Collaborative Robots – Power and Force LimitingABB collaborative industrial robot concept Measures...
Transcript of Collaborative Robots – Power and Force LimitingABB collaborative industrial robot concept Measures...
Collaborative Robots – Power and Force Limiting
Björn Matthias ABB Corporate Research
Overview • Human-Robot Collaboration (HRC) • Power and Force Limiting (PFL) • Possible Misconceptions • Biomechanical Limit Criteria • Risk Reduction Measures • Modeling Contact Events • Different Perspectives
– Robot Manufacturer – System Integrator – End-User
• Conclusions and Outlook
Human-Robot Collaboration (HRC)
Conventional industrial robots Collaborative industrial robots
Absolute separation Mixed environment
Discrete safety No HRC
Safety controllers Limited HRC
Harmless manipulators Full HRC
Power and Force Limiting (PFL)
• Form of collaborative operation in which incidental contact between moving robot and human body region can occur
• Contact is not excluded, thus it must be thoroughly understood and controlled to minimize risk
• Risk reduction uses – Suitable robot design – Suitable application design (tooling, work pieces, fixtures, motion patterns, etc.) – Biomechanical limit criteria on contact events – Design and control means to respect limit criteria
PFL – Possible Misconceptions • Use of collaborative robots featuring PFL
does not mean – Simply remove fences without further
considerations – “Safe” robot will render entire application “safe” – Requires a higher safety performance level than
standard industrial robots – Will be too slow for productive applications
Biomechanical Limit Criteria
vrel
F
Transient Contact Quasi-Static Contact
Description • Contact event is “short” (< 50 ms) • Human body part can usually recoil
• Contact duration is “extended” • Human body part cannot recoil, is
trapped
Limit Criteria • Peak forces, pressures, stresses • Energy transfer, power density
• Peak forces, pressures, stresses
Accessible in Design or Control
• Effective mass (robot pose, payload) • Speed (relative) • Contact area, duration
• Force (joint torques, pose) • Contact area, duration
ISO / TS 15066 – clause 5.5.4 “Power and force limiting”
Thre
shol
d fo
r to
uch
sens
atio
n
Thre
shol
d fo
r pai
n se
nsat
ion
Thre
shol
d fo
r low
-le
vel i
njur
y
Thre
shol
d fo
r “S1
” re
vers
ible
inju
ry
Thre
shol
d fo
r “S2
” irr
ever
sible
inju
ry
Collaborative operation
DGUV/IFA, U of Mainz, Faunhofer IFF experiments
DGUV/IFA literature survey; Fraunhofer IFF experiments
Biomechanical Limit Criteria
forces pressure
speed
J. Fryman, B. Matthias, Proceedings of ROBOTIK 2012, Munich http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6309480
mass
torques
Biomechanical Limit Criteria • Safety = freedom from injury
– “A bruise a day …” is unacceptable. • Safety ≠ ergonomically agreeable workplace
– Non-injuring but subjectively painful contacts are not forbidden, but not recommended from an ergonomics perspective.
• Design measures for collaborative robots using PFL technology – Must include sufficient risk reduction to prevent
injury – Can include ergonomics and usability features to
bolster acceptance by the workforce
Risk Reduction Measures Transient Contact Quasi-Static Contact
Mechanical Design • Reduce effective mass • Increase contact area • Increase contact
duration
• Increase contact area
Control Design • Reduce relative speed • (Reduce effective
mass by suitable choices of pose)
• Decrease maximum joint torques, forces
• Decrease contact duration
• Design choices are a balance between performance characteristics and safety requirements
• Safety-related control functions must be designed and implemented according to appropriate choice of safety performance level (PL) / safety integrity level (SIL) and designated architecture (ISO 13849-1, IEC 62061)
Level 1
Level 2
Level 3
Level 4
Robot system – mechanical hazards
Low payload and low robot inertia
Injury-avoiding mechanical design and soft padding
Power and speed limitation
Software-based collision detection, manual back-drivability
Level 5 Personal protective equipment
ABB
colla
bora
tive
indu
stria
l rob
ot c
once
pt
Mea
sure
s fo
r ris
k re
duct
ion
an
d er
gono
mic
s im
prov
emen
t Level 6 Perception-based real-time
adjustment to environment
Oth
er, a
pplic
atio
n-sp
ecifi
c
Qua
si-s
tatic
con
tact
Tran
sien
t co
ntac
t
ABB YuMi® Safety Concept
Modeling Contact Events • Properties of robot • Properties of human
body region • Characteristics of
contact • Describe as inelastic
2-body collision • Estimate worst case
forces and energy transfer
𝑚𝑅 𝑚𝐻
�⃗�𝑟𝑟𝑟
Robot Human body region
Properties of Robot • Effective moving mass at
contact location (reflected inertia) – 𝑚𝑅
• Speed of contact location – �⃗�𝑅
• Material properties of contact location – E.g. padding
• Compliance of kinematic chain – Can reduce effective mass
𝑚1
𝑚2 𝑚3
�⃗�1
�⃗�2 �⃗�3
Example for stiff 3 DOF robot
�⃗�𝑅 = �𝑚𝑖�⃗�𝑖𝑖
𝑚𝑅 =�⃗�𝑅 ∙ �⃗�𝑅𝑣𝑅2
Properties of Human Body Region • Effective mass at
contact location – 𝑚𝐻 • Speed of contact
location – �⃗�𝐻 • Material properties of
contact location – Nearly elastic at
relevant speeds • Compliance of
kinematic chain
5th 50th 95thpercentile percentile percentile
[kg] [%] [kg] [%] [kg] [%]
0 whole body 65.8 100% 82.2 100% 98.5 100%1 head 4.26 6.5% 4.4 5.4% 4.55 4.6%2 neck 0.93 1.4% 1.1 1.3% 1.27 1.3%3 thorax 20.42 31.0% 26.11 31.8% 31.76 32.2%4 abdomen 2.03 3.1% 2.5 3.0% 2.96 3.0%5 pelvis 9.42 14.3% 12.3 15.0% 15.15 15.4%6 upper arm 1.6 2.4% 2.5 3.0% 2.5 2.5%7 forearm 1.18 1.8% 1.45 1.8% 1.72 1.7%8 hand 0.46 0.7% 0.53 0.6% 0.61 0.6%9 hip flap 2.89 4.4% 3.64 4.4% 4.38 4.4%10 thigh minus flap 5.48 8.3% 6.7 8.2% 7.92 8.0%11 calf 3.32 5.0% 4.04 4.9% 4.76 4.8%12 foot 0.84 1.3% 1.01 1.2% 1.18 1.2%
sum 65.71 100% 82.51 100% 97.45 99%5+4+3 torso 31.87 48.4% 40.91 49.8% 49.87 50.6%9+10 thigh 8.37 12.7% 10.34 12.6% 12.3 12.5%7+8 forearm + hand 1.64 2.5% 1.98 2.4% 2.33 2.4%
40 year old American male
http://msis.jsc.nasa.gov/sections/section03.htm
Body Segment Masses
Characteristics of Contact • Duration ∆𝑡 • Compression,
displacement 𝑥 • Contact area and
shape 𝐴 𝑥 • Contact force 𝐹 𝑥 ,
pressure 𝑝 𝑥 • Energy transfer 𝑊(𝑥),
power 𝑃(𝑥)
𝑣0
𝑡 < 𝑡0 = 0
Before contact
𝑡 = 𝑡0
𝑣 = 𝑣0
Contact
𝑡0 < 𝑡 < 𝑡1
𝑣 < 𝑣0
Compression
𝑡 = 𝑡1
𝑣 = 0
Max. compression
Displacement + Contact Area Spherical contact
Cylindrical contact
𝑥 0 𝑥
Contact area 𝐴 𝑥
𝑅
Radius 𝑅 = 𝑅𝑠
𝐴𝑠 𝑥 = 2𝜋𝑅𝑠 𝑥
Radius 𝑅 = 𝑅𝑐
𝐴𝑐 𝑥 = ℓ𝑅𝑐 acos 1 −𝑥𝑅𝑐
;
Length ℓ
𝐴𝑠 𝑅𝑠 = 2𝜋𝑅𝑠2
𝐴𝑐 𝑅𝑐 = 𝜋ℓ𝑅𝑐
Inelastic 2-Body Collision
𝜇�̈� + 𝑏�̇� + 𝑘𝑥 = 0
𝜔02 =𝑘𝜇
𝛾 =𝑏2𝜇
�̈� + 2𝛾�̇� + 𝜔02 𝑥 = 0
Solution, assuming: 𝑘 = 𝑐𝑐𝑐𝑐𝑡., 𝑥 0 = 0, �̇� 𝑡 = 𝑣0
𝑥 𝑡 = 𝑒−𝛾𝛾𝑣0𝜔 sin𝜔𝑡
𝜔2 = 𝜔02 − 𝛾2
�̇� 𝑡 = 𝑣0 𝑒−𝛾𝛾 cos𝜔𝑡 −𝛾𝜔 sin𝜔𝑡
𝜇 =1𝑚𝑅
+1𝑚𝐻
−1
�⃗� = �⃗�𝑅 − �⃗�𝐻
𝑚𝑅 𝑚𝐻
𝑏
𝑘
𝑥
Solve in center-of-mass coordinates:
𝑀 = 𝑚𝑅 + 𝑚𝐻
𝑋 =𝑚𝑅�⃗�𝑅 + 𝑚𝐻�⃗�𝐻𝑚𝑅 + 𝑚𝐻
Total mass:
Reduced mass:
Center-of-mass coordinate:
Relative coordinate:
Differential Equation
Substitutions
Inelastic 2-Body Collision Assuming 𝑘 = 𝜆𝐴 𝑥 with 𝜆 = 𝑐𝑐𝑐𝑐𝑡.
𝜇�̈� + 𝑏�̇� + 𝜆𝐴 𝑥 𝑥 = 0
Spherical contact Cylindrical contact
𝐹𝑠 𝑥 = 𝜋𝜆 𝑅𝑠 𝑥2 𝐹𝑐 𝑥 = 𝜆ℓ𝑅𝑐2 1 − 𝑢2 − 𝑢 acos𝑢
𝑊𝑠 𝑥 =13𝜋𝜆 𝑅𝑠 𝑥3 𝑊𝑐 𝑥 =
14𝜆ℓ 𝑅𝑐3
𝜋2− 3𝑢 1 − 𝑢2 − asin𝑢 + 2𝑢2 acos𝑢
𝑑𝐹𝑑𝑥
= 𝑘 = 𝜆𝐴 𝑥 = 2𝜋𝑅𝑠 𝑥 ; sphericalℓ𝑅𝑐 acos𝑢 ; cylindrical
𝑝𝑠 𝑥 =𝐹𝑠 𝑥𝐴𝑠 𝑥
=12 𝜆𝑥 𝑝𝑐 𝑥 =
𝐹𝑐 𝑥𝐴𝑐 𝑥
= 𝜆𝑅𝑐1 − 𝑢2
acos𝑢− 𝑢
𝑢 = 1 −𝑥𝑅𝑐
𝑃 𝑥 =𝑑𝑊𝑑𝑡
=𝑑𝑊𝑑𝑥
𝑑𝑥𝑑𝑡
= 𝐹 𝑣 power
force
pressure
energy
Force + Pressure Spherical contact Cylindrical contact
v0 = 1000 mm/sλ = 10 N/mm/cm2
µ = 1000 gRsph = 5 mm
v0 = 1000 mm/sλ = 10 N/mm/cm2
µ = 1000 gRcyl = 5 mm
lcyl = 50 mm
Energy Transfer + Power Spherical contact Cylindrical contact
v0 = 1000 mm/sλ = 10 N/mm/cm2
µ = 1000 gRsph = 5 mm
v0 = 1000 mm/sλ = 10 N/mm/cm2
µ = 1000 gRcyl = 5 mm
lcyl = 50 mm
Comparison to Criteria Spherical contact Cylindrical contact
max As = 3.09 cm2 max Fs = 152.28 N max ps = 49.23 N/cm2
max Ws = 0.500 J max Ps = 68.65 W
max Ws/As = 0.1616 J/cm2 max Ps/As = 28.06 W/cm2
max Ac = 4.74 cm2 max Fc = 194.50 N max pc = 41.00 N/cm2
max Wc = 0.500 J max Pc = 89.66 W
max Wc/Ac = 0.1054 J/cm2 max Pc/Ac = 22.87 W/cm2
Example: back of hand, non-dominant
Example: lower arm, muscle
Example: back of hand, non-dominant
Example: lower arm, muscle
Static F < 81 N transient F < 162 N
Static F < 42 N transient F < 84 N
Static F < 81 N transient F < 162 N
Static F < 42 N transient F < 84 N
Static p < 214 N/cm2 transient p < 428 N/cm2
Static p < 162 N/cm2 transient p < 324 N/cm2
Static p < 214 N/cm2 transient p < 428 N/cm2
Static p < 162 N/cm2 transient p < 324 N/cm2
Criteria Criteria
Comparison to Criteria • Factor of ≥ 2
between quasi-static (∆t∞) and transient (∆t ≤ 100 ms)
Y. Yamada et al., IEEE International Conference on Robotics and Automation (ICRA) 1995
Robot Manufacturer – The PFL Robot
• Follow ISO 10218-1 • Follow ISO/TS 15066
– Experimental type tests of PFL robots – Simulation tools for assessment of biomechanical criteria
on PFL robot • Carry out risk assessment on generic level with
respect to robot alone • Generate user information on configuration of safety
functions to meet biomechanical criteria and on exclusions
• Machinery Directive: incomplete machinery, no CE marking
System Integrator – The PFL System
• Rely on work of robot manufacturer • Follow ISO 10218-2 • Follow ISO/TS 15066
– Computation of mechanical loading of exposed body parts in use cases
– Determination of speed limits to respect biomechanical limits • Do application level risk assessment with respect to robot,
tooling, work pieces, … – Intended use – Reasonably foreseeable misuse
• Generate user information on intended use of equipment • Machinery Directive: completed machinery, needs CE
marking in EU
End-User – The PFL Application
• Rely on work of integrator • Observe user information on intended
use and on exclusions • Machinery Directive: need CE marked
equipment
Conclusions and Outlook • PFL-type collaborative robots are on the verge of larger scale
deployment • Biomechanical limit criteria for transient and quasi-static
contact are under development • Important risk reduction methods for PFL robots have been
identified • Model-based estimations of transient contact events can be
very helpful in estimating whether or not a particular case is permissible or not
• Roles of manufacturer, integrator and end-user are different in their responsibilities and in their needs for data and calculation methods to estimate biomechanical loading data
Bjoern Matthias Senior Principal Scientist ABB Corporate Research Tel: +49 6203 716145 Email: [email protected]