Goal - Robotics - University of Michigan

advertisement
Robotics for Manufacturing
A Michigan Robotics Focus Area:
Manufacturing research has long been a strength at the University of
Michigan, and robotics is already being widely used in automotive and
other manufacturing plants. Continuing research is needed to expand the
role of robotics in manufacturing applications by improving robot
capabilities and safety and reducing cost and energy consumption.
Contributing Faculty:
Kira Barton
Chinedum Okwudire
Kazuhiro Saitou
Dawn M. Tilbury
A. Galip Ulsoy
Introduction
Industrial Robots
• There are approximately 1.5 million
industrial robots in operation today
world-wide (e.g., approx. 300,000 in
Japan, 200,000 in North America,
125,000 in S. Korea)
• They perform tasks such as welding,
spray paining and assembly in a
variety of industries.
• They are typically preprogrammed
to repeat the same task, are among
the most reliable machines
available, and are operated in
isolation from humans for safety.
• Automated Guided Vehicles (AGVs)
are also widely used in
manufacturing plants for material
handling.
Automated Guided Vehicles (AGVs)
Research Needs
• Safety research is needed to
enable operation with and
around humans.
• Cost reduction will enable use
for smaller volume production
and by small companies.
• Energy consumption is high,
and needs to be reduced.
• Dexterity and precision must
be improved for many
assembly tasks.
• Coordination among robots
and with other automated
machines.
• Flexibility and Autonomy are
important to move beyond
mass production.
Kiva Systems: Flexible Material Handling
Rethink Robotics: Low-Cost Assembly Robot
Baxter
Precision Motion Control for High-Speed, HighResolution Manufacturing (Barton)
• Goal: Design advanced sensing
and controls algorithms for high
precision motion control
• Iterative Learning Control
o
o
o
Flexible learning strategies
Robust learning for a range of
applications
Cooperative learning control strategies
• Advanced Sensing Strategies
o
o
o
High-resolution sensing techniques
Atomic force microscopy for
topographical and charge density
sensing
Vision-based detection
Fig. 1: Cooperative learning strategies. Develop
cooperative learning control techniques to enable efficient
and effective surveillance and monitoring operations.
• Applications
o
o
o
o
Emerging manufacturing processes
High-resolution, high-speed
manufacturing systems
Rehabilitation robotics
UAVs and other autonomous systems
Fig. 2: ILC process. As the number of iterations increases, the
feedforward time domain control signal is determined and the
error signal is minimized.
Design and Control of Cartesian Robots (Feed Drives) for
Improved Performance and Energy Efficiency (Okwudire)
Guideway
•
Goal: Improve performance and
energy efficiency of feed drives
•
Dynamically adaptive feed drives
o
Design feed drives such that dynamic
properties change based on
manufacturing operation
o
Integrally design time-varying controllers
to ensure stability and performance
under various dynamic configurations
o
•
Determine optimal dynamic
configurations/controllers to ensure
desired performance at minimal energy
consumption
Example: Hybrid feed drive
o
Feed drive is driven by linear and/or
rotary motors depending on manuf.
operation
o
Moving mass, drive point, sensing
location, etc. change dynamically
o
Up to 25% improvement in energyefficiency anticipated
Motor
Table
Screw
(smooth shaft)
Dynamically Adaptive
Hybrid Feed Drive
Linear Motor
Engagement/disengageme
nt
mechanism
(located under table)
Disengagable
Pneumatic pistons
Roh’lix nut
Toggle
arms
Improving Energy Efficiency by Multi-Robot Coordination and Task
Scheduling (Saitou)
• Goal: Reduce total energy
consumption and peak
energy demands in multirobot cells
o Arm posture optimization to
minimize idle time energy
consumption
o Multi-robot coordination to
maximize the use of
regenerative energy from one
robot in other robots
o Task scheduling to reduce the
need of rapid acceleration
o Task scheduling to spread
energy peaks in multiple
robots across cycle time
Power profile of typical operation exhibiting multiple
energy peaks during cycle time (Duflou, et al, 2012)
Virtual Fusion for Robotic System Design, Evaluation
and Monitoring (Tilbury)
• Goal: Use high-fidelity
simulation models running in
parallel with physical system
to evaluate new system
designs
• Validate performance by
integration of simulated
systems with physical systems
• Quickly evaluate multiple
scenarios for reconfigurability
• Operator training with highfidelity models
• Run high-fidelity simulation
models in parallel with
physical system for on-line
monitoring
Seamlessly swap a virtual robot for real :
Identical controls and networking interface
DeviceNet/ Ethernet
Evaluate new robotic concepts (e.g. Motoman
for material handling + assembly)
Design for Improved Reliability and Efficiency (Ulsoy)
•
Goal: Design robots to be
energy efficient, reliable and
safe
•
Passive-Assist Design
o
•
For each joint motor, design
parallel and/or series spring for
passive assist
o
Typical trajectory and load
o
Optimize spring design
(b)
(a)
(c)
Single-Link Manipulator
o
Experiments w/ and w/o spring
o
Properly designed torsional spring
reduces max. torque by 50% and
energy consumption by up to 25%
Planned Extensions
o
Robust design for family of
trajectories and loads
o
Multi-link robot arm
o
Co-design of robot and controller
0.2
Electrical Power [W]
•
Experimental setup: (a) Quanser DC motor
and controller, (b) worm gear transmission,
and (c) single link arm .
Model - No Spring
Model - With Spring
Experiment - No Spring
Experiment - With Spring
0.15
0.1
0.05
0
0
5
10
15
Time, t [s]
20
25
30
Download