The literature review reveals that existing parallel-serial machining manipulators differ mainly in their underlying parallel mechanisms. Therefore, we found it convenient to classify the manipulators according to this mechanism. The rest of the subsection will illustrate this classification and cover various architectures of hybrid machining robots.
3.1.1. Manipulators Based on the Tricept Parallel Mechanism
Manipulators in this family are based on a 3-DOF parallel mechanism called Tricept and patented by Neumann [
31]. This mechanism has one UP branch, which defines the motion type of the output link, and three U
PS branches, which include actuators. Based on this scheme, scholars have proposed different parallel-serial manipulators considered next.
Tetrabot (tetrahedral robot) [
32] was probably the first ever parallel-serial manipulator, which was used for assembly operations. In this robot, the output link of the Tricept parallel mechanism is augmented with an RRR serial spherical wrist (
Figure 4a). This 6-DOF kinematics has become suitable for other applications as well. For example, Choi and Lee [
33] considered using this parallel-serial manipulator for propeller grinding and developed its physical prototype. In the prototype, the serial wrist was designed as a geared mechanism. The authors analyzed the kinematics, dynamics, and position/velocity control of this manipulator in their paper [
34]. Zhao et al. [
35] designed another prototype of such a parallel-serial manipulator and applied it to polishing operations. In their later works, the authors focused on kinematic calibration [
36], stiffness [
37] and repeatability [
38] analysis, and different control strategies [
39,
40] of this robot.
While 6-DOF manipulators can reach any desired end-effector posture, most machining operations only need five DOFs to position and align the tool. In this regard, most Tricept-based parallel-serial robots are equipped with a 2-DOF RR serial wrist (
Figure 4b). The obtained (1-UP/3-U
PS)RR architecture is the one that is actually called Tricept in most cases. Parallel-serial manipulators with this architecture have been the subject of numerous articles and successfully applied in practice [
41]. For example, among the pioneering studies are the works of Siciliano [
42,
43], who analyzed the inverse and direct kinematics of this robot but focused mainly on the 3-DOF parallel part. Milutinović et al. [
44] continued these studies and considered the kinematics of the serial wrist, while Eastwood and Webb [
45] used these kinematic models and discussed how thermal deformations affect the positioning accuracy of the robot. Besides conventional milling and drilling operations, the Tricept parallel-serial manipulator has been used for incremental sheet forming [
46,
47], friction stir welding [
48,
49], and optical polishing [
50,
51]. Most recent applications include processing large optical mirrors [
52,
53,
54].
The actuated branches of the Tricept manipulator have a U
PS architecture where the universal joint is located at the base. An S
PU branch with the spherical joint placed at the base is a natural kinematic equivalent of this architecture, and other scholars have also considered Tricept manipulators with these branches (
Figure 4c). Thus, Kim et al. [
55] developed such a (1-UP/3-S
PU)RR parallel-serial manipulator, analyzed its kinematics [
56] and dynamics [
57,
58], and performed a kinematic calibration of the physical prototype [
55]. Another kinematically equivalent design was proposed by Tönshoff et al. [
59], who developed a Georg V manipulator with three
PUS branches, featuring actuated prismatic joints at the base (
Figure 4d).
Inspired by the successful design of Tricept, scholars proposed different Tricept-based architectures of parallel-serial manipulators. For example, Li et al. [
60] considered a 4-DOF (1-RP/2-R
PS)RR robot called Bicept (
Figure 4e). This robot includes a 2-DOF planar Tricept-like parallel mechanism where spherical joints are used to compensate for misalignment and can be replaced with revolute joints. The robot stiffness and performance were examined in papers [
61] and [
62], but all the analysis was done for the planar parallel mechanism.
Figure 4.
Parallel-serial machining manipulators of the Tricept family: (
a) (1-UP/3-U
PS)RRR (Tetrabot) [
32]; (
b) (1-UP/3-U
PS)RR (Tricept) [
42]; (
c) (1-UP/3-S
PU)RR [
55]; (
d) (1-UP/3-
PUS)RR (Georg V) [
59]; (
e) (1-RP/2-R
PS)RR (Bicept) [
60]; (
f) (1-U
P/2-U
PS)RR (TriVariant) [
63]; (
g) (1-U/2-U
PS)PRR (TriVariant B) [
64]; (
h) (1-UP/3-U
PS)PRR (Tricept IV) [
65]; (
i) (R(1-RP/2-R
PS)/1-U
PS)RR (TriMule) [
66]; (
j) (R(1-RP/1-R
PS)/2-U
PS)RR (Trifree) [
67]; (
k) (R(2-R
PR)U/1-U
P)RR [
68]; (
l) (R(1-R
P/1-R
PR)/1-U
PS)RR [
69]; (
m) (R(4-R
PR)/(2-R
PR)R)RR [
70]; (
n) PP(1-UP/3-U
PS) [
71]; (
o) PP(1-
PU/2-U
PS) [
72]; (
p) (1-
PU/2-
PSS)PP [
73].
Figure 4.
Parallel-serial machining manipulators of the Tricept family: (
a) (1-UP/3-U
PS)RRR (Tetrabot) [
32]; (
b) (1-UP/3-U
PS)RR (Tricept) [
42]; (
c) (1-UP/3-S
PU)RR [
55]; (
d) (1-UP/3-
PUS)RR (Georg V) [
59]; (
e) (1-RP/2-R
PS)RR (Bicept) [
60]; (
f) (1-U
P/2-U
PS)RR (TriVariant) [
63]; (
g) (1-U/2-U
PS)PRR (TriVariant B) [
64]; (
h) (1-UP/3-U
PS)PRR (Tricept IV) [
65]; (
i) (R(1-RP/2-R
PS)/1-U
PS)RR (TriMule) [
66]; (
j) (R(1-RP/1-R
PS)/2-U
PS)RR (Trifree) [
67]; (
k) (R(2-R
PR)U/1-U
P)RR [
68]; (
l) (R(1-R
P/1-R
PR)/1-U
PS)RR [
69]; (
m) (R(4-R
PR)/(2-R
PR)R)RR [
70]; (
n) PP(1-UP/3-U
PS) [
71]; (
o) PP(1-
PU/2-U
PS) [
72]; (
p) (1-
PU/2-
PSS)PP [
73].
Huang et al. [
63] introduced a TriVariant manipulator (
Figure 4f), where one U
PS was removed and the drive was displaced to the UP branch, which became the U
P actuated branch. This manipulator has been the subject of many studies, which considered its kinematics [
63,
74,
75], dynamics [
76], stiffness analysis [
77], dimensional synthesis [
78], and calibration [
79]. Most of these studies, however, focused on the 3-DOF parallel part. The TriVariant manipulator was later modified into TriVariant B [
64], where the actuated prismatic joint was moved to the serial chain, so the robot architecture became (1-U/2-U
PS)PRR (
Figure 4g). The authors analyzed its kinematics [
64], dynamics [
80], and stiffness [
81], with a major focus on the 2-DOF parallel mechanism.
Based on the Tricept and TriVariant B designs, Sun et al. [
65] proposed a novel parallel-serial manipulator, Tricept IV, with the (1-UP/3-U
PS)PRR architecture (
Figure 4h). In this manipulator, the prismatic joints in the UP branch and the PRR serial chain are coaxial, which results in kinematic redundancy and increases the manipulator workspace. The authors evaluated the stiffness of this robot [
65,
82], performed its dimensional synthesis [
83], and also considered the robot application for friction stir welding [
84].
In the Tricept-based parallel-serial manipulators discussed so far, the parallel mechanism had three or four branches attached to the base. Scholars soon discovered how to simplify and lighten the design by combining the base joints of these branches [
66]. This idea led to the development of the TriMule robot [
66,
85]—one of the most studied parallel-serial manipulators at the moment. In this robot, the universal joints of the UP branch and two U
PS branches share a yoke that rotates relative to the base (
Figure 4i). We can notate the TriMule architecture as (R(1-RP/2-R
PS)/1-U
PS)RR and see it has only two joints placed on the base, simplifying the robot design. The TriMule manipulator has become the subject of many papers devoted to its kinematics [
86,
87], dynamics and gravity compensation [
88,
89,
90,
91,
92], stiffness [
93,
94,
95,
96], performance evaluation [
97,
98], optimal design [
99], path planning [
100,
101,
102,
103], feed rate scheduling [
104,
105,
106], calibration [
107,
108,
109,
110,
111,
112,
113,
114,
115], and control [
116,
117,
118,
119,
120]. Diverse applications of the TriMule robot include helical milling [
121], collaborative mirror milling of large thin-walled parts [
13,
122,
123,
124,
125], grinding [
126], friction stir welding [
127,
128,
129,
130,
131,
132], and trimming unidirectional carbon fiber-reinforced polymers [
133]. Apart from these, the TriMule robot was mounted on a mobile platform and applied for large-scale machining of aerospace components [
134,
135]. There also exists a 6-DOF TriMule design with a 3-DOF RRR spherical wrist, as in Tetrabot [
136].
The TriMule success inspired scholars to develop other similar parallel-serial manipulators. For example, Wang et al. [
67] proposed a Trifree robot with an (R(1-RP/1-R
PS)/2-U
PS)RR architecture (
Figure 4j), where the universal joints of two branches had a common yoke. The authors analyzed the kinematics and dynamics of this robot in their works [
137] and [
138]. Li et al. [
68] presented an (R(2-R
PR)U/1-U
P)RR manipulator (
Figure 4k), whose parallel mechanism had only three branches, and performed dimensional synthesis of this mechanism. Another design with three branches was introduced by Dong et al. [
69], who used an asymmetrical (R(1-R
P/1-R
PR)/1-U
PS)RR architecture (
Figure 4l). Finally, Liu et al. [
70] considered a super redundantly actuated (R(4-R
PR)/(2-R
PR)R)RR manipulator, whose parallel mechanism had three DOFs and six drives (
Figure 4m). To our knowledge, all these novel manipulators exist only as computer models; no physical prototypes have been created.
All previous parallel-serial manipulators were equipped with a 2- or 3-DOF spherical wrist. There also exist Tricept-based architectures without this wrist but with a PP serial kinematic chain. Thus, Luo et al. [
71] proposed a PP(1-UP/3-U
PS) robot with a parallel mechanism placed on the PP moving frame (
Figure 4n). A similar design with a 1-
PU/2-U
PS parallel mechanism was introduced by Shan et al. [
72], who placed this mechanism on the XY moving table and used it as a spindle posture alignment device (
Figure 4o). One more design was devised in paper [
73], where the PP serial kinematic chain was placed on the platform of the 1-
PU/2-
PSS parallel mechanism (
Figure 4p). This manipulator is intended to process machine components with complex and/or longitudinal shapes, and studies [
139] and [
140] analyzed its kinematics, including workspace and singularities. All these listed manipulators, however, are only available as computer models.
3.1.2. Manipulators Based on the Sprint Z3 Parallel Mechanism
The parallel-serial manipulators discussed in this subsection rely on the Sprint Z3 parallel machine tool patented by Wahl [
141].
1 Unlike Tricept, this parallel mechanism does not have a passive branch but includes three symmetrically arranged
PRS branches. Each branch imposes one constraint, leaving the output link with two rotational and one translational DOFs. Numerous papers have studied this parallel mechanism [
10], but we will focus on parallel-serial manipulators based on its kinematics.
The first example is the Ecospeed machine center [
143], which includes this 3-
PRS parallel mechanism mounted on a 2-DOF PP serial chain (
Figure 5a). San et al. analyzed the kinematics and error sources of this parallel-serial manipulator in their recent work [
152]. Other scholars also considered a 4-DOF P(3-
PRS) architecture, which was used with a 1-DOF moving table. Multiple studies examined the error sensitivity [
153,
154], stiffness analysis [
155], calibration [
156,
157,
158,
159], and control [
160,
161,
162] of this robot. Huang et al. [
23] developed a similar manipulator where the revolute joints were replaced with parallelogram joints (
Figure 5b). The authors considered the dynamics [
163], calibration [
164,
165], and control [
166] of the manipulator, focusing on its 3-DOF parallel part. The actuated parallelogram joints were used in an RRRP(3-
PaRS) manipulator [
144] (
Figure 5c), based on a 4-DOF SCARA-type serial chain and a 3-DOF CaPaMan (Cassino Parallel Manipulator). The authors made a prototype of this manipulator for wood drilling and considered its position and force control. A SCARA-type serial chain was also used in a (3-
PSP)RRRP manipulator [
145], attached to the moving platform of the parallel mechanism (
Figure 5d). This 7-DOF manipulator was designed for welding operations inside a steam generator, and the authors analyzed the robot workspace and control in the cited paper.
The architectures mentioned above included actuated prismatic or parallelogram joints near the base of the parallel mechanisms. There are also manipulators where these joints are placed in the middle of the branches. For example, Wang et al. [
146] developed a 6-DOF PPP(3-R
PS) robot (
Figure 5e) for marine propeller processing and addressed its kinematic calibration in paper [
167]. A similar parallel mechanism was used in a P(3-R
PS)R manipulator (
Figure 5f) designed for blade machining and polishing [
147]. The authors applied this manipulator with a 1-DOF moving platform and performed its calibration in work [
168]. Li et al. [
148] “inverted” the branches of the parallel mechanism and introduced a (3-S
PR)RR architecture (
Figure 5g). The authors analyzed the performance [
169] and stiffness [
170,
171] of this manipulator and discussed its calibration [
172], with the main focus on the parallel part. Li et al. [
149] proposed a robot with one R
PS and two S
PR branches (
Figure 5h) and examined its kinematics and performance. This manipulator was designed for friction stir welding and was supposed to operate with a 1-DOF moving table, but as far as we know, there is no physical prototype of this robot.
There also exist several architectures where a parallel mechanism has a universal joint in its branches. For example, Chong et al. [
24] introduced a PaPa(3-R
CU) manipulator (
Figure 5i) aimed at polishing wind turbine blades. Here,
C means passive revolute and actuated prismatic joints with collinear axes. The 2-DOF PaPa serial chain had a complex actuation scheme, omitted in the figure, and was placed on a mobile platform to increase the robot workspace. The authors performed the elasto-geometrical calibration of this robot [
173] and considered its control strategies [
174]. Another polishing robot was designed by Xu et al. [
150] and included a 3-DOF 3-U
PR parallel mechanism attached to a 6-DOF serial manipulator (
Figure 5j). Unlike the previous architectures discussed in this subsection, the parallel mechanism is overconstrained. The authors studied the robot dynamics and control and performed its dimensional synthesis in the cited paper, focusing on the parallel part. Finally, we would like to mention a PP(1-PUS/3-R
HUR) manipulator (
Figure 5k) proposed by Tian et al. [
151]. The parallel mechanism of the manipulator includes a PUS branch to increase the large-stroke stiffness of this mechanism. The authors analyzed kinematics and performed a multi-objective design optimization of this robot, which currently exists only as a computer model.
3.1.3. Manipulators Based on the Exechon Parallel Mechanism
Manipulators discussed in the two previous subsections were based on 3-DOF non-overconstrained parallel mechanisms. Scholars also considered the use of overconstrained parallel mechanisms in parallel-serial manipulators. One of the most notable manipulators in this family is the 5-DOF Exechon robot proposed by Neumann [
175]. It includes a 3-DOF overconstrained parallel mechanism with one S
PR central branch and two 2-U
PR lateral branches, where the base axes of the universal joints are collinear (
Figure 6a). The output link of the parallel mechanism is equipped with an RR spherical wrist. This manipulator has been successfully applied in industry [
176], including optical [
177] and large-scale [
178,
179,
180] machining. Numerous articles performed its mobility [
181], kinematic [
182,
183,
184], and stiffness [
185,
186,
187,
188] analyses, as well as calibration [
189]. Some scholars studied Exechon manipulators with offsets in the base joints [
190,
191] and both the base and the wrist joints [
192,
193,
194,
195]. Sagar et al. [
196] also considered a 7-DOF kinematically redundant design with a 4-DOF serial chain developed for placing fixture heads along sheet metal (
Figure 6b).
Inspired by the Exechon success, scholars proposed different modifications of its design. For example, Li et al. [
197] inverted the central branch of the parallel mechanism and considered the (1-R
PS/2-U
PR)RR architecture (
Figure 6c). In their study, the authors performed a multi-objective dimensional synthesis of this robot. Tengfei et al. [
198] inverted the lateral chains and developed a (1-R
PS/2-R
PU)RR parallel-serial manipulator called Exe-Variant (
Figure 6d). Paper [
207] evaluated its workspace and performance. The authors also considered a variation of this design where this parallel mechanism was placed on a PR serial chain [
199] (
Figure 6e), whose revolute joint represented a carriage moving along a circular rail. This architecture provides a large workspace essential for machining aeronautic components, as the authors discussed in their paper. Jin et al. [
200] proposed a PP(1-S
PR/2-R
PU) parallel-serial robot for similar applications (
Figure 6f). However, as we know, all the manipulators discussed in this paragraph exist only as computer models.
The aforementioned manipulators were developed by inverting the central and/or lateral branches of the 3-DOF parallel mechanism. Other scholars also modified the structure of these branches. For example, Xu et al. [
201] introduced a 4-DOF manipulator with the 1-U
PR central branch (
Figure 6g). This manipulator, operating with a 1-DOF moving table, has been the subject of multiple studies that have focused on its kinematics [
208], dynamics [
209], stiffness [
210], optimal design [
211,
212], and calibration [
213]. Zhang et al. [
202] changed the structure of all branches and proposed a PP(1-
PR
PS/2-
PRU) manipulator with redundant actuation (
Figure 6h). The authors analyzed the mobility and workspace of this robot, focusing on its parallel part. Another original design with one S
P and two U
PU branches and an RR spherical wrist was introduced in work [
203] (
Figure 6i). This manipulator has also become the subject of many studies devoted to its kinematics [
214,
215], dynamics [
216,
217,
218], stiffness [
219,
220,
221], and calibration [
222]. Finally, Xu et al. [
204] developed an RP(1-
RRS/2-
RRU) inner-cavity machining robot without prismatic joints in the parallel mechanism (
Figure 6j). The authors studied the robot kinematics and dynamics, performed its dimensional synthesis, and designed a 3D-printed prototype.
There are also Exechon-inspired parallel-serial manipulators whose parallel mechanism has four branches. Thus, Yue et al. [
205] proposed a (1-SPR/1-S
PR/2-U
PU)RR architecture with one passive branch (
Figure 6k). In their work, the authors analyzed the mobility and kinematics of this robot and performed its dimensional optimization. Zhang et al. [
206] presented a PP(2-R
PU/2-S
PR) redundantly actuated manipulator (
Figure 6l) and considered its kinematics [
206], dynamics [
223], stiffness analysis [
224], and optimal design [
225]. The authors focused mainly on the 3-DOF parallel mechanism and developed a physical prototype only for this mechanism.
3.1.4. Manipulators Based on Planar Parallel Mechanisms
This subsection will cover parallel-serial machining manipulators based on parallel mechanisms whose end-effector performs planar motion. We can classify these manipulators into two groups, as discussed next.
The first group includes manipulators with 2-DOF parallel mechanisms. For example, Li et al. [
226] proposed a 3-DOF (1-
PR/1-
PRR)P architecture (
Figure 7a). The authors analyzed the kinematics of this hybrid robot and developed its prototype. Ma et al. [
227] introduced a 5-DOF P(1-R
P/1-R
PR)RR manipulator with an RR spherical wrist (
Figure 7b). Unlike the previous design, the prismatic actuators of the parallel mechanism were connected to its base by revolute joints. A similar architecture was considered by Wang et al. [
228] who used a 1-
RR/1-
RRR parallel mechanism as a parallel module (
Figure 7c). The authors applied this robot for machining large casting parts and discussed simulations and experiments in paper [
229]. Uchiyama et al. [
230] designed a (1-
PR/1-
PRP)PRR manipulator for robotized deburring operations (
Figure 7d). The authors studied the kinematics of this robot and proposed a continuous path control algorithm, which was successfully applied in practice. Another 5-DOF manipulator with an identical PRR serial chain was developed by Guo et al. [
231]. Unlike the architectures above, the 2-DOF planar parallel mechanism had one PR passive branch and two
PRP branches with prismatic actuators at the base (
Figure 7e). In their later works, the authors analyzed workspace [
232], dynamics [
233,
234], and control [
235] of this robot.
The second group includes manipulators with 3-DOF parallel mechanisms, and there are just a couple of architectures within this group. The first is a (3-
RRR)PRR manipulator developed by Yang et al. [
236] for deburring operations (
Figure 7f). The authors created a robot prototype with a modular design, and its kinematic analysis can be found in studies [
237] and [
17]. Finally, paper [
73] introduced a (2-R
PR/2-
RRR)PR redundantly actuated manipulator for processing elongated objects (
Figure 7g). At the moment, this manipulator exists only as a computer model, and its kinematics has been analyzed recently in works [
238,
239].
3.1.5. Manipulators Based on the Delta Parallel Mechanism
The Delta mechanism, proposed by Clavel [
240], has become one of the most successful parallel architectures. The output link of this mechanism has three translational DOFs provided by parallelogram joints in its branches. However, there are just a few examples of Delta-based parallel-serial manipulators applied for machining, possibly because of their lower rigidity compared to other architectures discussed in the previous subsections.
The first example is the JDYP51 polishing machine tool [
241] based on the Delta mechanism with prismatic actuators (
Figure 8a). The output link of the parallel mechanism is equipped with an RR spherical wrist. Paper [
245] analyzed manipulator stiffness, and the authors of [
246] optimized the polishing process performed with this robot. Xu et al. [
242] introduced another polishing manipulator based on the Delta mechanism. Unlike the previous design, the branches of this mechanism use spatial parallelogram joints with spherical joints (
Figure 8b). A 1-DOF serial chain connects to the moving platform by a revolute joint, actuated by an auxiliary
RUPU branch, which is omitted in the figure for clarity. This 4-DOF parallel-serial manipulator moves the processed object, while the polishing tool is placed on an independent 2-DOF serial mechanism. In the later works, the authors considered path planning [
247], dynamics [
248,
249], stiffness [
250], optimal design [
251], calibration [
252], and control [
253] of this robot, focusing on the parallel part.
Other examples of Delta-based manipulators include kinematically redundant architectures, which, to our knowledge, exist only as computer models. Thus, Mohammadipanah and Zohoor [
243] proposed an 8-DOF hybrid robot with an RRRRR serial chain (
Figure 8c). The authors considered its application for arc welding and analyzed the kinematics and dynamics of this manipulator. Nguyen et al. [
244] studied a 9-DOF macro-mini milling robot where the Delta mechanism was attached to the end-effector of the 6-DOF Kawasaki RS030N industrial manipulator (
Figure 8d). The main focus of the authors’ work was on the compliance analysis and error compensation of this robot.
3.1.7. Other Manipulators
This subsection will cover parallel-serial manipulators, which we cannot put in any of the previous groups. It is interesting that there are just a few manipulators in this family, and we have found only two architectures with working prototypes used in practice.
The first example is the 6-DOF RNT (Robot of a New Type) developed by Mianowski et al. [
273] for milling and polishing large objects. This robot features a 2-DOF 1-RR/2-U
PS parallel mechanism placed on the rotating base, and the output link of this mechanism is equipped with an RRR spherical wrist (
Figure 10a). The authors considered the design, control, and applications of RNT in their later studies [
281] and [
282]. The second example is the 7-DOF CraftsRobot [
274], which was also applied for processing large-scale objects. The robot includes a 5-DOF 1-U
CR/4-U
CU parallel mechanism attached to the PaPa serial chain with the same architecture as in the previously discussed PaPa(3-R
CU) manipulator [
24] (
Figure 10b). Similar to that work, the parallel-serial manipulator was installed on the moving platform, which increased its workspace. In recent works, scholars have considered the stiffness [
283], optimal design [
284], and error compensation [
285] of this hybrid robot.
Other manipulators considered below exist only as computer models. There are several architectures based on 4-DOF parallel mechanisms. For example, Lu et al. [
275] introduced a PP(1-
PRU/3-
PUS) manipulator and analyzed the workspace and dynamics of its parallel part (
Figure 10c). Another design was considered by Tian et al. [
276], where one
PUS branch was replaced with the second
PRU branch (
Figure 10d). The authors performed the mobility analysis of this robot and found that it had no singular configurations. Li et al. [
277] studied a similar P(2-R
PU/2-U
PS) architecture where prismatic actuators of the parallel mechanism were displaced to the middle of its branches (
Figure 10e). In their paper, the authors considered the kinematics and dynamics of this parallel-serial manipulator. Finally, Tian et al. [
278] modified their previous design and proposed a robot where all the universal joints of the parallel mechanism were placed on the base (
Figure 10f). The authors studied the robot kinematics and performed its dimensional synthesis.
To conclude this subsection, we would like to mention two manipulators based on 3-DOF parallel mechanisms. The first one is a 5-DOF manipulator with a 3-
PUU translational parallel mechanism considered by Wang et al. [
279] (
Figure 10g). The manipulator was intended for polishing, and the authors optimized the polishing process in their study. The second example is a (3-U
PU)RRR hybrid machining robot introduced by Dou et al. [
280] (
Figure 10h). In their work, the authors performed a comprehensive stiffness analysis of this manipulator.