1. Abstract— Extensive cost, time and effort associated with
setting up a production line often inhibits transitioning novel
ideas into commercialized products. Sustainable revenue
generation in such enterprises requires production in large
quantities in order to lower the unit cost, and thus improving
the marketability. Although this model has been successful and
is being followed by most of the industries today, it also limits
the scope for new, non-conventional products; especially in the
early stages of development where manufacturing risks are
higher. One way of reducing this risk can be through improving
the reusability of the manufacturing hardware, thus allowing
quick and inexpensive transition among different iterations and
even different products with the same set of hardware. In this
paper, we present such a novel and revolutionary solution for
flexible manufacturing. Consisting of a unique set of hardware
and software innovations, our proposed system offers a viable
way to low volume manufacturing and low risk prototyping of
novel product ideas. Called as the “Modular and
Reconfigurable Manufacturing Cell (MRMC)”, this proposed
system enables quick and easy setting up of a fully automated
robotic manipulation and assembly platform, optimized for
specific products. The system relies on a novel and proprietary
multifunctional interconnect design, built-in to various
hardware modules of the system, and a distributed intelligence
based self-locating software architecture to achieve almost any
possible assembler configuration that is suitable for a specific
set of tasks. Necessary and sufficient precision level is
maintained via a novel precision optimized hybrid controller
and path planner throughout the automation. Competitive
specifications in terms of travel range, resolution, accuracy,
repeatability, force output, size, weight, power ratings etc., as
compared to standard commercial manipulators, has been
experimentally verified for the proposed Modular and
Reconfigurable Manufacturing Cell (MRMC).
I. INTRODUCTION
Today’s industries heavily depend on automation, mass
production and reliable output for the commercial success of
their products. These factors, however, do not come cheap as
they require highly sophisticated equipment in order to
maintain the desired yield, throughput and reliability.
Generally, almost all of such equipment are dedicated
systems for specific tasks; extremely expensive; and require
advanced skills to operate them. Veering from conventional
and already well-established technologies with such an
This research is supported by US Office of Naval Research (ONR).
Aditya N. Das, PhD (corresponding author) is a senior research scientist
at the University of Texas at Arlington Research Institute (UTARI), Fort
Worth, TX 76118, USA. (Phone: +1-817-272-5970; Fax: +1-817-272-
5946; E-mail: aditya@uta.edu). Dr. Das is also a special member to the
graduate faculty at the University of Texas at Arlington.
Stephen Savoie is with University of Texas at Arlington Research
Institute (UTARI), Fort Worth, TX, USA.
infrastructure is a highly risky endeavor that few are
prepared to take upon.
One example can be the almost non-existing assembly
and packaging infrastructure for 3D and heterogeneous
micro/nano systems. Although 3D heterogeneous micro/nano
systems promise huge potential, revolutionizing
miniaturization technology and availing many novel classes
of products with advanced functionalities, there are no
standard off-the-shelf equipment to construct them and,
therefore, new micro/nano system developments are
prohibitively slow and expensive. Another example is the
area of low to medium volume productions. In such cases,
where only a few tens to a few thousands of units are needed
on-demand as opposed to millions of units, it is not
commercially viable to establish dedicated manufacturing
facilities. Military products and parts, high tech medical
research components, novel technology R&D facilities etc.
are some of the examples of areas requiring low to medium
volume production capability.
In the above described example cases, and many more
similar scenarios, flexible manufacturing is the only
compelling solution that can offer the pathway to rapid, cost-
effective productization. This flexibility is generally
considered to fall into two categories: machine flexibility and
routing flexibility. The first category, machine flexibility,
refers to the system's ability to be changed to produce new
product types, and ability to change the order of operations
executed on a part. The second category, routing flexibility,
offers the ability to use multiple machines to perform the
same operation on a part, as well as the system's ability to
absorb large-scale changes, such as in volume, capacity, or
capability. Among the major advantages of flexible
manufacturing system the following are noteworthy: reduced
manufacturing times; lower cost per unit; greater labor
productivity; greater machine efficiency; reduced parts
inventories; adaptability to operations; and shorter lead times
[1, 2]. Flexible manufacturing systems are, however,
significantly difficult to achieve due to the cost to implement
and also the requirement of substantial pre-planning. These
are also the reasons for their slow acceptance by industry.
In this paper, we present a flexible manufacturing system
that is low cost and implements an automated pre-planning
approach. The proposed modular and reconfigurable
manufacturing cell (MRMC) primarily offers machine
flexibility, as described above. The modules of the proposed
MRMC can be configured and reconfigured for various pick
and place, bonding and probing applications, as needed
during typical assembly and packaging operations. In this
paper we will be discussing a prototype of the MRMC that is
Quasi-static Evaluation of a Modular and Reconfigurable
Manufacturing Cell
Aditya N. Das, Member, IEEE, Stephen Savoie
2013 IEEE International Conference on Robotics and Automation (ICRA)
Karlsruhe, Germany, May 6-10, 2013
978-1-4673-5643-5/13/$31.00 ©2013 IEEE 258

2. best suited for meso scale system integration. The concept,
however, is scalable to higher or lower scales.
The paper has been organized as follows: in section II we
present the background work leading to the development of
MRMC. Section III presents the technical details for the
MRMC prototype. In section IV, the analytical evaluation of
the MRMC prototype has been presented. Section V presents
the experimental results. Finally, section VI concludes the
paper with notes on future direction.
II. BACKGROUND
The fundamental principles of the proposed modular and
reconfigurable manufacturing cell (MRMC) have been built
upon the findings and requirements from our previous
research in heterogeneous microsystems development via
automated 3D microassembly. In past, we have demonstrated
that complex microsystems [3, 4, 5], consisting of
heterogeneous 3D parts, can be built reliably with high yield
and throughput via a combination of custom manipulation
hardware [6] and custom automation software that
implements special motion planning [7] and control [8]
algorithms. We also demonstrated that necessary and
sufficient precision in custom configured manipulation
hardware can not only be controlled in real-time via the high
speed hybrid controller but also can be predicted, and
thereby compensated for beforehand during the automation
execution, through a unique kinematic analysis [9] that
quantitatively predicts the uncertainty in end-effector
position due to individual motion errors in the manipulation
hardware and their location in the overall kinematic chain.
Built upon these know-hows, we have also presented
one-of-its-kind pre-planning software [10] and assembly-
process simulation software [11], which aid optimization of
manufacturing processes that use custom equipment for
system integration. These software help optimizing three
general manufacturing metrics, i.e. production yield, overall
cost, cycle time, and one product specific metric i.e. device
performance.
Conceptually, manufacturability ‘M’, involving assembly
and packaging via custom hardware, is a function of product
complexity ‘Ω’, assembler reconfigurability ‘Λ’ and the
production volume ‘v’. Mathematically:
M = f (Ω, Λ, v) (1)
The expression in equation (1) represents the governing
dynamics of a flexible manufacturing system.
The complexity index ‘Ω’ is a binary value that primarily
suggests if a specific task can be automated using a simple
open loop control (Ω = 0) or it requires complex closed loop
control with active feedback based manipulation (Ω = 1).
The derivation is based on a statistical model which suggests
that if the combined uncertainty of locating a part, grasping it
with an end-effector, manipulating it to the destination is
lower than the designed tolerance in the mating mechanism
at the destination then there is a high probability (> 99%) of
successful assembly and thus the operation can be executed
via open loop control.
On the other hand, the reconfigurability index ‘Λ’ is a
value in between 0 and 10, with 10 representing the highest
reconfigurability, which suggests if a robotic manipulation
system is easy to reconfigure or not, as a function of the
percentages of cost and time associated with the
reconfiguration. For absolutely fixed equipment, such as
single function tools and certain off-the-shelf hardware, the
associate cost and time for reconfiguration is assumed to be
infinity and thus the reconfigurability index for such a system
is taken as zero.
The production volume ‘v’ is incorporated in the form of
a histogram data where the number of bins and bin size are
determined based on the product type.
Finally the manufacturability index ‘M’ is estimated as a
value in between 0 and 10, with 10 representing the highest
manufacturability.
Development of the proposed modular and
reconfigurable manufacturing cell (MRMC) is motivated by
the need for a truly flexible robotic hardware system that can
serve as an application platform and experimental test-bed
for the above mentioned innovations.
III. SYSTEM DESCRIPTION
The modular and reconfigurable manufacturing cell
(MRMC) prototype consists of several linear and rotational
manipulation modules, different connecting fixtures, and
multiple end-effectors. Figure 1 shows the 3D renderings of
two of the many possible configurations of the MRMC
prototypes.
Figure 1. 3D renderings of different configuration for the manipulation
modules of the modular and reconfigurable manufacturing cell (MRMC)
259

3. A. Manipulation modules
The prototype system consists of linear manipulation
modules of three different ranges of motion i.e. 30mm,
50mm and 70mm. The rotation module is capable of full
360o
rotation. Among the different modular fixtures there are
quick-build base plates, sample holders, and 90o
angle
brackets. Two different types of modular end-effectors, an
electromagnetic gripper and a dispenser, have been designed
and developed to achieve typical pick & place and bonding
operations. Figure 2 shows the actual hardware test-bed at
our laboratory.
Figure 2. MRMC prototype at the UTARI facility
These manipulation units use a 2.4Watt stepper motor with
a lead screw that moves the stage. Standard 1/4th
inch
graphite code bearings house the two stainless steel shafts
that guide the motion of the stage. The aluminum body of the
manipulation stage has been CNC machined.
Apart from these components, each manipulation module
also consists of a novel and proprietary multifunctional
interconnect that allows mechanical as well as electrical
connectivity between modules, fixtures and end-effectors.
This interconnect component consists of an orientation
independent magnetic locking mechanism, which allows
mounting of the manipulation modules in any of the four
orthogonal orientations i.e. at 0o
, 90o
, 180o
or 270o
.
Furthermore, the multifunctional interconnect component
also includes a microcontroller circuitry to provide local
processing capability. The end-effectors and other fixture
also incorporate of the multifunctional interconnect in their
design for seamless and rapid integration with the
manipulation system.
Several end-effectors also are a part of the MRMC. Among
them there is an electromagnetic gripper, which uses an
electromagnet and a permanent magnet. A balanced force
based actuation principle allows smooth, bi-directional
operation of the gripper. Among other types of end-effectors,
the MRMC also has a quick-change dispensing head.
B. Motion controller
A custom designed motion control PCB interfaces the
manipulation modules with motor drives, peripherals such as
a touch screen panel and a vacuum generator, and the
computer. As opposed to the 1:1 mapping between motor
drives and manipulation units in case of commercial systems,
this controller enables 1:8 mapping of motor drive and
manipulator. Through a multiplexed signaling architecture,
the MRMC controller enables a single motor drive to run up
to 8 manipulators. Figure 3 shows a picture of the controller
along with a few snapshots of the touch screen interface.
(a)
(b) (c)
Figure 3. (a) MRMC controller box and (b, c) touch panel snapshots
C. Communication protocols
The MRMC uses four types of communication protocols to
interface various modules of the system. The touch panel has
been interfaced with the main controller via a full-duplex
serial communication. Different manipulation modules are
connected to each other and to the main controller over an
I2
C bus. USB connectivity has been established in between
the main controller and computer. Lastly a few time-critical
communication channels have been hard-wired.
D. Other peripheral units
Different peripheral units such as a microscope, a vacuum
pump, a data acquisition module etc. have also been
integrated with the MRMC. These peripheral modules are
used in different types of operations executing the custom
automation plans.
260

4. E. Unique features
The following features are unique in the presented modular
and reconfigurable manufacturing cell:
1. Fully automated robotic prototyping system
2. Intelligent robotic modules with distributed intelligence
a. Modules can automatically identify their position and
orientation in the system
b. Multiple modules can be operated via single
controller via bi-directional communication channel
c. Entire system can be programmed & controlled by a
master controller without the need for a computer
3. Extremely portable architecture
a. No nuts and bolts needed
b. No complicated wiring needed
c. Low form factor controller unit
4. Commercialization friendly, as compared to conventional
manipulation modules in market
a. Lower cost
b. Competitive precision output
c. Improved force output
5. Revolutionary design enabled extreme flexibility for rapid
product transition
IV. ANALYTICAL EVALUATION
One of the major challenges in flexible manufacturing
architecture, where system components are frequently
reorganized to accommodate changes in tasks, is to
guarantee necessary and sufficient precision metrics such as
resolution, repeatability and accuracy. In our past work [12],
we had investigated the effect of parametric uncertainties in
a serial robot chain composed of prismatic or rotary modules
on the overall positioning uncertainty at the end-effector.
Two types of errors were considered: static errors due to
misalignment and link parameter uncertainties, and dynamic
errors due to inaccurate motion of individual links. Using
common uncertainty metrics, we had compared the precision
of different robot kinematic chain configurations and had
selected the best suited ones for a generic peg-in-hole
assembly task.
A. Problem statement
Using the above technique, in this work, we analyzed the
MRMC, as shown in figure 1. A simple two-step pick-and-
place task was chosen as the case study. The task has been
depicted in figure 4.
Figure 4. Assembly task for evaluating the MRMC
B. Machine setup
For the analysis purpose, we tested two different
configuration of the modular and reconfigurable
manufacturing cell (MRMC) to accomplish the task as
shown in Figure 4. Both configurations are shown in Figure
5. The first configuration is a RPP robot with ψ-y-z degrees
of freedom, whereas the second configuration is a PPP robot
with x-y-z degrees of freedom. (R: revolute, P: prismatic)
(a) (b)
Figure 5. MRMC configurations for executing the pick and place task
Robot configuration 1 assembles the product by moving
individual parts from one station to another, as shown in
Figure 5(a), with the rotation stage acting as a turret to
position the end-effector over the three stations. On the other
hand, robot configuration 2 (Figure 5(b)) assembles the
product on a single platform by positioning the end effector
in 3D space. In both cases, for the sake of simplicity, it is
assumed that the parts are fixture in the sample holder(s) in
such a way that there is no misalignment along the rotation
axes, as shown in Figure 4. Kinematic details for the two
robot configurations are given in Tables I and II.
TABLE I. KINEMATIC SETUP OF ROBOT CONFIGURATION 1
Degree of
Freedom
(DoF)
Order from
the root of
the chain
DoF type DoF
range of
motion
Accuracy
ψ 1 Revolute 360o
7.5o
y 2 Prismatic 50mm 8µm
z 3 Prismatic 30mm 5µm
TABLE II. KINEMATIC SETUP OF ROBOT CONFIGURATION 2
Degree of
Freedom
(DoF)
Order from
the root of
the chain
DoF type DoF
range of
motion
Accuracy
x 2 Prismatic 50mm 8µm
y 1 Prismatic 70mm 12µm
z 3 Prismatic 30mm 5µm
As mentioned in the above Tables, different manipulation
modules have different precision values. Furthermore, the
position of these modules in the serial robot kinematic chain
also impacts the error at the tip of the end-effector.
C. Analytical model
Assembly feasibility estimation was computed using the
following analytical model (see [9] for detailed derivation).
( ) ( )[ ] ( )[ ]00
1
0
TeeT N
n
i
iiN
iiii








⋅+≅ ∏=
δθξθξ
θξδθ
 
, (2)
Part 1
Part 2
Part 3
Tolerance: σ12 = 15µm
Tolerance: σ23 = 40µm
x
y
z
θ
φ
ψ
261

5. where
θ  joint angle in case of revolute joints and displacement
in case of prismatic joints;
ξ  twist vector representing the instantaneous motion of a
link;
T  transformation matrix
In equation (2), the additive term is the “static error” or
error due to link misalignment, whereas the multiplicative
term is the “dynamic error” or error due to joint motion.
We used our proprietary iterative analysis software called
“Design for Multiscale Manufacturability (DfM2
)”, as
mentioned in [10], in order to build a statistical model of the
common manufacturing metrics, such as production yield,
cycle time and overall cost, for the two assembler
configurations shown in Figure 5. A snapshot of the DfM2
software running the robot analysis, using the equation (2), is
shown in figure 6.
Figure 6. Assembly task for evaluating the MRMC
D. Results
After 1000 iterations in the analyzer, the statistical data for
the two robot configurations suggest the following
manufacturing metrics as shown in table III.
TABLE III. MANUFACTURING METRICS OBTAINED FROM
COMPUTATIONS VIA THE ANALYTICAL MODEL
Robot
configuration
No. of
iterations
Overall
Yield b
Cost (% of
optimum) a
Time (% of
optimum) a
1 (Figure 5(a)) 1000 83% 50% 77%
2 (Figure 5(b)) 1000 92% 48% 80%
a. optimum desired cost: $3,600; optimum desired time taken in assembly: 4 minutes
b. using the design tolerance values as mentioned in the Figure 4 for the parts
As evident from the data in Table III, robot configuration
2 offers better yield and cost efficiency at a marginal
increase in cycle time. Therefore, although both
configurations are capable of executing the specified task,
robot configuration 2 is more suited for it.
Furthermore, as the tolerance levels for the assembly task
changes it also affects the manufacturing metrics. This is
shown in table IV.
TABLE IV. PERFORMANCE WITH VARING ASSEMBLY TOLERANCE
Tolerance Configuration 1 yield Configuration 2 yield
σ12:15µm, σ23:40µm 83% 92%
σ12:10µm, σ23:30µm 54% 88%
σ12:5µm, σ23:20µm 7% 76%
It is observed from the Table IV that configuration 2 is a
better option as the tolerance for the assembly gets tighter.
V. EXPERIMENTAL VALIDATION
A. MRMC hardware specifications
In order to validate the analytical results, as mentioned in
the previous section, we use the prototype modules for the
modular and reconfigurable manufacturing cell (MRMC), as
shown in Figure 2 and also in the video attachment. The
prototype module specifications are given in Table V.
TABLE V. MRMC PROTOTYPE MODULE SPECIFICATIONS
Parameter Value Unit
Resolution 4 to 15 µm
Range of motion (linear module) 30 to 70 mm
Range of motion (rotation module) 360 degrees
Maximum thrust 10 lb
Maximum speed 3 mm/sec
Pull force limit of interconnects 20 lb
Motor type Stepper -
Motor power rating 5/0.25 Volt/Ampere
System power rating 24/0.9 Volt/Ampere
Typical configuration time < 2 minutes
Typical calibration/ program time < 5 minutes
Manipulation module cost ~ 300 $
Controller system cost ~ 500 $
Manipulators per controller 8 -
Size (length x width x height) (90-185)x90x35 mm3
Weight 420 to 780 grams
Individual cabling to manipulator Not required -
Communication frequency 10 KHz
Computation frequency 20 MHz
Stand-alone interface Touch panel -
PC connectivity USB -
Configuration identification Automatic -
Assembly automation mode Programmable -
B. Experimentation steps
The experimentation conducted to validate the analysis
consists of the following steps:
i. The parts are pre-fixtured on the sample holder prior
to the assembly.
ii. The experimentation begins with a blank base plate
mounted to the optical table; this base plate will
eventually hold the robot.
iii. The system with the master controller is powered on.
iv. The manipulation system is configured by placing the
robotic modules in a serial order, starting with placing
the first one on the base plate and consequent ones on
to the previous ones. During this step, the master
controller automatically identifies the module’s
position and orientation with respect to a global
coordinate frame.
v. Once the desired robot configuration is achieved, a
calibration command is sent from the controller by
pressing a button, which initiates a multi-point
calibration technique by each of the robotic modules
under the field of view (FoV) of a fixed camera. It is
assumed, and also experimentally verified that, during
configuration of the manipulation system, the locking
error in each robotic module is well within the size of
the FoV and thus these errors are observable.
262

6. vi. After the calibration, the automation program is either
retrieved from an on-board memory card of the master
controller unit or coded by using the appropriate
console buttons.
vii. Finally, the program is executed by pressing the RUN
button which moves the calibrated robotic modules to
carry out the assembly task. During this step, the
master controller implements a precision optimized
path planning and control algorithm for the
automation.
viii. Success percentage of the assembly is the ratio of the
number of parts assembled vs. total number parts in
the device.
ix. After assembly is completed, the manipulation
modules are dismantled and main power is turned off.
x. For each repetitions in the experiment, we go back to
step ii and repeat the steps up to step ix.
xi. Finally, after the desired number of iterations, the
compiled data on manufacturing performances is
statistically analyzed.
C. Experimentation results
Table VI shows the data from the experimentations
conducted on the modular and reconfigurable manufacturing
cell (MRMC) according to steps mentioned above.
TABLE VI. MANUFACTURING METRICS OBTAINED EXPERIMENTALLY
USING ROBOT CONFIURATION 2 (FIGURE 5-B) FOR 10 ITERATIONS
Parameter Value Deviation from
analytical model
Overall yield 90% - 2%
Average time/assembly 3 min. 36 sec. + 6%
Total manufacturing cost $1,850 + 3.3%
As seen in Table VI, the experimental finding for
assembly using the proposed MRMC is close to the
analytical predictions. The higher cycle time can be due to
delays in image stabilization/processing during the
calibration steps. The marginal increase in the actual cost is
due to the labor associated with additional cycle time.
VI. CONCLUSION
In this paper, we presented a novel flexible manufacturing
system offering rapid prototyping capability at low cost,
which is ideal for low to medium volume manufacturing. The
unique hardware, presented in this work, are of much lower
cost in comparison to similar commercially available
manipulation modules. Furthermore, they significantly
reduce reconfiguration complexity; offer competitive
performance in terms of range of motion, force output and
precision; and allow easy transition between products. A
novel and proprietary multifunction interconnect design
enables quick-change mounting and untethered connection
between different modules of the system. A 1:8 mapping in
controller and manipulation module significant reduces the
cost and also make the overall system much more compact
and portable as compared commercial manipulators.
Future directions in this research will focus on rigorous
reliability and performance tests on the hardware modules
and extending the software analyzer to also generate
automation programs along with predicting manufacturing
metrics. The discussed modular and reconfigurable
manufacturing cells are envisioned to enable manufacturing
at remote sites such as on-board ships and aircrafts; make
low volume production of highly specialized and advanced
products sustainable; and encourage new product
development endeavors.
ACKNOWLEDGMENT
The authors are thankful to the research and support staff
at the University of Texas at Arlington Research Institute
(UTARI) for their invaluable help in this work. The authors
would also like to extend their gratitude towards office of
naval research (ONR) for supporting this work.
REFERENCES
[1] G. Chryssolouris, “ Manufacturing Systems – Theory and Practice,”
in NY: Springer Verlag, 2nd edition, 2005.
[2] T. Tolio, “Design of Flexible Production Systems – Methodologies
and Tools,” in Berlin: Springer, 2009.
[3] A. N. Das, J. Sin, D. O. Popa, and H. E. Stephanou, “Precision
Alignment and Assembly for a Fourier Transform
Microspectrometer,” in International Journal of Micro-Nano
Mechatronics (JMNM), Vol. 5, No. 1-2, Springer Berlin, pp. 15-28,
2009.
[4] R. Murthy, A. N. Das, D. O. Popa, “ARRIpede: An Assembled Die
Scale Microcrawler,” in Journal of Advanced Robotics, The Robotics
Society of Japan (RSJ), 2010.
[5] D. O. Popa, R. Murthy, A. N. Das, “M³ - Deterministic, Multiscale,
Multirobot Platform for Microsystems Packaging: Design and Quasi-
Static Precision Evaluation,” in IEEE Transactions on Automation
Science and Engineering (T-ASE), Vol. 6, Issue: 2, pp. 345–361,
2009.
[6] A. N. Das, R. Murthy, D. O. Popa, H. E. Stephanou, “A Multiscale
Assembly & Packaging System for Manufacturing of Complex Micro-
nano Devices,” in IEEE transactions on Automation Science and
Engineering (T-ASE), vol. 9, Issue 1, pp. 15–64, Jan 2011.
[7] A. N. Das, D. O. Popa, “Precision-based Robot Path Planning for
Microassembly,” in Proceedings of the 6th IEEE Conference on
Automation Science and Engineering (CASE), Toronto, Canada, pp.
527-532, Aug 2010.
[8] A. N. Das, D. O. Popa, H. E. Stephanou, “Automated Microassembly
using Precision based Hybrid Control,” in Proceedings of IEEE
International Conference on Robotics and Automation (ICRA),
Anchorage, Alaska, page(s): 4106-4112, May 2010.
[9] A. N. Das, D. O. Popa, “Precision-based Robot Kinematics Design for
Microassembly Applications,” in Proceedings of the ASME
International Design Engineering Technical Conferences &
Computers and Information in Engineering Conference
(IDETC/CIE), Montréal, Quebec, Canada, pp. 857-862, Aug 2010.
[10] A. N. Das, H. E. Stephanou, “Concurrent Engineering in A
Microfactory,” in Proceedings of Commercialization of Micro-Nano
Systems Conference (COMS), Greensboro, North Carolina, USA,
August, 2011.
[11] A. N. Das, H. E. Stephanou, “Design of Microassembly through
Process Modeling in Virtual Reality,” in Proceedings of Microtech
conference and expo, Boston, MA, USA, June, 2011.
[12] A. N. Das, D. Popa, H. E. Stephanou, “Precision Evaluation of
Modular Multiscale Robots for Peg-in-Hole Microassembly Tasks,”
in Proceedings of International Conference on Intelligent Robots and
Systems (IROS), San Francisco, California, USA, pp. 1699 – 1704,
Sep. 2011.
263