Automatic Control: From Mechanical and Electrical Engineering to - - PowerPoint PPT Presentation

Automatic Control:


From Mechanical and Electrical Engineering to Computer Science

Dawn Tilbury

Professor, Mechanical Engineering Department Associate Dean for Research and Graduate Education University of Michigan

1

Control systems

• Plant to be controlled, defined by

differential equations:

– Actuation (input) – u(t) – Sensing (output) – y(t) – Desired behavior (reference) – r(t)

• Design a controller to achieve desired

behavior

– Given the error e = r-y, compute the input u

2

Challenges to be addressed

• How to find the reference input for the low-

level controller?

• How to represent the desired action?

– Can physics be leveraged to improve performance?

• How to interpret the information that is

coming from the low-level sensors?

– Aggregation and abstraction

3

Information based control

• Control at every

level is built on models

– How can they be adapted as things change?

• Different levels of

abstraction

– What are the appropriate interfaces 
 between levels?

4

˙ x = f (x,u)

y = Ax + Be

Pre-process data Model system behavior Make decisions Interpret models Update models Understand anomalous data Guide data collection Predict

• utcomes

Crossing boundaries

• ASME DSCD TC on Robotics: Control

applications in robotics; mobile robots and mobile robot collectives; biomimetics, bio-mechanical, humanoid robots, human-machine interface

• ASME DED TC on Mechanisms and

Robotics: design and applications of mechanisms,

machines, robots and mechatronics systems.

• IEEE CSS TC on Manufacturing

Automation and Robotic Control

• IEEE Robotics and Automation Society

5

h"p://robo)cs.umich.edu ¡

• New MS/PhD program in Robotics: Fall 2014

– 5 new PhD students, 3 new MS students – Two new graduate courses: Math for Robotics and Robot Systems (hands-on)

• Robotics Day: 5th annual – April 9, 2015

– More than 500 attendees in 2014, including FIRST robotics, industry members, university students

• Connected and autonomous vehicles,

rehabilitation robotics, and many more…

Robotics Initiative at Michigan