RMI: Evaluation metrics and

design methods
Prof. Dr.-Ing. habil. Philipp Beckerle
Pose eines starren Körpers
Recap

• Fester Referenzrahmen 𝑂𝑥𝑦𝑧

• Pose des starren Körpers 𝑂𝑥′ ′ 𝑦 ′ 𝑧 ′ :

Position (Vektor) Orientierung (Rotationsmatrix)

𝑝𝑥 𝑥𝑥′ 𝑦𝑥′ 𝑧𝑥′
𝐩 = 𝑝𝑦 𝐑 = 𝐱′ 𝐲′ 𝐳 ′ = 𝑥𝑦′ 𝑦𝑦′ 𝑧𝑦′
𝑝𝑧
𝑥𝑧′ 𝑦𝑧′ 𝑧𝑧′

(Siciliano et al., 2010)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 2
Ausrichtung von aufeinander folgenden [Link]
Recap

Aus DH-Parametern Aus den Vektorparametern

−𝑐1 𝑠4 − 𝑠1 𝑐2 𝑐4 −𝑐1 𝑐4 + 𝑠1 𝑐2 𝑠4 𝑠1 𝑠2 𝑝𝑥 𝑐1 𝑐4 − 𝑠1 𝑐2 𝑠4 −𝑐1 𝑠4 − 𝑠1 𝑐2 𝑐4 𝑠1 𝑠2 𝑝𝑥

−𝑠1 𝑠4 + 𝑐1 𝑐2 𝑐4 −𝑠1 𝑐4 − 𝑐1 𝑐2 𝑠4 −𝑐1 𝑠2 𝑝𝑦 𝑠 𝑐 + 𝑐1 𝑐2 𝑠4 −𝑠1 𝑠4 + 𝑐1 𝑐2 𝑐4 −𝑐1 𝑠2 𝑝𝑦
𝐀𝟎𝟒 = 𝐀𝟎𝟓 = 1 4
𝑠2 𝑐4 −𝑠2 𝑠4 𝑐2 𝑝𝑧 𝑠2 𝑠4 𝑠2 𝑐4 𝑐2 𝑝𝑧
0 0 0 1 0 0 0 1
𝑝𝑥 = − 𝑙1 𝑠1 − 𝑐2 𝑠1 𝑙2 + 𝑙3 + 𝑞3 − ℎ3 𝑠1 𝑠2 − 𝑙4 𝑐1 𝑠4 + 𝑐2 𝑐4 𝑠1
𝑠𝑖 = sin 𝑞𝑖
𝑝𝑦 = 𝑙1 𝑐1 + 𝑐1 𝑐2 𝑙2 + 𝑙3 + 𝑞3 + ℎ3 𝑐1 𝑠2 − 𝑙4 𝑠1 𝑠4 − 𝑐1 𝑐2 𝑐4 𝑐𝑖 = cos 𝑞𝑖
𝑝𝑧 = ℎ0 + ℎ1 + 𝑠2 (𝑙2 + 𝑙3 + 𝑞3 ) − ℎ3 𝑐2 + 𝑙4 𝑐4 𝑠2

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 3
 Jacobi-Matrix
Recap

• Die Geschwindigkeit der externen Koordinaten kann wie folgt ausgedrückt werden:

𝒅 𝝏𝒑
𝒑ሶ = 𝒑 𝒒 = 𝒒ሶ = 𝑱𝒒ሶ
𝒅𝒕 𝝏𝒒

𝜕𝑝1 𝜕𝑝1 𝜕𝑝1

𝜕𝑞1 𝜕𝑞2
⋯ 𝜕𝑞𝑛
𝜕𝑝2 𝜕𝑝2 𝜕𝑝2
𝝏𝒑 ⋯
Dabei ist 𝐉 = = 𝜕𝑞1 𝜕𝑞2 𝜕𝑞𝑛 die 𝑚 × 𝑛 Jacobi-Matrix
𝝏𝒒
⋮ ⋮ ⋱ ⋮
𝜕𝑝𝑚 𝜕𝑝𝑚 𝜕𝑝𝑚
⋯
𝜕𝑞1 𝜕𝑞2 𝜕𝑞𝑛
Die Jacobi-Matrix ist die Grundlage für die
Differentialkinematik, d. h. die Beziehung zwischen
den Geschwindigkeiten der inneren und äußeren
Koordinaten.

(Siciliano et al., 2010; Lenarcic et al., 2012)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 4
Singuläre Posen
Recap

a) Singularität der ersten Art : det 𝐽 = 0

Die interne Bewegung wird nicht vollständig auf die externe Bewegungen bzw. die Ausgangsbewegung
übertragen, d. h. ein oder mehrere Freiheitsgrade gehen verloren.

b) Singularität der zweiten Art: det 𝐽−1 = 0

Die inneren Kräfte oder Momente werden nicht vollständig auf die äußeren Kräfte oder Momente bzw.
die abtriebsseitigen Kräfte oder Momente übertragen.

passives Gelenk
aktives Gelenk/Aktor
(Hatzfeld & Kern, 2016)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 5
Agenda

01 Workspace analysis

02 Kinematic flexibility and singularity

03 Manipulability

04 Design approaches

05 Summary

This lecture is being recorded. The unedited video will be made available to all students through [Link] and StudOn.

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 6
Workspace definition
01 Workspace analysis

• The workspace of a mechanism is the spatial volume containing all

possible poses of the end-effector.

• In the workspace region, the mechanism possesses certain

properties, e.g., the ability to transfer loads or reach certain velocities.

• The mechanism workspace depends on its structure, i.e., lengths and

arrangement of segments and constraints in joint motion.

Reachable workspace → all reachable positions ([Link])

Dexterous workspace → all reachable positions with all possible orientations

(Lenarcic et al., 2012)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 7
Reachable workspace
01 Workspace analysis

• The reachable workspace of a mechanism is the spatial region

containing all positions 𝒓 which can be reached by the end-effector
of a mechanism, regardless of its orientation.

• Reachability is an important kinematic property for

robot/mechanism design.

• Even despite having a large reachable workspace, motion 𝐫

possibilities might be limited in some areas as orientation is not 𝑧
guaranteed.
𝑥 𝑦
([Link])

(Lenarcic et al., 2012)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 8
Reachable workspace based on forward kinematics
01 Workspace analysis

• Generally, the workspace of a mechanism is determined by calculating forward kinematics 𝒓 𝒒 for all
possible values of internal (joint) coordinates 𝒒.

• Practically, such calculations are not feasible as they would require solving nested infinite loops.

Solution: discretize joint coordinates 𝑞𝑖 with a finite number of steps:

𝑞𝑖,𝑚𝑎𝑥 −𝑞𝑖,𝑚𝑖𝑛
𝑞𝑖 𝑘𝑖 = 𝑞𝑖,𝑚𝑖𝑛 + 𝑘𝑖 , 𝑘𝑖 = 0,1, … , 𝐾𝑖
𝐾𝑖

𝐾𝑖 + 1 different values are assigned to each joint coordinate,

and 𝑖 = 1,2, … 𝑛 corresponds to each joint position.

• The number of combinations of 𝒒 is 𝐾1 + 1 𝐾2 + 1 … 𝐾𝑛 + 1

• The reachable workspace comprises all vectors r for the examined joint configuration
(Lenarcic et al., 2012)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 9
Reachable workspace based on inverse kinematics
01 Workspace analysis

• Only practical if a closed-form solution of the inverse kinematics problem exists.

• The space around the mechanism, where the reachable workspace is expected, is divided into a finite
number of candidate positions 𝒓𝒊 = 𝑟𝑥,𝑖 𝑟𝑦,𝑖 𝑟𝑧,𝑖 𝑇 :
𝑟𝑥𝑚𝑎𝑥 − 𝑟𝑥𝑚𝑖𝑛
𝑟x, i = 𝑟𝑥𝑚𝑖𝑛 + 𝑘𝑥 , 𝑘𝑥 = 0,1, … . 𝐾𝑥
𝐾𝑥
𝑟𝑦𝑚𝑎𝑥 − 𝑟𝑦𝑚𝑖𝑛
𝑟y, i = 𝑟𝑦𝑚𝑖𝑛 + 𝑘𝑦 , 𝑘𝑦 = 0,1, … . 𝐾𝑦
𝐾𝑦
𝑟𝑧𝑚𝑎𝑥 − 𝑟𝑧𝑚𝑖𝑛
𝑟𝑧, 𝑖 = 𝑟𝑧𝑚𝑖𝑛 + 𝑘𝑧 , 𝑘𝑧 = 0,1, … . 𝐾𝑧
𝐾𝑧
1. Solve inverse kinematics problem 𝒒(𝒓) for each candidate 𝒓𝐢 .
2. Check if inverse kinematics solution exists within joint
constraints 𝑞𝑖,𝑚𝑖𝑛 ≤ 𝑞𝑖 ≤ 𝑞𝑖,𝑚𝑎𝑥 .
3. 𝑟 is an element of the reachable workspace.
(Lenarcic et al., 2012)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 10
Dexterous workspace
01 Workspace analysis

• The dexterous workspace is a region encompassing all positions

𝒓 which can be reached by the end-effector of a mechanism with
all possible orientations (𝜓, 𝜃, 𝜙).

• It is a subarea of the reachable workspace.

• The requirement that orientation in all directions must be possible (𝜓, 𝜃, 𝜙)

is very strict for practical situations. Accordingly, other
understandings of dexterous workspace consider more relaxed r
criteria, e.g., orientations in certain directions. Reachable 𝑧
workspace

Dexterous
workspace 𝑥 𝑦
([Link])

(Lenarcic et al., 2012; Bajd et al., 2010)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 11
Dexterous workspace based on FK and IK
01 Workspace analysis

Orientation criteria: A position is considered to be within the dexterous workspace if it attains all orientational
angles within the following intervals:
𝜓𝑚𝑎𝑥 − 𝜓𝑚𝑖𝑛 = 2𝜋, 𝜃𝑚𝑎𝑥 − 𝜃𝑚𝑖𝑛 = 𝜋, 𝜙𝑚𝑎𝑥 − 𝜙𝑚in = 2𝜋.
• The dexterous workspace can be determined by forward kinematics
𝒑 𝒒 = 𝒓, 𝜓, 𝜃, 𝜙 (as for reachability) and checking the orientation criteria.
• The dexterous workspace can be determined via inverse kinematics (similar to reachability) adding a
discretization of Euler angles:
𝜓𝑚𝑎𝑥 − 𝜓𝑚𝑖𝑛
𝜓 = 𝑘𝜓 , 𝑘𝜓 = 0,1 … 𝐾𝜓
𝐾𝜓
𝜃𝑚𝑎𝑥 − 𝜃𝑚𝑖𝑛
𝜃 = 𝑘𝜃 , 𝑘𝜃 = 0,1 … 𝐾𝜃
𝐾𝜃 Particular positions belong to
𝜙𝑚𝑎𝑥 − 𝜙𝑚𝑖𝑛 dexterous workspace
𝜙 = 𝑘𝜙 , 𝑘𝜙 = 0,1 … 𝐾𝜙 if an IK solution exists including
𝐾𝜙
all discrete orientations.
(Lenarcic et al., 2012)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 12
Agenda

01 Workspace analysis

02 Kinematic flexibility and singularity

03 Manipulability

04 Design approaches

05 Summary

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 13
Kinematic flexibility I
02 Kinematic flexibility and singularity

The kinematic flexibility 𝑓𝑎 of a serial • Kinematic flexibility of discrete poses can be

mechanism can be evaluated for each end- determined:
effector pose as the number of joint • via forward kinematics by grouping
configurations that lead to the desired r. configurations that attain approximately the
same pose

• via analytical inverse kinematics by

determining the number of possible
solutions for each pose

The aim may be to design mechanisms with a high

flexibility in order to more efficiently comply with the
requirements of the given tasks.

(Lenarcic et al., 2012)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 14
Kinematic flexibility II
02 Kinematic flexibility and singularity

As a global measure, the relative kinematic flexibility evaluates the entire workspace. It averages kinematic
flexibility in different subvolumes 𝑉1 , 𝑉2 , … of the workspace:

1
𝑓𝑟 = 𝑓 𝑉 + 𝑓𝑎,2 𝑉2 + ⋯
𝑉 𝑎,1 1

𝑟𝑥𝑚𝑎𝑥 − 𝑟𝑥𝑚𝑖𝑛
Workspace can be discretized in equally-sized 𝑟x, i = 𝑟𝑥𝑚𝑖𝑛 + 𝑘𝑥 , 𝑘𝑥 = 0,1, … . 𝐾𝑥
volumes to compute kinematic flexibility for one 𝐾𝑥
representative position in each discrete volume. 𝑟𝑦𝑚𝑎𝑥 − 𝑟𝑦𝑚𝑖𝑛
𝑟y, i = 𝑟𝑦𝑚𝑖𝑛 + 𝑘𝑦 , 𝑘𝑦 = 0,1, … . 𝐾𝑦
𝐾𝑦
Since all volumes are equal, the relative kinematic
𝑟𝑧𝑚𝑎𝑥 − 𝑟𝑧𝑚𝑖𝑛
flexibility is the average of the kinematic flexibility 𝑓𝑎 𝑟𝑧, 𝑖 = 𝑟𝑧𝑚𝑖𝑛 + 𝑘𝑧 , 𝑘𝑧 = 0,1, … . 𝐾𝑧
for all 𝐾𝑥 + 1 𝐾𝑦 + 1 𝐾𝑧 + 1 positions. 𝐾𝑧

(Lenarcic et al., 2012)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 15
Singularity
02 Kinematic flexibility and singularity

Singularity decreases the mobility and controllability of a mechanism and is an important

kinematic property .

At singular configurations, the mechanism loses the ability to move in particular directions or
transmit forces/torques from the internal to the external coordinates.

Mathematically, it occurs when the Jacobian matrix degenerates, i.e., becomes singular.

Singularity of the first kind (𝑑𝑒𝑡 𝑱 = 0):

Internal motion is not completely transferred.

Singularity of the second kind (𝑑𝑒𝑡 𝑱−1 = 0):

Internal forces or torques are not completely transmitted.

(Hatzfeld & Kern, 2016)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 16
Example: Singularity of serial mechanism
02 Kinematic flexibility and singularity

r3
• Consider the geometric Jacobian of the 3-DoF anthropomorphic arm:
𝒛 × (𝒓𝟑 − 𝒓𝟎 ) 𝒛𝟏 × (𝒓𝟑 − 𝒓𝟏 ) 𝒛𝟐 × (𝒓𝟑 − 𝒓𝟐 )
𝐉= 𝟎
𝒛𝟎 𝒛𝟏 𝒛𝟐
𝟎 𝒂𝟐 𝒄𝟏 𝒄𝟐 𝟎 r2
𝒓𝟎 = 𝒓𝟏 = 𝟎 𝒓𝟐 = 𝒂𝟐 𝒔𝟏 𝒄𝟐 𝒛𝟎 = 𝟎
𝟎 𝒂𝟐 𝒔𝟐 𝟏

𝒄𝟏 (𝒂𝟐 𝒄𝟐 + 𝒂𝟑 𝒄𝟐𝟑 ) 𝒔𝟏
𝐫𝟑 = 𝒔𝟏 (𝒂𝟐 𝒄𝟐 + 𝒂𝟑 𝒄𝟐𝟑 ) 𝐳𝟏 = 𝐳𝟐 = −𝒄𝟏
𝒂𝟐 𝒔𝟐 + 𝒂𝟑 𝒔𝟐𝟑 𝟎 Cause of independence

Cause of −𝑠1 (𝑎2 𝑐2 + 𝑎3 𝑐23 ) −𝑐1 (𝑎2 𝑠2 + 𝑎3 𝑠23 ) −𝑎3 𝑐1 𝑠23

independence 𝑐1 (𝑎2 𝑐2 + 𝑎3 𝑐23 ) −𝑠1 (𝑎2 𝑠2 + 𝑎3 𝑠23 ) −𝑎3 𝑠1 𝑠23
Linearly independent
0 𝑎 𝑐
2 2 + 𝑎 𝑐
3 23 𝑎 𝑐
3 23
J=
0 𝑠1 𝑠1
0 −𝑐1 −𝑐1
1 0 0 (Siciliano et al., 2010)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 17
In-class exercise: Singularity of serial mechanism
02 Kinematic flexibility and singularity

−𝑠1 (𝑎2 𝑐2 + 𝑎3 𝑐23 ) −𝑐1 (𝑎2 𝑠2 + 𝑎3 𝑠23 ) −𝑎3 𝑐1 𝑠23

𝐉𝑃 = 𝑐1 (𝑎2 𝑐2 + 𝑎3 𝑐23 ) −𝑠1 (𝑎2 𝑠2 + 𝑎3 𝑠23 ) −𝑎3 𝑠1 𝑠23
0 𝑎2 𝑐2 + 𝑎3 𝑐23 𝑎3 𝑐23
Determine the singular configurations of the serial mechanism

𝑑𝑒𝑡 𝐽𝑃 = −𝑎2 𝑎3𝑠3(𝑎2𝑐2 + 𝑎3𝑐23) = 0

(a)

The determinant vanishes if:

(a) 𝜗3 = 0 or 𝜗3 = 𝜋
(outstretched or retracted elbow → elbow singularity)

(b) 𝑎2 𝑐2 + 𝑎3 𝑐23 = 0 (b)

(end-effector lies on axis 𝑧𝑜 → shoulder singularity)
(Siciliano et al., 2010)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 18
Exploiting singularity: clutches
02 Kinematic flexibility and singularity

• Singularity generally causes undesired behaviors in mechanisms.

• Singular positions are usually avoided in motion and force control.

Unlocked Locked
However, singular poses can be exploited for
certain applications, e.g., by singular clutches.

These locking devices make use of the fact that

singularities cause a transfer ratio going towards
infinity. Hence, in such a singular position, the
locking device has an infinitely high locking force
and an infinitely small unlocking force.

(Plooij et al., 2015)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 19
Exploiting singularity: manipulator
02 Kinematic flexibility and singularity

The task of a mechanism is to bring an object from a pose 𝑃1 into a pose 𝑃2 . With only 2 DoFs, this mechanism
cannot attain arbitrary orientations and the desired orientation of 𝑃2 would normally be unreachable. However,
we can do the following:

• The object is moved to 𝑃0 (singular configuration) where the mechanism can rotate around the axis of 𝑞1
without changing the position of the end-effector.
• The robot releases the object in 𝑃0 and after rotating the mechanism by an angle of ∆𝑞1 , the object is re-
grasped and taken to 𝑃2 .
• The final orientation at 𝑃2 is different from that which would have been obtained without releasing and re-
grasping the object in the singular configuration by ∆𝑞1 .

(Lenarcic et al., 2012)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 20
Singularity example: DELTA mechanism I
02 Kinematic flexibility and singularity

• From inverse kinematics (lecture 3), we know that:

𝑥𝐵𝑖 𝑎 cos 𝜃1𝑖 + 𝑏 sin 𝜃3𝑖 cos 𝜃1𝑖 + 𝜃2𝑖
𝑦𝐵𝑖 = 𝑏 cos 𝜃3𝑖
𝑧𝐵𝑖 𝑎 sin 𝜃1𝑖 + 𝑏 sin 𝜃3𝑖 sin 𝜃1𝑖 + 𝜃2𝑖

𝑥𝐵𝑖 cos −𝜙𝑖 sin −𝜙𝑖 0 𝑥𝑃 𝑟𝑇𝐶𝑃 − 𝑟𝑏𝑎𝑠𝑒

𝑦𝐵𝑖 = −sin −𝜙 cos −𝜙𝑖 0 ∙ 𝑦𝑃 + 0
𝑖
𝑧𝐵𝑖 0 0 1 𝑧𝑃 0
• The IK solution is:
𝑦𝐵𝑖
𝜃3𝑖 = cos −1
𝑏
2 + 𝑦 2 + 𝑧 2 − 𝑎2 − 𝑏2
𝑥𝐵𝑖 𝐵𝑖 𝐵𝑖
𝜃2𝑖 = cos −1
2𝑎𝑏 sin 𝜃3𝑖

−1
𝑥𝐵𝑖 − 𝑏 sin 𝜃3𝑖 cos( 𝜃1𝑖 + 𝜃2𝑖 )
𝜃1𝑖 = tan
𝑧𝐵𝑖 − 𝑏 sin 𝜃3𝑖 sin( 𝜃1𝑖 − 𝜃2𝑖 ) (Hatzfeld & Kern, 2016)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 21
Singularity example: DELTA mechanism II
02 Kinematic flexibility and singularity

The inverse Jacobian can then be computed as:

𝑐𝑜𝑠(𝜃1𝑖 + 𝜃2𝑖 ) 𝑠𝑖𝑛 𝜃3𝑖 𝑐𝑜𝑠 𝜙𝑖 − 𝑐𝑜𝑠 𝜃3𝑖 𝑠𝑖𝑛 𝜙𝑖
𝑗𝑖1 =
𝑗11 𝑗12 𝑗13 𝑎 𝑠𝑖𝑛 𝜃2𝑖 𝑠𝑖𝑛 𝜃3𝑖
−1
𝐽 = 𝑗21 𝑗22 𝑗23
𝑗31 𝑗32 𝑗33 𝑐𝑜𝑠(𝜃1𝑖 + 𝜃2𝑖 ) 𝑠𝑖𝑛 𝜃3𝑖 𝑐𝑜𝑠 𝜙𝑖 +𝑐𝑜𝑠 𝜃3𝑖 𝑠𝑖𝑛 𝜙𝑖
𝑗𝑖2 =
𝑎 𝑠𝑖𝑛 𝜃2𝑖 𝑠𝑖𝑛 𝜃3𝑖
𝑑𝑒𝑡 𝐽−1 = 𝑗11 𝑗22 𝑗33 + 𝑗12 𝑗23 𝑗31 + 𝑗13 𝑗21 𝑗32 − 𝑗11 𝑗23 𝑗32 − 𝑗12 𝑗21 𝑗33 − 𝑗13 𝑗22 𝑗31
𝑠𝑖𝑛( 𝜃1𝑖 + 𝜃2𝑖 )
𝑗𝑖3 =
𝑎 𝑠𝑖𝑛 𝜃2𝑖
Kinematic singularity of the first kind occurs when 𝑑𝑒𝑡 𝐽−1 = ∞ at
• 𝜃2𝑖 = 0 or 𝜃2𝑖 = 𝜋 for any 𝑖
• 𝜃3𝑖 = 0 or 𝜃3𝑖 = 𝜋 for any 𝑖

Kinematic singularity of the second kind occurs when 𝑑𝑒𝑡 𝐽−1 = 0 at

• 𝜃1𝑖 + 𝜃2𝑖 = 0 or 𝜃1𝑖 + 𝜃2𝑖 = 𝜋 for all 𝑖

(Hatzfeld & Kern, 2016)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 22
Agenda

01 Workspace analysis

02 Kinematic flexibility and singularity

03 Manipulability

04 Design approaches

05 Summary

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 23
Manipulability ellipsoids
03 Manipulability

For given FK 𝒑𝟎 (𝒒𝟎 ), a displacement 𝑑q leads to the displacement 𝑑p in external coordinates.

When 𝑑q is of equal length for all directions, i.e., when the vector 𝑑𝒒 lies on the surface of an n-
dimensional sphere (𝑑𝒒𝑇 𝑑𝒒 = 1), then 𝑑𝒑 lies on the surface of an m-dimensional ellipsoid,
called the manipulability ellipsoid.

• The singular values 𝜎𝑖 , 𝑖 = 1,2, … , 𝑚 of

the Jacobian matrix J determine the
width of the manipulability ellipsoid in
the direction of the eigenvectors v𝑖 .
• The manipulability of a mechanism is
defined as the product of the singular
values of the Jacobian matrix:
𝑚

𝑀 = ෑ 𝜎𝑖 = |𝑑𝑒𝑡(𝐽)|
𝑖=1

(Lenarcic et al., 2012)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 24
Kinematic index
03 Manipulability

• Manipulability 𝑀 informs about the volume of the manipulability

ellipsoid. The extend of the ellipsoid can be evaluated based on the
singular values with the kinematic index is defined as:

𝜎min (𝐉)
𝜅=
𝜎max (𝐉)

• 𝜅 is a normalized magnitude which describes the sphericity of

the manipulability ellipsoid.

• The ellipsoid is closer to spherical (isotropic) for 𝜅 → 1 and ([Link])

flattens for 𝜅 → 0 (singular).
(Lenarcic et al., 2012)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 25
Velocity ellipsoid
03 Manipulability

• If the joint velocity vector 𝒒ሶ lies on the surface of an n-dimensional sphere:

𝒒ሶ 𝑻 𝒒ሶ = 𝟏

• Given the relationship between internal and external velocities:

𝒒ሶ = 𝑱(𝒒)−𝟏 𝒑ሶ

𝑻 −𝟏
• Combining both we obtain: 𝒑ሶ 𝑻 𝑱 𝒒 𝑱 𝒒 𝒑ሶ = 𝟏 (velocity ellipsoid)

• The velocity ellipsoid is defined only in 3-dimensional

space. The axes and lengths coincide with the eigenvectors
and singular values of the translational part of the Jacobian.

• The velocity ellipsoid coincides with the manipulability

ellipsoid in translational coordinates.
(Lenarcic et al., 2012)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 26
Force ellipsoid
03 Manipulability

• If the internal force/torque vector 𝝉 lies on the surface of an n-dimensional sphere:

𝝉𝑻 𝝉 = 𝟏

• Given the relationship between internal and external forces/torques 𝑭:

𝝉 = 𝑱(𝒒)𝑻 𝑭

• Combining both we obtain: 𝑭𝑻 𝑱 𝒒 𝑱 𝒒 𝑻

𝑭 = 𝟏 (force ellipsoid)

• The axis of the force ellipsoid (shaded area) coincide with

those of the velocity ellipsoid (dashed area).

• However, the lengths are reciprocal to the singular value of

the translational singular values of the Jacobian.

This also means that we need to make design decisions here, i.e., usually a compromise.
(Lenarcic et al., 2012)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 27
Agenda

01 Workspace analysis

02 Kinematic flexibility and singularity

03 Manipulability

04 Design approaches

05 Summary

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 28
Product planning procedure - VDI 2220
04 Design approaches

Market Environment Company

1 Analyse the situation

Situation analysis
2 Formulate search strategies
Product planning

Search field
3 Find product ideas

Product ideas
4 Select product ideas

Selected product idea

5 Define products

Product proposal
6 Clarify and elaborate
Conceptual
design

Requirement list
Design
(Pahl et al., 2007)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 29
Requirement list
04 Design approaches

Requirements are obtained from the design specification.

Possible requirement categories:

• Kinematics • Production
• Geometry • Assembly
• Forces • Transport
• Energy • Operation
• Material • Maintenance
• Safety • Expenditure
• Ergonomics • Environmental impact
Actually, mechanism design would touch many
of those and not only kinematics.
(Pahl et al., 2007)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 30
Conceptual design
04 Design approaches

Requirements list
Information
Abstract to identify the essential problems
Definition
Establish function structures / overall function & subfunctions

Conceptual design
Search for working principles Creation

Combine working principles into working structures

Select suitable combinations

Elaboration into principle solution variants

Evaluate variants against technical and economic criteria Evaluation

Decision
Principle solution (Concept)
(Pahl et al., 2007)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 31
Establishing function structures
04 Design approaches

• Once the task has been defined it is possible to represent the overall function that relate the inputs and
outputs of a plant, machine, or assembly.

• This overall function can be represented graphically with a block diagram that indicates the function
structure, divided in several subfunctions, and the flow of energy, material, and signals.

• The function structure should be as accurate as possible.

(Pahl et al., 2007)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 32
Functional structure example: active knee prosthesis
04 Design approaches

Electrical Power Management Actuators

energy store, provide, manage Torque generation, Heat loss
variable stiffness
Femural
motion
Mechanics
Ground reaction Kinematics/geometr Movement
forces Control y, statics, dynamics
Data collection, processing,
Data management
Damping Interface
radial, axial
Sensors
Measurement

Electrical Energy
Signal / Data
Mechanical energy
Thermal energy
Energy recuperation

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 33
Developing working structures I
04 Design approaches

Once the task is well-defined, working principles need to be found for the different subfunctions,
and these principles are combined into a working structure.

• A working principle implements a function through • A working structure is a combination of several

an appropriate physical principle, geometry and working principles to generate an overall function.
material.

Combinations can be developed considering the

Principles can be found by literature research,
function structures, practical associations, etc.
analyzing natural and technical systems,
Elaborating a working structure yields a solution
intuition-based methods, and design catalogues.
variant.

• This design step intents to find multiple solution variants by varying design features that might fulfill the
required function.
(Pahl et al., 2007)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 34
Developing working structures II
04 Design approaches

• Selecting solution variants can be challenging

depending on their number.

• A systematic selection approach, e.g., selection

charts, could facilitate eliminating unsuitable
solutions and identifying the “best” solutions.

• At this step, the selection is done based on the

function structure to ensure fulfilling technical
functionality (leaving out secondary objectives).

(Pahl et al., 2007)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 35
Developing concepts
04 Design approaches

• Solution variants are elaborated into principle solutions. Their evaluation demands for further:
− Technical analyses, e.g., basic calculations and sketches
− Research, e.g., preliminary experiments, testing, modelling, simulation, market research, etc.,
to define complex and suitable solutions that fulfill all (or most) of the requirements.

• A concept is a solution variant that fulfills not only the technical function, but also secondary
requirements, e.g., economic feasibility and safety demands.

• The solution variants should be systematically evaluated considering the fulfillment of technical and
economic criteria, e.g., VDI 2225.

(Pahl et al., 2007)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 36
Correlation of methods with conceptual design steps
04 Design approaches

Product Clarifying the task Abstracting to Establishing Searching Combining Selecting Elaborating Evaluating
Steps
planning, Setting up the identify essential function for working working suitable principle principle
Methods
Task selection requirement list problems structures principles principles combinations solutions solutions
Trend studies
MAIN SUPPORT
Market analysis
Requirement list MAIN SUPPORT

Abstraction MAIN SUPPORT

Black box representation
SUPPORT MAIN
Function structure
Literature search SUPPORT SUPPORT MAIN SUPPORT

Analysis: natural systems SUPPORT MAIN

Analysis: known solutions SUPPORT MAIN MAIN MAIN SUPPORT

Analysis: mathematical,
MAIN MAIN
physical relationships
Tests, measurements MAIN MAIN MAIN
Brainstorming, Gallery
SUPPORT MAIN
method, Synectics
Systematic study of
MAIN
physical processes
Classification schemes MAIN MAIN

Design catalogues MAIN MAIN

Sketches
SUPPORT MAIN MAIN
Intuitive improvements
Selection methods SUPPORT SUPPORT MAIN MAIN SUPPORT

(Pahl et al., 2007)

Evaluation methods MAIN
Quality assurance
SUPPORT SUPPORT MAIN MAIN
methods
Costing methods SUPPORT MAIN

Value analysis SUPPORT SUPPORT

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 37
General design approach - VDI 2221
04 Design approaches

Stages Results Phases

Task
Iterate towards and backwards between previous and following stages

1 Clarify and define the task Phase I

Specification

2 Determine functions and their structure

Fulfill and Adapt Requirements

Function structure
Phase II
3 Search for solution principles and their combinations
Principle solutions

4 Divide into realizable modules

Module structures
Phase III
5 Develop layout of key modules
Preliminary layouts

6 Complete overall layout

Definitive layouts
Phase IV
7 Prepare production and operating instructions
Product documents

Further realization
(Jänsch & Birkhofer, 2006; Pahl et al., 2007)

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 38
Agenda

01 Workspace analysis

02 Kinematic flexibility and singularity

03 Manipulability

04 Design approaches

05 Summary

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 39
Your point of view?
05 Summary

What are the main takeaways from this lecture?

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 40
Take Home Message
05 Summary

1. When selecting a mechanism we might consider its workspace, either reachable or dexterous. Shape
and size should be aligned to the desired task.
2. Mechanism design should aim for high kinematic flexibility. Singular configurations can both be
challenging or beneficial for particular tasks. Generally, kinematic properties are design-specific.
3. Manipulability describes how motion is transferred from joint space to workspace. Velocity and force
ellipsoids do the same for the corresponding quantities.
4. Selecting a particular inverse kinematic solution might be done by considering different criteria, e.g.,
kinematic index, manipulability, joint limits, obstacle avoidance, etc. Nevertheless, mechanism analysis
should consider static and dynamic properties.
5. Engineering design methods provide various approaches to implement systematic mechanism design.
Appropriate methods can yield an efficient and objective design process and outcome.

This lecture is being recorded. The unedited video will be made available to all students through [Link] and StudOn.

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 41
Literature
05 Summary

• Lenarcic, J., Bajd, T., & Stanišić, M. M. (2012). Robot mechanisms (Vol. 60). Springer Science & Business Media.
• Bajd, T., Mihelj, M., Lenarčič, J., Stanovnik, A., & Munih, M. (2010). Robotics (Vol. 43). Springer Science & Business
Media.
• Hatzfeld, C., & Kern, T. A. (2016). Engineering Haptic Devices. Springer London Limited.
• Siciliano, B., Sciavicco, L., Villani, L., & Oriolo, G. (2010). Robotics: modelling, planning and control. Springer
Science & Business Media.
• Jänsch, J., & Birkhofer, H. (2006). The development of the guideline VDI 2221-the change of direction. In DS 36:
Proceedings DESIGN 2006, the 9th International Design Conference, Dubrovnik, Croatia.
• Norton, R. L. (2004). Design of machinery: an introduction to the synthesis and analysis of mechanisms and
machines. Boston: McGraw-Hill Higher Education.
• Pahl, G., Beitz, W., Feldhusen, J., & Grote, K. (2007). Engineering Design: A Systematic Approach (K. Wallace & L.
Blessing, Trans.). Springer Science and Business Media.
• Plooij, M., Mathijssen, G., Cherelle, P., Lefeber, D., & Vanderborght, B. (2015). Lock your robot: A review of locking
devices in robotics. IEEE Robotics & Automation Magazine, 22(1), 106-117.

Evaluation metrics and design methods | Robot mechanisms and user interfaces | ASM @ FAU Erlangen-Nürnberg | P. Beckerle 18. November 2024 42