COM2009-3009 Robotics Lecture 5

© 2023 The University of Sheffield COM2009-3009 Robotics Lecture 5 Control Theory https://youtu.be/YNhXYgm7ebs COM2009-3009 Robotics: Lecture 5 slide 1 1 © 2023 The University of Sheffield COM2009-3009 Robotics Lecture 5 Control Theory COM2009-3009 Robotics: Lecture 5 slide 2 2 1 © 2023 T...

© 2023 The University of Sheffield COM2009-3009 Robotics Lecture 5 Control Theory https://youtu.be/YNhXYgm7ebs COM2009-3009 Robotics: Lecture 5 slide 1 1 © 2023 The University of Sheffield COM2009-3009 Robotics Lecture 5 Control Theory COM2009-3009 Robotics: Lecture 5 slide 2 2 1 © 2023 The University of Sheffield DBA FEE What’s going on here? CK Regulatory/Control System Closed-loop feedback control (‘homeostasis’) CONTROL COM2009-3009 Robotics: Lecture 5 slide 3 3 © 2023 The University of Sheffield Regulatory/Control Systems https://youtu.be/sI8sF3Ls8As COM2009-3009 Robotics: Lecture 5 slide 4 4 2 © 2023 The University of Sheffield Regulatory/Control Systems COM2009-3009 Robotics: Lecture 5 slide 5 5 © 2023 The University of Sheffield Actions & Consequences WORLD (w) Actions Transform (fw) Consequences Continuous cycle of cause and effect COM2009-3009 Robotics: Lecture 5 slide 6 6 3 © 2023 The University of Sheffield Intentions, Actions & Consequences AGENT (a) Intentions WORLD (w) Transform (ga) Actions Consequences Transform (fw) ‘Open-loop’ configuration (also known as ‘dead reckoning’) Requires extensive trial-and-error to find a good ga, but will be affected by disturbances to fw and/or the output COM2009-3009 Robotics: Lecture 5 slide 7 7 © 2023 The University of Sheffield Open-Loop Control AGENT (a) Intentions Transform-1 (fw-1) WORLD (w) Actions Transform (fw) Consequences Open-loop configuration using an ‘inverse model’ (also known as ‘inverse kinematics’) The forward transform fw-1 can be difficult to calculate, and the system will still be affected by disturbances to fw and/or the output COM2009-3009 Robotics: Lecture 5 slide 8 8 4 © 2023 The University of Sheffield Closed-Loop Control AGENT (a) Intentions - WORLD (w) Controller Consequences Actions Consequences Transform (fw) Observations Closed-loop configuration using ‘negative-feedback’ (sometimes referred to as a ‘servomechanism’) The system automatically compensates for disturbances to fw and/or the output COM2009-3009 Robotics: Lecture 5 slide 9 9 © 2023 The University of Sheffield Open- vs. Closed-Loop Control • ‘Open-loop’ (dead reckoning) 1. 2. 3. issue command action initiated effect may or may not be as intended (even after extensive calibration) • ‘Closed-loop’ (feedback control) 1. 2. 3. 4. 5. set intended result sense current situation compare the two if different, act to bring current situation closer to intention loop to 2 until consequence = intention COM2009-3009 Robotics: Lecture 5 slide 10 10 5 © 2023 The University of Sheffield Open-Loop Control https://youtu.be/b1tc6OMGmOQ COM2009-3009 Robotics: Lecture 5 slide 11 11 © 2023 The University of Sheffield Closed-Loop Control COM2009-3009 Robotics: Lecture 5 slide 12 12 6 © 2023 The University of Sheffield Demo: Open/Closed-Loop Control Objective = 50 cm from wall 1. Open-loop Control Control-x 2. Closed-loop 3. 2x closed-loops – loop-1: proximity – loop-2: orientation COM2009-3009 Robotics: Lecture 5 slide 13 13 © 2023 The University of Sheffield Negative-Feedback Control controller actuator negative feedback provides stability sensor DiStefano III, J. J., Stubberud, A. R., & Williams, I. J. (1990). Feedback and Control Systems (2nd ed.). New York: McGraw-Hill. COM2009-3009 Robotics: Lecture 5 slide 14 14 7 © 2023 The University of Sheffield Example: A ‘Thermostat’ desired temperature difference heater/cooler open/closed windows/doors ambient temperature measured temperature controller temperature sensor DiStefano III, J. J., Stubberud, A. R., & Williams, I. J. (1990). Feedback and Control Systems (2nd ed.). New York: McGraw-Hill. COM2009-3009 Robotics: Lecture 5 slide 15 15 © 2023 The University of Sheffield Negative-Feedback Control Also know as a ‘servomechanism’ DiStefano III, J. J., Stubberud, A. R., & Williams, I. J. (1990). Feedback and Control Systems (2nd ed.). New York: McGraw-Hill. COM2009-3009 Robotics: Lecture 5 slide 16 16 8 © 2023 The University of Sheffield Example: A ‘Servo’ Motor reference input actuator feedback sensor COM2009-3009 Robotics: Lecture 5 slide 17 17 © 2023 The University of Sheffield Demo: Servos in Action https://www.petoi.com COM2009-3009 Robotics: Lecture 5 slide 18 18 9 © 2023 The University of Sheffield Demo: Servos in Action https://www.petoi.com COM2009-3009 Robotics: Lecture 5 slide 19 19 © 2023 The University of Sheffield Feedback-Control in Living Systems https://youtu.be/YpIm39GHeLw COM2009-3009 Robotics: Lecture 5 slide 20 20 10 © 2023 The University of Sheffield Multiple Control Loops • Most control problems cannot be solved using a single controller • Recall the Lego demo that required control loops for proximity and orientation; these loops ran in parallel • Some problems require control loops to be stacked in a hierarchy • Hierarchical control is not very common in engineering systems (due to the lack of good design principles) • Hierarchical control is a key tenet of ‘Perceptual Control Theory’ (PCT) for living systems • PCT emerged from Cybernetics in the 1960s (pioneered by William T. Powers) COM2009-3009 Robotics: Lecture 5 slide 21 21 © 2023 The University of Sheffield Perceptual Control Theory (PCT) W. T. Powers (1973). Behaviour: The Control of Perception, Aldine, Chicago. COM2009-3009 Robotics: Lecture 5 slide 22 22 11 © 2023 The University of Sheffield Perceptual Control Theory (PCT) W. T. Powers (1973). Behaviour: The Control of Perception, Aldine, Chicago. COM2009-3009 Robotics: Lecture 5 slide 23 23 © 2023 The University of Sheffield Hierarchical PCT (HPCT) W. T. Powers (1973). Behaviour: The Control of Perception, Aldine, Chicago. COM2009-3009 Robotics: Lecture 5 slide 24 24 12 © 2023 The University of Sheffield Hierarchical PCT (HPCT) W. T. Powers (1973). Behaviour: The Control of Perception, Aldine, Chicago. COM2009-3009 Robotics: Lecture 5 slide 25 25 © 2023 The University of Sheffield Hierarchical PCT (HPCT) W. T. Powers (1973). Behaviour: The Control of Perception, Aldine, Chicago. • • • • • • • • • • 1st-order: intensity 2nd-order:sensation 3rd-order: configuration 4th-order: transitions 5th-order: sequence 6th-order: relationships 7th-order: program 8th-order: principles 9th-order: system concepts etc. COM2009-3009 Robotics: Lecture 5 slide 26 26 13 © 2023 The University of Sheffield Hierarchical Control Grand, S. (2003). Growing Up With Lucy: Phoenix. Grand, S. (2000). Creation: Life and How to Make it: Phoenix. COM2009-3009 Robotics: Lecture 5 Hawkins, J. (2004). On Intelligence: Times Books. slide 27 27 © 2023 The University of Sheffield Steve Grand’s ‘Creatures’ COM2009-3009 Robotics: Lecture 5 slide 28 28 14 © 2023 The University of Sheffield Demo: HPCT PCT-AWE PCT-AWE-WANDER COM2009-3009 Robotics: Lecture 5 slide 29 29 © 2023 The University of Sheffield Layered Control • Subsumption • Hierarchical FB-Control – established the principles of behaviour-based robotics – tasks are accomplished sequentially – open-loop(s) – susceptible to disturbances COM2009-3009 Robotics: Lecture 5 – aligns with the principles of behaviour-based robotics – tasks are accomplished in parallel – closed-loop(s) – protected from disturbances slide 30 30 15 © 2023 The University of Sheffield This lecture has covered … • • • • • • • • Regulatory/control systems Open-loop vs. closed-loop control Control theory Servos Multiple control loops Perceptual control theory Hierarchical control Layered control COM2009-3009 Robotics: Lecture 5 slide 31 31 © 2023 The University of Sheffield Any Questions ? (or you can post on the ‘Lecture’ Discussion Forum) COM2009-3009 Robotics: Lecture 5 slide 32 32 16 © 2023 The University of Sheffield Next lecture … PID Control COM2009-3009 Robotics: Lecture 5 slide 33 33 © 2023 The University of Sheffield COM2009-3009 Robotics: Lecture 5 slide 34 34 17