1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2: %---------------------------------------------------------------------

3: \subsection{Differential Vector Method}

4: \label{section:differential-vector-method}

5: %---------------------------------------------------------------------

6: In this section, we perfect a sensitivity analysis method of the

7: Orthoglide, which complements the previous one. This method is

8: used to analyze the sensitivity of the position and the

9: orientation of the end-effector to dimensional and angular

10: variations, and particularly to the variations in the

11: parallelograms. Moreover, it allows us to distinguish the

12: variations which are responsible for the position errors of the

13: end-effector from the ones which are responsible for its

14: orientation errors. To develop this method, we were inspired by a

15: Huang \& al. work on a parallel kinematic machine, which is made

16: up of parallelogram joints too \cite{HUANG03}.

17: 

18: First, we express the dimensional and angular variations in

19: vectorial form. Then, a relation between the position and the

20: orientation errors of the end-effector is obtained from the

21: closed-loop kinematic equations. The expressions of the

22: orientation and the position errors of the end-effector, with

23: respect to the variations in the design parameters, are deduced

24: from this relation. Finally, we introduce two sensitivity indices

25: to assess the sensitivity of the position and the orientation of

26: the end-effector to dimensional and angular variations, and

27: particularly to the parallelism errors of the bars of the

28: parallelograms.

29: 

30: \begin{figure}[!h]

31: \begin{center}

32: \includegraphics[width=80mm]{F16_ch5_OAi3}

33: \caption{Variations in $O-A_i$ chain} \label{fig:OAi}

34: \end{center}

35: \end{figure}

36: 

37: %---------------------------------------------------------------------

38: \subsubsection{Formulation}

39: \label{section:differential-vector-method-theory}

40: %---------------------------------------------------------------------

41: The schematic drawing of the $i^{th}$ leg of the Orthoglide

42: depicted in Fig.\ref{fig:schema-cinematique-jambe} is split in

43: order to depict the variations in design parameters by a vectoriel

44: form. The closed-loop kinematic chains

45: $O-A_i-B_i-B_{ij}-C_{ij}-C_{i}-P$, $i=1,2,3$, $j=1,2$, are

46: depicted by Figs.\ref{fig:OAi}-\ref{fig:CijCiP}. ${\cal R}_i$ is

47: the coordinate frame attached to the $i^{th}$ prismatic joint.

48: ${\bf o}, {\bf a}_i, {\bf b}_i, {\bf b}_{ij}, {\bf c}_{ij}, {\bf

49: c}_{i}, {\bf p}$, are the Cartesian coordinates of points $O, A_i,

50: B_i, B_{ij}, C_{ij}, C_{i}, P$, respectively, expressed in ${\cal

51: R}_i$ and depicted in Fig.\ref{fig:schema-cinematique-jambe}.

52: 

53: According to Fig.\ref{fig:OAi},

54: \begin{equation}\label{eq:OAi}

55:   {\bf a}_i - {\bf o} = {\bf R}_i ({\bf a}_0 + \delta{\bf a}_i)

56: \end{equation}

57: 

58: \noindent where ${\bf a}_0$ is the nominal position vector of

59: $A_i$ with respect to $O$ expressed in ${\cal R}_i$, $\delta{\bf

60: a}_i$ is the positioning error of $A_i$. ${\bf R}_i$ is the

61: transformation matrix from ${\cal R}_i$ to ${\cal R}_b$. ${\bf

62: I}_3$ is the ($3\times3$) identity matrix and

63: \begin{displaymath} \left.

64: \begin{array}{l} {\bf R}_1 = {\bf I}_3 \quad ; \quad  {\bf R}_2 = \mat{ccc}{0 & 0 &

65: -1 \\ 1 & 0 & 0 \\ 0 & -1 &  0} \quad ; \quad {\bf R}_3 =

66: \mat{ccc}{0 & 1 & 0 \\ 0 & 0 & 1 \\ 1 & 0 &

67:  0}

68: \end{array}

69: \right.

70: \end{displaymath}

71: 

72: \begin{figure}[!h]

73: \begin{center}

74: \includegraphics[width=80mm]{F17_ch5_AiBi3}

75: \caption{Variations in $A_i-B_i$ chain} \label{fig:AiBi}

76: \end{center}

77: \end{figure}

78: 

79: According to Fig.\ref{fig:AiBi},

80: \begin{equation}\label{eq:AiBi} {\bf b}_i - {\bf a}_i = {\bf R}_i

81: ( \rho_i + \delta \rho_i ) {\bf e}_1 + {\bf R}_i {\delta

82: \theta}_{Ai} \times ( \rho_i + \delta \rho_i ) {\bf e}_1

83: \end{equation}

84: 

85: \noindent where $\rho_i$ is the displacement of the $i^{th}$

86: prismatic joint, $\delta \rho_i$ is its displacement error,

87: $\delta \theta_{Ai}= \mat{ccc}{\delta \theta_{Aix} & \delta

88: \theta_{Aiy} & \delta \theta_{Aiz}}^{T}$ is the angular variation

89: of its direction, and

90: \begin{displaymath}

91: \left.

92: \begin{array}{l}

93: {\bf e}_1 = \mat{c}{1 \\ 0 \\ 0} \quad ; \quad  {\bf e}_2 = \mat{c}{0 \\ 0 \\ 1} ;

94:   \quad \xi(j) = \left\{\begin{array}{cc} 1 & \textrm{if } j=1 \\ -1 & \textrm{if } j=2

95:   \end{array}\right.

96: \end{array}

97: \right.

98: \end{displaymath}

99: 

100: \begin{figure}[!h]

101: \begin{center}

102: \includegraphics[width=80mm]{F18_ch5_BiBijCij3}

103: \caption{Variations in $B_i-B_{ij}-C_{ij}$ chain}

104: \label{fig:BiBijCij}

105: \end{center}

106: \end{figure}

107: 

108: According to Fig.\ref{fig:BiBijCij},

109: \begin{eqnarray}\label{eq:BiBijCij}

110: &{\bf b}_{ij}& - {\bf b}_i = {\bf R}_i [{\bf I}_3 + {\delta

111: \theta}_{Ai} \times] \big( \xi(j) (d/2 + \delta {b}_i /2)[{\bf

112: I}_3 + \delta {\theta}_{Bi} \times] {\bf e}_2 \big)\\ &{\bf

113: c}_{ij}& - {\bf b}_{ij} = L_i {\bf w}_i + \delta L_{ij} {\bf w}_i

114: + L_i \delta {\bf w}_i

115: \end{eqnarray}

116: 

117: \noindent where $d$ is the nominal width of the parallelogram,

118: $\delta {b}_i$ is the variation in the length of link

119: $\overline{B_{i1}B_{i2}}$ and is supposed to be equally shared by

120: each side of $B_i$. $\delta \theta_{Bi}= \mat{ccc}{\delta

121: \theta_{Bix} & \delta \theta_{Biy} & \delta \theta_{Biz}}^{T}$ is

122: the orientation error of link $\overline{B_{i1}B_{i2}}$ with

123: respect to the direction of the $i^{th}$ prismatic joint, $L_i$ is

124: the length of the $i^{th}$ parallelogram, $\delta L_{ij}$ is the

125: variation in the length of link $\overline{B_{ij}C_{ij}}$, of

126: which ${\bf w}_i$ is the direction, and $\delta {\bf w}_i$ is the

127: variation in this direction, orthogonal to ${\bf w}_i$.

128: \begin{figure}[!h]

129: \begin{center}

130: \includegraphics[width=80mm]{F19_ch5_CijCiP3}

131: \caption{Variations in $C_{ij}-C_i-P$ chain} \label{fig:CijCiP}

132: \end{center}

133: \end{figure}

134: 

135: According to Fig.\ref{fig:CijCiP},

136: \begin{eqnarray} \label{eq:PCij} &{\bf c}_{ij}&

137: - {\bf c}_{i} = {\bf R}_i [{\bf I}_3 + \delta \theta \times] \big(

138: \xi(j)(d/2 + \delta{c}_i /2)[{\bf I}_3 + \delta {\theta}_{Ci}

139: \times]{\bf e}_2 \big)\\ &{\bf c}_{i}& - {\bf p} = [{\bf I}_3 +

140: \delta \theta \times] {\bf R}_i ({\bf c}_0 + \delta{\bf c}_i)

141: \end{eqnarray}

142: 

143: \noindent where $\delta {c}_i$ is the variation in the length of

144: link $\overline{C_{i1}C_{i2}}$, which is supposed to be equally

145: shared by each side of $C_i$. $\delta \theta_{Ci}=

146: \mat{ccc}{\delta \theta_{Cix} & \delta \theta_{Ciy} & \delta

147: \theta_{Ciz}}^{T}$ is the orientation error of link

148: $\overline{C_{i1}C_{i2}}$ with respect to link $\overline{C_iP}$.

149: ${\bf c}_0$ is the nominal position vector of $C_i$ with respect

150: to end-effector $P$, expressed in ${\cal R}_i$, $\delta{\bf c}_i$

151: is the position error of $C_i$ expressed in ${\cal R}_i$, and

152: $\delta \theta = \mat{ccc}{\delta \theta_x & \delta \theta_y &

153: \delta \theta_z}^{T}$ is the orientation error of the

154: end-effector, expressed in ${\cal R}_b$.

155: 

156: Implementing linearization of eqs.(\ref{eq:OAi}-\ref{eq:PCij}) and

157: removing the components associated with the nominal constrained

158: equation ${\bf p}_0 = {\bf R}_i ({\bf a}_0 + \rho_i {\bf e}_1 -

159: {\bf c}_0) + L_i {\bf w}_i$, yields

160: \begin{eqnarray}\label{eq:delta-r} \delta {\bf p} & = & {\bf p} - {\bf p}_0 \nonumber\\ & = & {\bf R}_i

161: \big( \delta {\bf e}_i + \rho_i (\delta \theta_{Ai} \times {\bf

162: e}_1) + \xi(j) \ d/2 \ ( \delta \theta_{Ai} \times {\bf e}_2) + \\

163: & & \xi(j) \ d/2 \

164: (\delta \gamma_i \times {\bf e}_2) + \xi(j) \ \delta{m_i}/2 \ {\bf e}_2 \big) + \nonumber\\

165: & & \delta L_{ij} {\bf w}_i + L_{i} \delta {\bf w}_i - \delta

166: \theta \times {\bf R}_i \big( {\bf c}_0 + d/2 \ \xi(j) \ {\bf e}_2

167: \big) \nonumber

168: \end{eqnarray}

169: 

170: \noindent where

171: \begin{description}

172: \item $\delta {\bf p}$ is the position error of the end-effector

173: of the manipulator.

174: \item $\delta {\bf e}_i = \delta {\bf a}_i +

175: \delta{\rho}_i {\bf e}_1 - \delta {\bf c}_i$ is the sum of the

176: position errors of points $A_i$, $B_i$, and $C_i$ expressed in

177: ${\cal R}_i$.

178: \item $\delta \gamma_i = \delta \theta_{Bi} - \delta

179: \theta_{Ci}$ is the sum of the orientation errors of the $i^{th}$

180: parallelogram with respect to the $i^{th}$ prismatic joint and the

181: end-effector.

182: \item $\delta{m_i} = \delta{b}_i - \delta{c}_i $

183: corresponds to the parallelism error of links

184: $\overline{B_{i1}C_{i1}}$ and $\overline{B_{i2}C_{i2}}$, which is

185: depicted by Fig.\ref{fig:parallelogramme}.

186: \end{description}

187: 

188: \begin{figure}[!htbp]

189: \begin{center}

190: \centering\includegraphics[width=80mm]{F20_ch5_parall_var_english}

191: \caption{Variations in the $i^{th}$ parallelogram}

192: \label{fig:parallelogramme}

193: \end{center}

194: \end{figure}

195: 

196: Equation (\ref{eq:delta-r}) shows the coupling of the position and

197: orientation errors of the end-effector. Contrary to the

198: orientation errors, the position errors can be compensated because

199: the manipulator is a translational 3-DOF PKM. Thus, it is more

200: important to minimize the geometrical variations, which are

201: responsible for the orientation errors of the end-effector than

202: the ones, which are responsible for its position errors.

203: 

204: The following equation is obtained by multiplying both sides of

205: eq.(\ref{eq:delta-r}) by ${\bf w}_i^T$ and utilizing the

206: circularity of hybrid product.

207: \begin{eqnarray}\label{eq:delta-r-2}

208: {\bf w}_i^T \delta {\bf p} & = & {\bf w}_i^T {\bf R}_i {\delta

209: {\bf e}_i} + \rho_i ({\bf R}_i {\bf e}_1 \times {\bf w}_i)^T {\bf

210: R}_i \delta \theta_{Ai} + \xi(j) \ d/2  \\ &  & ({\bf R}_i {\bf

211: e}_2 \times {\bf w}_i)^T {\bf R}_i (\delta \theta_{Ai} + \delta

212: \gamma_i) + \xi(j) \ \delta{m_i}/2 \ {\bf w}_i^T {\bf R}_i {\bf

213: e}_2 \nonumber \\ & & + \delta L_{ij} - \big( {\bf R}_i ({\bf c}_0

214: + \xi(j) \ d/2 \ {\bf e}_2) \times {\bf w}_i \big)^T \delta \theta

215: \nonumber

216: \end{eqnarray}

217: %---------------------------------------------------------------------

218: 

219: {\bf Orientation Error Mapping Function:} By substraction of

220: eqs.(\ref{eq:delta-r-2}) written for $j=1$ and $j=2$, and for the

221: $i^{th}$ kinematic chain, a relation is obtained between the

222: orientation error of the end-effector and the variations in design

223: parameters, which is independent of the position error of the

224: end-effector.

225: \begin{equation}\label{eq:orientation-error-mapping-function}

226: d({\bf R}_i {\bf e}_2 \times {\bf w}_i)^T \delta \theta = \delta

227: l_i + d({\bf R}_i {\bf e}_2 \times {\bf w}_i)^T {\bf R}_i (\delta

228: \theta_{Ai} + \delta \gamma_i) + \delta{m_i} \ {\bf w}_i^T {\bf

229: R}_i {\bf e}_2

230: \end{equation}

231: 

232: \noindent where $\delta l_i = \delta L_{i1} - \delta L_{i2}$, the

233: relative length error of links $\overline{B_{i1}C_{i1}}$ and

234: $\overline{B_{i2}C_{i2}}$, depicts the parallelism error of links

235: $\overline{B_{i1}B_{i2}}$ and $\overline{C_{i1}C_{i2}}$ as shown

236: in Fig.\ref{fig:parallelogramme}. Equation

237: (\ref{eq:orientation-error-mapping-function}) can be written in

238: matrix form:

239: \begin{equation}\label{eq:expression-Jtheta-theta}

240:   \delta \theta = {\bf J}_{\theta \theta} {\gbf \epsilon}_{\theta}

241: \end{equation}

242: 

243: \noindent with

244: \begin{eqnarray*}

245:   {\bf J}_{\theta \theta} &=& {\bf D}^{-1}{\bf E} \\

246:   {\bf D} &=& d \mat{c}{({\bf R}_1 {\bf e}_2 \times {\bf w}_1)^T \\ ({\bf R}_2 {\bf e}_2 \times {\bf w}_2)^T \\ ({\bf R}_3 {\bf e}_2 \times {\bf

247:   w}_3)^T}  \quad ; \quad {\bf E} = \mat{ccc}{{\bf E}_1 & \cdots & \cdots \\ \cdots  & {\bf E}_2 & \cdots \\ \cdots & \cdots & {\bf

248:   E}_3}\\

249:   {\bf E}_i &=& \mat{cccc}{1 & {\bf w}_i^T {\bf R}_i{\bf e}_2 & d({\bf R}_i{\bf e}_2 \times {\bf w}_i)^T{\bf R}_i & d({\bf R}_i{\bf e}_2 \times {\bf w}_i)^T{\bf R}_i}

250: \end{eqnarray*}

251: $\delta \theta$ is the orientation error of the end-effector

252: expressed in ${\cal R}_b$, and ${\gbf \epsilon}_{\theta} = ({\gbf

253: \epsilon}_{\theta1}^T, {\gbf \epsilon}_{\theta2}^T, {\gbf

254: \epsilon}_{\theta3}^T )^T$ such that ${\gbf \epsilon}_{\theta i} =

255: (\delta l_i, \delta m_i, \delta \theta_{Ai}^T, \delta

256: \gamma_i^T)^T$. The determinant of ${\bf D}$ will be null if the

257: normal vectors to the plans, which contain the three

258: parallelograms respectively, are collinear, or if one

259: parallelogram is flat. Here, this determinant is not null when $P$

260: covers $C_u$ because of the geometry of the manipulator.

261: Therefore, ${\bf D}$ is nonsingular and its inverse ${\bf D}^{-1}$

262: exists.

263: 

264: As $({\bf R}_i {\bf e}_2 \times {\bf w}_i)^T \bot {\bf R}_i {\bf

265: e}_2$, $\delta \theta_{Aiz}$ and $\delta \gamma_{iz}$, the third

266: components of $\delta \theta_{Ai}$ and $\delta \gamma_{i}$

267: expressed in ${\cal R}_i$, have no effect on the orientation of

268: the end-effector. Thus, matrix ${\bf J}_{\theta \theta}$ can be

269: simplified by eliminating its columns associated with $\delta

270: \theta_{Aiz}$ and $\delta \gamma_{iz}$, $i=1,2,3$. Finally,

271: eighteen variations: $\delta l_i$, $\delta m_i$, $\delta

272: \theta_{Aix}$, $\delta \theta_{Aiy}$, $\delta \gamma_{ix}$,

273: $\delta \gamma_{iy}$, $i=1,2,3$, should be responsible for the

274: orientation error of the end-effector.\\

275: 

276: %---------------------------------------------------------------------

277: {\bf Position Error Mapping Function:} By addition of

278: eqs.(\ref{eq:delta-r-2}) written for $j=1$ and $j=2$, and for the

279: $i^{th}$ kinematic chain, a relation is obtained between the

280: position error of the end-effector and the variations in design

281: parameters, which does not depend on $\delta \gamma_{i}$.

282: \begin{equation}\label{eq:position-error-mapping-function}

283: {\bf w}_i^T \delta{\bf p} = \delta L_i + {\bf w}_i^T {\bf R}_i

284: \delta {\bf e}_i + \rho_i({\bf R}_i {\bf e}_1 \times {\bf w}_i)^T

285: {\bf R}_i \delta \theta_{Ai} - ({\bf R}_i {\bf c}_0 \times {\bf

286: w}_i)^T \delta \theta

287: \end{equation}

288: 

289: \noindent Equation (\ref{eq:position-error-mapping-function}) can

290: be written in matrix form:

291: \begin{equation}\label{delta-r}

292:   \delta{\bf p} = {\bf J}_{pp} {\gbf \epsilon}_{p} + {\bf J}_{p \theta} {\gbf

293:   \epsilon}_{\theta} = [{\bf J}_{pp} {\bf J}_{p \theta}]\left[\begin{array}{c} {\gbf \epsilon}_{p} \\ {\gbf

294:   \epsilon}_{\theta} \end{array} \right]

295: \end{equation}

296: 

297: \noindent with

298: \begin{eqnarray*}

299: {\bf J}_{pp} &=& {\bf F}^{-1}{\bf G} \quad ; \quad {\bf J}_{p

300: \theta} = {\bf F}^{-1}{\bf H}{\bf J}_{\theta \theta} \\ {\bf F}

301: &=& [{\bf w}_1 {\bf w}_2 {\bf w}_3]^T \\ {\bf G} &=& \mbox{diag}

302: ({\bf G}_i) \quad ; \quad {\bf G}_i = \mat{ccc}{1 & {\bf w}_i^T

303: {\bf R}_i & \rho_i ({\bf R}_i {\bf e}_1 \times {\bf w}_i )^T {\bf

304: R}_i}

305: \\

306: {\bf H} &=& - \mat{ccc}{{\bf R}_1 {\bf c}_0 \times {\bf w}_1 &

307: {\bf R}_2 {\bf c}_0 \times {\bf w}_2 & {\bf R}_3 {\bf c}_0 \times

308: {\bf w}_3} \\ {\gbf \epsilon}_{p} &=& ({\gbf \epsilon}_{p1}^T,

309: {\gbf \epsilon}_{p2}^T, {\gbf \epsilon}_{p3}^T)^T, \ {\gbf

310: \epsilon}_{pi} = (\delta L_i, \delta {\bf e}_i^T, \delta

311: \theta_{Ai}^T)^T

312: \end{eqnarray*}

313: 

314: $\delta L_i = (\delta L_{i1} + \delta L_{i2})/2$ is the mean value

315: of the variations in links $\overline{B_{i1}C_{i1}}$ and

316: $\overline{ B_{i2}C_{i2}}$, {\it i.e.}: the variation in the

317: length of the $i^{th}$ parallelogram. ${\gbf \epsilon}_{p}$ is the

318: set of the variations in design parameters, which should be

319: responsible for the position errors of the end-effector, except

320: the ones which should be responsible for its orientation errors,

321: {\it i.e.}: ${\gbf \epsilon}_{\theta}$. ${\gbf \epsilon}_{p}$ is

322: made up of three kinds of errors: the variation in the length of

323: the $i^{th}$ parallelogram, {\it i.e.}: $\delta L_i, i=1,2,3$, the

324: position errors of points $A_i$, $B_i$, and $C_i$, {\it i.e.}:

325: $\delta {\bf e}_i, i=1,2,3$, and the orientation errors of the

326: directions of the prismatic joints, {\it i.e.}: $\delta

327: \theta_{Ai}$, $i=1,2,3$. Besides, ${\bf F}$ is nonsingular and its

328: inverse ${\bf F}^{-1}$ exists because ${\bf F}$ corresponds to the

329: Jacobian kinematic matrix of the manipulator, which is not

330: singular when $P$ covers $C_u$, \cite{CHABLAT03}.

331: 

332: According to eq.(\ref{eq:position-error-mapping-function}) and as

333: $({\bf R}_i {\bf e}_1 \times {\bf w}_i)^T \bot {\bf R}_i {\bf

334: e}_1$, matrix ${\bf J}_{pp}$ can be simplified by eliminating its

335: columns associated with $\delta \theta_{Aix}, \ i=1,2,3$. Finally,

336: thirty-three variations: $\delta L_i$, $\delta e_{ix}$, $\delta

337: e_{iy}$, $\delta e_{iz}$, $\delta \theta_{Aix}$, $\delta

338: \theta_{Aiy}$, $\delta \theta_{Aiz}$, $\delta l_i$, $\delta m_i$,

339: $\delta \gamma_{ix}$, $\delta \gamma_{iy}$, $i=1,2,3$, should be

340: responsible for the position error of the end-effector.

341: 

342: Rearranging matices ${\bf J}_{pp}$ and ${\bf J}_{p \theta}$, the

343: position error of the end-effector can be expressed as:

344: \begin{equation}\label{delra-r2}

345:   \delta{\bf p} = {\bf J} \ {\gbf \epsilon}_q = \mat{ccc}{{\bf J}_1 & {\bf J}_2 & {\bf J}_3}({\gbf \epsilon}_{q1}^T \ {\gbf \epsilon}_{q2}^T

346: \ {\gbf \epsilon}_{q3}^T)^T

347: \end{equation}

348: 

349: \noindent with ${\gbf \epsilon}_{qi} = (\delta L_i, \delta e_{ix},

350: \delta e_{iy}, \delta e_{iz}, \delta \theta_{Aix}, \delta

351: \theta_{Aiy}, \delta \theta_{Aiz}, \delta l_i, \delta m_i, \delta

352: \gamma_{ix},

353: \delta \gamma_{iy})^T$, and ${\bf J} \in {\field{R}}^{3 \times 33}$.\\

354: 

355: %---------------------------------------------------------------------

356: {\bf Sensitivity Indices:} In order to investigate the influence

357: of the design parameters errors on the position and the

358: orientation of the end-effector, sensitivity indices are required.

359: According to section

360: \ref{section:linkage-kinematic-analysis-results}, variations in

361: the design parameters of the same type from one leg to the other

362: have the same influence on the location of the end-effector. Thus,

363: assuming that variations in the design parameters are independent,

364: the sensitivity of the position of the end-effector to the

365: variations in the $k^{th}$ design parameter responsible for its

366: position error, {\it i.e.}: ${\epsilon}_{q(1,2,3)k}$, is called

367: ${\mu}_{k}$ and is defined by eq.(\ref{eq:sensi-pos}).

368: 

369: \begin{equation}\label{eq:sensi-pos}

370:   {\mu}_{k} = \sqrt{\sum_{i=1}^{3} \sum_{m=1}^{3}J_{imk}^{2}} \ ,

371:   \ k = 1, \cdots, 11

372: \end{equation}

373: 

374: Likewise, the sensitivity of the orientation of the end-effector

375: to the variations in the $r^{th}$ design parameter responsible for

376: its orientation error, {\it i.e.}: ${\epsilon}_{\theta (1,2,3)

377: r}$, is called $\nu_r$ and follows from

378: eq.(\ref{eq:expression-Jtheta-theta}).

379: \begin{equation}\label{eq:sensi-orientation}

380:   {\nu}_{r} = \sqrt{\sum_{j=0}^{2} \sum_{i=1}^{3}J_{\theta \theta

381:   i(6j+r)}^{2}} \ , \ r = 1, \cdots, 6

382: \end{equation}

383: 

384: Finally, ${\mu}_{k}$ can be employed as a sensitivity index of the

385: position of the end-effector to the $k^{th}$ design parameter

386: responsible for the position error. Likewise, ${\nu}_{r}$ can be

387: employed as a sensitivity index of the orientation of the

388: end-effector to the $r^{th}$ design parameter responsible for the

389: orientation error. It is noteworthy that these sensitivity indices

390: depend on the location of the end-effector.

391: 

392: %---------------------------------------------------------------------

393: \subsubsection{Results of the Differential Vector Method}

394: \label{section:differential-vector-analysis-results}

395: %--------------------------------------------------------------------

396: The sensitivity indices defined by eqs.(\ref{eq:sensi-pos}) and

397: (\ref{eq:sensi-orientation}) are used to evaluate the sensitivity

398: of the position and orientation of the end-effector to variations

399: in design parameters, particularly to variations in the

400: parallelograms.

401: 

402: \begin{figure}[!h]

403: \includegraphics[width=44mm]{F21a_ch5_deltap_vardim4} \hfill

404: \includegraphics[width=44mm]{F21b_ch5_deltap_varang3}

405: \caption{Sensitivity of the position of the end-effector along

406: $Q_1Q_2$, (a): to dimensional variations, (b): to angular

407: variations} \label{fig:deltar-vardim-varang}

408: \end{figure}

409: 

410: %\begin{figure}[!h]

411: %\psfrag{deix}{$\delta e_{ix}$} \psfrag{dLi}{$\delta L_{i}$}

412: %\psfrag{dli}{$\delta l_{i}$} \psfrag{dmi}{$\delta m_{i}$}

413: %\psfrag{deiydeiz}{$\delta e_{iy},\delta e_{iz}$}

414: %\psfrag{Q1}{$q_1$} \psfrag{Q2}{$q_2$} \psfrag{a}{(a)}

415: %\psfrag{p}{$p$}

416: %\includegraphics[width=44mm]{deltar_vardim2} \hfill

417: %\psfrag{thix}{$\delta \gamma_{ix},\delta \theta_{Aix}$}

418: %\psfrag{thiy}{$\delta \theta_{Aiy}$} \psfrag{thiz}{$\delta

419: %\theta_{Aiz}$} \psfrag{giy}{$\delta \gamma_{iy}$}

420: %\psfrag{Q1}{$q_1$} \psfrag{Q2}{$q_2$} \psfrag{b}{(b)}

421: %\psfrag{p}{$p$}

422: %\includegraphics[width=44mm]{deltar_varang4}

423: %\caption{Sensitivity of the position of the end-effector along

424: %$Q_1Q_2$, (a): to dimensional variations, (b): to angular variations}

425: %\label{fig:deltar-vardim-varang}

426: %\end{figure}

427: 

428: Figures \ref{fig:deltar-vardim-varang}(a-b) depict the sensitivity

429: of the position of the end-effector along the diagonal $Q_1Q_2$ of

430: $C_u$, to dimensional variations and angular variations,

431: respectively. According to Fig.\ref{fig:deltar-vardim-varang}(a),

432: the position of the end-effector is very sensitive to variations

433: in the lengths of the parallelograms, $\delta L_i$, and to the

434: position errors of points $A_i$, $B_i$, and $C_i$ along axis $x_i$

435: of ${\cal R}_i$, {\it i.e}: $\delta e_{ix}$. Conversely, the

436: influence of $\delta l_i$ and $\delta m_i$, the parallelism errors

437: of the parallelograms, is low and even negligible in the kinematic

438: isotropic configuration. According to

439: Fig.\ref{fig:deltar-vardim-varang}(b), the orientation errors of

440: the prismatic joints depicted by $\delta \theta_{Aiy}$ and $\delta

441: \theta_{Aiz}$ are the most influential angular errors on the

442: position of the end-effector. Besides, the position of the

443: end-effector is not sensitive to angular variations in the

444: isotropic configuration.

445: 

446: \begin{figure}[!h]

447: \includegraphics[width=44mm]{F22a_ch5_deltatheta_vardim2} \hfill

448: \includegraphics[width=44mm]{F22b_ch5_deltatheta_varang2}

449: \caption{Sensitivity of the orientation of the end-effector along

450: $Q_1Q_2$, (a): to dimensional variations, (b): to angular

451: variations} \label{fig:deltatheta-vardim-varang}

452: \end{figure}

453: 

454: Figures \ref{fig:deltatheta-vardim-varang}(a-b) depict the

455: sensitivity of the orientation of the end effector, along

456: $Q_1Q_2$, to dimensional and angular variations. According to

457: Fig.\ref{fig:deltatheta-vardim-varang}(a), $\delta l_i$ and

458: $\delta m_i$ are the only dimensional variations, which are

459: responsible for the orientation error of the end-effector.

460: However, the influence of the parallelism error of the small sides

461: of the parallelograms, depicted by $\delta l_i$, is more important

462: than the one of the parallelism error of their long sides,

463: depicted by $\delta m_i$.

464: 

465: According to figures \ref{fig:deltar-vardim-varang} and

466: \ref{fig:deltatheta-vardim-varang}, the sensitivity of the

467: position and the orientation of the end-effector is generally null

468: in the kinematic isotropic configuration ($p = 0$), and is a

469: maximum when the manipulator is close to a kinematic singular

470: configuration, {\it i.e.}: $P \equiv Q_2$. Indeed, only two kinds

471: of design parameters variations are responsible for the variations

472: in the position of the end-effector in the isotropic

473: configuration: $\delta L_i$ and $\delta e_{ix}$. Likewise, two

474: kinds of variations are responsible for the variations in its

475: orientation in this configuration: $\delta l_i$, the parallelism

476: error of the small sides of the parallelograms, $\delta

477: \theta_{Aiy}$ and $\delta \gamma_{iy}$. Moreover, the

478: sensitivities of the pose (position and orientation) of the

479: end-effector to these variations are a minimum in this

480: configuration, except for $\delta l_i$. On the contrary, $Q_2$

481: configuration, {\it i.e.}: $P \equiv Q_2$, is the most sensitive

482: configuration of the manipulator to variations in its design

483: parameters. Indeed, as depicted by

484: Figs.\ref{fig:deltar-vardim-varang} and

485: \ref{fig:deltatheta-vardim-varang}, variations in the pose of the

486: end-effector depend on all the design parameters variations and

487: are a maximum in this configuration.

488: 

489: Moreover, figures \ref{fig:deltar-vardim-varang} and

490: \ref{fig:deltatheta-vardim-varang} can be used to compute the

491: variations in the position and orientation of the end-effector

492: with knowledge of the amount of variations in the design

493: parameters. For instance, let us assume that the parallelism error

494: of the small sides of the parallelograms, $\delta l_i$, is equal

495: to $10 \mu$m. According to

496: Fig.\ref{fig:deltatheta-vardim-varang}(a), the position error of

497: the end-effector will be equal about to $3\mu$m in $Q_1$

498: configuration ($P \equiv Q_1$). Likewise, according to

499: Fig.\ref{fig:deltar-vardim-varang}(b), if the orientation error of

500: the direction of the $i^{th}$ prismatic joint round axis $y_i$ of

501: ${\cal R}_i$ is equal to 1 degree, {\it i.e.}: $\delta

502: \theta_{Aiy} = 1$ degree, the position error of the end-effector

503: will be equal about to 4.8 mm in $Q_2$ configuration.

504: