A Multi-Bennett 8R Mechanism Obtained From Factorization of Bivariate Motion Polynomials

Johanna Frischauf Martin Pfurner
Daniel F. Scharler Hans-Peter Schröcker
Department of Basic Sciences in Engineering Sciences
University of Innsbruck Austria

Abstract

Wepresentaclosed-loop8Rmechanismwithtwodegreesoffreedomwhosemotionexhibitscuriousproperties.Inanypointofatwo-dimensionalcomponentofitsconfigurationvarietyitispossibletofixeverysecondjointwhileretainingonedegreeoffreedom.Thisshowsthattheevenandtheoddaxes,respectively,alwaysformaBennettmechanism.Inthismechanism,oppositedistancesandanglesareequalandalloffsetsarezero.The8Rmechanismhasfour“totallyaligned”configurationsinwhichthecommonnormalsofanypairofconsecutiveaxescoincide.

\setkomafont

title \setkomafontsectioning

1 Introduction

Overconstrainedlinkagesisalong-lastingbutstillhighlyactivetopicofresearchinmechanismscience.Forseveraldecades,researchersfocusedonoverconstrainedmechanismsconsistingofasingleloopof $n \leq 6$ revolutejoints(R),prismaticjoints(P),or,sometimes,helicaljoints(H).Newlinkagesofthattypearecontinuouslybeingdiscovered,oftenbycraftilycombiningknownlinkages[28, 3, 2],sometimesvianovelconceptsfortheirconstruction.Oneoftheseconceptsisthefactorizationofmotionpolynomials[8].Itgaverisetotheconstructionoftheonlyclassofoverconstrained6RlinkageswithstillunknownrelationsbetweenitsDenavit-Hartenbergparameters.In[9, 5, 17, 16, 19, 18],motionpolynomialfactorizationwasexploitedforthesynthesisoflinkages.Inspiteofsomeattempts,acompleteclassificationofoverconstrainedsingle-looplinkagesiscurrentlyoutofreach.Itisthusnaturalthatresearcheffortsshiftedtowardstheinvestigationofsingle-looplinkagesconsistingof $n \geq 7$ linkswith,generically, $n - 6 \geq 1$ degreesoffreedom.(Theclassificationofsingle-looplinkageswith $n \geq 7$ linksandmorethan $n - 6$ degreesoffreedomhasrecentlybeencompletedin[7].)Aguidingprinciplefortheirconstructionisexistenceof“interesting”propertiesofthemechanism’smotionoritsconfigurationvariety.Oneexampleis[12],where7Rlinkageswhoseconfigurationvarietycontainsirreduciblecomponentsofdifferentdimensions–apropertythathasbeennamedkinematotropicin[27]–areconstructed.[23]combinesmobile4Rlinkages(Bennettlinkages)orRPRPlinkagestoloopsof7R/Pjointswhoseconfigurationvarietyisreducible.Themotionoftheoriginal4R/Plinkagesisobtainedbylockingofjointsincertainconfigurations.Analogically[22]restrictedaspeciallydesignedsingle-loop8Rmechanismtoitspossiblesub-motions.[19]and[18]pursuesimilaraimsbutusemotionpolynomialfactorizationtechniques.Jointlockingisalsousedin[11]forrestrictingamechanismtoacertainsubvarietyofitstotalconfigurationspace,althoughforaclassofparallelmechanism.Ourcontributioninthisarticleisofsimilarspiritastheworkscitedabovebutalsodiffersinseveralaspects.Wepresentan8Rlinkagewithtwodegreesoffreedomthathastheweirdpropertythatitretainsonedegreeoffreedomwhenlockingeverysecondjointinanyconfigurationofatwo-dimensionalsubvarietyofitstotalconfigurationspace.Thispropertyimmediatelyimpliesthatthequadruplesof“even”or“odd”axesformrespectiveBennettlinkagesinanyconfiguration.Wethereforerefertothismechanismbythename“multi-Bennett8Rmechanism”.WhilecombinationofBennettlinkagesisacommontechniqueinthisarea[6, 26, 1, 23, 12],ourexampleseemstobenovel.Itisnotgeometricallymotivated–atleastinthecurrentstateofourunderstanding–butratherbasedonanalgebraicconstruction.Assuggestedbyexamplesin[13],thereexistbivariatemotionpolynomialsthatadmit,inanon-trivialway,twofactorizationsintoproductsoflinearunivariatefactorswithalternatingindeterminates.Thesetwotimesfourfactorsgiverisetotherevoluteaxesofthe8Rmechanismanddescribetheirrelativemotions.Theunderlyingbivariatefactorizationtheoryiscurrentlybeingexplored[15, 13]andisconsiderablyharderthanintheunivariatecase.ThisiswitnessedbyourproofofexistenceinTheorem 3.3.Inspiteofitsalgebraicconstruction,the8Rlinkageissubjecttoseveregeometricconstraints.WedemonstratethisbycomputingsimplenecessaryrelationsbetweenitsDenavit-HartenbergparametersinTheorem 4.2.InTheorems 4.1and4.3wedescriberemarkablepropertiesofseveraldiscreteconfigurations.Wefeelthatitsnumerousspecialproperties(simpleDenavit-Hartenbergparameters,twodegreesoffreedomthatareeasytocontrolvialow-degreerationalparametrization,existenceofspecialconfigurations)makeourmechanismapromisingcandidateforyettobeexploredapplications.WecontinuethistextbyrecallingsomebasicfactsaboutmotionpolynomialsandtheirrelationtomechanismscienceinSection 2.InSection 3weprovideaproofforexistenceofquaternionpolynomialswithtwonon-trivialunivariatefactorizations.Theproofisconstructiveandprovidesagoodmethodtodirectlycomputetheunderlying8Rlinkage.Nonetheless,wefoundtheprocedureinsufficientforobtainingresultsthataresuitableforfurtherprocessingandinparticularforthecomputationofDenavit-Hartenbergparameters.Thus,ourfurtherinvestigationofthemulti-Bennett8RmechanisminSection 4isbasedoncarefullyselectedcoordinateframesandconfigurations.Thissimplificationresultsinformulasthataretractablebymeansofcomputeralgebraand,ultimately,providesthedesirednecessaryrelationsamongtheDenavit-Hartenbergparameters(Theorem 4.2).Thispaperisacontinuationof[14],aconferencepaperwhichverifiesmostoftheclaimsmadeinthisarticleathandofaconcretenumericexample.Strictmathematicalproofsoftheclaimedfactsarepresentedhereinforthefirsttime.

2 Preliminaries

Ourconstructionofthemulti-Bennettmechanismisbasedoncertainfactorizationsofbivariatequaternionpolynomials.Inthissectionweprovideabriefintroductiontosomefundamentalconceptsthatwillbeusedlaterinthistextandwesettleournotation.Denoteby $H$ thefour-dimensionalassociativerealalgebraofquaternions.Itisgeneratedbybasiselements $i$ , $j$ ,and $k$ viatherelations

i^{2} = j^{2} = k^{2} = i j k = - 1.

Aquaternion $q \in H$ canbewrittenas $q = q_{0} + q_{1} i + q_{2} j + q_{3} k$ withrealnumbers $q_{0}$ , $q_{1}$ , $q_{2}$ , $q_{3}$ .Extendingrealscalarstodualnumbers $a + b ε$ with $a$ , $b \in R$ and $ε^{2} = 0$ yieldsthealgebra $D H$ ofdualquaternions

D H = {h = p + ε q ∣ p, q \in H} .

Theconjugatequaternionorconjugatedualquaternion $h^{*}$ isobtainedbyreplacing $i$ , $j$ ,and $k$ with $- i$ , $- j$ ,and $- k$ ,respectively,the $ε$ -conjugate $h_{ε}$ ofadualquaternionisobtainedbyreplacing $ε$ with $- ε$ .Given $h = h_{0} + h_{1} i + h_{2} j + h_{3} k + ε (h_{4} + h_{5} i + h_{6} j + h_{7} k)$ ,thevalue $Scal(h)\coloneqq12(h+h∗)=h0+εh4$ iscalledthescalarpartand $Vect(h)\coloneqq12(h−h∗)=h1i+h2j+h3k+ε(h5i+h6j+h7k)$ thevectorpartof $h$ .Thequaternionordualquaternionnormis $h h^{*}$ .Forgeneric $h = p + ε q = h_{0} + h_{1} i + h_{2} j + h_{3} k + ε (h_{4} + h_{5} i + h_{6} j + h_{7} k) \in D H$ ,itisthedualnumber

h h^{*} = p p^{*} + ε (p q^{*} + q p^{*}) = h_{0}^{2} + h_{1}^{2} + h_{2}^{2} + h_{3}^{2} + 2 ε (h_{0} h_{4} + h_{1} h_{5} + h_{2} h_{6} + h_{3} h_{7}) \in D .

(1)

Dualquaternionssatisfying $h h^{*} = 1$ aresaidtobenormalizedorunit.Inthiscase,thedualpartin(1)vanishes,thatis

h_{0} h_{4} + h_{1} h_{5} + h_{2} h_{6} + h_{3} h_{7} = 0.

(2)

Thisiswell-knownunderthenameStudycondition.Thequaternion $p$ andthedualquaternion $h = p + ε q$ areinvertibleifandonlyif $p \neq 0$ .Inthiscase,wehave

p^{- 1} = \frac{p^{*}}{p p^{*}}, h^{- 1} = p^{- 1} (1 - ε q p^{- 1}) .

If $h$ isunit,then $h^{- 1} = h^{*}$ .Themultiplicativesub-group $DH×\coloneqq{h∈DH∣hh∗∈R∖{0}}$ modulotherealmultiplicativegroup $R^{\times}$ isisomorphicto $SE (3)$ ,thegroupofrigidbodydisplacements.Usinghomogeneouscoordinatesintheprojectivespace $P^{3} (R) = P (⟨ 1, ε i, ε j, ε k ⟩)$ ,theactionof $h \in D H^{\times}$ on $x = x_{0} + ε (x_{1} i + x_{2} j + x_{3} k) \in P^{3} (R)$ isgivenby

x↦y\coloneqqhεxh∗.

(3)

2.1 DualQuaternionsandLineGeometry

Inthispaper,rotationsaroundafixedaxisbutwithvariablerotationanglewillplayanimportantrole.Wethereforehaveacloserlookattherepresentationofstraightlines(revoluteaxes)androtationswithintheframeworkofdualquaternions.IdentifyingtheorientedrevoluteaxiswithnormalizedPlückercoordinates $(p_{1}, p_{2}, p_{3}, q_{1}, q_{2}, q_{3})$ inthesenseof[24, Section 2]withtheunitdualquaternion $r = p_{1} i + p_{2} j + p_{3} k + ε (q_{1} i + q_{2} j + q_{3} k)$ ,therotationwithangle $φ$ around $r$ isgivenbytheunitdualquaternion

h\coloneqqcos(φ2)+sin(φ2)r,φ∈[0,2π)

(4)

or,becauseweusehomogeneouscoordinates,byanyofitsnon-zerorealmultiples.Notethatthedualquaternion $h$ of(4)satisfiestheStudycondition(2)because $r$ satisfiesthePlückercondition $r r^{*} \in R ∖ {0}$ .Theaction(3)onpointscanbeusedtotransformstraightlinesbytransformingpointsonthem.PointsonastraightlinegivenbyitsPlückercoordinatescanbefound,forexample,by[24, Equation (2.4)].Astraightforwardcalculationalsoprovidesuswithadirectformulafordisplacingastraightline $r$ whosePlückercoordinatesaregivenasvectorialdualquaternions:

r↦q\coloneqq(hrh∗)ε.

(5)

2.2 DualQuaternionPolynomials

Therepresentation(4)ofarotationaroundanorientedgeneralaxisisonlyuniqueuptomultiplicationwitharealscalar.Assuming,forthetimebeing, $φ \neq 0$ ,wecandivide(4)by $sin \frac{φ}{2}$ ,substitute $cot \frac{φ}{2}$ with $- t$ andmultiplytheresultwith $- 1$ toseethatthelineardualquaternionpolynomial

h = t - r, t \in R

(6)

parametrizesallrotationswithnon-vanishingrotationanglearound $r$ aswell.Inordertoalsoaccountfor $φ = 0$ ,weshouldextendtheparameterrangein(6)to $R \cup {\infty}$ .Withthenaturalunderstandingthat $h(∞)\coloneqqlimt→∞1th(t)=1$ ,theparametervalue $t = \infty$ indeedcorrespondstotherotationangle $φ = 0$ ,thatis,theidentitytransformation.Moregenerally,wecanconsiderarbitrarypolynomials $C = \sum_{i = 0}^{d} c_{i} t^{i}$ withcoefficients $c_{i} \in D H$ .Sincetheindeterminate $t$ typicallyservesasarealparameterinourcontext,multiplication,conjugationandevaluationatrealvaluesofpolynomialsaredefinedbytheconventionsthat $t$ commuteswithallcoefficientsand $t^{*} = t$ .Thethusobtainedringofpolynomialsisdenotedby $D H [t]$ .Similarly,wecanalsoconsidertheringofbivariatedualquaternionpolynomials $D H [s, t]$ in $t$ and $s$ .Itsmultiplication,conjugationandevaluationatrealvaluesisdefinedbysimilarconventionsandtheassumptionthat $t$ and $s$ commutewithallcoefficientsandwitheachother.Thelinearpolynomial $t - r$ fromEquation(6)satisfies $(t - r) {(t - r)}^{*} \in R [t]$ andalso $(t - r) {(t - r)}^{*} \neq 0$ (notethat $r$ satifiesthePlückercondition).Ageneralizationofthispropertyleadsto

Definition 1.

Apolynomial $C \in D H [t]$ orin $D H [s, t]$ iscalledamotionpolynomialif $C C^{*} \in R [t]$ orin $R [s, t]$ ,respectively,and $C C^{*} \neq 0$ .

Thename“motionpolynomial”isjustifiedbytheobservationthattheaction(3)onpointsallowstheparametricversion

x \mapsto y (s, t) = C_{ε} (s, t) x C^{*} (s, t), s, t \in (R \cup {\infty}) \times (R \cup {\infty}) .

(7)

Equation (7)isapolynomialmapinhomogeneouscoordinates.TheCartesiancoordinatesof $y (s, t)$ arerationalfunctionssothat(7)describesarigidbodymotionwithrationalsurfacesastrajectories.Univariatemotionpolynomialshavebeenoriginallydefinedin[8].There,itwasimplicitlyassumedthatmotionpolynomialsaremonic.Weratherreplacethisassumptionbythecondition $C C^{*} \neq 0$ which,togetherwithaproperevaluationat $s = \infty$ or $t = \infty$ andthepossibilityofrationalre-parametrizations,sufficesforourpurpose.

Definition 2.

Thevalueofthemotionpolynomial $C \in D H [t]$ at $t = \infty$ isdefinedas $C(∞)\coloneqqlimt→∞t−degCC(t)$ .Itistheleadingcoefficientof $C$ .Thevalueof $C \in D H [s, t]$ at $(s, t)$ where $s = \infty$ or(notexclusively) $t = \infty$ isdefinedbysimilarlimits.Itistheleadingcoefficientin $s$ or $t$ (orinboth),respectively.

Re-parametrizationsthatpreservepolynomialityanddegreeofunivariatemotionpolynomialsaremapsoftheform

τ \mapsto t = \frac{a τ + b}{c τ + d}, a, b, c, d \in R, a d - b c \neq 0,

(8)

combinedwithmultiplyingawaydenominators.With $C = \sum_{i = 0}^{d} c_{i} t^{i}$ wehave

C (τ) = d \sum i = 0 c_{i} (a τ + b)^{i} (c τ + d)^{d - i} .

Itisnoteworthythat(8)naturallyisamapfrom $R \cup {\infty}$ to $R \cup {\infty}$ .Assuming $c \neq 0$ wehave

\infty \mapsto \frac{a}{c} and - \frac{d}{c} \mapsto \infty .

If $c = 0$ ,then $\infty$ isafixpointof(8).Re-parametrizationsoftype(8)donotchangethepropertyofbeingamotionpolynomial.Anextensionof(8)tobivariatepolynomialsisstraightforward.Letusillustratesomedefinitionsandconceptssofarforlinearmotionpolynomials,whichconstituteanimportantspecialexample.

Example 2.1.

Thelinearpolynomial $C = t - h$ with $h = h_{0} + h_{1} i + h_{2} j + h_{3} k + ε (h_{4} + h_{5} i + h_{6} j + h_{7} k)$ isamotionpolynomialbyDefinition 1if

C C^{*} = (t - h) {(t - h)}^{*} = t^{2} - (h + h^{*}) t + h h^{*}

isreal.Thisisequivalentto $h + h^{*} = 2 (h_{0} + ε h_{4})$ and $h h^{*} = h_{0}^{2} + h_{1}^{2} + h_{2}^{2} + h_{3}^{2} + 2 ε (h_{0} h_{4} + h_{1} h_{5} + h_{2} h_{6} + h_{3} h_{7})$ bothbeingrealwhence

h_{4} = 0 and h_{1} h_{5} + h_{2} h_{6} + h_{3} h_{7} = 0.

(9)

Inthiscasethemotionpolynomial $C$ describesarotationaroundthestraightlinewithPlückercoordinates $Vect (h) = h_{1} i + h_{2} j + h_{3} k + ε (h_{5} i + h_{6} j + h_{7} k)$ ,afactwhichshouldnotbesurprising.Wealreadydemonstratedtherelationbetweenlinearmotionpolynomialsandrotations.Thesecondequationin(9)isjustthePlückercondition.ByDefinition 2,thevalue $C (\infty)$ equals ${lim}_{t \to \infty} t^{- 1} C (t) = 1$ whichistheidentitydisplacement.Obviously, $C (0) = - h$ .There-parametrization $τ \mapsto \frac{1}{τ}$ isoftype(8)with $a = d = 0$ and $b = c = 1$ .Itinterchanges $0$ and $\infty$ .Indeed, $C (τ) = 1 - τ h$ and

C (τ) |_{τ = 0} = 1, C (τ) |_{τ = \infty} = lim τ \to \infty τ^{- 1} C (τ) = - h,

asexpected.

Inthenextsectionwewillstudybivariatemotionpolynomialswhichcanbewrittenasproductsoflinearmotionpolynomials.

3 AlternatingFactorizationsofBivariateQuaternionPolynomials

Givenabivariatedualquaternionpolynomial $C \in D H [s, t]$ ,wedenoteitsbi-degreeby $bdeg (C)$ .Wewishtofindamotionpolynomial $C \in D H [s, t]$ with $bdeg (C) = (2, 2)$ thatadmitstwodifferentfactorizationswithalternatingunivariatelinearfactors,i.e.,

C = (t - h) (s - ℓ) (t - m) (s - n) = (s - n^{'}) (t - m^{'}) (s - ℓ^{'}) (t - h^{'}),

(10)

where $h$ , $ℓ$ , $m$ , $n$ , $n^{'}$ , $m^{'}$ , $ℓ^{'}$ , $h^{'} \in D H ∖ {0}$ and $(t - r) {(t - r)}^{*} = (t - r^{'}) {(t - r^{'})}^{*} \in R [t]$ , $(s - u) {(s - u)}^{*} = (s - u^{'}) {(s - u^{'})}^{*} \in R [s]$ for $r \in {h, m}$ and $u \in {ℓ, n}$ ,arequirementthatisseentobenecessarybytakingnormsonbothsidesof(10).Wecallthesefactorizationsalternatingsincethe $s$ -and $t$ -factorsappearinalternatingorder.BytheconsiderationsinSection 2,thelinearfactorswillrepresentrotationsaroundfixedaxes.Motionpolynomialsofshape(10)immediatelyleadtoclosed-loop8Rmechanismswiththepropertiesmentionedintheintroduction:

Eachfactorizationgivesrisetoatwo-parametricmotionofanopen4Rchain.Sincethefactorizationsagree,thetwodistallinkscanberigidlyconnectedtoformaclosed-loop8Rlinkagewiththesametwodegreesoffreedom.
Inthistwo-dimensionalmotioncomponent(weconjecturethatothercomponentsexistaswell),themotionofanyaxisisdeterminedbyeither $s$ or $t$ .Lockingoneaxis,thatis,fixing $s$ or $t$ ,automaticallylockseverysecondaxiswhiletheaxesparametrizedbytheotherparameterstillmove.

Uptonow,onlyisolatedexamplesofthiskindofpolynomialshavebeenknown(c. f.[13, 14]).Wewillpresentasystematicconstructionofthesepolynomials,andthusofmulti-Bennett8Rmechanisms,andstartwithasimpleyetcruciallemma:

Lemma 3.1.

Let $C \in D H [s, t]$ beadualquaternionpolynomialthatadmitstwoalternatingfactorizations:

C = (t - h) (s - ℓ) (t - m) (s - n) = (s - n^{'}) (t - m^{'}) (s - ℓ^{'}) (t - h^{'})

(11)

Then

(s - ℓ) (s - n) = (s - n^{'}) (s - ℓ^{'}) and (t - h) (t - m) = (t - m^{'}) (t - h^{'}) .

Proof.

Wemayview $C$ asapolynomialin $t$ withcoefficientsinthering $D H [s]$ .Comparingthecoefficientof $t^{2}$ ontheleft-handandtheright-handsideofEquation (11)showsthat $(s - ℓ) (s - n) = (s - n^{'}) (s - ℓ^{'})$ .Thesecondstatementfollowsbyinterchangingtherolesof $s$ and $t$ .∎

Remark 3.2.

Ifthepolynomial $C = (t - h) (s - ℓ) (t - m) (s - n)$ admitsasecondalternatingfactorizationasin(11),itcanalwaysbecomputedbyso-calledBennettflips[17, Definition 4].Thenameismotivatedbytheobservationthattherevoluteaxesto $ℓ$ , $n$ , $ℓ^{'}$ ,and $n^{'}$ (andalsoto $h$ , $m$ , $h^{'}$ and $m^{'}$ )form,inthatorder,aBennettlinkage.Moreprecisely,thequaternions $n^{'}$ , $ℓ^{'}$ , $m^{'}$ , $h^{'} \in D H$ canbecomputedbyreplacingtheunivariatepolynomials $(s - ℓ) (s - n)$ and $(t - h) (t - m)$ bytheirsecondfactorizationwithlinearfactors.In[17, Definition 4],itisshownthatthesecondfactorizationofaunivariatepolynomial $(u - h_{1}) (u - h_{2}) \in D H [u]$ isobtainedviatheformulas

k_{2} = - ({h_{1}}^{*} - h_{2})^{- 1} (h_{1} h_{2} - h_{1} {h_{1}}^{*}) and k_{1} = h_{1} + h_{2} - k_{2},

(12)

where $(u - h_{1}) (u - h_{2}) = (u - k_{1}) (u - k_{2})$ and $(u - h_{1}) {(u - h_{1})}^{*} = (u - k_{2}) {(u - k_{2})}^{*}$ , $(u - h_{2}) {(u - h_{2})}^{*} = (u - k_{1}) {(u - k_{1})}^{*}$ .

Thefollowingtheoremisthecenterpieceofthepresentsection.Itpresentsamethodthatcanbeusedtoconstructquaternionpolynomialsofbi-degree $(2, 2)$ thatadmittwodifferentfactorizationswithlinearfactors.Thetheoremonlytalksaboutquaternions.Anextensiontodualquaternionswillbediscussedlater.

Theorem 3.3.

Let $h$ , $m$ , $n \in H$ bequaternions.Moreover,assumethateither $Scal (h) \neq Scal (m)$ or,if $Scal (h) = Scal (m)$ ,that $h h^{*} \neq m m^{*}$ .Thenthereexistsasuitablequaternion $ℓ \in H$ suchthatthepolynomial

C\coloneqq(t−h)(s−ℓ)(t−m)(s−n)∈H[s,t]

(13)

admitsasecondfactorizationwithunivariatelinearfactors.

Proof.

Webrieflyexplainthemainideaoftheproof:Accordingto(13),thepolynomial $C$ hasaleftfactoroftheform $t - h$ .Bychoosingthequaternion $ℓ$ inaspecialway,weforcethepolynomial $C$ toadmitanotherfactorizationwitharightfactor $t - h^{'}$ ofthesamenorm,thatis

(t - h) (s - ℓ) (t - m) (s - n) = A (t - h^{'}) with A \in H [s, t], bdeg (A) = (1, 2),

(14)

and $(t - h) {(t - h)}^{*} = (t - h^{'}) {(t - h^{'})}^{*}$ .In[25, 13],itisshownthatpolynomialsofdegreeonein $t$ admitfactorizationswithunivariatelinearfactorsaslongasthecorrespondingnormpolynomialsplitsintoaproductofrealunivariatepolynomials.¹¹1Theoriginalreferenceis[25, Lemma 2.9],butin[13, p. 9]weprovideanalgorithmthatcanbeusedtocomputeafactorizationofthedesiredform.Thisisindeedthecaseforthepolynomial $A$ in(14)since $A A^{*} = P R$ with $P = (t - m) {(t - m)}^{*} \in R [t]$ and $R = (s - ℓ) {(s - ℓ)}^{*} (s - n) {(s - n)}^{*} \in R [s]$ .Therefore,

A = (u_{1} - h_{1}) (u_{2} - h_{2}) (u_{3} - h_{3}) with u_{i} \in {s, t} and h_{i} \in H for i = 1, 2, 3

(15)

and $C$ admitsasecondfactorizationwithunivariatelinearfactors.Allpossiblecombinationsoflinear $s$ -and $t$ -factorswillbediscussedintheproofofCorollary 3.4.Inordertoshow(14),wedefine $M\coloneqq(t−h)(t−h)∗∈R[t]$ ,view $C$ and $M$ asunivariatepolynomialswithcoefficientsin $H [s]$ andapplydivisionwithremainderof $C$ by $M$ :

C = (t - h) (s - ℓ) (t - m) (s - n) = T M + R,

(16)

where $T$ , $R \in H [s, t]$ and $bdeg (R) = (d_{t}, d_{s})$ with $d_{t} \leq 1$ and $d_{s} \leq 2$ .Wecomparethecoefficientsof $t^{2}$ ontheleft-handandright-handsideofequation(16)andconclude $T = (s - ℓ) (s - n)$ since $R$ isatmostlinearin $t$ .Therefore,thelinearfactor $s - n$ isarightfactorofboth $C$ and $T$ .Representation(16)thenshowsthatitisalsoarightfactorof $R$ (weusedthefact $T M = M T$ sincethepolynomial $M \in R [t]$ isrealandcommuteswithotherpolynomials).Similarly,thelinearfactor $t - h$ isaleftfactorofboth $C$ and $M = (t - h) {(t - h)}^{*}$ andhencealsoaleftfactorof $R$ .Wecanwrite $R = (t - h) R^{'} = R^{'} t - h R^{'} \in H [s, t]$ with $R^{'} \in H [s]$ .Since $s - n$ divides $R$ fromtherightitneedstodivideeachcoefficientof $R$ whenviewedaspolynomialin $t$ withcoefficientsin $H [s]$ .Weconcludethat $s - n$ isarightfactorof $R^{'}$ .Hence $R$ isnecessarilyoftheform

R = (t - h) (r_{1} s + r_{0}) (s - n)

(17)

with $r_{1}, r_{0} \in H$ .Thequaternions $r_{1}$ and $r_{0}$ areobtainedbycomparingappropriatecoefficientsin(16):Uptonow,wealwaysconsidered $C$ and $R$ asunivariatepolynomialswithcoefficientsin $H [s]$ .Wenowviewthemasbivariatepolynomials,whichallowsustocomparethecoefficientsof $t s^{2}$ and $t$ :Comparingcoefficientsof $t s^{2}$ in(16)yields

- h - m = - h - h^{*} + r_{1}

andhence $r_{1} = h^{*} - m$ .Comparingcoefficientsof $t$ leadsto

- ℓ m n - h ℓ n = - ℓ n h - ℓ n h^{*} - r_{0} n

andhence

r_{0} = [(ℓ m + h ℓ) n - ℓ n (h + h^{*})      (*) = (h + h^{*}) n] n^{- 1} = ℓ m + h ℓ - ℓ (h + h^{*}) = h ℓ - ℓ (r_{1} + h) .

In $(*)$ weusedthefact $h + h^{*} \in R$ .Letusrecallthemainideaoftheproof:Weneedtoforce $C$ toadmitafactorizationwiththerightfactor $t - h^{'}$ ,where $h^{'}$ isyettobedetermined.Alternatively,wecanforce $R$ tohavetherightfactor $t - h^{'}$ .By(16),itisthenalsoarightfactorof $C$ (notethatwerequire $M = {(t - h^{'})}^{*} (t - h^{'})$ ).Wewrite

R = r_{1} (t - r_{1}^{- 1} h r_{1}) (s + r_{1}^{- 1} r_{0}) (s - n) = r_{1} (t - h^{'}) (s + r_{1}^{- 1} r_{0}) (s - n),

where $h′\coloneqqr−11hr1$ .Thepolynomial $t - h^{'}$ indeedsatisfiestherequiredcondition $(t - h^{'}) {(t - h^{'})}^{*} = M$ .Ifthepolynomial $S\coloneqq(s+r−11r0)(s−n)∈H[s]$ wasarealpolynomial,thefactor $t - h^{'}$ wouldcommutewith $S$ andhencebearightfactorof $R$ .Incase $- r_{1}^{- 1} r_{0} = n^{*}$ weobtain $S = (s - n^{*}) (s - n) \in R [s]$ .Therefore,weneedtofindaquaternion $ℓ \in H$ suchthat $- r_{0} = r_{1} n^{*}$ ,thatis

ℓ (r_{1} + h) - h ℓ = r_{1} n^{*} .

Aboveequationisalinearequationinthequaternionunknown $ℓ \in H$ .By[10, Theorem2.3]itisuniquelysolvableifandonlyif $Scal (A) \neq - Scal (B)$ or $Scal (A) = - Scal (B)$ and $A A^{*} \neq B B^{*}$ ,where $A\coloneqq−h$ and $B\coloneqqh+r1$ .Thisisequivalenttoourtheorem’sassumption $Scal (h) \neq Scal (m)$ or $Scal (h) = Scal (m)$ and $h h^{*} \neq m m^{*}$ .InthereferencedTheorem 2.3of[10],anexplicitformulaforthesolution $ℓ \in H$ isprovided:

ℓ = (2 (2 Scal (h) - Scal (m)) - h - (h + r_{1}) {(h + r_{1})}^{*} h^{- 1})^{- 1} (r_{1} n^{*} - h^{- 1} r_{1} n^{*} {(h + r_{1})}^{*}) .

Thisprovestheclaim.∎

Inordertoconstructmechanisms,weneedtoguaranteethatthesecondfactorizationinTheorem3.3isalternatingaswell.Thisisensuredbysomeadditionalassumptionsstatedinthe(rathertechnical)Corollary3.4.

Corollary 3.4.

Let $h$ , $m$ , $n \in H$ bequaternionsand $h^{'}$ , $m^{'} \in H$ besuchthat $(t - h) (t - m) = (t - m^{'}) (t - h^{'})$ (c. f.Remark3.2).If $m n \neq n m$ and $h^{'} n \neq n h^{'}$ ,thesecondfactorizationinTheorem3.3isalternatingaswell,thatis

C = (t - h) (s - ℓ) (t - m) (s - n) = (s - n^{'}) (t - m^{'}) (s - ℓ^{'}) (t - h^{'}),

where $(t - r) {(t - r)}^{*} = (t - r^{'}) {(t - r^{'})}^{*}$ , $(s - u) {(s - u)}^{*} = (s - u^{'}) {(s - u^{'})}^{*}$ for $r \in {h, m}$ and $u \in {ℓ, n}$ .

Proof.

Wehave $3! = 6$ possibilitiesforfactorizationsof $A$ withunivariatelinearfactors,where $A$ isdefinedin(15).Weuserepresentation(14)andobtain

$C$	$= (t - h) (s - ℓ) (t - m) (s - n) = (s - ℓ) (t - m^{'}) (s - n) (t - h^{'}),$	(18)
$C$	$= (t - h) (s - ℓ) (t - m) (s - n) = (t - m^{'}) (s - ℓ) (s - n) (t - h^{'}),$	(19)
$C$	$= (t - h) (s - ℓ) (t - m) (s - n) = (s - ℓ) (s - n) (t - m^{'}) (t - h^{'}),$	(20)
$C$	$= (t - h) (s - ℓ) (t - m) (s - n) = (t - m^{'}) (s - n^{'}) (s - ℓ^{'}) (t - h^{'}),$	(21)
$C$	$= (t - h) (s - ℓ) (t - m) (s - n) = (s - n^{'}) (s - ℓ^{'}) (t - m^{'}) (t - h^{'}), or$	(22)
$C$	$= (t - h) (s - ℓ) (t - m) (s - n) = (s - n^{'}) (t - m^{'}) (s - ℓ^{'}) (t - h^{'}) .$	(23)

Wehighlightedthedifferentpossibilitiesforfactorizationsof $A$ byusingboldletters.Notethatthelinear $s$ -factorsonthelefthavethesamenormsasthelinear $s$ -factorsontherightbutthe $s$ -factorspossiblyappearinadifferentorder.Iftheorderisdifferent,the $s$ -factorsmustcorrespondinBennettflipsbyargumentsasintheproofofLemma 3.1andaredenotedbyaprime,i.e. $n^{'}$ , $ℓ^{'}$ .Iftheorderisthesame,the $s$ -factorsareequalbyargumentssimilartoLemma 3.1and[8, Lemma 3].Thesameargumentsapplytolinear $t$ -factors.Thetwofactorizationsin(18)are $t$ -equivalentinthesenseof[13, Definition 4.3].²²2In[13, Definition 4.3],twodifferentfactorizationsofbivariatequaternionpolynomialswithlinearfactorsarecalled $t$ -equivalent,ifthelinear $s$ -factorsappearinthesameorder.Thisisthecasein(18)and(19)since $s - ℓ$ isthefirstand $s - n$ thesecond $s$ -factorinbothfactorizations.SuchfactorizationsarespecialsincetheycanbetransferredintoeachotherbyapplyingBennettflipsandlettingappropriate $s$ -and $t$ -factorscommutewitheachother(c. f.[13, Proposition 4.6]).By[13, Proposition 4.6],weconclude $h^{'} n = n h^{'}$ (andalso $h ℓ = ℓ h$ ),acasewhichisexcludedbyassumption.Thesamecanbesaidforthetwofactorizationsin (19).Thesecondfactorizationin(20)canberewrittenas $(s - ℓ) (s - n) (t - h) (t - m)$ andthereforeturnsouttobe $s$ -equivalenttothefirstfactorizationin(20).Weagainuse[13, Proposition 4.6]andconclude $m n = n m$ ,whichisalsoexcludedbyassumption.Thesecondfactorizationsin(21)andalsoin(22)arecoincidentwiththesecondfactorizationsin(19)and(20)afterapplyingBennettflipsof $(s - n^{'}) (s - ℓ^{'})$ .Therefore, $C$ needstoadmittwodifferentfactorizationsoftheform (23).∎

Remark 3.5.

UndertheweakassumptionsofCorollary 3.4,Theorem 3.3guaranteesexistenceofaquaternion $ℓ \in H$ suchthat $C = (t - h) (s - ℓ) (t - m) (s - n)$ admitsasecondalternatingfactorization.Whileourproofsareconstructive,theactualcomputationofthesecondfactorizationcanbesimplifiedalotwiththehelpofRemark 3.2.Atfirst,wecomputequaternions $h^{'}$ , $ℓ^{'}$ , $m^{'}$ , $n^{'} \in H$ viaBennettflips(12)oftheunivariatepolynomials $(t - h) (t - m)$ and $(s - ℓ) (s - n)$ ,respectively.Thesecondfactorizationisthengivenby $C = (s - n^{'}) (t - m^{'}) (s - ℓ^{'}) (t - h^{'})$ .PseudocodeforthisapproachisgiveninAlgorithm 1.

1:Realquaternions

h

m

n \in H

with

Scal (h) \neq Scal (m)

Scal (h) = Scal (m)

and

h h^{*} \neq m m^{*}

2:Twotuples

(t - h, s - ℓ, t - m, s - n)

and

(s - n^{'}, t - m^{'}, s - ℓ^{'}, t - h^{'})

suchthat

(t - h) (s - ℓ) (t - m) (s - n) = (s - n^{'}) (t - m^{'}) (s - ℓ^{'}) (t - h^{'})

r_{1} \leftarrow h^{*} - m

ℓ \leftarrow (2 (2 Scal (h) - Scal (m)) - h - (h + r_{1}) {(h + r_{1})}^{*} h^{- 1})^{- 1} (r_{1} n^{*} - h^{- 1} r_{1} n^{*} {(h + r_{1})}^{*})

h^{'} \leftarrow - (h^{*} - m)^{- 1} (h m - h h^{*}), m^{'} \leftarrow h + m - h^{'}

ℓ^{'} \leftarrow - (ℓ^{*} - n)^{- 1} (ℓ n - ℓ ℓ^{*}), n^{'} \leftarrow ℓ + n - ℓ^{'}

7:return

(t - h, s - ℓ, t - m, s - n), (s - n^{'}, t - m^{'}, s - ℓ^{'}, t - h^{'})

Algorithm 1 Polynomialswithtwoalternatingfactorizations

Example 3.6.

Setting

h = 2 i - j - 3 k, m = - 6 - 2 i + 3 j - 3 k, n = - j

andapplyingAlgorithm 1yieldsthetwoalternatingfactorizations

(t - 2 i + j + 3 k) (s + i - j) (t + 6 + 2 i - 3 j + 3 k) (s + j) = (s - j) (t - 2 i - 3 j + 3 k + 6) (s + i + j) (t + 2 i + j + 3 k) .

Whenitcomestoapplicationsinspacekinematics,itisnecessarytoformulateourstatementsfordualquaternionpolynomials.Thealgebraofdualquaternions $D H$ containszerodivisorsandnon-invertibleelements,hencefactorizationtheoryforbivariatedualquaternionpolynomialsturnsouttobeevenmoreinvolvedandwearenotawareofanyresultsinthatdirection.Weobserve,however,thattheextensionoftwodifferentalternatingfactorizationsoverthequaternionstodualquaternionsisstraightforwardbyusingthefollowingapproach:Foranydualquaternion $h \in D H$ ,wedenoteitsprimalpartby $h_{p}$ anditsdualpartby $h_{d}$ ,thatis, $h = h_{p} + ε h_{d}$ for $h_{p}, h_{d} \in H$ .Similarnotationisusedformotionpolynomials.Wefindtwoalternatingfactorizationsofamotionpolynomialbyfollowingthreesteps:

Step1:

Westartwithtwodualquaternions $h = h_{p} + ε h_{d}$ and $m = m_{p} + ε m_{d}$ suchthat $t - h$ and $t - m$ satisfythemotionpolynomialconditionofDefinition 1andwecomputeBennettflipsof $h$ and $m$ toobtaindualquaternions $h^{'} = h_{p}^{'} + ε h_{d}^{'}$ and $m^{'} = m_{p}^{'} + ε m_{d}^{'}$ ,respectively.

Step2:

Wechoosearealquaternion $n_{p} \in H$ suchthat $h_{p}$ , $m_{p}$ ,and $n_{p}$ satisfytheconditionsofCorollary 3.4andapplyTheorem 3.3tothequaternions $h_{p}$ , $m_{p}$ ,and $n_{p}$ .Thisgivesaquaternionpolynomial $C_{p} \in H [s, t]$ thatadmitsthetwodifferentalternatingfactorizations

C_{p} = (t - h_{p}) (s - ℓ_{p}) (t - m_{p}) (s - n_{p}) = (s - n_{p}^{'}) (t - m_{p}^{'}) (s - ℓ_{p}^{'}) (t - h_{p}^{'}) .

(24)

Step3:

Finally,wehavetodeterminetherespectivedualparts $ℓ_{d}$ , $n_{d}$ , $ℓ_{d}^{'}$ , $n_{d}^{'} \in H$ ofthedualquaternions $ℓ = ℓ_{p} + ε ℓ_{d}$ , $n = n_{p} + ε n_{d}$ , $ℓ^{'} = ℓ_{p}^{'} + ε ℓ_{d}^{'}$ ,and $n^{'} = n_{p}^{'} + ε n_{d}^{'}$ ,toallowfortwofactorizationsof

C\coloneqq(t−hp−εhd)(s−ℓp−ε\blld)(t−mp−εmd)(s−np−εnd)=(s−n′p−εn′d)(t−m′p−εm′d)(s−ℓ′p−ε\bll′d)(t−h′p−εh′d).

(25)

Theyetunknownquaternionsarehighlightedinboldletters.Comparingcoefficientsin $t$ and $s$ forallquaternioncoefficientsontheleft-handandright-handsideofequation(25)yieldsasystemof $32$ equationsin $16$ unknowns.(Notethattheprimalpartsareequalbyconstruction.)Additionally,wehavetoimposethemotionpolynomialconditionsofDefinition 1onthelinear $s$ -polynomials,leadingtoeightfurtherlinearequationsin $16$ unknowns:

Intotal,wehavetosolveasystemof $40$ linearequationsin $16$ unknowns.Therespectivelinearsystemofequationsseemstobehighlyoverconstrained.Quitesurprisingly,itturnsouttoalwaysadmitasolutions.ThiswillbeprovedbyastraightforwardcomputationinSection 4.1sothatwehave:

Theorem 3.7.

TheconstructionoutlinedinaboveSteps 1to3genericallyyieldsamotionpolynomial $C$ satisfying

C = (t - h) (s - ℓ) (t - m) (s - n) = (s - n^{'}) (t - m^{'}) (s - ℓ^{'}) (t - h^{'})

withlinearmotionpolynomials $t - h$ , $s - ℓ$ ,…, $t - h^{'}$ .

Example 3.8.

WebuildonExample3.6andadditionallychoosequaternions

hd\coloneqq23i−74j+40kandmd\coloneqq−45i−66j−36k.

Thepolynomials $t - h_{p} - ε h_{d}$ and $t - m_{p} - ε m_{d}$ aremotionpolynomials:

	$(t - h_{p} - ε h_{d}) {(t - h_{p} - ε h_{d})}^{*}$	$= t^{2} + 14 \in R [t] ∖ {0},$
	$(t - m_{p} - ε m_{d}) {(t - m_{p} - ε m_{d})}^{*}$	$= t^{2} + 12 t + 58 \in R [t] ∖ {0} .$

WecomputeBennettflipsof $h\coloneqqhp+εhd$ and $m\coloneqqmp+εmd$ andobtain

	$m^{'} =$	$- 6 + 2 i + 3 j - 3 k - ε (21 i + 22 j + 36 k)$
	$h^{'} =$	$- 2 i - j - 3 k - ε (i + 118 j - 40 k) .$

Theunknowns $ℓ_{d}$ , $n_{d}$ , $ℓ_{d}^{'}$ , $n_{d}^{'}$ areobtainedbysolvingtherespectivesystemoflinearequations:

	$ℓ_{d}$	$= - 11 i - 11 j + 2 k,$	$n_{d}$	$= - 3 i - 2 k,$
	$ℓ_{d}^{'}$	$= 11 i - 11 j - 2 k,$	$n_{d}^{'}$	$= - 25 i + 2 k .$

Finally,wegetamotionpolynomialin $D H [s, t]$ withtwoalternatingfactorizations:

(t - 2 i + j + 3 k + ε (- 23 i + 74 j - 40 k)) (s + i - j + ε (11 i + 11 j - 2 k)) (t + 6 + 2 i - 3 j + 3 k + ε (45 i + 66 j + 36 k)) (s + j + ε (3 i + 2 k)) = (s - j + ε (25 i - 2 k)) (t + 6 - 2 i - 3 j + 3 k + ε (21 i + 22 j + 36 k)) (s + i + j - ε (11 i - 11 j - 2 k)) (t + 2 i + j + 3 k + ε (i + 118 j - 40 k)) .

4 TheMulti-Bennett8RMechanism

Intheprecedingsectionweprovedexistenceofbivariatequaternionpolynomials $C \in H [s, t]$ thatadmittwofactorizationswithlinearquaternionpolynomialsandwehintedatthepossibilitytoextendthistomotionpolynomialsoftheshape

C = (t - h) (s - ℓ) (t - m) (s - n) = (s - n^{'}) (t - m^{'}) (s - ℓ^{'}) (t - h^{'})

withdualquaternions $h$ , $ℓ$ , $m$ , $n$ , $n^{'}$ , $m^{'}$ , $ℓ^{'}$ ,and $h^{'}$ .Byconstruction,eachlinearfactorin $t$ orin $s$ parametrizesarotationaroundastraightlineinspacesothateachofthetwofactorizationsgivesrisetoanopen4Rchainwhoseend-effectorssharethetwo-parametricrationalmotionparametrizedby $C$ .Thus,thismotioniscontainedintheconfigurationvarietyoftheclosed-loop8Rlinkageformedbythetwoopen4Rchains.Investigationofpropertiesofthis8Rlinkageisthetopicofthissection.Indoingso,weonlyconsiderthegenericcase,i.e.,weassumethatnospecialalgebraicrelationsbetweentheinputparametersarefulfilled.Atpresent,acomprehensivediscussionofallspecialcasesseemsoflittlevalue.Thelinkage’szeroconfigurationisgivenby $t = s = \infty$ becausethenthereiszerorotationinalljoints(c.f.Definition 2).Theaxes’Plückercoordinatesinthiszeroconfigurationaresimplytherespectivevectorparts

Vect (h), Vect (ℓ), Vect (m), Vect (n), Vect (n^{'}), Vect (m^{'}), Vect (ℓ^{'}), Vect (h^{'})

ofthelinearfactors.Thepositionsoftheseaxesintheconfigurationdeterminedbyageneralparameterpair $(s, t)$ canbecomputedvia(5)as

$H (s, t)$	$= Vect (h),$	(26)
$L (s, t)$	$= {((t - h) Vect (ℓ) (t - h^{*}))}_{ε},$
$M (s, t)$	$= {((t - h) (s - ℓ) Vect (m) (s - ℓ^{}) (t - h^{}))}_{ε},$
$N (s, t)$	$= {((t - h) (s - ℓ) (t - m) Vect (n) (t - m^{}) (s - ℓ^{}) (t - h^{*}))}_{ε},$
$N^{'} (s, t)$	$= Vect (n^{'}),$
$M^{'} (s, t)$	$= {((s - n^{'}) Vect (m^{'}) (s - {n^{'}}^{*}))}_{ε},$
$L^{'} (s, t)$	$= {((s - n^{'}) (t - m^{'}) Vect (ℓ^{'}) (t - {m^{'}}^{}) (s - {n^{'}}^{}))}_{ε},$
$H^{'} (s, t)$	$= {((s - n^{'}) (t - m^{'}) (s - ℓ^{'}) Vect (h^{'}) (s - {ℓ^{'}}^{}) (t - {m^{'}}^{}) (s - {n^{'}}^{*}))}_{ε} .$

Notethat $H (s, t)$ and $N^{'} (s, t)$ areindependentof $t$ and $s$ , $L (s, t)$ dependsonlyon $t$ and $M^{'} (s, t)$ dependsonlyon $s$ .Forfixed $t = t_{0}$ ,theaxes $L (s, t_{0})$ , $N (s, t_{0})$ , $L^{'} (s, t_{0})$ ,and $N^{'} (s, t_{0})$ form,inthatorder,aBennettlinkagewhosemotionisparameterizedby $s$ .Wecallitthe $s$ -Bennettlinkageat $t_{0}$ .Similarly,forfixed $s = s_{0}$ weobtaina $t$ -Bennettlinkageat $s_{0}$ ,formedby $H (s_{0}, t)$ , $M (s_{0}, t)$ , $H^{'} (s_{0}, t)$ ,and $M^{'} (s_{0}, t)$ .Itiswell-known(andfollowsfromBennett’soriginaldescriptionofhismechanismasisogram,c.f.[21, Section 10.3])thatforgiven $t_{0}$ thereexisttwovaluesfor $s \in R \cup {\infty}$ atwhichthe $t_{0}$ -Bennettmechanismisinaconfigurationwhereitsfouraxeshavethesamecommonperpendicular.Wecallthisanalignedconfiguration.Asimilarstatementholdstrueforevery $s_{0}$ -Bennettmechanism.Thealignedconfigurationswillplayacrucialroleinourcomputationofthe8R-linkage’sDenavit-Hartenbergparametersinthenextsection.The8R-linkageitselfexhibitsaninterestingaligningbehavioraswellthatwillbeinvestigatedinmoredetailintheforthcomingSection 4.2.

4.1 Denavit-HartenbergParameters

Theaimofthissectionistheproofofsimplerelationsamongthemulti-Bennett’sDenavit-Hartenbergparameters.Inordertodoso,wewillcomputeparametrizationsofitsmovingaxeswithrespecttospecialcoordinates.NoneoftheseassumptionsisalossofgeneralitysothattheresultingstatementsareofgeneralvalidityandaresuitableforprovingthemissingpieceinTheorem 3.7.Ourcomputationofthe8R-linkage’sDenavit-Hartenbergparameterswillprofitalotfromthegeometryofits $t$ -and $s$ -Bennettlinkages.Accordingto[21, Section 10.3],theaxesofanyBennettlinkagecanbecomputedby

pickingtwoarbitrarypoints $F_{h}$ , $F_{m}$ andastraightline $z$ ,
rotating $F_{h}$ and $F_{m}$ ,respectively,around $z$ byarotationangleof $180\lx@arcdegree$ toobtainaspatialquadrilateral $F_{h}$ , $F_{m}$ , $F_{h^{'}}$ , $F_{m^{'}}$ withequaloppositesides,and
selectingtheaxes $H$ , $M$ , $H^{'}$ , $M^{'}$ astheperpendicularstothequadrilateral’ssidesat $F_{h}$ , $F_{m}$ , $F_{m^{'}}$ ,and $F_{h^{'}}$ ,respectively(Figure 1).

Figure 1: AxisconfigurationofaBennettlinkage

Inordertocomputethelinkage’sDenavit-Hartenbergparameters,weassumethatthe $t$ -Bennettlinkageat $s_{0} = \infty$ isalignedfor $t_{0} = \infty$ .Thisisnolossofgeneralityasitcanbeachievedviare-parametrizationsoftype(8).Moreover,weassumethattheaxesinthisconfigurationintersectthefirstcoordinateaxisperpendicularlyandthattheaxisofhalf-turnsymmetryisthethirdcoordinateaxis.Thisentailsaslightalterationoftheconstructionfromabove.Weassigncoordinates

F_{h} = (h_{1}, 0, 0), F_{m} = (m_{1}, 0, 0), F_{h^{'}} = (- h_{1}, 0, 0), F_{m^{'}} = (- m_{1}, 0, 0)

tothecommonnormalfeetand

D_{h} = (0, h_{2}, h_{3}), D_{m} = (0, m_{2}, m_{3}), D_{h^{'}} = (0, - h_{2}, h_{3}), D_{m^{'}} = (0, - m_{2}, m_{3}) .

tothecorrespondingaxisdirections.Bythischoice,weensureequaloppositedistanceandanglesbutnotequalityofBennettratios.³³3AnimportantcharacteristicofaBennettmechanismisitsBennettratio,theratiobetweensineofangleanddistanceoftwoconsecutiveaxes,whichisindependentofthechosenpairofconsecutiveaxes[20, Equation (11.69)].Astraightforwardcomputationyieldsthatthiscanbesatisfiedby

m_{1} = \frac{h_{1} h_{3} m_{2}}{h_{2} m_{3}} or m_{1} = \frac{h_{1} h_{2} m_{3}}{h_{3} m_{2}} .

Sincebothexpressionsareequaluptointerchanging $h_{2}$ with $h_{3}$ and $m_{2}$ with $m_{3}$ wecanuseeitherofthem.Thefollowingcomputationsuse $m_{1} = h_{1} h_{3} m_{2} / (h_{2} m_{3})$ .Now,wecomputethePlückercoordinates,viewedasdualquaternions,oftheaxesinthezeroconfigurationas

$Vect (h)$	$= D_{h} + ε (F_{h} \times D_{h}) = h_{2} j + h_{3} k - h_{1} ε (h_{3} j - h_{2} k),$	(27)
$Vect (m)$	$= D_{m} + ε (F_{m} \times D_{m}) = m_{2} j + m_{3} k - \frac{}{h_{1} h_{3} m_{2}} h_{2} m_{3} ε (m_{3} j - m_{2} k),$
$Vect (h^{'})$	$= D_{h^{'}} + ε (F_{h^{'}} \times D_{h^{'}}) = - h_{2} j + h_{3} k + h_{1} ε (h_{3} j + h_{2} k),$
$Vect (m^{'})$	$= D_{m^{'}} + ε (F_{m^{'}} \times D_{m^{'}}) = - m_{2} j + m_{3} k + \frac{h_{1} h_{3} m_{2}}{h_{2} m_{3}} ε (m_{3} j + m_{2} k) .$

Here,weidentifiedintheusualwayvectorsin $R^{3}$ withvectorialquaternions.Thecoefficients $h$ , $m$ , $h^{'}$ ,and $m^{'}$ inthefactors $t - h$ , $t - m$ , $t - h^{'}$ , $t - m^{'}$ ofthesoughtmotionpolynomialarelinearcombinationsof $1$ and $Vect (h)$ , $Vect (m)$ , $Vect (h^{'})$ ,and $Vect (m^{'})$ ,respectively.Thecoefficientscannotbechosenarbitrarilybutaresubjecttotheclosurecondition $(t - h) (t - m) = (t - m^{'}) (t - h^{'})$ .Thisisensuredbyhaving

	$h$	$= μ - ν m_{2} (j + \frac{h_{3}}{h_{2}} k) + \frac{ν h_{1} m_{2}}{h_{2}} ε (h_{3} j - h_{2} k),$
	$m$	$= μ + ν (m_{2} j + m_{3} k) - \frac{ν h_{1} h_{3} m_{2}}{h_{2} m_{3}} ε (m_{3} j - m_{2} k),$
	$h^{'}$	$= μ + ν m_{2} (j - \frac{h_{3}}{h_{2}} k) - \frac{ν h_{1} m_{2}}{h_{2}} ε (h_{3} j + h_{2} k),$
	$m^{'}$	$= μ - ν (m_{2} j - m_{3} k) + \frac{ν h_{1} h_{3} m_{2}}{h_{2} m_{3}} ε (m_{3} j + m_{2} k)$

withparameters $ν$ , $μ \in R$ .Sofar,wehavefollowedStep 1ofSection 3andcomputed,infullgeneralitybutataspecialconfiguration,theaxesandcorrespondingdualquaternionsthatmovewithparameter $t$ .ForSteps 2and3wemakethegeneralansatz

ℓ = ℓ_{p} + ε ℓ_{d}, n = n_{p} + ε n_{d}, ℓ^{'} = ℓ_{p}^{'} + ε ℓ_{d}^{'}, n^{'} = n_{p}^{'} + ε n_{d}^{'}

with $ℓ_{p}$ , $ℓ_{d}$ , $n_{p}$ , $n_{d}$ , $ℓ_{p}^{'}$ , $ℓ_{d}^{'}$ , $n_{p}^{'}$ , $n_{d}^{'} \in H$ .Step 2givestheprimalparts $ℓ_{p}$ , $ℓ_{p}^{'}$ ,and $n^{'}$ intermsoftheindeterminedcoefficientsof $n_{p} = n_{0} + n_{1} i + n_{2} j + n_{3} k$ :

$ℓ_{p}$	$= \frac{1}{h_{2} m_{3} + h_{3} m_{2}} (h_{2} (m_{3} n_{0} - 2 m_{2} n_{1}) + h_{3} m_{2} n_{0} + n_{1} (h_{2} m_{3} - h_{3} m_{2}) i$	(28)
	$+ (h_{2} (m_{3} n_{2} - 2 m_{2} n_{3}) - h_{3} m_{2} n_{2}) j - n_{3} (h_{2} m_{3} + h_{3} m_{2}) k),$
$ℓ_{p}^{'}$	$= \frac{1}{h_{2} m_{3} + h_{3} m_{2}} (h_{2} (m_{3} n_{0} - 2 m_{2} n_{1}) + h_{3} m_{2} n_{0} + n_{1} (h_{2} m_{3} - h_{3} m_{2}) i$
	$+ (h_{2} (m_{3} n_{2} - 2 m_{2} n_{3}) - h_{3} m_{2} n_{2}) j + n_{3} (h_{2} m_{3} + h_{3} m_{2}) k),$
$n_{p}^{'}$	$= n_{0} + n_{1} i + n_{2} j - n_{3} k .$

Thisensuresthattheprimalpartsonbothsidesof

(t - h) (s - ℓ) (t - m) (s - n) = (s - n^{'}) (t - m^{'}) (s - ℓ^{'}) (t - h^{'})

agree.Equalityoftherespectivedualpartstogetherwiththemotionpolynomialconditionboilsdowntoasystemoflinearequations(Step 3)fortherealcoefficientsof $ℓ_{d}$ , $ℓ_{d}^{'}$ , $n_{d}$ ,and $n_{d}^{'}$ whichwesolvewithacomputeralgebrasystem.Thereis,indeed,auniquesolutionwhencewehaveprovidedthemissingpieceintheproofofTheorem 3.7.Thesolutionsarejustabittoolongtobedisplayedhere.Therefore,andalsohavinginmindforthcomingcomputations,westriveforfurthersimplifications.Byarationalre-parametrizationwecanachievethattherevoluteaxis $Vect (n)$ isperpendiculartothefirstcoordinateaxisinthezeroconfiguration,at $s_{0} = \infty$ .Thishavingdone,weseethatnecessarily $n_{1} = 0$ .Withthisadmissiblesimplification,thesolutionsforthedualpartsare

ℓ_{d} = \frac{h_{1} n_{2}}{Δ} (n_{3} (h_{2} m_{3} + h_{3} m_{2}) (h_{2} m_{2} n_{3} - h_{2} m_{3} n_{2} + h_{3} m_{2} n_{2} + h_{3} m_{3} n_{3}) j - (h_{2} m_{2} n_{3} - h_{2} m_{3} n_{2} + h_{3} m_{2} n_{2} + h_{3} m_{3} n_{3}) (2 h_{2} m_{2} n_{3} - h_{2} m_{3} n_{2} + h_{3} m_{2} n_{2}) k),

n_{d} = \frac{h_{1}}{Δ} (- n_{3} (2 h_{2} m_{2} n_{3} - h_{2} m_{3} n_{2} + h_{3} m_{2} n_{2}) (h_{2} m_{2} n_{3} - h_{2} m_{3} n_{2} + h_{3} m_{2} n_{2} + h_{3} m_{3} n_{3}) j + n_{2} (2 h_{2} m_{2} n_{3} - h_{2} m_{3} n_{2} + h_{3} m_{2} n_{2}) (h_{2} m_{2} n_{3} - h_{2} m_{3} n_{2} + h_{3} m_{2} n_{2} + h_{3} m_{3} n_{3}) k),

ℓ_{d}^{'} = \frac{h_{1} n_{2}}{Δ} (n_{3} (h_{2} m_{3} + h_{3} m_{2}) (h_{2} m_{2} n_{3} - h_{2} m_{3} n_{2} + h_{3} m_{2} n_{2} + h_{3} m_{3} n_{3}) j + (h_{2} m_{2} n_{3} - h_{2} m_{3} n_{2} + h_{3} m_{2} n_{2} + h_{3} m_{3} n_{3}) (2 h_{2} m_{2} n_{3} - h_{2} m_{3} n_{2} + h_{3} m_{2} n_{2}) k)

n_{d}^{'} = \frac{h_{1}}{Δ} (- n_{3} (2 h_{2} m_{2} n_{3} - h_{2} m_{3} n_{2} + h_{3} m_{2} n_{2}) (h_{2} m_{2} n_{3} - h_{2} m_{3} n_{2} + h_{3} m_{2} n_{2} + h_{3} m_{3} n_{3}) j - n_{2} (2 h_{2} m_{2} n_{3} - h_{2} m_{3} n_{2} + h_{3} m_{2} n_{2}) (h_{2} m_{2} n_{3} - h_{2} m_{3} n_{2} + h_{3} m_{2} n_{2} + h_{3} m_{3} n_{3}) k)

where

Δ = h_{2} m_{3} ((h_{3} m_{2} - h_{2} m_{3}) (n_{2}^{2} - n_{3}^{2}) + 2 (h_{2} m_{2} + h_{3} m_{3}) n_{2} n_{3}) .

Buthaving $n_{1} = 0$ hasfurtherconsequences:

Aglanceat(28)immediatelyconfirmsthatallcoefficientsof $i$ vanishforallrevoluteaxesinthezeroconfiguration.Therefore,allrevoluteaxesinthezeroconfigurationareperpendiculartothefirstcoordinateaxis.
Itcanreadilybeverifiedthattheintersectionconditions

$Vect (ℓ) i - i {V e c t (ℓ)}^{*} = Vect (n) i - i {V e c t (n)}^{*} = Vect (ℓ^{'}) i - i {Vect (ℓ^{'})}^{*} = Vect (n^{'}) i - i {Vect (n^{'})}^{*} = 0.$

betweenthefirstcoordinateaxis(withPlückercoordinates $i$ )andallmechanismaxesthatmovewithparameter $s$ inthezeroconfigurationaresatisfied.

Thismeansthatinthezeroconfigurationallrevoluteaxesintersectthefirstcoordinateaxisperpendicularly.Weinferthatnotonlythe $t$ -Bennettmechanismbutalsothe $s$ -Bennettmechanismalignsandbothsharethecommonperpendicularoftheiraxis.SinceeachBennettmechanismhastwoalignedconfigurationsandthereisnothingspecialaboutourzeroconfiguration,wecansay:

Theorem 4.1.

Themulti-Bennett8Rmechanismhasfouralignedconfigurationsinwhichalleightrevoluteaxesshareacommonperpendicularline.

ThefouralignedconfigurationsofanexamplecanbeseeninthecornersofFigure 2.Fromtherepresentations(27)and(28)oftheaxes’Plückercoordinates,itisstraightforwardtocomputethemechanism’sDenavit-Hartenbergparameters.Theinformationgivenin[24, Section 2.1.2]issufficientforthatpurposebutmoreexplicitformulasarealsoavailable,forexamplein[4].Usingcomputeralgebra,itiseasytoverify

Theorem 4.2.

Theoffsetsofamulti-Bennett8Rmechanismareallzero.Oppositedistancesaswellasoppositeanglesareequal.

Remarkably,thefourdistancesarerationalexpressionsintheinputparameters,nosquarerootsappear:

	$d_{1}$	$= \frac{1}{Φ} (h_{1} (m_{2} n_{3} - m_{3} n_{2}) (2 h_{2}^{2} m_{2} n_{3} - h_{2}^{2} m_{3} n_{2} + h_{2} h_{3} m_{2} n_{2} + h_{2} h_{3} m_{3} n_{3} + h_{3}^{2} m_{2} n_{3})),$
	$d_{2}$	$= \frac{1}{Φ} (h_{1} (m_{2} n_{3} - m_{3} n_{2}) (h_{2} n_{2} - h_{3} n_{3}) (h_{2} m_{3} - h_{3} m_{2})),$
	$d_{3}$	$= \frac{1}{Φ} (h_{1} (m_{2} n_{2} + m_{3} n_{3}) (h_{2} n_{3} + h_{3} n_{2}) (h_{2} m_{3} - h_{3} m_{2})),$
	$d_{4}$	$= \frac{1}{Φ} (h_{1} (h_{2} n_{3} + h_{3} n_{2}) (2 h_{2} m_{2}^{2} n_{3} - h_{2} m_{2} m_{3} n_{2} + h_{2} m_{3}^{2} n_{3} + h_{3} m_{2}^{2} n_{2} + h_{3} m_{2} m_{3} n_{3})),$

where

Φ = h_{2} m_{3} ((h_{3} m_{2} - h_{2} m_{3}) (n_{2}^{2} - n_{3}^{2}) + 2 n_{2} n_{3} (h_{2} m_{2} + h_{3} m_{3})) .

Thesquaredcosinesofthecorrespondinganglesare

	${cos}^{2} α_{1}$	$= \frac{(m_{2} n_{2} + m_{3} n_{3})^{2}}{(n_{2}^{2} + n_{3}^{2}) (m_{2}^{2} + m_{3}^{2})},$
	${cos}^{2} α_{2}$	$= \frac{(2 h_{2} m_{2}^{2} n_{3} - h_{2} m_{2} m_{3} n_{2} + h_{2} m_{3}^{2} n_{3} + h_{3} m_{2}^{2} n_{2} + h_{3} m_{2} m_{3} n_{3})^{2}}{(m_{2}^{2} + m_{3}^{2}) Ψ}$
	${cos}^{2} α_{3}$	$= \frac{(2 h_{2}^{2} m_{2} n_{3} - h_{2}^{2} m_{3} n_{2} + h_{2} h_{3} m_{2} n_{2} + h_{2} h_{3} m_{3} n_{3} + h_{3}^{2} m_{2} n_{3})^{2}}{(h_{2}^{2} + h_{3}^{2}) Ψ},$
	${cos}^{2} α_{4}$	$= \frac{(h_{2} n_{2} - h_{3} n_{3})^{2}}{(h_{2}^{2} + h_{3}^{2}) (n_{2}^{2} + n_{3}^{2})}$

where

Ψ = 4 h_{2}^{2} m_{2} n_{3} (m_{2} n_{3} - m_{3} n_{2}) + (h_{2}^{2} m_{3}^{2} + h_{3}^{2} m_{2}^{2}) (n_{2}^{2} + n_{3}^{2}) + 2 h_{2} h_{3} m_{2} (2 m_{2} n_{2} n_{3} - m_{3} (n_{2}^{2} - n_{3}^{2})) .

WeconjecturethatthenecessaryconditionsofTheorem 4.2onthemechanism’sDenavit-Hartenbergparametersarenotsufficienttocharacterizeamulti-Bennett8Rmechanism.

4.2 BennettSub-Mechanisms

Wehavealreadymentionedthatforfixed $s = s_{0}$ theaxes $H (s, t)$ , $M (s, t)$ , $H^{'} (s, t)$ ,and $M^{'} (s, t)$ totherespectivefactors $t - h$ , $t - m$ , $t - h^{'}$ ,and $t - m^{'}$ formaBennettmechanism.Thesameistrueforfixed $t = t_{0}$ andtheaxes $L (s, t)$ , $N (s, t)$ , $L^{'} (s, t)$ , $N^{'} (s, t)$ totherespectivefactors $s - ℓ$ , $s - n$ , $s - ℓ^{'}$ , $s - n^{'}$ .WerefertotherespectiveBennettmechanismsas $t$ -Bennettmechanismat $s_{0}$ andas $s$ -Bennettmechanismat $t_{0}$ .The $t$ -Bennettmechanismalignsforpreciselytwoparametervalues $t^{'}$ , $t^{''}$ .Bymeansof(26)itcanreadilybeverifiedthataligningof $t$ -Bennettlinkagehappensat

t^{'} = \infty, t^{''} = μ

(29)

whilean $s$ -Bennettlinkagealignsat

s^{'} = \infty, s^{''} = n_{0} .

(30)

ThemostremarkablethingaboutEquations (29)and(30)isthatthat $t^{'}$ and $t^{''}$ donotdependon $s$ and $s^{'}$ , $s^{''}$ donotdependon $t$ .Abstractingfromourspecialgeometricdescriptiontothegeneralcase,wecanthusstate:

Theorem 4.3.

Inamulti-Bennett8Rmechanism,the $t$ -Bennettsub-mechanismsalignpreciselyfortwofixedparametervalues $t^{'}$ , $t^{''}$ andthe $s$ -Bennettsub-mechanismsalignpreciselyfortwofixedparametervalues $s^{'}$ , $s^{''}$ .Thepoints $(t^{'}, s^{'})$ , $(t^{'}, s^{''})$ , $(t^{''}, s^{'})$ ,and $(t^{''}, s^{''})$ intheconfigurationspacecorrespondtothefouralignedstatesofthecompletemechanism,c.f.Theorem 4.1.

Theorem 4.3isillustratedinFigure 2.There,theeightlinksarevisualizedbycylindersaroundthecommonnormalsofconsecutivejointaxes.Thisisclearlyvisibleinthefourtotallyalignedconfigurationsinthecorners.Themotionsbetweenneighbouringcornershave $t = t^{'}$ , $t = t^{''}$ , $s = s^{'}$ ,or $s = s^{''}$ .Figure 2alsoillustratesthemulti-Bennett’sconfigurationspace,atorus,andthefourcurves,meridianandlateralcirclesonthetorus,alongwhichBennettsub-mechanismsalign.

{tblr}

colspec=X[c]X[c]X[c]X[c],colsep=1mm, Configurationspaceofmulti-Bennett8Rmechanismandanimatedtransitionsbetweenfourtotallyalignedconfigurationsinthecorners;pointscorrespondtodepictedconfigurations. &&&
&\SetCell[r=2,c=2]c&&
&&&
&&&

Figure 2: Configurationspaceofmulti-Bennett8Rmechanismandanimatedtransitionsbetweenfourtotallyalignedconfigurationsinthecorners;pointscorrespondtodepictedconfigurations.

Asexpected,theBennettratioisnotconstantwithinthefamilyof $t$ -Bennettmechanismbutdependson $s$ (andviceversafor $s$ -Bennettmechanisms).However,anoteworthypropertyis:

Theorem 4.4.

TheBennettratiowithinthefamilyof $t$ -Bennettlinkagesisarationalfunctionofdegreefourin $s$ andviceversaforthe $s$ -Bennettlinkages.

Proof.

Adirectcomputationusingcomputeralgebrayieldsthevalue

τ = \frac{h_{2} m_{2} ((h_{2} m_{3} + h_{3} m_{2})^{2} (s^{4} - 4 n_{0} s^{3}) + c_{2} s^{2} + c_{1} s + c_{0})}{h_{1} \sqrt{h_{2}^{2} + h_{3}^{2}} \sqrt{m_{2}^{2} + m_{3}^{2}} (s^{2} - 2 n_{0} s + n_{0}^{2} + n_{2}^{2} + n_{3}^{2}) ((h_{3}^{2} m_{2}^{2} - h_{2}^{2} m_{3}^{2}) s (s - 2 n_{0}) + d_{0})}

forthe $t$ -Bennettratiowhere

c_{2} = 6 (h_{2} m_{3} + h_{3} m_{2})^{2} n_{0}^{2} + (2 h_{2}^{2} m_{3}^{2} + 2 h_{3}^{2} m_{2}^{2} + 4 h_{3}^{2} m_{3}^{2}) n_{2}^{2} + 2 (2 h_{2}^{2} m_{2}^{2} + h_{2}^{2} m_{3}^{2} + h_{3}^{2} m_{2}^{2}) n_{3}^{2} - 4 n_{2} n_{3} (h_{2} m_{2} - h_{3} m_{3}) (h_{2} m_{3} - h_{3} m_{2}),

c_{1} = - 4 n_{0}^{3} (h_{2} m_{3} + h_{3} m_{2})^{2} - 4 n_{0} n_{2}^{2} (h_{2}^{2} m_{3}^{2} + h_{3}^{2} m_{2}^{2} + 2 h_{3}^{2} m_{3}^{2}) + 8 n_{0} n_{2} n_{3} (h_{2} m_{2} - h_{3} m_{3}) (h_{2} m_{3} - h_{3} m_{2}) - 4 n_{0} n_{3}^{2} (2 h_{2}^{2} m_{2}^{2} + h_{2}^{2} m_{3}^{2} + h_{3}^{2} m_{2}^{2}),

c_{0} = n_{0}^{4} (h_{2} m_{3} + h_{3} m_{2})^{2} + 2 n_{0}^{2} n_{2}^{2} (h_{2}^{2} m_{3}^{2} + h_{3}^{2} m_{2}^{2} + 2 h_{3}^{2} m_{3}^{2}) + 4 n_{0}^{2} n_{2} n_{3} (h_{2} m_{2} - h_{3} m_{3}) (h_{3} m_{2} - h_{2} m_{3}) + 2 n_{0}^{2} n_{3}^{2} (2 h_{2}^{2} m_{2}^{2} + h_{2}^{2} m_{3}^{2} + h_{3}^{2} m_{2}^{2}) + n_{2}^{4} (h_{2} m_{3} - h_{3} m_{2})^{2} + 4 n_{2}^{3} n_{3} (h_{2} m_{2} + h_{3} m_{3}) (h_{3} m_{2} - h_{2} m_{3}) + 2 n_{2} n_{3}^{2} (2 h_{2}^{2} m_{2}^{2} - h_{2}^{2} m_{3}^{2} + 6 h_{2} h_{3} m_{2} m_{3} - h_{3}^{2} m_{2}^{2} + 2 h_{3}^{2} m_{3}^{2}) + 4 n_{2} n_{3}^{3} (h_{2} m_{2} + h_{3} m_{3}) (h_{2} m_{3} - h_{3} m_{2}) + n_{3}^{4} (h_{2} m_{3} - h_{3} m_{2})^{2},

and

d_{0} = - 4 h_{2} n_{2} n_{3} (2 h_{2} m_{2} m_{3} - h_{3} m_{2}^{2} + h_{3} m_{3}^{2}) + n_{0}^{2} (h_{3}^{2} m_{2}^{2} - h_{2}^{2} m_{3}^{2}) + n_{2}^{2} (3 h_{2} m_{3} - h_{3} m_{2}) (h_{2} m_{3} - h_{3} m_{2}) + n_{3}^{2} (4 h_{2}^{2} m_{2}^{2} - h_{2}^{2} m_{3}^{2} + 4 h_{2} h_{3} m_{2} m_{3} + h_{3}^{2} m_{2}^{2}) .

Asimilarformulacanbederivedforthe $s$ -Bennettratio.∎

5 ConclusionandFutureResearch

Wepresentedthefirstexampleofamechanismconstructedfromthefactorizationofbivariatemotionpolynomialsanddescribedsomeofitsfundamentalproperties.Ofcourse,openquestionsremain.Thesimpleconditionsonthemechanism’sDHparameterswhichwedescribeinTheorem 4.2arenecessarybut,sowebelieve,notsufficient.Itwouldbedesirabletoaugmentthemwithfurtherconditionstoobtainasetofsufficientconditions.Wefurtherbelievethattheconfigurationspaceparameterizedbytheunderlyingmotionpolynomial $C (s, t)$ isonlyapartofthemechanism’scompleteconfigurationspace.Obtainingaclearerpictureofpossibleassemblymodesorbifurcationsofthemotioniscertainlyaworthytopicoffutureresearch.Theconfigurationspacecomponentdescribedby $C (s, t)$ hasmanyattractivefeaturesforpotentialapplications:Ithasarationalparametrizationwithlowdegreeparameterlines.Themotionalongaparameterlineisthewell-understoodcouplermotionofaBennettlinkage.Simpleparametrizationbutalsotheunusualseparationintojointsthatonlymovewithparameter $t$ andjointsthatonlymovewithparameter $s$ isexpectedtobebeneficialforthecontrolofamulti-Bennett8Rmechanism.

References

Baker [1993] J. EddieBaker. Acomparativesurveyofthebennett-based,6-revolutekinematicloops. Mech.Mach.Theory,28(1):–96,1993. ISSN0094-114X. doi:10.1016/0094-114X(93)90049-2.
Baker [2005] J. EddieBaker. Ongeneratingaclassoffoldablesix-barspatiallinkages. ASMEJ.Mech.Design,128(2):–383,062005. ISSN1050-0472. doi:10.1115/1.2166853.
ChenandYou [2007] Y. ChenandZ. You. Spatial6RlinkagesbasedonthecombinationoftwoGoldberg5Rlinkages. Mech.Mach.Theory,42(11):–1498,2007. ISSN0094-114X. doi:10.1016/j.mechmachtheory.2006.12.008.
Fariaet al. [2019] CarlosFaria,João L.Vilaca,SérgioMonteiro,WolframErlhagen,andEstelaBicho. AutomaticDenavit-Hartenbergparameteridentificationforserialmanipulators. InIECON2019-45thAnnualConferenceoftheIEEEIndustrialElectronicsSociety,pages610–617,2019. doi:10.1109/iecon.2019.8927455.
Galletet al. [2017] M. Gallet,Ch.Koutschan,Z. Li,G. Regensburger,J. Schicho,andN. Villamizar. Planarlinkagesfollowingaprescribedmotion. Math.Comp.,86:–506,2017. doi:10.1090/mcom/3120.
Goldberg [1943] MichaelGoldberg. Newfive-barandsix-barlinkagesinthreedimensions. TransactionsoftheAMS,65:–661,1943.
Guerreiroet al. [2021] TiagoGuerreiro,ZijiaLi,andJosefSchicho. Classificationofhighermobilitylinkages. arXiv:2103.04799,2021.
Hegedüset al. [2013] GáborHegedüs,JosefSchicho,andHans-PeterSchröcker. FactorizationofrationalcurvesintheStudyquadricandrevolutelinkages. Mech.Mach.Theory,69(1):–152,2013. doi:10.1016/j.mechmachtheory.2013.05.010.
Hegedüset al. [2015] GáborHegedüs,JosefSchicho,andHans-PeterSchröcker. Four-posesynthesisofangle-symmetric6Rlinkages. ASMEJ.MechanismsRobotics,7(4),2015. doi:10.1115/1.4029186.
JanovskáandOpfer [2008] DrahoslavaJanovskáandGerhardOpfer. Linearequationsinquaternionicvariables. Mitt.Math.Ges.Hamburg,27(15):–234,2008.
KongandJin [2016] XianwenKongandYanJin. Typesynthesisof3-DOFmulti-modetranslational/sphericalparallelmechanismswithlockablejoints. Mech.MachineTheory,96:–333,2016. doi:10.1016/j.mechmachtheory.2015.04.019.
KongandPfurner [2015] XianwenKongandMartinPfurner. Typesynthesisandreconfigurationanalysisofaclassofvariable-DOFsingle-loopmechanisms. Mech.Mach.Theory,85:–128,2015. doi:10.1016/j.mechmachtheory.2014.10.011.
LercherandSchröcker [2022] JohannaLercherandHans-PeterSchröcker. Amultiplicationtechniqueforthefactorizationofbivariatequaternionicpolynomials. Adv.Appl.CliffordAlgebras,32(1),2022. doi:10.1007/s00006-021-01194-9.
Lercheret al. [2022a] JohannaLercher,Daniel F.Scharler,andHans-PeterSchröcker. Aremarkable8R-mechanism. InW. HolderbaumandJ. M.Selig,editors,2ndIMAConferenceonMathematicsofRobotics.IMA2020.,number 21inSpringerProceedingsinAdvancedRobotics,pages107–114.Springer,Cham,2022a. doi:10.1007/978-3-030-91352-6˙11.
Lercheret al. [2022b] JohannaLercher,Daniel F.Scharler,Hans-PeterSchröcker,andJohannesSiegele. Factorizationofquaternionicpolynomialsofbi-degree $(n, 1)$ . Beitr.AlgebraGeom.,2022b. doi:10.1007/s13366-022-00629-z.
Liet al. [2020] Z. Li,G. Nawratil,F. Rist,andM. Hensel. Invertibleparadoxicloopstructuresfortransformabledesign. ComputerGraphicsForum,39(2):–275,2020. doi:10.1111/cgf.13928.
Liet al. [2018] ZijiaLi,JosefSchicho,andHans-PeterSchröcker. Kempe’suniversalitytheoremforrationalspacecurves. Found.Comput.Math.,18(2):–536,2018. doi:10.1007/s10208-017-9348-x.
Liuet al. [2021a] K. Liu,J. Yu,andX. Kong. Structuresynthesisandreconfigurationanalysisofvariable-degree-of-freedomsingle-loopmechanismswithprismaticjointsusingdualquaternions. J.Mech.Robot,14(2),2021a. ISSN1942–4310. doi:10.1115/1.4052338.
Liuet al. [2021b] KaiLiu,JingjunYu,andXianwenKong. Synthesisofmulti-modesingle-loopBennett-basedmechanismsusingfactorizationofmotionpolynomials. Mech.Mach.Theory,155:,2021b. doi:10.1016/j.mechmachtheory.2020.104110.
McCarthyandSoh [2011] J. MichaelMcCarthyandGim SongSoh,editors. GeometricDesignofLinkages. Springer,2edition,2011.
Odehnalet al. [2020] BorisOdehnal,HellmuthStachel,andGeorgGlaeser. TheUniverseofQuadrics. Springer-Verlag,Berlin,Heidelberg,2020. ISBN978-3-662-61052-7. doi:10.1007/978-3-662-61053-4.
Pfurner [2018] MartinPfurner. Synthesisandmotionanalysisofasingle-loop $8 R$ -chain. InProceedingsofthe4thIEEE/IFToMMInternationalConferenceonReconfigurableMechanisms&Robots(ReMAR),Delft,TheNetherlands,June2018.
Pfurneret al. [2014] MartinPfurner,XianwenKong,andChintienHuang. Completekinematicanalysisofsingle-loopmultiple-mode7-linkmechanismsbasedonBennettandoverconstrainedRPRPmechanisms. Mech.Mach.Theory,73:–129,2014. doi:10.1016/j.mechmachtheory.2013.10.012.
PottmannandWallner [2010] HelmutPottmannandJohannesWallner. ComputationalLineGeometry. MathematicsandVisualization.Springer,2edition,2010.
SkopenkovandKrasauskas [2019] MikhailSkopenkovandRimvydasKrasauskas. Surfacescontainingtwocirclesthrougheachpoint. Math.Ann.,373:–1327,2019. doi:10.1007/s00208-018-1739-z.
Waldron [1968] K.J.Waldron. Hybridoverconstrainedlinkages. JournalofMechanisms,3(2):–78,1968. ISSN0022-2569. doi:10.1016/0022-2569(68)90016-5.
Wohlhart [1996] K. Wohlhart. Kinematotropiclinkages. InJ. LenarčičandV. Parenti-Castelli,editors,RecentAdvancesinRobotKinematics,pages359–368.Springer,Dordrecht,1996. doi:10.1007/978-94-009-1718-7˙36.
Wohlhart [1991] KarlWohlhart. MergingtwogeneralGoldberg5Rlinkagestoobtainanew6Rspacemechanism. Mech.Mach.Theory,26:–668,1991.