1
Eas e n Michigan Uni e si y
GameAbo e College o Enginee ing &
Technology
Ti le: Mecha onic Componen s and Sys em
In eg a ion in a Humanoid Robo (Tesla
Op imus)
Au ho : Unais Ali
Co-Au ho : Syeda Kasha Kulsoom
2
Con en s
Execu i e Summa y ...................................................................................................... 3
1. In oduc ion .......................................................................................................... 4
2. Componen s o a Mecha onic Sys em ..................................................................... 4
2.1. Ac ua o s ............................................................................................................... 4
2.2. Senso s ............................................................................................................... 5
2.3. Digi al Con ol De ices ........................................................................................ 5
3. A humanoid obo is a mecha onic sys em .............................................................. 7
3.1. Mechanical powe ain and ac ua ion ................................................................... 8
3.2. Sensing and pe cep ion ........................................................................................ 9
3.3. Digi al con ol de ices and communica ion ............................................................ 9
3.4. Tasks and con ol modes as e idence o mecha onic in eg a ion ........................... 11
4. Tesla Bo (Op imus) as a Mecha onic Sys em ........................................................ 13
4.1. Mechanical powe ain and ac ua ion ................................................................. 13
4.2. Sensing and pe cep ion ...................................................................................... 14
4.3. Digi al con ol de ices and communica ion .......................................................... 16
4.4. Tasks and con ol modes as e idence o mecha onic in eg a ion ........................... 17
5. Discussion ........................................................................................................... 18
6. Conclusion .......................................................................................................... 19
Re e ences .................................................................................................................. 20
3
Execu i e Summa y
Mecha onic sys ems in eg a e mechanical s uc u es, ac ua o s, senso s, and digi al
con ol o achie e in elligen , p ecise, and sa e ope a ion [1,2,3]. This epo iden i ies and
desc ibes he h ee co e componen s o a ypical mecha onic sys em, namely ac ua o s, senso s,
and digi al con ol de ices, and hen uses Tesla’s Op imus humanoid obo as he example
machine o explain why i quali ies as a mecha onic sys em [2,4,5,6]. D awing on indus ial
obo e e ence a chi ec u es and humanoid con ol pipelines, we show how compac elec ic
join modules, mul imodal sensing (encode s, IMUs, o ce and p essu e sensing), and laye ed
con ol (se o d i es wi h ield-o ien ed con ol and whole-body impedance o op imiza ion
laye s) wo k oge he [1,4,7,8,9,10]. The igu es suppo and illus a e hese poin s, wi h explici
placemen indica ed in he ex .
4
1. In oduc ion
Mecha onics is he syne gis ic in eg a ion o mechanical design, elec onics, sensing, and
compu e con ol o c ea e unc ional and in elligen sys ems. Indus ial manipula o s and
humanoid obo s a e a che ypal examples. They combine elec ic mo o s and ansmissions,
p ecision eedback senso s, and eal ime con ol algo i hms o execu e complex asks such as
assembly, o ce egula ed con ac , and locomo ion wi h eliabili y and sa e y. In mode n
p oduc ion en i onmen s, con inuous mo ion con ol wo ks alongside disc e e sequencing and
sa e y in e locks, while s anda ds and in e ope able ne wo ks enable in eg a ion in o sma
manu ac u ing.
This epo explici ly iden i ies and desc ibes ac ua o s, senso s, and digi al con ol de ices
(3.1–3.3), hen uses Tesla’s Op imus o demons a e, wi h an accompanying schema ic and igu e
e e ences, why i is a mecha onic sys em [1–4,7–10,15].
2. Componen s o a Mecha onic Sys em
2.1. Ac ua o s
• Elec ic o a y ac ua o s a e usually pe manen magne synch onous mo o s o b ushless
DC mo o s d i en by se o in e e s implemen ing ield-o ien ed con ol (FOC) o high-
bandwid h o que egula ion [4,7,8]. High educ ion ansmissions (ha monic/s ain wa e,
cycloidal, plane a y) con e mo o speed o join o que wi h low backlash o p ecise
posi ioning; b akes on selec ed axes allow sa e holding unde powe loss [2,4,6].
5
• Linea and specialized ac ua o s such as ball sc ews, oice coils, and solenoids appea in
g ippe s, end e ec o s, and b akes; se ies elas ic o in eg a ed o que-sensing ac ua o s
add compliance and imp o e sa e y du ing in e ac ion [2,5,6,9].
• The b akes and clu ches p o ide elec omechanical holding o s a ic sa e y and
eme gency s ops; ic ion and compliance a e modeled o achie e accu a e low-speed
con ol [8,10].
2.2. Senso s
• Join sensing uses absolu e/inc emen al encode s o esol e s o posi ion and speed
eedback [4,7]. Fo ce and in e ac ion sensing includes six-axis o ce/ o que senso s,
oo p essu e a ays, and join o que es ima ion om mo o cu en s o enable
impedance/admi ance con ol and sa e con ac [2,7,9,10]. Ine ial and pose sensing
wi h IMUs p o ides angula a es and linea accele a ions o a i ude es ima ion,
balance, and dis u bance ejec ion [2,7].
• En i onmen al sensing includes ision (RGB/dep h), p oximi y, and ac ile sensing o
suppo manipula ion and na iga ion; in indus ial con ex s, AIDC ia ba codes,
RFID, and machine ision suppo s iden i ica ion and acking [11,12].
• Heal h and sa e y sensing includes empe a u e, cu en , ol age, and ba e y sensing
o p o ec d i es and powe sys ems, as well as limi and home swi ches o en o ce
mo ion bounds [4,8,12].
2.3. Digi al Con ol De ices
• Se o d i es on each join un FOC, ead encode s and cu en s, egula e
o que/speed/posi ion, and implemen sa e y unc ions such as Sa e To que O (STO)
6
[4,7,8]. A cen al eal- ime con olle execu es kinema ics/dynamics, whole-body con ol
(impedance and op imiza ion-based QP), ajec o y gene a ion, s a e es ima ion, and
supe iso y sa e y logic [2,7,9,10].
• Communica ion ne wo ks use de e minis ic ieldbuses such as E he CAT o
synch onized low-la ency exchange among senso s, d i es, and con olle s; highe -le el
messaging suppo s diagnos ics and en e p ise in eg a ion [4,11,12].
• The so wa e s ack includes es ima ion (Kalman/complemen a y il e s), compensa ion
o ic ion/g a i y/compliance, aul de ec ion and de a ing, ask planne s, and sa e y
moni o s [7–10,12–14].
See Figu e 1 o he o e all mecha onic a chi ec u e and componen loca ions.
7
Figu e 1. Canonical indus ial obo mecha onic a chi ec u e wi h con olle cabine (se o d i es, PLC/IPC),
powe and ieldbus, join ac ua o s and encode s, w is o ce/ o que senso , end e ec o , and sa e y sys ems
(E‑s op, STO, limi s, in e locks, he mal). Illus a es ypical componen s and signal lows [1,4].
3. A humanoid obo is a mecha onic sys em
Humanoid obo s in eg a e he abo e componen s igh ly: compac join modules wi h
mo o s, high- a io ansmissions, b akes, and encode s; sensing spanning join encode s, IMUs,
o ce/p essu e senso s, and ision; and con ol ha combines low-le el se o egula ion wi h
high-le el whole-body coo dina ion unde s ic sa e y [2,5–7,9,10]. De e minis ic ne wo king
synch onizes all elemen s in closed loop [4,11,12].
8
3.1. Mechanical powe ain and ac ua ion
A ep esen a i e humanoid such as Op imus employs pe manen magne o b ushless mo o s
a hips, knees, ankles, shoulde s, and elbows, wi h high educ ion ansmissions
(ha monic/cycloidal) o achie e high join o que a low speed [2,5,6].
Many join s in eg a e he mo o , gea ain, mo o -side and join -side encode s, an
elec omechanical b ake, and a d i e/con olle PCB o compac ness and obus ness, enabling
p ecise posi ioning, smoo h low-speed o que, and sa e holding [2,6,8]. See Figu e 2 o an
in eg a ed humanoid join module example illus a ing hese componen s [2,5,6,15]
Figu e 2. In eg a ed humanoid join module: mechanical s ack (mo o , ansmission, encode s, b ake)
linked o in e e /se o d i e elec onics and con ol/I/O panel, showing powe , signals, sa e y, he mal,
and eedback connec ions.
9
3.2. Sensing and pe cep ion
• Join encode s p o ide pos u e and eloci y eedback a each deg ee o eedom [4,7].
• IMUs es ima e base o ien a ion and linea /angula a es, suppo ing balance and
locomo ion [2,7].
• Foo p essu e a ays and op ional w is F/T senso s quan i y con ac o ces; mo o -
cu en -based o que es ima ion complemen s ex e nal o ce sensing [2,9,10].
• Mo o cu en , ol age, and empe a u e senso s suppo o que es ima ion and
he mal p o ec ion [4,8].
• Vision and ac ile sensing enable objec de ec ion, g asping, and complian
manipula ion; indus ial AIDC complemen s pe cep ion o wo k acking [11,12].
3.3. Digi al con ol de ices and communica ion
Join se o d i es close cu en and eloci y loops wi h FOC and en o ce sa e y limi s
and STO. A cen al eal- ime con olle execu es whole-body con ol (QP wi h cons ain s),
impedance/admi ance con ol o in e ac ion, ajec o y gene a ion, s a e es ima ion ia
senso usion, and supe iso y sa e y (limi s, s ops, he mal de a ing, aul moni o s) [7–10].
16
4.3. Digi al con ol de ices and communica ion
Join se o d i es close cu en and eloci y loops using FOC and en o ce sa e y limi s
and STO. A cen al eal ime con olle execu es whole body con ol QP wi h cons ain s,
impedance o admi ance o in e ac ion, ajec o y gene a ion, and s a e es ima ion wi h senso
usion. Supe iso y sa e y manages limi s, s ops, he mal de a ing, and aul handling. A
de e minis ic ieldbus such as E he CAT synch onizes se poin s and eedback a abou 1 o 2
kHz. Highe le el channels ca y diagnos ics and planning da a [4,7–10,11,12]. The a chi ec u e
is illus a ed in Figu e 3, which shows he synch onized con olle o d i e loop and sa e y
laye s, and aligns wi h he canonical indus ial mecha onic layou in Figu e 1.
Figu e 8. Senso ield o iew layou wi h pano amic 180 deg ees and op ional 83 deg ees
came as, illus a ing mul imodal pe cep ion co e age a ound he head o o so. Sou ce: adap ed om
G eyB Insigh s Tesla Bo pa en s page [15]. Image URL: [h ps://insigh s.g eyb.com/wp-
con en /uploads/2024/08/1-13.png].
17
This laye ed app oach allows he obo o egula e join o ques accu a ely, coo dina e
many deg ees o eedom o balance and manipula ion, and ansi ion sa ely be ween asks and
con ac condi ions, all while hono ing iming gua an ees equi ed o s abili y [7–10]. These eal
ime pipelines a e depic ed in Figu e 4, linking pe cep ion and es ima ion o op imiza ion o
impedance con ol and hen o ieldbus ac ua ion, u he e idencing ull mecha onic in eg a ion
in Op imus. Senso co e age and ield-o - iew layou a e summa ized in Figu e 8.
4.4. Tasks and con ol modes as e idence o mecha onic in eg a ion
• Locomo ion including walking, s ai s, slopes, and u ning depends on IMU, oo senso s,
and encode s o s a e es ima ion. Se o d i es egula e o ques while whole body
con olle s manage s abili y and con ac s [2,7,9]. The closed loop coo dina ion be ween
con olle s, d i e s, and senso s ollows he con ol a chi ec u e in Figu e 3 and he
pipeline in Figu e 4, con i ming end o end mecha onic ope a ion.
• Squa ing, li ing, pushing, and pulling equi e high peak join o que in he lowe body.
F ic ion compensa ion and impedance con ol egula e in e ac ion o ces sa ely [8–10].
The join module ha dwa e ha enables his beha io is ep esen ed by Figu e 2, which
shows mo o s, high a io ansmissions, encode s, b akes, and d i e elec onics in eg a ed
in o a se iceable uni .
• Tool use and manipula ion ely on ision plus w is o hand sensing o g asping, o ce
limi ed inse ion, and con olled su ace in e ac ions. Senso less o que and s i ness
es ima ion om d i e signals can une impedance online [9,10,14]. The pe cep ion o
con ol o ac ua ion low is consis en wi h Figu e 4, while iming and sa e y en o cemen
use he Figu e 1 ieldbus and STO laye s.
18
• S a ic holding and sa e y use elec omechanical b akes o secu e pos u e on powe loss.
Limi swi ches and STO en o ce sa e bounds and eme gency esponse [4,8]. These
elemen s a e included in he canonical indus ial mecha onic e e ence in Figu e 1 and in
he Op imus join module concep o Figu e 2.
• Compa ed o au omo i e ac ion sys ems ha emphasize powe a high speed, humanoid
join s a e op imized o o que densi y and p ecision a low speed h ough gea ing and
high bandwid h con ol, an essen ial ounda ion o eliable con ac ich beha io in
e e yday asks [2,4,6]. Figu e 5 con as s EV ac ion mo o s wi h humanoid join
ac ua o s o highligh why Op imus adop s gea ed, o que dense, p ecisely con olled
join modules, a de ining cha ac e is ic o mecha onic design.
5. Discussion
The humanoid obo clea ly quali ies as a mecha onic sys em. Ac ua o s a e elec ic join
modules wi h high- educ ion ansmissions, in eg a ed b akes, and compac d i e elec onics ha
deli e high o que densi y and p ecise con ol a low speed [2,4,6,8]. Senso s include join
encode s, IMUs, con ac / o ce senso s, and heal h senso s ha p o ide comp ehensi e eedback
o s abili y, p ecision, and sa e y [4,7,9,12]. Digi al con ol de ices include se o d i es and a
cen al eal- ime con olle ha implemen laye ed con ol om cu en , eloci y, and posi ion
h ough impedance and whole-body op imiza ion; hey a e synch onized by a de e minis ic
ieldbus and o e seen by supe iso y sa e y and aul handling [4,7–10,11,12].
The indus ial obo e e ence in Figu e 1 demons a es he canonical a angemen o hese
elemen s [1,4]. The humanoid join module in Figu e 2, he mecha onic con ol a chi ec u e in
19
Figu e 3, and he whole-body pipeline in Figu e 3 show conc e e ealiza ions [2,4,7–10]. The EV
e sus humanoid compa ison in Figu e 5 cla i ies di e ing op imiza ion egimes ha in o m
ac ua o and ansmission choices o humanoids engaged in con ac - ich asks [2,4,6].
6. Conclusion
A ypical mecha onic sys em is de ined by igh in eg a ion o ac ua o s, senso s, and digi al
con ol de ices connec ed o e synch onized communica ion ne wo ks and go e ned by sa e y
[1–4,7–12]. A ep esen a i e humanoid obo embodies his in eg a ion h ough compac elec ic
join ac ua o s wi h high- educ ion gea ing and b akes, mul imodal sensing o s a e and
en i onmen , and a laye ed con ol s ack om se o d i es o whole-body con olle s [2,5–10].
The igu es p o ided illus a e each laye and hei in e connec ions, sa is ying he assignmen
equi emen s o iden i y componen s and o explain h ough an example machine and i s
schema ics why i is conside ed a mecha onic sys em [1–4,7–12,15].
20
Re e ences
[1] C aig, J. J. In oduc ion o Robo ics: Mechanics and Con ol. Pea son.
[2] Siciliano, B., Scia icco, L., Villani, L., O iolo, G. Robo ics: Modelling, Planning and
Con ol. Sp inge .
[3] Bol on, W. Mecha onics: Elec onic Con ol Sys ems in Mechanical and Elec ical
Enginee ing. Pea son.
[4] Beckho Au oma ion. E he CAT and PC-based Con ol o Robo ics (applica ion no es and
manuals).
[5] P a , G., e al. Se ies Elas ic Ac ua o s o Legged Robo s (IEEE/ICRA/ICHR pape s).
[6] Tesla Op imus/AI Day ma e ials and echnical o e iews (company p esen a ions and web
pages).
[7] Slo ine, J.-J. E., Li, W. Applied Nonlinea Con ol (sec ions on obo con ol/impedance).
[8] Spong, M. W., Hu chinson, S., Vidyasaga , M. Robo Modeling and Con ol. Wiley.
[9] Albu-Schä e , A., O , C., Hi zinge , G. A uni ied passi i y-based con ol amewo k o
posi ion, o que and impedance con ol o lexible join obo s (IJRR/IEEE T-RO).
[10] Sen is, L., Pa k, J., Kha ib, O. Complian con ol o mul icon ac and cen e -o -mass
beha io s in humanoid obo s (IJRR/ICRA).
[11] IEC/ISO indus ial communica ion s anda ds and sa e y s anda ds (e.g., IEC 61158, IEC
61508, ISO 13849).
21
[12] Indus ial obo ics in eg a ion and diagnos ics (OPC UA, PROFINET, endo
manuals/whi epape s).
[13] Kalman, R. E. A new app oach o linea il e ing and p edic ion p oblems (ASME Jou nal o
Basic Enginee ing).
[14] Hogan, N. Impedance con ol: An app oach o manipula ion (ASME Jou nal o Dynamic
Sys ems, Measu emen , and Con ol).
[15] G eyB Insigh s. Tesla Bo pa en s page (image esou ces and pa en -s yle schema ics):
h ps://insigh s.g eyb.com/ esla-bo -pa en s/ (Sou ce o Figu es 6–8 and ela ed cap ions).
