In thi[P2(US)] re[P2(US)]earch, Artificial Ne[P2(US)]ral Network[P2(US)] [P2(US)]ANN[P2(US)][P2(US)] have been [P2(US)][P2(US)]ed a[P2(US)] a [P2(US)]owerf[P2(US)]l tool to [P2(US)]olve the inver[P2(US)]e kinematic eq[P2(US)]ation[P2(US)] of a [P2(US)]arallel robot. For thi[P2(US)] [P2(US)][P2(US)]r[P2(US)]o[P2(US)]e, we have develo[P2(US)]ed the kinematic eq[P2(US)]ation[P2(US)] of a Trice[P2(US)]t [P2(US)]arallel kinematic mechani[P2(US)]m with two rotational and one tran[P2(US)]lational degree[P2(US)] of freedom [P2(US)]DoF[P2(US)]. [P2(US)][P2(US)]ing the analytical method, the inver[P2(US)]e kinematic eq[P2(US)]ation[P2(US)] are [P2(US)]olved for [P2(US)][P2(US)]ecific trajectory, and [P2(US)][P2(US)]ed a[P2(US)] in[P2(US)][P2(US)]t[P2(US)] for the a[P2(US)][P2(US)]lied ANN[P2(US)]. The re[P2(US)][P2(US)]lt[P2(US)] of both a[P2(US)][P2(US)]lied network[P2(US)] [P2(US)]M[P2(US)]lti-Layer [P2(US)]erce[P2(US)]tron and Redial Ba[P2(US)]i[P2(US)] F[P2(US)]nction[P2(US)] [P2(US)]ati[P2(US)]fied the req[P2(US)]ired [P2(US)]erformance in [P2(US)]olving com[P2(US)]lex inver[P2(US)]e kinematic[P2(US)] with [P2(US)]ro[P2(US)]er acc[P2(US)]racy and [P2(US)][P2(US)]eed.

A novel [P2(US)][P2(US)]atial [P2(US)]arallel mani[P2(US)][P2(US)]lator de[P2(US)]igned to a[P2(US)][P2(US)]emble diagno[P2(US)]tic in[P2(US)]tr[P2(US)]ment[P2(US)] in [P2(US)]G-III i[P2(US)] introd[P2(US)]ced in thi[P2(US)] [P2(US)]a[P2(US)]er. Fir[P2(US)]tly, re[P2(US)]orting to [P2(US)]crew theory, mobility analy[P2(US)]i[P2(US)] i[P2(US)] [P2(US)]re[P2(US)]ented for thi[P2(US)] mani[P2(US)][P2(US)]lator. Then, the inver[P2(US)]e kinematic[P2(US)] [P2(US)]roblem i[P2(US)] determined by the method of R[P2(US)]Y tran[P2(US)]formation with the [P2(US)]ing[P2(US)]larity analyzed. A[P2(US)] a key i[P2(US)][P2(US)][P2(US)]e in [P2(US)]arallel mani[P2(US)][P2(US)]lator[P2(US)], it i[P2(US)] more di c[P2(US)]lt to [P2(US)]olve the forward kinematic[P2(US)] [P2(US)]roblem, [P2(US)]ince it i[P2(US)] highly nonlinear and co[P2(US)][P2(US)]led. In thi[P2(US)] work, three di erent a[P2(US)][P2(US)]roache[P2(US)] are [P2(US)]re[P2(US)]ented to deal with thi[P2(US)] i[P2(US)][P2(US)][P2(US)]e, namely, the back [P2(US)]ro[P2(US)]agation ne[P2(US)]ral network, the [P2(US)]im[P2(US)]li ed ant colony o[P2(US)]timization, and the [P2(US)]ro[P2(US)]o[P2(US)]ed im[P2(US)]roved Newton iterative method. [P2(US)]im[P2(US)]lation of each a[P2(US)][P2(US)]roach i[P2(US)] cond[P2(US)]cted, and their merit[P2(US)] and demerit[P2(US)] are com[P2(US)]ared in detail. It i[P2(US)] concl[P2(US)]ded that the im[P2(US)]roved Newton iterative method, which can [P2(US)]rovide good initial iteration val[P2(US)]e[P2(US)], [P2(US)]how[P2(US)] the be[P2(US)]t [P2(US)]erformance in e[P2(US)]timation of the nonlinear forward kinematic ma[P2(US)][P2(US)]ing of the con[P2(US)]idered [P2(US)]arallel mani[P2(US)][P2(US)]lator.

In thi[P2(US)] re[P2(US)]earch work, a novel [P2(US)]arallel mani[P2(US)][P2(US)]lator with high [P2(US)]o[P2(US)]itioning and orienting rate i[P2(US)] introd[P2(US)]ced. Thi[P2(US)] mechani[P2(US)]m ha[P2(US)] two rotational and one tran[P2(US)]lational degree of freedom. Kinematic[P2(US)] and Jacobian analy[P2(US)]i[P2(US)] are inve[P2(US)]tigated.Moreover, work[P2(US)][P2(US)]ace analy[P2(US)]i[P2(US)] and o[P2(US)]timization ha[P2(US)] been [P2(US)]erformed by [P2(US)][P2(US)]ing genetic algorithm toolbox in Matlab [P2(US)]oftware. Beca[P2(US)][P2(US)]e of red[P2(US)]cing moving element[P2(US)], it i[P2(US)] ex[P2(US)]ected m[P2(US)]ch more better dynamic [P2(US)]erformance with re[P2(US)][P2(US)]ect to other co[P2(US)]nter[P2(US)]art mechani[P2(US)]m[P2(US)] with the [P2(US)]ame degree[P2(US)] of freedom. In addition, [P2(US)][P2(US)]ing co[P2(US)][P2(US)]le of cylindrical and revol[P2(US)]te joint[P2(US)], increa[P2(US)]ed mechani[P2(US)]m ability re[P2(US)][P2(US)]lted to have more extended work[P2(US)][P2(US)]ace.

The forward [P2(US)]o[P2(US)]ition kinematic[P2(US)] [P2(US)]F[P2(US)]K[P2(US)] of a [P2(US)]arallel mani[P2(US)][P2(US)]lator with new architect[P2(US)]re [P2(US)][P2(US)][P2(US)][P2(US)]o[P2(US)]ed to be [P2(US)][P2(US)]ed a[P2(US)] a moving mechani[P2(US)]m in a flight [P2(US)]im[P2(US)]lator [P2(US)]roject i[P2(US)] di[P2(US)]c[P2(US)][P2(US)][P2(US)]ed in thi[P2(US)] [P2(US)]a[P2(US)]er. The clo[P2(US)]ed form [P2(US)]ol[P2(US)]tion for the F[P2(US)]K [P2(US)]roblem of the mani[P2(US)][P2(US)]lator i[P2(US)] fir[P2(US)]t determined. It ha[P2(US)], then, been [P2(US)]hown that there are at mo[P2(US)]t [P2(US)]4 [P2(US)]ol[P2(US)]tion[P2(US)] for F[P2(US)]K [P2(US)]roblem. Thi[P2(US)] re[P2(US)][P2(US)]lt ha[P2(US)] been verified by [P2(US)][P2(US)]ing other techniq[P2(US)]e[P2(US)] [P2(US)][P2(US)]ch a[P2(US)] geometric a[P2(US)][P2(US)]roach and a n[P2(US)]merical method known a[P2(US)] [P2(US)]olynomial contin[P2(US)]ation. N[P2(US)]merical exam[P2(US)]le[P2(US)] are [P2(US)]erformed to di[P2(US)][P2(US)]lay all [P2(US)]o[P2(US)][P2(US)]ible [P2(US)]ol[P2(US)]tion[P2(US)] available for the devi[P2(US)]ed mani[P2(US)][P2(US)]lator.

Thi[P2(US)] [P2(US)]a[P2(US)]er [P2(US)]how[P2(US)] the coordinate[P2(US)] infl[P2(US)]ence on [P2(US)]ing[P2(US)]larity of a three degree-of-freedom [P2(US)]tr[P2(US)]ct[P2(US)]re, namely, three-[P2(US)]niver[P2(US)]al-[P2(US)]ri[P2(US)]matic-[P2(US)][P2(US)]herical [P2(US)]3-[P2(US)][P2(US)][P2(US)][P2(US)] [P2(US)]arallel mani[P2(US)][P2(US)]lator. Rotational coordinate[P2(US)], which are cho[P2(US)]en to define the orientation of the [P2(US)]latform, affect the [P2(US)]ing[P2(US)]larity of the mani[P2(US)][P2(US)]lator. E[P2(US)]ler [P2(US)]arameter[P2(US)], which don't have any inherent geometrical [P2(US)]ing[P2(US)]larity are [P2(US)]tilized, however they are de[P2(US)]endent coordinate[P2(US)]. Thi[P2(US)] [P2(US)]a[P2(US)]er [P2(US)]how[P2(US)] the advantage of E[P2(US)]ler [P2(US)]arameter[P2(US)] rather than E[P2(US)]ler angle[P2(US)] a[P2(US)] the rotational coordinate[P2(US)] for the mani[P2(US)][P2(US)]lator. Additionally, the real loci of [P2(US)]ing[P2(US)]larity for the mani[P2(US)][P2(US)]lator d[P2(US)]e to it[P2(US)] [P2(US)]tr[P2(US)]ct[P2(US)]re are [P2(US)]redicted.

Thi[P2(US)] [P2(US)]a[P2(US)]er [P2(US)]re[P2(US)]ent[P2(US)] a new architect[P2(US)]re de[P2(US)]ign for [P2(US)]lanar [P2(US)]arallel mani[P2(US)][P2(US)]lator[P2(US)], which hel[P2(US)][P2(US)] greatly to increa[P2(US)]e their mane[P2(US)]verability and enlarging their work[P2(US)][P2(US)]ace. In many ca[P2(US)]e[P2(US)] it i[P2(US)] very diffic[P2(US)]lt to ex[P2(US)]re[P2(US)][P2(US)] interlace of link[P2(US)] a[P2(US)] con[P2(US)]traint and it i[P2(US)] an ob[P2(US)]tacle, which doe[P2(US)]n"t let the com[P2(US)][P2(US)]tation[P2(US)] for kinematic[P2(US)], or work[P2(US)][P2(US)]ace i[P2(US)] the [P2(US)]ame a[P2(US)] in [P2(US)]ractical ca[P2(US)]e. Thi[P2(US)] new architect[P2(US)]ral de[P2(US)]ign make[P2(US)] thi[P2(US)] wi[P2(US)]h come[P2(US)] tr[P2(US)]e for com[P2(US)][P2(US)]tation and [P2(US)]ractical ca[P2(US)]e[P2(US)]. To [P2(US)]how how it work[P2(US)], a 3-DOF [P2(US)]lanar [P2(US)]arallel mani[P2(US)][P2(US)]lator i[P2(US)] [P2(US)]elected. In the mechani[P2(US)]m, the [P2(US)]ri[P2(US)]matic act[P2(US)]ator[P2(US)] are fixed to the ba[P2(US)]e which lead[P2(US)] to a red[P2(US)]ction of the moving link[P2(US)] inertia, hence make[P2(US)] it attractive, [P2(US)]artic[P2(US)]larly when high [P2(US)][P2(US)]eed[P2(US)] are req[P2(US)]ired and electric act[P2(US)]ation i[P2(US)] con[P2(US)]idered. Fir[P2(US)]t, the mechani[P2(US)]m i[P2(US)] introd[P2(US)]ced; then a new de[P2(US)]ign i[P2(US)] introd[P2(US)]ced ba[P2(US)]ed on the geometry of the mani[P2(US)][P2(US)]lator to increa[P2(US)]e flexibility and acce[P2(US)][P2(US)]ibility. Finally a work[P2(US)][P2(US)]ace analy[P2(US)]i[P2(US)] i[P2(US)] [P2(US)]erformed. O[P2(US)]timization of work[P2(US)][P2(US)]ace i[P2(US)] con[P2(US)]idered with 3 different method[P2(US)]: Monte Carlo, Exact Method [P2(US)]introd[P2(US)]ced for the fir[P2(US)]t time[P2(US)] and Global Condition Index. Then the re[P2(US)][P2(US)]lt[P2(US)] are com[P2(US)]ared.