s eaming modes wi h ec o quan iza ion comp ession me hod; and yea s o ba e y li e. Abs ac : Ex ensi e unsupe ised li es ock a ming is a habi ual echnique in many places a ound he globe. Animal elease can be done o mon hs, in la ge a eas and wi h di e en species packing and beha ing e y di e en ly. Ne e heless, he a me ’s needs a e simila : whe e li es ock is (and whe e has been) and how heal hy hey a e. The geog aphical a eas in ol ed usually ha e di icul access wi h ha sh o og aphy and lack o communica ions in as uc u e. This pape p esen s he design o a solu ion o ex ensi e li es ock moni o ing in hese a eas. Ou p oposal is based in a wea able equipped wi h ine ial senso s, global posi ioning sys em and wi eless communica ions; and a Low-Powe Wide A ea Ne wo k in as uc u e ha can un wi h and wi hou in e ne connec ion. Using adap i e analysis and da a comp ession, we p o ide eal- ime moni o ing and logging o ca le’s posi ion and ac i i ies. Ha dwa e and i mwa e design achie e e y low ene gy consump ion allowing mon hs o ba e y li e. We ha e ho oughly es ed he de ices in di e en labo a o y se ups and e alua ed he sys em pe o mance in eal scena ios in he moun ains and in he o es . Keywo ds: animal moni o ing; low-powe wide a ea ne wo ks; LoRaWAN; wea able de ices design 1. In oduc ion Ex ensi e li es ock a ming is a ypical echnique in many places a ound he wo ld. I occu s on 25% o global land su ace and suppo s a ound 200 million subsis ence pas o al households [1]. In A ica, 40% o he land is dedica ed o ex ensi e pas o alism [2]. Many habi a s impo an o biodi e si y conse a ion ha e been c ea ed by and a e s ill inhe - en ly linked o ex ensi e li es ock p oduc ion, in pa icula g azing. Fo ins ance, ex ensi e g azing is conside ed i al o main aining many biodi e si y- ich habi a s in Eu ope. I is sugges ed as op imum managemen o de-in ensi ied g assland o enhance biodi e si y [ 3 ]. Ex ensi e g azing was epo ed o posi i ely in luence swa d species composi ion and s uc u e which, in u n, p o ided a ou able condi ions o colonizing auna. Fu he mo e, ex ensi e li es ock and he ela ed g azing is also c i ical o main aining many o Eu ope’s cul u al landscapes and sus aining u al communi ies. O e he cen u ies, pas o alism, and anshumance (seasonal mo emen o li es ock be ween g azing a eas) c ea ed a wide a ie y o speci ic cul u al landscapes. The la ges emaining ex ensi e pas o al sys ems on pe manen wood pas u es in Eu ope a e Dehesa in Spain and Mon ado in Po ugal [ 4 ]. G azing and anshumance a e o pa icula impo ance o he p ese a ion o open landscapes in he Eu opean moun ains. In addi ion, socie al conce n o e he Appl. Sci. 2021,11, 1240. h ps://doi.o g/10.3390/app11031240 h ps://www.mdpi.com/jou nal/applsci Appl. Sci. 2021,11, 1240 2 o 18 wel a e o a m animals has ecen ly inc eased and a g owing numbe o ci izens in many coun ies hink ha i is impo an o p o ec he wel a e o a m animals [ 5 ]; ex ensi e li es ock is belie ed o be ad an ageous in e ms o animal wel a e. Ex ensi e li es ock a ming is, in mos cases in Eu ope, unsupe ised. Ca le, mainly bo ine, a e ee in la ge enclosed emo e a eas o mos o he yea . Al hough shephe ds do con ol he ca le pe iodically, whene e he wea he allows i , animals a e collec ed only o one mon h o ake a e e ina y con ol. This implies mo ing o he a ea whe e he animals a e li ing and looking o hem. We a e alking abou hund eds o hec a es, in mos cases wi hou any kind o oad. Consequen ly, i is usual ha shephe ds spend mos o hei ime walking h ough he ield while hey a e looking o he animals. In o de o assis hem in his ask, some Global Na iga ion Sa elli e Sys em (GNSS)-based solu ions ha e been de eloped o con olling he mo emen s o he animals and epo ing hem o he a me s. Fo ins ance, Pé ez e al. [ 6 ] use a GPRS-based sys em o moni o ing lidia ca le, meanwhile La ia e al. [7] use a SigFox-based one. In Eu ope, his ype o li es ock a ming is usually es ic ed o emo e a eas wi h low popula ion densi y. This makes hese a eas less in e es ing o elecommunica ion se ice p o ide (TSP), o o e high wand-wind se ices: high in es men o small e u n (see Figu es 1and 2). E en o echnologies de eloped o he imp o emen o In e ne o Things (IoT) canno o e co e age o all hese emo e a eas (see Figu e 2). Appl. Sci. 2021, 11, x FOR PEER REVIEW 3 o 19 Figu e 1. 3G co e age p o ided by he main elecommunica ion se ice p o ide (TSP) in he Py enees a ea in Spain (sou ce websi es om Mo is a , O ange, Voda one and MasMo il, Decembe 2020). Figu e 1. 3G co e age p o ided by he main elecommunica ion se ice p o ide (TSP) in he Py enees a ea in Spain (sou ce websi es om Mo is a , O ange, Voda one and MasMo il, Decembe 2020). The e a e se e al o he al e na i es o cellula ne wo ks o IoT moni o ing. These can be mesh ne wo ks, whe e link quali y indica o s could be used o loca e animals wi hou GNSS [ 8 ]. This kind o deploymen is hea ily dependen o he ga eway which usually becomes he bo leneck and, depending on he o og aphy, equi es a ou e in as uc u e ha needs conside able amoun o ene gy [ 9 ]. The e a e imes whe e he link be ween he wi eless senso ne wo k (WSN) and he in e ne is a Wi-Fi anscei e , which is mean o inc ease he ange in zones wi hou GPRS co e age [10]. Appl. Sci. 2021,11, 1240 3 o 18 Appl. Sci. 2021, 11, x FOR PEER REVIEW 3 o 19 Figu e 1. 3G co e age p o ided by he main elecommunica ion se ice p o ide (TSP) in he Py enees a ea in Spain (sou ce websi es om Mo is a , O ange, Voda one and MasMo il, Decembe 2020). Figu e 2. Co e age p o ided by SigFox (up) and NB-IoT by Voda one (down, and p obably he bes one in Spain) in he Py enees a ea in Spain (sou ce websi es om SigFox and Voda one, Decembe 2020). Low-Powe Wide A ea Ne wo ks (LPWAN) a e also sui able op ions, b inging o- ge he he low-powe quali ies o WSN and long- ange capaci ies o cellula ne wo ks. The e a e no el modula ions such as LoRa ha can be used on mesh opologies o ex end ange, al hough i is usually used in s a opology. Gi en he bene i s such as long ange, low-powe and cos sa ings, LoRa is ideal o ca le moni o ing [11–13]. Ano he appealing al e na i e is he use o UAVs, which is an eme ging applica ion in he ield o p ecision ag icul u e [ 14 – 16 ]. Speci ic usage o UAVs o ca le moni o ing is desc ibed in [ 17 – 20 ] which elay on image p ocessing echniques o iden i y animals and accoun o hem. Howe e , ha app oach ha e some limi a ions when we a e conce ned abou acking speci ic indi iduals (a shephe d would be in e es ed in loca ing his animals only) and does no allow con inuous acking as GNSS solu ions do. Combined usage o UAVs and GNSS is also p oposed in [ 21 ], which conside d ones o easing he e ie e o he acking in o ma ion acqui ed by GPS colla s wo n by he animals. Ne e heless, a numbe o echnical and adminis a i e issues a ise ha s ill possess some di icul ies o using UAVs in ex ensi e ca le moni o ing, such as he cos o he need o specialis ope a o s (i is equi ed mid/high-size UAVs o co e ing big ex ensions), which may also equi e a lying license and special pe mission in some coun ies [22]. Besides localiza ion acking o he animals using GNSS, egis e ing hei s a us and ac i i ies can be o in e es o shephe ds [ 23 ]. Ac i i y de ec ion can be pe o med wi h di e en s a egies such as adio equency iden i ica ion (RFID) o in e ca le d inking beha io and wa e in ake by de ec ing p oximi y o speci ic places [ 24 ]. Using empe - a u e [ 25 ] and o he biosenso s [ 26 ] can help o de ec a e e s a e and o asse animal well-being. The mos common a e ine ial senso s based on accele ome e s and gy oscopes ha a e able o iden i y animal ac i i ies such as es ing, walking, ma ing o eeding among o he s [27,28]. Appl. Sci. 2021,11, 1240 4 o 18 Using ine ial da a o ex ac ac i i ies is a e y powe ul, cheap, ene gy-e icien and, consequen ly, b oadly adop ed s a egy. Ne e heless, ine ial senso s canno be di ec ly used, due o he high h oughpu equi ed: a 12-bi 3-axis accele ome e sampling a 10 Hz gene a es mo e han 1.25 Mb o da a e e y hou . Sending his amoun o da a h ough LPWAN is oo cos ly in bo h economic and ene ge ic e ms, some imes iola es egional es ic ions (e.g., LoRaWAN) o i is jus echnically impossible (e.g., Sig ox). Thus, aw da a needs o be p ocessed in o de o ex ac he mos impo an ea u es, which also helps sa ing and sending way less da a han gene a ed. The e a e wo main s a egies ha can be applied. Cus om da a p ocessing echniques o speci ically de ec he equi ed ac i i ies pe o med by animals [ 29 ] and humans [ 30 ] is a ending esea ch opic. I is also possible o use da a comp ession s a egies o educe he amoun o in o ma ion o be sen and hen decomp ess and p ocess i in he cloud. Usual echniques o accomplishing da a comp ession a e P incipal Componen Analysis (PCA) [ 31 – 34 ], Sequen ial Fo wa d Selec ion (SFS) [ 35 ], Random Subse Fea u e selec ion (RSFS) [ 35 ], Independen Componen Analysis (ICA) [ 34 ], I-PCA [ 34 ], Vec o Quan iza ion (VQ) [36], and F equency Sensi i e Compe i i e Lea ning (FSCL) [37,38]. In his con ex , we aim o moni o animals in ex ensi e a ming in he no h pa o Spain (in he Py enees a ea, nea he bo de wi h F ance). We ha e wo main objec i es: Fi s ly, we need o ack he mo emen o he animals o e alua e hei impac on he plan biodi e si y. This could be done jus by using a de ice i ed wi h a GNSS ecei e ha would collec he posi ions du ing he ime he animals a e in he moun ain and download hem when hey a e e u ned o he ba ns. Secondly, we in end o i ually educe he dis ance be ween he animals and he shephe ds. Fo his eason, i is necessa y o elay he posi ions and s a uses o he animals in eal ime. As he selec ed a ea has no gua an eed co e age om any TSP, i has been necessa y o ocus on a di e en app oach. Ou p oposal is a wea able equipped wi h ine ial senso s, GNSS and wi eless com- munica ions. Including adap i e analysis and VQ comp ession model in he wea able achie ing high comp ession a ios wi h good peak signal- o-noise a ios (PSNR) [ 39 ] allows o eal- ime moni o ing and logging o ca le’s posi ion and ac i i ies. We ha e chosen a anscei e wi h bo h FSK and LoRa capabili ies. As i has been men ioned be o e, he e is no co e age by TSPs and links based on hese modula ions ha e p o en i s u ili y in open a ming scena ios. We ocused ou elec onic design on achie ing e y low ene gy consump ion, allowing o se e al mon hs o ba e y un ime. We ha e e alua ed bo h he wea able and he sys em pe o mance o acking cows in eal scena ios, such as moun ains and o es . The es o he pape is s uc u ed as ollows: i s , he cha ac e is ics o he wea able, i s capabili ies and he ope a ion modes a e de ined; hen, labo a o y expe imen a ion and ield es s a e p esen ed; and las ly, sys em cha ac e is ics and wo k esul s a e discussed. 2. Ma e ials and Me hods 2.1. Wea able Design The wea able in eg a es se e al senso s o measu ing empe a u e, accele a ion, and magne ic ields. I also ea u es a GNSS posi ioning module and a LoRa anscei e able o ansmi using FSK and LoRa modula ions o communica ion pu poses. In he p ocessing side he e is a 64 Mbi NOR lash o s o ing acking da a, a 16-bi MCU wi h low-powe ea u es and a powe managemen ci cui y ha allows being powe ed wi h any ba e y chemis y and echa ging i om di e en powe sou ces like sola panels, induc ion, he moelec ic and piezoelec ic (Figu e 3[40]). Appl. Sci. 2021,11, 1240 5 o 18 Appl. Sci. 2021, 11, x FOR PEER REVIEW 5 o 19 The wea able in eg a es se e al senso s o measu ing empe a u e, accele a ion, and magne ic ields. I also ea u es a GNSS posi ioning module and a LoRa anscei e able o ansmi using FSK and LoRa modula ions o communica ion pu poses. In he p o- cessing side he e is a 64 Mbi NOR lash o s o ing acking da a, a 16-bi MCU wi h low- powe ea u es and a powe managemen ci cui y ha allows being powe ed wi h any ba e y chemis y and echa ging i om di e en powe sou ces like sola panels, induc- ion, he moelec ic and piezoelec ic (Figu e 3 [40]). Figu e 3. Ha dwa e block diag am. The design objec i es conside ed se e al pe spec i es, mainly ha o he shephe d and ha o he animal, as wo ypes o use s wi h di e en needs. Thus, he design speci- ica ions we e aimed a ensu ing high esis ance o shock and wa e ; compa ibili y wi h di e en exis ing s ap a achmen me hods; ease o a achmen and emo al; and ac- cep ance o he de ice by he hos . We designed, es ed and edesigned a ious 3D enclo- su es o e i y di e en o maliza ions, inishing angles, and manu ac u ing ma e ials. Finally, we selec ed a clam design wi h an o- ing ha s ays in place wi h ou sc ews. The enclosu e has wo slo s wi h di e en sizes allowing di e en -sized animals o com o a- bly wea hem (e.g., sheep and cows). We ha e success ully pe o med wa e p oo IP68 es ing. Final design can be seen in Figu e 4. Figu e 4. 3D and inal p o o ype. 2.2. Logic and Communica ions We conside se e al modes each wi h di e en con igu a ion pa ame e s: ● Beacon: GNSS is sampled a TGNSS and once posi ion is acqui ed, i is sa ed in lash memo y, sen ia LoRaWAN and hen he anscei e wai s o a con igu a ion Figu e 3. Ha dwa e block diag am. The design objec i es conside ed se e al pe spec i es, mainly ha o he shephe d and ha o he animal, as wo ypes o use s wi h di e en needs. Thus, he design speci ica ions we e aimed a ensu ing high esis ance o shock and wa e ; compa ibili y wi h di e en exis ing s ap a achmen me hods; ease o a achmen and emo al; and accep ance o he de ice by he hos . We designed, es ed and edesigned a ious 3D enclosu es o e i y di e en o maliza ions, inishing angles, and manu ac u ing ma e ials. Finally, we selec ed a clam design wi h an o- ing ha s ays in place wi h ou sc ews. The enclosu e has wo slo s wi h di e en sizes allowing di e en -sized animals o com o ably wea hem (e.g., sheep and cows). We ha e success ully pe o med wa e p oo IP68 es ing. Final design can be seen in Figu e 4. Appl. Sci. 2021, 11, x FOR PEER REVIEW 5 o 19 The wea able in eg a es se e al senso s o measu ing empe a u e, accele a ion, and magne ic ields. I also ea u es a GNSS posi ioning module and a LoRa anscei e able o ansmi using FSK and LoRa modula ions o communica ion pu poses. In he p o- cessing side he e is a 64 Mbi NOR lash o s o ing acking da a, a 16-bi MCU wi h low- powe ea u es and a powe managemen ci cui y ha allows being powe ed wi h any ba e y chemis y and echa ging i om di e en powe sou ces like sola panels, induc- ion, he moelec ic and piezoelec ic (Figu e 3 [40]). Figu e 3. Ha dwa e block diag am. The design objec i es conside ed se e al pe spec i es, mainly ha o he shephe d and ha o he animal, as wo ypes o use s wi h di e en needs. Thus, he design speci- ica ions we e aimed a ensu ing high esis ance o shock and wa e ; compa ibili y wi h di e en exis ing s ap a achmen me hods; ease o a achmen and emo al; and ac- cep ance o he de ice by he hos . We designed, es ed and edesigned a ious 3D enclo- su es o e i y di e en o maliza ions, inishing angles, and manu ac u ing ma e ials. Finally, we selec ed a clam design wi h an o- ing ha s ays in place wi h ou sc ews. The enclosu e has wo slo s wi h di e en sizes allowing di e en -sized animals o com o a- bly wea hem (e.g., sheep and cows). We ha e success ully pe o med wa e p oo IP68 es ing. Final design can be seen in Figu e 4. Figu e 4. 3D and inal p o o ype. 2.2. Logic and Communica ions We conside se e al modes each wi h di e en con igu a ion pa ame e s: ● Beacon: GNSS is sampled a TGNSS and once posi ion is acqui ed, i is sa ed in lash memo y, sen ia LoRaWAN and hen he anscei e wai s o a con igu a ion Figu e 4. 3D and inal p o o ype. 2.2. Logic and Communica ions We conside se e al modes each wi h di e en con igu a ion pa ame e s: • Beacon: GNSS is sampled a T GNSS and once posi ion is acqui ed, i is sa ed in lash memo y, sen ia LoRaWAN and hen he anscei e wai s o a con igu a ion message om he se e be o e going o sleep. I no GNSS signal is de ec ed o T TIMEOUT , las known coo dina es a e sen . I can ollow a sma beha io i desi ed. Then, GNSS is acqui ed only i mo emen has been de ec ed be ween wo T GNSS pe iods. Mo emen condi ion is asse ed by he ine ial measu emen uni (IMU) i sel , as i ac i a es an in e up when an A TRHESOHLD is exceeded. I no sma beha io is se , he de ice sends GNSS ega dless o he mo emen condi ion. Appl. Sci. 2021,11, 1240 6 o 18 • Mo emen : on op o he beacon sma mode, he wea able s eams IMU da a a T SENSOR pe iod. As IMU aw da a s eam would lead o a du y cycle policy iola ion, we ha e implemen ed ou di e en beha io s o p e en exceeding i : # Con inuous: The easies way o ge maximum bi a e is o use FSK modula ion o achie e 50 Kbps. Conside ing ha he accele ome e has 14 bi p ecision measu emen s, a EU868 (869.525 MHz) equency, he one wi h he bes du y cycle (10%), we can s eam a : FSAMPLE =ACCPRECISION ×NAXIS DATARATE ×DUTY CYCLE ×PRATIO =14 b ×3 50 kb/s ×10% ×64 B +8 B 64 B =105.82 Hz, (1) Un o una ely, his is no LoRaWAN complian as du y cycle and bi a e a e abo e lim- i s. Thus, assuming du y cycle policies om EU868 egion (1%), usual DR5 LoRaWAN modula ion, acco ding o Equa ion (2), we can s eam a ound one sample pe second. FSAMPLE =14 b ×3 5.47 kb/s ×1% ×358 B 242 B =1.14 Hz, (2) I we jus send one piece o axis da a, he FSAMPLE can each 3.4 Hz, which could lead o sending mo e meaning ul da a. # Coded: his mode beha es simila ly o he p e ious, bu i includes an addi- ional s ep whe e he da a is comp essed p io o be sen [ 30 ]. VQ comp ession de ines a NxM look-up able (codebook), being N he numbe o cen oids, i.e., he numbe o di e en comp essed samples ha can be chosen o being ansmi ed, and M he size o he signal ha each cen oid ep esen s. The comp ession algo i hm accumula es a window o M samples o he signal and picks he cen oid ha be e i s he o iginal signal (minimizing mean squa ed e o ). Then, jus he index o he cen oid is ansmi ed, and he signal can be econs uc ed a he ecep ion by accessing he codebook wi h he index. The numbe o cen oids N and he size o he window M allows es ablishing in ad ance he comp ession a e o be achie ed (and hus inc easing e ec i e FSAMPLE): comp ession_ a e =M×sample_size log2(N)/8 , (3) The codebook has o be gene a ed p e iously wi h a machine lea ning algo- i hm ha is ained wi h samples o he da a o iden i y he cen oids ha will lead o be e PSNR igu es wi h he desi ed comp ession a e, Ou majo limi a ion is he memo y a ailable in he de ice. Using his mode, we can go up o 20 Hz sampling a DR5 o e LoRaWAN p o ocol wi h good pe o mance. # Bu s : his mode de ec s mo emen and hen accumula es se e al seconds o da a. Then, in o de o comply wi h LoRaWAN egula ion, we spli he packe be ween he a ailable channels using a schedule ha moni o s channel usage and p e en s messages o be sen when he du y cycle is abou o be exceeded. When his happens, new da a is disca ded un il we a e eady o send again. Using his mode, we can send 3-s bu s s a 10 Hz sampling a DR5 o e LoRaWAN p o ocol. # Sma : The sma s a egy is like bu s mode bu be o e sending da a, i ana- lyzes i o check i no mo emen is de ec ed; in ha case, mo emen da a is disca ded, and no sen . This condi ion is de e mined when he maximum peak o peak alue o he las i e seconds o da a is below a h eshold A NO-MOVE . This h eshold is con igu able o adjus sensi i i y o he de ice. • Con ig: his mode sends LoRaWAN beacons e e y 10 s, allowing downlink messages ha con igu e he wea able. Con ig mode is au oma ically en e ed a e a ha d ese Appl. Sci. 2021,11, 1240 7 o 18 and las s T CONFIG . In his mode we can e ie e he GNSS samples s o ed in lash memo y and send hem h ough FSK packe s. • S andby: his is he lowes ene gy consump ion mode whe e he wea able jus sleeps. The only way o exi his s a e is ia ha d ese . Figu e 5illus a es i mwa e’s low diag am in eg a ing all modes. To cus omize he se ings o his de ice, we de ine a con igu a ion message ia Lo- RaWAN, al hough i is also possible o do so ia USB. Acco ding o LoRaWAN speci ica ion, as a class A de ice, each ime an uplink message is sen , he e a e wo ecei ing windows when he de ice can ecei e downlink messages. This is he eason why in con ig mode he de ices send beacon messages each 10 s du ing 1 min. Appl. Sci. 2021, 11, x FOR PEER REVIEW 7 o 19 alue o he las i e seconds o da a is below a h eshold ANO-MOVE. This h eshold is con igu able o adjus sensi i i y o he de ice. ● Con ig: his mode sends LoRaWAN beacons e e y 10 s, allowing downlink messages ha con igu e he wea able. Con ig mode is au oma ically en e ed a e a ha d ese and las s TCONFIG. In his mode we can e ie e he GNSS samples s o ed in lash memo y and send hem h ough FSK packe s. ● S andby: his is he lowes ene gy consump ion mode whe e he wea able jus sleeps. The only way o exi his s a e is ia ha d ese . Figu e 5 illus a es i mwa e’s low diag am in eg a ing all modes. To cus omize he se ings o his de ice, we de ine a con igu a ion message ia Lo- RaWAN, al hough i is also possible o do so ia USB. Acco ding o LoRaWAN speci ica- ion, as a class A de ice, each ime an uplink message is sen , he e a e wo ecei ing win- dows when he de ice can ecei e downlink messages. This is he eason why in con ig mode he de ices send beacon messages each 10 s du ing 1 min. Figu e 5. Fi mwa e’s low diag am. Thus, he e is one downlink message used o wea able con igu a ion and wo pos- sible payloads o upload: one wi h GNSS da a and he o he is he IMU da a a ays. Bo h GNSS da a and con igu a ion messages use LoRaWAN in as uc u e because hey a e sho messages ha a e no sen e y o en and occu a de e minis ic imes. This allows us o mee ai use policies [41] needed o use a ailable LoRaWAN in as uc u e in he ield, being he bes -e o app oach, as he e a e many a ailable ga eways ha can ou e he packe s o he cloud. Besides ea u es such as secu i y, enc yp ion, e y coun e o moni o o un eachable nodes, LoRaWAN allows us o send con igu a ion packe s when- e e a de ice becomes a ailable and he ne wo k laye allows o acknowledged packe s. Meanwhile, IMU da a is hea y, occu s non-de e minis ically and needs o be ime s amped and managed as i occu s. We ha e conside ed using wo di e en modula ions as we ha e wo use cases: bu s and con inuous ansmission. The i s one is used when we migh be in e es ed in spo adic da a sampled and low equencies (less han 5 Hz), and i is possible o ge bene i om he ex a ange LoRa p o ides on noisy en i onmen s. This kind o beha iou is accep able o he deploymen in open ield a eas whe e he e a e animals ee o mo e long dis ances. Fi mwa e needs o check whe he he payloads Figu e 5. Fi mwa e’s low diag am. Thus, he e is one downlink message used o wea able con igu a ion and wo possible payloads o upload: one wi h GNSS da a and he o he is he IMU da a a ays. Bo h GNSS da a and con igu a ion messages use LoRaWAN in as uc u e because hey a e sho messages ha a e no sen e y o en and occu a de e minis ic imes. This allows us o mee ai use policies [ 41 ] needed o use a ailable LoRaWAN in as uc u e in he ield, being he bes -e o app oach, as he e a e many a ailable ga eways ha can ou e he packe s o he cloud. Besides ea u es such as secu i y, enc yp ion, e y coun e o moni o o un eachable nodes, LoRaWAN allows us o send con igu a ion packe s whene e a de ice becomes a ailable and he ne wo k laye allows o acknowledged packe s. Meanwhile, IMU da a is hea y, occu s non-de e minis ically and needs o be ime s amped and managed as i occu s. We ha e conside ed using wo di e en modula ions as we ha e wo use cases: bu s and con inuous ansmission. The i s one is used when we migh be in e es ed in spo adic da a sampled and low equencies (less han 5 Hz), and i is possible o ge bene i om he ex a ange LoRa p o ides on noisy en i onmen s. This kind o beha iou is accep able o he deploymen in open ield a eas whe e he e a e animals ee o mo e long dis ances. Fi mwa e needs o check whe he he payloads mee he egula ion du y cycle; in case we gene a e mo e da a han he numbe o by es we can send, he de ice will send da a un il i eaches he maximum ai ime allowed and hen i will disable he anscei e un il i could send da a again. In case ha con inuous animal moni o ing is needed, he as es bi a e is a ailable h ough FSK modula ion. The only downsides a e he sho ange and he compa ibili y Appl. Sci. 2021,11, 1240 8 o 18 wi h comme cial LoRaWAN ga eways. Addi ionally, no ne wo k laye is p o ided so ou ing is up o he so wa e implemen a ion. 2.3. In as uc u e and Backend Wea able de ices should wo k ega dless o he in as uc u e a ailable and use i s ene gy wisely in any condi ion. We conside wo di e en scena ios: a comple ely online in as uc u e when bo h beacon and s eaming messages can be ecei ed by s a iona y IP-connec ed ga eways and a po able o line in as uc u e when he e a e no s a iona y ga eways in ange and jus a shephe d wi h a ga eway ha can ecei e beacons. The sys em in as uc u e and communica ions low a e shown in Figu e 6. Appl. Sci. 2021, 11, x FOR PEER REVIEW 8 o 19 mee he egula ion du y cycle; in case we gene a e mo e da a han he numbe o by es we can send, he de ice will send da a un il i eaches he maximum ai ime allowed and hen i will disable he anscei e un il i could send da a again. In case ha con inuous animal moni o ing is needed, he as es bi a e is a ailable h ough FSK modula ion. The only downsides a e he sho ange and he compa ibili y wi h comme cial LoRaWAN ga eways. Addi ionally, no ne wo k laye is p o ided so ou ing is up o he so wa e implemen a ion. 2.3. In as uc u e and Backend Wea able de ices should wo k ega dless o he in as uc u e a ailable and use i s ene gy wisely in any condi ion. We conside wo di e en scena ios: a comple ely online in as uc u e when bo h beacon and s eaming messages can be ecei ed by s a iona y IP-connec ed ga eways and a po able o line in as uc u e when he e a e no s a iona y ga eways in ange and jus a shephe d wi h a ga eway ha can ecei e beacons. The sys- em in as uc u e and communica ions low a e shown in Figu e 6. Figu e 6. Sys em in as uc u e and communica ion low. In he i s case scena io, we use a comme cial ga eway ha connec s o a TSP and send messages o a LoRaWAN Se e . Then, we use MQTT in eg a ions o deli e he da a o a sel -hos ed da abase, and we p o ide a use dashboa d whe e da a can be isualized. I is he mos s aigh o wa d way o deploy a LoRaWAN ne wo k. We use an addi ional LoRa ga eway o ou e he FSK a ic ha canno be decoded by comme cial ga eways di ec ly o ou da abase. Fo he second scena io, we de eloped a ga eway using an embedded sys em (Rasp- be y Pi) wi h headless Linux, whe e we ins all a cus om LoRaWAN se e ha allows o o line logging and s o age o payloads. GNSS payloads a e decoded and sen o a web applica ion ha uns an o line maps iewe . I consis s o a Lea le map wi h OSM iles and Flask se e . Use s can check he las epo s ecei ed on in e ac i e b owse window in hei mobile o able de ices. A WiFi AP is deployed by he ga eway o p o ide TCP/IP connec ion wi h he use mobile de ice. Bo h ga eways use an 8-channel concen a o ca d ha demodula es all channels a he same ime and p o ides he enc yp ed payload o he hos sys em. LoRaWAN payload is end- o-end enc yp ed wi h AES c yp og aphic algo i hms [42], based on he AppKey Figu e 6. Sys em in as uc u e and communica ion low. In he i s case scena io, we use a comme cial ga eway ha connec s o a TSP and send messages o a LoRaWAN Se e . Then, we use MQTT in eg a ions o deli e he da a o a sel -hos ed da abase, and we p o ide a use dashboa d whe e da a can be isualized. I is he mos s aigh o wa d way o deploy a LoRaWAN ne wo k. We use an addi ional LoRa ga eway o ou e he FSK a ic ha canno be decoded by comme cial ga eways di ec ly o ou da abase. Fo he second scena io, we de eloped a ga eway using an embedded sys em (Rasp- be y Pi) wi h headless Linux, whe e we ins all a cus om LoRaWAN se e ha allows o o line logging and s o age o payloads. GNSS payloads a e decoded and sen o a web applica ion ha uns an o line maps iewe . I consis s o a Lea le map wi h OSM iles and Flask se e . Use s can check he las epo s ecei ed on in e ac i e b owse window in hei mobile o able de ices. A WiFi AP is deployed by he ga eway o p o ide TCP/IP connec ion wi h he use mobile de ice. Bo h ga eways use an 8-channel concen a o ca d ha demodula es all channels a he same ime and p o ides he enc yp ed payload o he hos sys em. LoRaWAN payload is end- o-end enc yp ed wi h AES c yp og aphic algo i hms [ 42 ], based on he AppKey pa ame e , known o node and se e . In he online scena io he payload is sen o he se e o decode i , bu in he o line scena io he hos sys em mus use he key o dec yp he payload on he ly. The ga eway also un a ull ea u ed LoRaWAN se e o managemen o de ices, da abase in eg a ion and inpu /ou pu pipelines. Appl. Sci. 2021,11, 1240 9 o 18 3. Resul s We chose o spli es in o wo di e en s ages. The i s one will include all es o cha ac e ize he elec onics, such as powe consump ion, RF pe o mance and digi al modula ion. A e e i ying ha all hypo heses a e co ec , we ake he senso s o he ields so we can pe o m eal wo ld measu emen s. 3.1. Labo a o y Tes s 3.1.1. Ene gy and Timing Analysis The idea behind he p og amming done in his de ice is o ine une pa ame e s o achie e he be e ange o each use case while main aining he lowes powe igu e. P e iously we s a ed ha we we e using bo h LoRa and FSK modula ions. Addi ionally, LoRa packe s a e coded in o LoRaWAN ames, ha allows an easy deploymen in mos cases. In hese es s we will e i y he comp omise be ween speed and powe consump ion in wo di e en da a a es (DR1 and DR5) and FSK modula ion. La e , we will check wha in luences does i ha e in ange i any. Labo a o y es se up comp ise se e al senso nodes all p og ammed o do he ollowing Sma IMU asks: - Fi s (Figu e 7), he senso samples da a om IMU a 12.5 Hz when an e en is de ec ed, un il desi ed payload leng h is achie ed. Da a a e sen h ough LoRaWAN a DR5, he as es a ailable a any channel. Then, GNSS ecei e is u ned on un il ix is achie ed, and i is sen h ough LoRaWAN a DR1, he second slowes . We chose DR1 o loca ion da a because i le us send da a each minu e i IMU s eaming is u ned o , o comply wi h he du y cycle shows he cu en while doing a basic Sma IMU ask. - Second (Figu e 8), we s eam accele ome e da a a he highes possible esolu ion wi h FSK modula ion. This mode is used o de ine ac i a ion h esholds and deciding on he il e cons an s o he embedded IMU uni egis e s. Appl. Sci. 2021, 11, x FOR PEER REVIEW 10 o 19 Figu e 7. Cu en consump ion while sampling IMU a 12.5 Hz and ac i a ing GNSS ecei e . Figu e 8. Cu en consump ion while s eaming IMU da a sampled a 100 Hz. The de ice mus be usable on animals ha usually spend a whole yea g azing on he ield. Acco ding o he p e ious powe consump ion equi emen s we calcula ed he ba e y li e o each mode as shown in Table 2. Table 1. Communica ion modes pe o mance analysis. Modula ion Payload Time Th oughpu Ene gy pe By e Link Budge LoRa (DR1) 8 by es 495 ms 129.3 b/s 712 uJ 148 dB LoRa (DR5) 228 by es 358 ms 5.1 kb/s 18.1 uJ 138 dB FSK 61 by es 11 ms 44.4 kb/s 2.1 uJ 123 dB Table 2. De ice mode ene ge ical analysis. Mode TGNSS- SAMPLE FSENSOR- SAMPLE TREPORT Ene gy pe Day 1 Li-SOCl2 Li e ime 2 Alkaline Li e ime 3 Li-Ion Li e ime 4 Beacon 1 h - 1 h 31.2 J 9.0 yea s 2.8 yea s 4.2 yea s Beacon 1 h - 10′ 37.9 J 7.4 yea s 2.3 yea s 3.4 yea s Figu e 7. Cu en consump ion while sampling IMU a 12.5 Hz and ac i a ing GNSS ecei e . Ou communica ion model consis o he modes p e iously explained and which pe o mance is summa ized in Table 1. We can obse e ha each mode is ma ched o he bes modula ion o he use case. The bes link budge is allowed in GNSS epo s, which equi es he bes sensi i i y. In mo ion epo s we alue speed o e ecep ion, so we chose a as e da a a e. Fo moni o ing pu poses when we a e close o he animal, we can use FSK o he as es a ailable da a a e. Appl. Sci. 2021,11, 1240 16 o 18 5. Conclusions In o ma ion and Communica ion Technologies (ICT-based) solu ions a e imp o ing he p oduc i i y o all indus ial sec o s, e en ag icul u e and li es ock. They should also imp o e he well-li ing o he people ac oss he Wo ld. Ne e heless, in many cases, he success o one o hese solu ions is linked o he a ailabili y o communica ion in as uc- u es ha could p o ide suppo o he “C” o ICT. In de eloped coun ies, whe e we could assume ha communica ion in as uc u es should ha e a high deg ee o deploymen , ag i- cul u e and li es ock a e usually ela ed wi h u al a eas whe e hese in as uc u es begin o loss hei capaci ies. Un o una ely, in he es o he Wo ld, his lack o communica ions capaci ies begins e en in he big ci ies. I we pu he ocus on ex ensi e li es ock, we can ind ha , in Eu ope, i is usually es ic ed o emo e a eas, whe e he communica ions in as uc u es a e no capable o p o ide any kind o se ices o deploying ICT solu ions. In o he coun ies, we can ind ex ensi e li es ock also nea he big ci ies, bu his does no gua an ee he a ailabili y o communica ions se ices (o hey a e bu a high p ice). This pape has p esen ed a solu ion o dealing wi h his p oblem o lack o s ablished communica ion in as uc u es o con olling he mo emen s o animals om ex ensi e li es ock a ms. I is based on he de elopmen o an animal wea able ha is equipped wi h ine ial senso s, GNSS and wi eless communica ions. I also includes enough compu a ion capaci y o making adap i e analysis and VQ comp ession ha p o ides high comp ession a ios wi h good peak signal- o-noise a ios, and allows o eal- ime moni o ing and logging o ca le’s posi ion and ac i i ies. This wea able has been e alua ed in eal scena ios, such as moun ains and o es , in he con ex o a p ojec ha aims o moni o animals in ex ensi e a ming in he no h pa o Spain (in he Py enees a ea, nea he bo de wi h F ance). This p ojec includes he necessi y o ack he mo emen o he animals o e alua e hei impac on he plan biodi e si y, as well as he in en ion o i ually educe he dis ance be ween he animals and he shephe ds. Au ho Con ibu ions: Concep ualiza ion, R.C. and F.J.Z.-S.; me hodology, F.J.Z.-S. and T.B.; ha d- wa e and so wa e, R.C., A.H. and Á.M.; design, T.B.; alida ion, R.C., A.H. and T.B.; w i ing, e iew and edi ing, All; p ojec adminis a ion and unding acquisi ion, R.C. and F.J.Z.-S. All au ho s ha e ead and ag eed o he published e sion o he manusc ip . Funding: This wo k has been pa ially suppo ed by he A agon egional Go e nmen hough he p og am o R&D g oups (T27_20R and T59_20R) and by he Ru al De elopmen P og am in conjunc ion wi h FEADER om EC (GCP2019004100). The wo k o Al a o Ma co and Te esa Blanco has been pa ially suppo ed by he Spanish Go e nmen , p og am To es Que edo (PTQ2017-09481 and PTQ2018-010045, espec i ely). Ins i u ional Re iew Boa d S a emen : E hical e iew and app o al we e wai ed o his s udy, due o i was accomplished wi hin he con ex o he wo ks ela ed o he Ru al De elopmen P og am, whe e shephe ds ag eed wi h hese ials. Only a limi ed numbe o animals we e in ol ed in he s udy, and hey we e no exposed o ha m ul le els o adia ion, as discussed in he ex . In o med Consen S a emen : No applicable. Con lic s o In e es : The au ho s decla e no con lic o in e es . Re e ences 1. No i, M.; Swi ze , J.; C aw o d, A. He ding on he B ink: Towa ds a Global Su ey o Pas o al Communi ies and Con lic ; In e na ional Insi u e o Sus ainable De elopmen : Winnipeg, MB, Canada, 2005. 2. Some, J. Can Pas o alism Su i e in he 21s Cen u y? A ailable online: h ps://www. henewhumani a ian.o g/news/2007/07/ 13/can-pas o alism-su i e-21s -cen u y (accessed on 14 Decembe 2020). 3. Issels ein, J.; Jeang os, B.; Pa lu, V. Ag onomic Aspec s o Biodi e si y Ta ge ed Managemen o Tempe a e G asslands in Eu ope—A Re iew. Ag on. Res. 2005,3, 139–151. 4. Finck, P.; Riecken, U.; Sch öde , E. Pas u e Landscapes and Na u e Conse a ion—New s a egies o he p ese a ion o open landscapes in Eu ope. In Pas u e Landscapes and Na u e Conse a ion; Sp inge : Be lin, Ge many, 2002; pp. 1–13. Appl. Sci. 2021,11, 1240 17 o 18 5. Temple, D.; Man eca, X. Animal Wel a e in Ex ensi e P oduc ion Sys ems Is S ill an A ea o Conce n. F on . Sus ain. Food Sys . 2020,4, 154. [C ossRe ] 6. Lomillos Pé ez, J.M.; de la Va ga, M.E.A.; Ga cía, J.J.; Gaudioso Lacasa, V.R. Moni o ing Lidia Ca le wi h GPS-GPRS Technology— A S udy on G azing Beha iou and Spa ial Dis ibu ion. Ve . México 2018,4, 1–17. [C ossRe ] 7. Lla ia, A.; Te asson, G.; A egui, H.; Hacala, A. Geoloca ion and Moni o ing Pla o m o Ex ensi e Fa ming in Moun ain Pas u es. In P oceedings o he IEEE In e na ional Con e ence on Indus ial Technology, Se ille, Spain, 17–19 Ma ch 2015; pp. 2420–2425. 8. Hui cán, J.I.; Muñoz, C.; Young, H.; Von Dossow, L.; Bus os, J.; Vi allo, G.; Tonea i, M. ZigBee-Based Wi eless Senso Ne wo k Localiza ion o Ca le Moni o ing in G azing Fields. Compu . Elec on. Ag ic. 2010,74, 258–264. [C ossRe ] 9. K ishnan, A.J.; Binu, G. Ene gy E icien T ee Cons uc ion o ZigBee Rou e Ne wo k. In P oceedings o he 2017 2nd In e na ional Con e ence on Communica ion and Elec onics Sys ems (ICCES), Coimba o e, India, 19–20 Oc obe 2017; pp. 746–751. 10. Nkwa i, P.K.M.; Rime , S.; Paul, B.; Fe ei a, H. He e ogeneous Wi eless Ne wo k Based on Wi-Fi and ZigBee o Ca le Moni o ing. In P oceedings o he 2015 IST-A ica Con e ence, Lilongwe, Malawi, 6–8 May 2015; pp. 1–9. 11. Sampa h, A.; Meena, P. A No el App oach o Ca le Heal h Moni o ing o Maximizing Dai y Ou pu Using LoRa IoT Technology. IJARCCE 2019,8, 94–102. [C ossRe ] 12. Zinas, N.; Kon ogiannis, S.; Kokkonis, G.; Valsamidis, S.; Kazanidis, I. P oposed Open Sou ce A chi ec u e o Long Range Moni o ing. The Case S udy o Ca le T acking a Pogoniani. In P oceedings o he Panhellenic Con e ence on In o ma ics, La issa, G eece, 28–30 Sep embe 2017; pp. 1–6. 13. Codeluppi, G.; Cil one, A.; Da oli, L.; Fe a i, G. LoRaFa M: A LoRaWAN-Based Sma Fa ming Modula IoT A chi ec u e. Senso s 2020,20, 2028. [C ossRe ] [PubMed] 14. Zhang, C.; Ko acs, J.M. The Applica ion o Small Unmanned Ae ial Sys ems o P ecision Ag icul u e: A Re iew. P ecis. Ag ic. 2012,13, 693–712. [C ossRe ] 15. Dapon e, P.; De Vi o, L.; Glielmo, L.; Iannelli, L.; Liuzza, D.; Pica iello, F.; Silano, G. A Re iew on he Use o D ones o P ecision Ag icul u e. In IOP Con e ence Se ies: Ea h and En i onmen al Science; IOP Publishing: B is ol, UK, 2019; p. 12022. 16. Mogili, U.R.; Deepak, B. Re iew on Applica ion o D one Sys ems in P ecision Ag icul u e. P ocedia Compu . Sci. 2018 ,133, 502–509. [C ossRe ] 17. Ba bedo, J.G.A.; Koenigkan, L.V.; San os, P.M.; Ribei o, A.R.B. Coun ing Ca le in UAV Images—Dealing wi h Clus e ed Animals and Animal/Backg ound Con as Changes. Senso s 2020,20, 2126. [C ossRe ] 18. Abu asain, R.; Edi isinghe, E.; Alba ay, A. D one-Based Ca le De ec ion Using Deep Neu al Ne wo ks; Sp inge : Be lin, Ge many, 2020; pp. 598–611. 19. Shao, W.; Kawakami, R.; Yoshihashi, R.; You, S.; Kawase, H.; Naemu a, T. Ca le De ec ion and Coun ing in UAV Images Based on Con olu ional Neu al Ne wo ks. In . J. Remo e Sens. 2020,41, 31–52. [C ossRe ] 20. Ri as, A.; Chamoso, P.; González-B iones, A.; Co chado, J.M. De ec ion o Ca le Using D ones and Con olu ional Neu al Ne wo ks. Senso s 2018,18, 2048. [C ossRe ] 21. Li, X.; Xing, L. Reac i e Deploymen o Au onomous D ones o Li es ock Moni o ing Based on Densi y-Based Clus e ing. In P oceedings o he In e na ional Con e ence on Robo ics and Biomime ics (ROBIO), Dali, China, 6–8 Decembe 2019; pp. 2421–2426. 22. Ba bedo, J.G.A.; Koenigkan, L.V. Pe spec i es on he Use o Unmanned Ae ial Sys ems o Moni o Ca le. Ou look Ag ic. 2018 ,47, 214–222. [C ossRe ] 23. Chapa, J.M.; Mascha , K.; Iwe sen, M.; Baumga ne , J.; D illich, M. Accele ome e Sys ems as Tools o Heal h and Wel a e Assessmen in Ca le and Pigs—A Re iew. Beha . P ocess. 2020,181. [C ossRe ] [PubMed] 24. Williams, L.; Fox, D.; Bishop-Hu ley, G.J.; Swain, D. Use o Radio F equency Iden i ica ion (RFID) Technology o Reco d G azing Bee Ca le Wa e Poin Use. Compu . Elec on. Ag ic. 2019,156, 193–202. [C ossRe ] 25. Gody´n, D.; He bu , P.; Ang ecka, S. Measu emen s o Pe iphe al and Deep Body Tempe a u e in Ca le—A Re iew. J. The m. Biol. 2018,79. [C ossRe ] 26. P iya, M.; Jaya am, B. WSN-Based Elec onic Li es ock o Dai y Ca le and Physical Pa ame e s Moni o ing. In Eme ging Resea ch in Elec onics, Compu e Science and Technology; Sp inge : Singapo e, 2019; pp. 37–45. ISBN 978-3-662-53832-6. 27. Wang, J.; He, Z.; Ji, J.; Zhao, K.; Zhang, H. IoT-Based Measu emen Sys em o Classi ying Cow Beha io om T i-Axial Accele ome e . Ciênc. Ru al 2019,49. [C ossRe ] 28. Tamu a, T.; Okubo, Y.; Deguchi, Y.; Koshikawa, S.; Takahashi, M.; Chida, Y.; Okada, K. Dai y Ca le Beha io Classi ica ions Based on Decision T ee Lea ning Using 3-Axis Neck-Moun ed Accele ome e s. Anim. Sci. J. 2019,90. [C ossRe ] 29. Riabo , L.; Aubin, S.; Bédè e, N.; Cou eu , S.; Madouasse, A.; Goumand, E.; Chau in, A.; Plan ie , G. E alua ion o P e- P ocessing Me hods o he P edic ion o Ca le Beha iou om Accele ome e Da a. Compu . Elec on. Ag ic. 2019. [C ossRe ] 30. Jalal, A.; Quaid, M.A.K.; Kim, K. A S udy o Accele ome e and Gy oscope Measu emen s in Physical Li e-Log Ac i i ies De ec ion Sys ems. Senso s 2020,20, 6670. 31. Ma qués, G.; Bas e e xea, K. E icien Algo i hms o Accele ome e -Based Wea able Hand Ges u e Recogni ion Sys ems. In P oceedings o he 13 h In e na ional Con e ence on Embedded and Ubiqui ous Compu ing, Po o, Po ugal, 21–23 Oc obe 2015; pp. 132–139. Appl. Sci. 2021,11, 1240 18 o 18 32. Suko , A.A.; Zaka ia, A.; Rahim, N.A. Ac i i y Recogni ion Using Accele ome e Senso and Machine Lea ning Classi ie s. In P oceedings o he 14 h In e na ional Colloquium on Signal P ocessing & I s Applica ions (CSPA), Penang, Malaysia, 9–10 Ma ch 2018; pp. 233–238. 33. Jahanjoo, A.; Tahan, M.N.; Rash i, M.J. Accu a e Fall De ec ion Using 3-Axis Accele ome e Senso and MLF Algo i hm. In P oceedings o he 3 d In e na ional Con e ence on Pa e n Recogni ion and Image Analysis (IPRIA), Shah eko d, I an, 19–20 Ap il 2017; pp. 90–95. 34. Sanche i, P.; Shedge, R.; Pulgam, N. Wo d-IPCA: An Imp o emen in Dimension Reduc ion Techniques. In P oceedings o he In e na ional Con e ence on Con ol, Powe , Communica ion and Compu ing Technologies, Kannu , India, 23–24 Ma ch 2018; pp. 575–578. 35. Padmaja, D.L.; Vishnu a dhan, B. Compa a i e S udy o Fea u e Subse Selec ion Me hods o Dimensionali y Reduc ion on Scien i ic Da a. In P oceedings o he 6 h In e na ional Con e ence on Ad anced Compu ing (IACC), Bhima a am, India, 27–28 Feb ua y 2016; pp. 31–34. 36. Choudhu y, S.; Bandyopadhyay, S.; Mukhopadhyay, S.; Mukhe jee, S. Vec o Quan iza ion and Mul i Class Suppo Vec o Ma- chines Based Finge p in Classi ica ion. In P oceedings o he In e na ional Con e ence on In en i e Compu a ion Technologies, Coimba o e, India, 26–27 Augus 2016; pp. 1–4. 37. Bane jee, A.; Ghosh, J. F equency-Sensi i e Compe i i e Lea ning o Scalable Balanced Clus e ing on High-Dimensional Hype sphe es. IEEE T ans. Neu al Ne w. 2004,15, 702–719. [C ossRe ] 38. Pelayo, E.; Buldain, D.; O i e, C. Magni ude Sensi i e Compe i i e Lea ning. Neu ocompu ing 2013,112, 4–18. [C ossRe ] 39. Vizá aga, J.; Casas, R.; Ma co, Á.; Buldain, J. Dimensionali y Reduc ion o Sma IoT Senso s. Elec onics 2020 ,9, 2035. [C ossRe ] 40. Blanco, T.; Casas, R.; Manchado-Pé ez, E.; Asensio, Á.; López-Pé ez, J.M. F om he Islands o Knowledge o a Sha ed Unde s and- ing: In e disciplina i y and Technology Li e acy o Inno a ion in Sma Elec onic P oduc Design. In . J. Technol. Des. Educ. 2017,27, 329–362. [C ossRe ] 41. LoRa Alliance. LoRaWAN Regional Pa ame e s. A ailable online: h ps://lo a-alliance.o g/si es/de aul / iles/2020-06/ p_2-1. 0.1.pd (accessed on 14 Decembe 2020). 42. LoRaWAN Secu i y Whi epape . A ailable online: h ps://lo a-alliance.o g/si es/de aul / iles/2019-05/lo awan_secu i y_ whi epape .pd (accessed on 14 Decembe 2020). 43. S isen, A.; Blunck, H.; Bha acha ya, S.; P en ow, T.S.; Kjæ gaa d, M.B.; Dey, A.; Sonne, T.; Jensen, M.M. Sma De ices A e Di e en : Assessing and Mi iga ingmobile Sensing He e ogenei ies o Ac i i y Recogni ion. In P oceedings o he 13 h ACM Con e ence on Embedded Ne wo ked Senso Sys ems, Seoul, Ko ea, 1–4 No embe 2015; pp. 127–140. 44. Weise, F.J.; Haup meie , H.; S a o d, K.J.; Haywa d, M.W.; Aal, K.; Heue , M.; Tomele so, M.; Wul , V.; Some s, M.J.; S ein, A.B. Lions a he Ga es: T ans-Disciplina y Design o an Ea ly Wa ning Sys em o Imp o e Human-Lion Coexis ence. F on . Ecol. E ol. 2019,6, 242. [C ossRe ] 45. Tang, Y.; Dananjayan, S.; Hou, C.; Guo, Q.; Luo, S.; He, Y. A Su ey on he 5G Ne wo k and I s Impac on Ag icul u e: Challenges and Oppo uni ies. Compu . Elec on. Ag ic. 2021,180, 105895. [C ossRe ] 46. ETSI UMTS Te es ial Radio Access Ne wo k (UTRAN); UTRA FDD; Radio T ansmission and Recep ion (UMTS XX.06 Ve sion 1.0.0). A ailable online: h ps://www.3gpp.o g/ p/ sg_ an/TSG_RAN/TSGR_02/Docs/pd s/RP-99028.pd (accessed on 22 Janua y 2021). 47. Yamaguchi, H.; Tsu i a, G.; Ueno, S.; Wa anabe, S.; Wake, K.; Taki, M.; Nagawa, H. 1439 MHz Pulsed TDMA Fields A ec Pe o mance o Ra s in a T-maze Task Only When Body Tempe a u e Is Ele a ed. Bioelec omagne ics 2003,24, 223–230. 48. Zook, B.C.; Simmens, S.J. The E ec s o 860 MHz Radio equency Radia ion on he Induc ion o P omo ion o B ain Tumo s and O he Neoplasms in Ra s. Radia . Res. 2001,155, 572–583. [C ossRe ] 49. La Regina, M.; Mo os, E.G.; Picka d, W.F.; S aube, W.L.; Ba y, J.; Ro i Ro i, J.L. The E ec o Ch onic Exposu e o 835.62 MHz FDMA o 847.74 MHz CDMA Radio equency Radia ion on he Incidence o Spon aneous Tumo s in Ra s. Radia . Res. 2003 ,160, 143–151. [C ossRe ] 50. Ziemann, C.; B ockmeye , H.; Reddy, S.B.; Vijayalaxmi, M.; P ihoda, T.J.; Kus e , N.; Tillmann, T.; Dasenb ock, C. Absence o Geno oxic Po en ial o 902 MHz (GSM) and 1747 MHz (DCS) Wi eless Communica ion Signals: In Vi o Two-Yea Bioassay in B6C3F1 Mice. In . J. Radia . Biol. 2009,85, 454–464. 51. U e idge, T.D.; Gebski, V.; Finnie, J.W.; Ve non-Robe s, B.; Kuchel, T.R. Long-Te m Exposu e o E µ -Pim1 T ansgenic Mice o 898.4 MHz Mic owa es Does No Inc ease Lymphoma Incidence. Radia . Res. 2002,158, 357–364. [C ossRe ] 52. Oli ei a, L.; Rod igues, J.J.; Kozlo , S.A.; Rabêlo, R.A.; Fu ado, V. Pe o mance Assessmen o Long- ange and Sig ox P o ocols wi h Mobili y Suppo . In . J. Commun. Sys . 2019,32, e3956. [C ossRe ] 53. E e yne . Co e age. A ailable online: h ps://www.e e yne .com/co e age (accessed on 22 Janua y 2021). 54. T as iña-Mo eno, C.A.; Blasco, R.; Ma co, Á.; Casas, R.; T as iña-Cas o, A. Unmanned Ae ial Vehicle Based Wi eless Senso Ne wo k o Ma ine-Coas al En i onmen Moni o ing. Senso s 2017,17, 460. 55. Sali, S.; Pa a hi, C. Heal h Moni o ing Sys em Using Wi eless Senso Ne wo k. In . J. Eng. Res. Appl. 2018,8, 4–12.