Project-Based Control Education

Transcription

1 Project-Based Control Education All learning is in the learner, not in the teacher Plato Mark W. Spong Donald Biggar Willett Professor Department of Electrical and Computer Engineering and The Coordinated Science Laboratory University of Illinois at Urbana-Champaign, USA Introduction In the book, Engineering in History*, one finds the following chronology of the major developments in engineering: ``[T]he presentation [of this book] is oriented about eight of the great events of history which totally changed the ways of human life. These are: Food-producing revolution (ca B.C.) Appearance of urban society (ca B.C.) Birth of Greek science ( B.C.) Revolution in power (middle ages) Rise of modern science (seventeenth century) Steam and the Industrial Revolution (eighteenth century) Electricity and the beginnings of applied science (nineteenth century) Age of automatic control (twentieth century) (Italics mine) *Kirby, et.al. Dover Publications, New York,

2 The 21 st Century The importance of Automatic Control will continue to increase in this century due to: advances in theory, computation, sensing new application domains in biological sciences, nanotechnology, robotics changing demographics, the need to secure critical infrastructure, the need for energy efficiency consumer demands in automotive, aerospace, electronics, home appliances Component Design vs. Systems Engineering "To develop a complete mind, study the science of art, study the art of science. Learn how to see. Realize that everything connects to everything else." Leonardo da Vinci A recent National Academy of Engineering report cites interdisciplinary systems engineering as an increasingly important aspect of modern engineering and the education of future engineers. Industrial applications of control are increasingly largescale, complex, and interdisciplinary How does this impact the way we teach control? 2

3 Component Design vs. Systems Engineering The reductionist approach to engineering and applied science has served us well and will continue to serve us. However, it is insufficient to prepare engineering students for industrial projects that are increasingly being conducted by multi-disciplinary teams of engineers Moreover, control is important enough that all engineering students and, indeed, all students should be exposed to it s fundamental concepts. My premise is that we generally do a good job preparing control engineers but we should think more broadly. Interdisciplinary Education Interdisciplinary engineering education can be addressed through projects that integrate elements of mechanical design, electronics design, modeling, control system design, and software implementation. Projects can also be used to motivate and teach concepts of control to non-engineering students and to education the general public about technology. Robotics and Mechatronics projects seem to be the most popular way to accomplish this. 3

4 Undergraduate Systems and Control Education at Illinois FEEDBACK CONTROL SYSTEMS SYSTEMS AND ENTREPRENEURIAL ENGINEERING ROBOTICS MECHATRONICS Mechatronics Projects Introduction to Mechatronics is a fourth year elective course that culminates in a team-based project Mechatronics Lab 4

5 The Mechatronics Course Concepts of Mechatronics putting intelligence onto physical systems sensing, actuation, design Real-Time Programming Electronics serial and parallel communication, I/O, interrupts board design Noise and filtering Modeling and Control PID, State Space Final Project First Example:The Segbot Project The Segbot is a small, twowheeled self-balancing robot The Segbot design team, David Chen, Erik Bettini, Chris Graesser, Andy Block, Carlos Montesinos 5

6 Some Features of the Segbot The Segbot incorporates: On-board DSP board/encoder interface board/pwm amplifier board CCD camera IR distance sensor Accelerometer Inclinometer motors and encoder Design,Modeling and Control The design team considered: standardization of components to reduce cost structural rigidity sensor and board placement A CAD model was created and inertia parameters determined A Model-Based Control was designed and implemented balance control using rate gyro and inclinometer velocity and heading control wall-following using IR sensors 6

7 generalized coordinates equations of motion where 7

8 Linearized Model and LQR Balance Control Vertical Angle Measurement and Sensor Fusion: The Complementary Filter One of the more interesting aspects of this project, and one that is rarely if ever taught in control courses, is the sensing of the vertical angle of the Segbot. This angle was determined by using two sensors: a tilt sensor and a solidstate rate gyro. The tilt sensor is a Crossbow CXTLA01, which is a single axis sensor with an accuracy of ±20 degrees. The rate gyro is a Silicon VSG CRS02 with an accuracy of ±150 degrees/sec. Either device alone can be used to sense the vertical angle in theory. However, in practice, both sensors are necessary to achieve acceptable performance. 8

9 Sensor Issues The tilt sensor measures an angle relative to the direction of gravity. It has excellent steady state characteristics, however, it develops significant phase lag with increasing frequency. The slow response of the tilt sensor is problematic for this system because of the need for accurate dynamic as well as static measurements. The rate gyro measures angular rate; the vertical angle is then determined by integration. While the gyroscope has excellent dynamic response, a problem with all gyroscopes is that there is a bias near DC that causes drift in the angular measurement. Sensor Fusion A sensor-fusion solution used by the students was to combine the measurements of the two sensors, taking advantage of the low frequency characteristics of the tilt sensor and the high frequency characteristics of the gyroscope. Two methods were used to determine the vertical angle from the two sensor measurements: A complementary filter A Kalman filter. 9

10 Complementary Filter Motion Control In order to control heading and speed the wheel variables are mapped to the linear and angular speeds and angular orientation: 10

11 These expressions can be inverted to yield offsets to the wheel variables to control speed and heading: The Results Spin in Place Wall Following using IR Sensor 11

12 Network of Second-Generation Segbots Tracking Using the CCD Camera 12

13 Second Example: The Segmonster The Segmonster is a robot designed to ride a Segway Human Transporter (SHT) The Design Team and Design Philosophy The Segmonster project was begun by two undergraduate students as a summer project following the mechatronics course and completed as an MS thesis project. The design philosophy is that the Segmonster should have access to only the information that is available to a human rider. Therefore no special instrumentation was added to the SHT 13

14 Segmonster The robot rider controls the Segway HT just as a human rider, by inclining its torso to control forward and reverse velocity and by rotating the handle grip to turn. This led to several important design constraints Design Constraints Attaching the robot firmly to the Segway so that it does not fall off or damage the Segway There are three standard mounting points on either side of the operator's platform of the Segway. These are normally used to hang bags or other Segway accessories over the wheels. These were the only mounting points that could be used so as not to void the manufacturer's warranty. 14

15 Design Constraints Engaging and disengaging the foot pressure sensors The footpads on the Segway contain pressure sensors that indicate the presence (or absence) of a rider. The Segway cannot be turned on if any weight is on these footpads. Therefore the Segmonster could not be designed simply to sit on the Segway footpads, but had to be mounted on a platform over the footpads, attached to the standard mounting point as above, in such a way that it could engage the pressure sensors (with sufficient weight) after power to the Segway is turned on. Design Constraints Selecting and placing sensors to obtain feedback on position and velocity The Segmonster is equipped with Four optical encoders One infrared (IR) range sensor One CCD color camera Two of the optical encoders are mounted to dummy wheels that ride on the Segway wheels to determine velocity and relative orientation of the Segway. A third encoder is mounted on the robot arm to determine its position and the fourth is used for the motor that actuates the handle grip for turning. The IR range sensor is used for obstacle detection and the CCD camera is used to collect visual data for navigation and terrain following. 15

16 Segmonster Control The Segmonster controller resides on a TI C6713 Digital Signal Processor. A pair of daughter cards designed at the University of Illinois Control Systems Lab provide interface to all sensors as well as digital-to-analog output to the motor amplifiers. The Segmonster can be controlled manually through a wireless communication card, or autonomously using it's on-board vision system. Segmonster Control The Segmonster has several integrated controllers A modified PID-velocity control loop A PID turning controller A sidewalk tracking controller using a vision sensor 16

17 Some Segmonster Videos Inside Fast Turn Slow Turn Some Conclusions Undergraduate students are an extremely capable and often untapped resource Solving real control problems rather than textbook problems is a tremendous motivator and well within the grasp of many students Project-Based Control is a holistic method of education that offers advantages that cannot be obtained in the classroom. 17

18 THANK YOU! QUESTIONS? 18