SlideShare a Scribd company logo
1 of 41
Control-theoretic approach to the
analysis and synthesis of
sensorimotor loops
A few main principles and connections to neuroscience
Neurotheory and Engineering seminar - 05/28/2013
Matteo Mischiati
1
โ€ข Control theory framework
- linear time-invariant (LTI) case
โ€ข Properties of feedback
- internal model principle
โ€ข Common control schemes
- forward and inverse models, Smith predictor
- state feedback, observers, optimal control
โ€ข A model of human response in manual
tracking tasks (Kleinman et al. 1970, Gawthrop et al. 2011)
2Matteo Mischiati Control theory primer
Control theory framework
๐‘ข = ๐‘–๐‘›๐‘๐‘ข๐‘ก (๐‘š๐‘œ๐‘ก๐‘œ๐‘Ÿ)
๐‘ฆ = ๐‘œ๐‘ข๐‘ก๐‘๐‘ข๐‘ก ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ
๐’™ = ๐‘–๐‘›๐‘ก๐‘’๐‘Ÿ๐‘›๐‘Ž๐‘™ ๐‘ ๐‘ก๐‘Ž๐‘ก๐‘’
๐‘‘ = ๐‘‘๐‘–๐‘ ๐‘ก๐‘ข๐‘Ÿ๐‘๐‘Ž๐‘›๐‘๐‘’
๐‘› = ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐‘›๐‘œ๐‘–๐‘ ๐‘’
Assume you have a system (PLANT) for which you can control certain variables (inputs) and sense
others (outputs). There may be disturbances on your inputs and your sensed outputs may be noisy.
๐’™ = ๐‘“ ๐’™, ๐‘ข
๐‘ฆ = ๐‘” ๐’™, ๐‘ข
๐‘ข ๐‘ฆ
PLANT
๐‘ข ๐‘
++
๐‘›
++
๐‘‘
๐‘ฆ ๐‘›
3Matteo Mischiati Control theory primer
Control theory framework
๐‘ข = ๐‘–๐‘›๐‘๐‘ข๐‘ก (๐‘š๐‘œ๐‘ก๐‘œ๐‘Ÿ)
๐‘ฆ = ๐‘œ๐‘ข๐‘ก๐‘๐‘ข๐‘ก ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ
๐’™ = ๐‘–๐‘›๐‘ก๐‘’๐‘Ÿ๐‘›๐‘Ž๐‘™ ๐‘ ๐‘ก๐‘Ž๐‘ก๐‘’
๐‘‘ = ๐‘‘๐‘–๐‘ ๐‘ก๐‘ข๐‘Ÿ๐‘๐‘Ž๐‘›๐‘๐‘’
๐‘› = ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐‘›๐‘œ๐‘–๐‘ ๐‘’
SYNTHESIS problem: design a controller that applies the right inputs to the plant, based on the
noisy outputs, to achieve a desired goal while satisfying one or more performance criteria
๐’™ = ๐‘“ ๐’™, ๐‘ข
๐‘ฆ = ๐‘” ๐’™, ๐‘ข
๐‘ข ๐‘ฆ
PLANT
๐‘ข ๐‘
++
๐‘›
++
๐‘‘
๐‘ฆ ๐‘›๐‘ฆ ๐’› = โ„Ž ๐’›, ๐‘ฆ, ๐‘ฆ๐‘›
๐‘ข ๐‘ = ๐‘– ๐’›, ๐‘ฆ, ๐‘ฆ ๐‘›
CONTROLLER
4Matteo Mischiati Control theory primer
Control theory framework
๐‘ข = ๐‘–๐‘›๐‘๐‘ข๐‘ก (๐‘š๐‘œ๐‘ก๐‘œ๐‘Ÿ)
๐‘ฆ = ๐‘œ๐‘ข๐‘ก๐‘๐‘ข๐‘ก ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ
๐’™ = ๐‘–๐‘›๐‘ก๐‘’๐‘Ÿ๐‘›๐‘Ž๐‘™ ๐‘ ๐‘ก๐‘Ž๐‘ก๐‘’
๐‘‘ = ๐‘‘๐‘–๐‘ ๐‘ก๐‘ข๐‘Ÿ๐‘๐‘Ž๐‘›๐‘๐‘’
๐‘› = ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐‘›๐‘œ๐‘–๐‘ ๐‘’
SYNTHESIS problem: design a controller that applies the right inputs to the plant, based on the
noisy outputs, to achieve a desired goal while satisfying one or more performance criteria
Possible Goals:
โ€ข Output regulation (disturbance rejection, homeostasis) : keep ๐‘ฆ constant despite disturbance
โ€ข Trajectory tracking : keep ๐‘ฆ ๐‘ก โ‰ˆ ๐‘ฆ ๐‘ก โˆ€ ๐‘ก
Performance criteria:
โ€ข Static performance (at steady state): e.g. lim
๐‘กโ†’โˆž
๐‘ฆ ๐‘ก โˆ’ ๐‘ฆ ๐‘ก
โ€ข Dynamic performance: transient time, etcโ€ฆ
โ€ข Stability: not blowing up!
โ€ข Robustness: amount of disturbance that can be tolerated
โ€ข Limited control effort
๐’™ = ๐‘“ ๐’™, ๐‘ข
๐‘ฆ = ๐‘” ๐’™, ๐‘ข
๐‘ข ๐‘ฆ
PLANT
๐‘ข ๐‘
++
๐‘›
++
๐‘‘
๐‘ฆ ๐‘›๐‘ฆ ๐’› = โ„Ž ๐’›, ๐‘ฆ, ๐‘ฆ๐‘›
๐‘ข ๐‘ = ๐‘– ๐’›, ๐‘ฆ, ๐‘ฆ ๐‘›
CONTROLLER
5Matteo Mischiati Control theory primer
Control theory framework
๐‘ข = ๐‘–๐‘›๐‘๐‘ข๐‘ก (๐‘š๐‘œ๐‘ก๐‘œ๐‘Ÿ)
๐‘ฆ = ๐‘œ๐‘ข๐‘ก๐‘๐‘ข๐‘ก ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ
๐’™ = ๐‘–๐‘›๐‘ก๐‘’๐‘Ÿ๐‘›๐‘Ž๐‘™ ๐‘ ๐‘ก๐‘Ž๐‘ก๐‘’
๐‘‘ = ๐‘‘๐‘–๐‘ ๐‘ก๐‘ข๐‘Ÿ๐‘๐‘Ž๐‘›๐‘๐‘’
๐‘› = ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐‘›๐‘œ๐‘–๐‘ ๐‘’
ANALYSIS problem: infer the functional structure of the controller, given the observed
performance of the overall system in multiple tasks
Goals:
โ€ข Output regulation (disturbance rejection, homeostasis) : keep ๐‘ฆ constant despite disturbance
โ€ข Trajectory tracking : keep ๐‘ฆ ๐‘ก โ‰ˆ ๐‘ฆ ๐‘ก โˆ€ ๐‘ก
Performance criteria:
โ€ข Static performance (at steady state): e.g. lim
๐‘กโ†’โˆž
๐‘ฆ ๐‘ก โˆ’ ๐‘ฆ ๐‘ก
โ€ข Dynamic performance: transient time, etcโ€ฆ
โ€ข Stability: not blowing up!
โ€ข Robustness: amount of disturbance that can be tolerated
โ€ข Limited control effort
๐’™ = ๐‘“ ๐’™, ๐‘ข
๐‘ฆ = ๐‘” ๐’™, ๐‘ข
++
๐‘ข
๐‘›
๐‘ฆ
PLANT
++
๐‘‘
๐‘ข ๐‘ ๐‘ฆ ๐‘›๐‘ฆ ๐’› = โ„Ž ๐’›, ๐‘ฆ, ๐‘ฆ๐‘›
๐‘ข ๐‘ = ๐‘– ๐’›, ๐‘ฆ, ๐‘ฆ ๐‘›
CONTROLLER
?
6Matteo Mischiati Control theory primer
Example of analysis problem:
uncovering the dragonfly control system
๐’— ๐‘ซ๐‘ญ
๐’‰๐’†๐’‚๐’…
๐’“
โ€ข We want to precisely characterize the foraging behavior of the dragonfly
(what it does) to gain insight on its neural circuitry (how it does it).
๐’— ๐‘ซ๐‘ญ
๐’‰๐’†๐’‚๐’…
๐’‡๐’๐’š
?
๐’“, relative to
dragonfly, in
๐’‰๐’†๐’‚๐’… ref. frame
dragonfly accel.
head rotation
dragonfly
head, body &
wing dynamics
dragonfly
visual
system
๐‘๐„๐“๐ˆ๐๐€ ? ๐“๐’๐ƒ๐s ?
๐–๐ˆ๐๐† ๐Œ๐”๐’๐‚๐‹๐„๐’
๐๐„๐‚๐Š ๐Œ๐”๐’๐‚๐‹๐„๐’
๐’ƒ๐’๐’…๐’š
๐’ƒ๐’๐’…๐’š
7Matteo Mischiati Control theory primer
Linear time-invariant systems
๐‘ข = ๐‘–๐‘›๐‘๐‘ข๐‘ก (๐‘š๐‘œ๐‘ก๐‘œ๐‘Ÿ)
๐‘ฆ = ๐‘œ๐‘ข๐‘ก๐‘๐‘ข๐‘ก ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ
๐’™ = ๐‘–๐‘›๐‘ก๐‘’๐‘Ÿ๐‘›๐‘Ž๐‘™ ๐‘ ๐‘ก๐‘Ž๐‘ก๐‘’
๐‘‘ = ๐‘‘๐‘–๐‘ ๐‘ก๐‘ข๐‘Ÿ๐‘๐‘Ž๐‘›๐‘๐‘’
๐‘› = ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐‘›๐‘œ๐‘–๐‘ ๐‘’
๐‘ƒ ๐‘  =
๐‘ โˆ’๐‘ง1 ๐‘ โˆ’๐‘ง2 โ€ฆ(๐‘ โˆ’๐‘ง ๐‘š)
๐‘ โˆ’๐‘1 ๐‘ โˆ’๐‘2 โ€ฆ(๐‘ โˆ’๐‘ ๐‘›)
, ๐‘ง๐‘–โˆˆ โ„‚ = ๐‘ง๐‘’๐‘Ÿ๐‘œ๐‘ , ๐‘๐‘– โˆˆ โ„‚ = ๐‘๐‘œ๐‘™๐‘’๐‘ 
โ€ข Stability (of a system) โ‡” ๐‘1, โ€ฆ , ๐‘ ๐‘› have negative real part
โ€ข Performance (of a system): depends on location of poles and zeros
โ€ข Related to transfer function in frequency domain:
๐‘ข ๐‘ก = sin ๐œ”๐‘ก โ‡’ ๐‘ฆ ๐‘ก = ๐‘ƒ ๐‘—๐œ” โ‹… sin(๐œ”๐‘ก + ฮฆ ๐‘ƒ ๐‘—๐œ” )
๐’™ = ๐ด๐’™ + ๐ต๐‘ข
๐‘ฆ = ๐ถ๐’™ + ๐ท๐‘ข
++
๐‘ข
๐‘›
๐‘ฆ
PLANT
++
๐‘‘
๐‘ข ๐‘ ๐‘ฆ ๐‘›๐‘ฆ ๐’› = ๐ป๐’› + ๐บ ๐‘ฆ
๐‘ข ๐‘ = ๐ผ๐’› + ๐ฟ ๐‘ฆ
CONTROLLER
๐‘ƒ(๐‘ )
++
๐‘ˆ(๐‘ )
๐‘(๐‘ )
๐‘Œ(๐‘ )
PLANT
++
๐ท(๐‘ )
๐‘ˆ ๐ถ(๐‘ ) ๐‘Œ๐‘›(๐‘ )๐‘Œ(๐‘ ) ๐ถ(๐‘ )
CONTROLLER
Laplace transform
Y ๐‘  = ๐‘ƒ ๐‘  โˆ™ ๐‘ˆ(๐‘ )
8Matteo Mischiati Control theory primer
Linear time-invariant systems
Laplace transform for signals:
โ€ข Final value theorem : lim
๐‘กโ†’โˆž
๐‘ฆ ๐‘ก = lim
๐‘ โ†’0
๐‘  โˆ™ ๐‘Œ ๐‘  (if limit exists)
โ€ข If u ๐‘ก โŸท ๐‘ˆ ๐‘  , then u ๐‘ก โŸท
๐‘ˆ ๐‘ 
๐‘ 
Typical reference/disturbance signals:
- Step ๐‘Œ ๐‘  = ๐‘Ž
1
๐‘ 
- Ramp ๐‘Œ ๐‘  = ๐‘Ž
1
๐‘ 2
- Sinusoid ๐‘Œ ๐‘  =
๐œ”
๐‘ 2+๐œ”2
๐‘ƒ(๐‘ )
++
๐‘ˆ(๐‘ )
๐‘(๐‘ )
๐‘Œ(๐‘ )
PLANT
++
๐ท(๐‘ )
๐‘ˆ ๐ถ(๐‘ ) ๐‘Œ๐‘›(๐‘ )๐‘Œ(๐‘ ) ๐ถ(๐‘ )
CONTROLLER
Y ๐‘  = ๐‘ƒ ๐‘  โˆ™ ๐‘ˆ(๐‘ )
๐‘ก
๐‘ฆ(๐‘ก)
๐‘Ž
๐‘ก
๐‘ฆ(๐‘ก)
๐‘Ž๐‘ก
๐‘ก
๐‘ฆ(๐‘ก)
sin(๐œ”๐‘ก)
9Matteo Mischiati Control theory primer
โ€ข Control theory framework
- linear time-invariant (LTI) case
โ€ข Properties of feedback
- internal model principle
โ€ข Common control schemes
- forward and inverse models, Smith predictor
- state feedback, observers, optimal control
โ€ข A model of human response in manual
tracking tasks (Kleinman et al. 1970, Gawthrop et al. 2011)
10Matteo Mischiati Control theory primer
Feedforward (inverse model)
Stability
Depends on poles (and zeros!) of ๐‘ƒ(๐‘ )
Performance (static and dynamic)
Arbitrarily good if ๐‘ƒ ๐‘  โ‰ˆ ๐‘ƒ ๐‘  and its inverse exists and is stable: Y ๐‘  โ‰ˆ
๐‘ƒ ๐‘  โˆ™ ๐‘ƒโˆ’1 ๐‘  โ‹… ๐‘Œ(๐‘ ) โ‰ˆ ๐‘Œ(๐‘ )
Robustness to disturbance (disturbance rejection)
None : Y ๐‘  = ๐‘ƒ ๐‘  โˆ™ ( ๐‘ƒโˆ’1 ๐‘  โ‹… ๐‘Œ ๐‘  + ๐ท(๐‘ )) โ‰ˆ ๐‘Œ(๐‘ ) + ๐‘ƒ(๐‘ ) โ‹… ๐ท(๐‘ )
๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ )
PLANT
++
๐ท(๐‘ )
๐‘ˆ ๐ถ(๐‘ )๐‘Œ(๐‘ ) ๐‘ƒโˆ’1
(๐‘ )
CONTROLLER
Y ๐‘  = ๐‘ƒ ๐‘  โˆ™ ๐‘ˆ(๐‘ )Uc ๐‘  = ๐‘ƒโˆ’1
๐‘  โˆ™ ๐‘Œ(๐‘ )
11Matteo Mischiati Control theory primer
Properties of Feedback
๐‘Œ ๐‘  = ๐‘ƒ ๐‘  ๐ท ๐‘  + ๐ถ ๐‘  ๐ธ ๐‘  = ๐‘ƒ ๐‘  ๐ท ๐‘  + ๐ถ ๐‘  ( ๐‘Œ ๐‘  โˆ’ ๐‘Œ ๐‘  ) โ‡’
๐‘Œ ๐‘  =
๐‘ƒ ๐‘  ๐ถ ๐‘ 
1 + ๐‘ƒ ๐‘  ๐ถ ๐‘†
๐‘Œ ๐‘  +
๐‘ƒ ๐‘ 
1 + ๐‘ƒ ๐‘  ๐ถ ๐‘†
๐ท(๐‘ )
Stability
Depends on 1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† . Can potentially stabilize unstable plants.
Disturbance rejection
Depends on 1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† . Can potentially attenuate/cancel effect of ๐ท ๐‘  .
๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ )
PLANT
++
๐ท(๐‘ )
๐‘ˆ ๐ถ(๐‘ )๐ธ(๐‘ ) ๐ถ(๐‘ )
CONTROLLER
๐‘Œ(๐‘ )
Y ๐‘  = ๐‘ƒ ๐‘  โˆ™ ๐‘ˆ(๐‘ )Uc ๐‘  = ๐ถ ๐‘  โˆ™ ๐ธ(๐‘ )
+
-
12Matteo Mischiati Control theory primer
Properties of Feedback
Static performance
lim
๐‘กโ†’โˆž
๐‘’ ๐‘ก = lim
๐‘ โ†’0
๐‘  โˆ™ ๐ธ ๐‘  , ๐ธ ๐‘  =
1
1 + ๐‘ƒ ๐‘  ๐ถ ๐‘†
๐‘Œ ๐‘  โˆ’
๐‘ƒ ๐‘ 
1 + ๐‘ƒ ๐‘  ๐ถ ๐‘†
๐ท(๐‘ )
Letโ€™s see how different controllers perform:
๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ )
PLANT
++
๐ท(๐‘ )
๐‘ˆ ๐ถ(๐‘ )๐ธ(๐‘ ) ๐ถ(๐‘ )
CONTROLLER
๐‘Œ(๐‘ )
+
-
๐‘’. ๐‘”.
1
1 + ๐œ๐‘ 
๐‘˜
13Matteo Mischiati Control theory primer
Properties of Feedback
Static performance
lim
๐‘กโ†’โˆž
๐‘’ ๐‘ก = lim
๐‘ โ†’0
๐‘  โˆ™ ๐ธ ๐‘  , ๐ธ ๐‘  =
1
1 + ๐‘ƒ ๐‘  ๐ถ ๐‘†
๐‘Œ ๐‘  โˆ’
๐‘ƒ ๐‘ 
1 + ๐‘ƒ ๐‘  ๐ถ ๐‘†
๐ท(๐‘ )
Proportional controller: ๐ถ ๐‘  = ๐‘˜
lim
๐‘กโ†’โˆž
๐‘’ ๐‘ก = lim
๐‘ โ†’0
๐‘  โˆ™
1
1+๐‘˜ ๐‘ƒ ๐‘ 
๐‘Ž
๐‘ 
=
๐‘Ž
1+๐‘˜
โ‰  0 errors in tracking a step
(but small if ๐‘˜ is large)
lim
๐‘กโ†’โˆž
๐‘’ ๐‘ก = lim
๐‘ โ†’0
๐‘  โˆ™
1
1+๐‘˜ ๐‘ƒ ๐‘ 
๐‘Ž
๐‘ 2 โ†’ โˆž cannot track a ramp at all
๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ )
PLANT
++
๐ท(๐‘ )
๐‘ˆ ๐ถ(๐‘ )๐ธ(๐‘ ) ๐ถ(๐‘ )
CONTROLLER
๐‘Œ(๐‘ )
+
-
๐‘Ž
๐‘ 
๐‘Ž
๐‘ 2
๐‘’. ๐‘”.
1
1 + ๐œ๐‘ 
๐‘˜
14Matteo Mischiati Control theory primer
Properties of Feedback
Static performance
lim
๐‘กโ†’โˆž
๐‘’ ๐‘ก = lim
๐‘ โ†’0
๐‘  โˆ™ ๐ธ ๐‘  , ๐ธ ๐‘  =
1
1 + ๐‘ƒ ๐‘  ๐ถ ๐‘†
๐‘Œ ๐‘  โˆ’
๐‘ƒ ๐‘ 
1 + ๐‘ƒ ๐‘  ๐ถ ๐‘†
๐ท(๐‘ )
Proportional+Integral (PI) controller: ๐ถ ๐‘  = ๐‘˜ ๐‘ +
๐‘˜ ๐‘–
๐‘ 
=
๐‘˜ ๐‘ ๐‘ +๐‘˜ ๐‘–
๐‘ 
lim
๐‘กโ†’โˆž
๐‘’ ๐‘ก = lim
๐‘ โ†’0
๐‘  โˆ™
1
1+
๐‘˜ ๐‘ ๐‘ +๐‘˜ ๐‘–
๐‘ 
๐‘ƒ ๐‘ 
๐‘Ž
๐‘ 
= 0 perfect in tracking a step
lim
๐‘กโ†’โˆž
๐‘’ ๐‘ก = lim
๐‘ โ†’0
๐‘  โˆ™
1
1+
๐‘˜ ๐‘ ๐‘ +๐‘˜ ๐‘–
๐‘ 
๐‘ƒ ๐‘ 
๐‘Ž
๐‘ 2 =
๐‘Ž
๐‘˜ ๐‘–
โ‰  0 errors in tracking a ramp
๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ )
PLANT
++
๐ท(๐‘ )
๐‘ˆ ๐ถ(๐‘ )๐ธ(๐‘ ) ๐ถ(๐‘ )
CONTROLLER
๐‘Œ(๐‘ )
+
-
๐‘Ž
๐‘ 
๐‘Ž
๐‘ 2
1
1 + ๐œ๐‘ 
๐‘˜ ๐‘ +
๐‘˜๐‘–
๐‘ 
15Matteo Mischiati Control theory primer
(but small if ๐‘˜๐‘– is large)
Properties of Feedback
Static performance
lim
๐‘กโ†’โˆž
๐‘’ ๐‘ก = lim
๐‘ โ†’0
๐‘  โˆ™ ๐ธ ๐‘  , ๐ธ ๐‘  =
1
1 + ๐‘ƒ ๐‘  ๐ถ ๐‘†
๐‘Œ ๐‘  โˆ’
๐‘ƒ ๐‘ 
1 + ๐‘ƒ ๐‘  ๐ถ ๐‘†
๐ท(๐‘ )
To perfectly track: We need:
Step
Ramp
How about sinusoid? We need:
๐‘Ž
๐‘ 
๐‘Ž
๐‘ 2
๐œ”
๐‘ 2 + ๐œ”2
๐‘ƒ(๐‘ )๐ถ ๐‘  =
1
๐‘ 
โ‹… (๐‘ƒ ๐‘  ๐ถ ๐‘  )โ€ฒ
๐‘ƒ(๐‘ )๐ถ ๐‘  =
1
๐‘ 2 โ‹… (๐‘ƒ ๐‘  ๐ถ ๐‘  )โ€ฒ
๐‘ฆ ๐‘ก = sin ๐œ”๐‘ก
โ‡“
๐‘’ ๐‘ก = ๐บ ๐‘—๐œ” โ‹… sin(๐œ”๐‘ก + ฮฆ ๐บ ๐‘—๐œ” )
๐บ ๐‘ 
๐‘ƒ(๐‘ )๐ถ ๐‘  =
1
๐‘ 2 + ๐œ”2
โ‹… (๐‘ƒ ๐‘  ๐ถ ๐‘  )โ€ฒ
โ‡“
๐บ ๐‘—๐œ” = 1 + ๐‘ƒ ๐‘—๐œ” ๐ถ ๐‘—๐œ” โˆ’1
โ†’ 0
16Matteo Mischiati Control theory primer
Internal model principle
To perfectly track: We need:
Step
Ramp
Sinusoid
Internal model principle: To achieve asymptotical tracking of a reference signal
(rejection of a disturbance signal) via feedback, the controller (or the plant) must
contain an โ€œinternal modelโ€ of the signal.
It is a necessary condition, not a sufficient condition (need also stability)
๐‘Ž
๐‘ 
๐‘Ž
๐‘ 2
๐œ”
๐‘ 2 + ๐œ”2
๐‘ƒ(๐‘ )๐ถ ๐‘  =
1
๐‘ 
โ‹… (๐‘ƒ ๐‘  ๐ถ ๐‘  )โ€ฒ
๐‘ƒ(๐‘ )๐ถ ๐‘  =
1
๐‘ 2
โ‹… (๐‘ƒ ๐‘  ๐ถ ๐‘  )โ€ฒ
๐‘ƒ(๐‘ )๐ถ ๐‘  =
1
๐‘ 2 + ๐œ”2
โ‹… (๐‘ƒ ๐‘  ๐ถ ๐‘  )โ€ฒ
17Matteo Mischiati Control theory primer
Feedback vs. Feedforward
Feedback
โ€ข is needed if plant is unstable or for disturbance rejection
โ€ข does not require full knowledge of the plant
โ€ข incorporating the knowledge of possible reference and disturbance
signals is very useful (internal model principle)
Feedforward
โ€ข if plant is known, and no disturbance, its performance canโ€™t be beat
โ€ข no sensory delays
๐‘Œ ๐‘  =
๐‘ƒ ๐‘  ๐ถ ๐‘ 
1 + ๐‘ƒ ๐‘  ๐ถ ๐‘†
๐‘Œ ๐‘  +
๐‘ƒ ๐‘ 
1 + ๐‘ƒ ๐‘  ๐ถ ๐‘†
๐ท(๐‘ ) ๐‘Œ ๐‘  = ๐‘ƒ ๐‘  โˆ™ ( ๐‘ƒโˆ’1
๐‘  โ‹… ๐‘Œ ๐‘  + ๐ท(๐‘ ))
โ‰ˆ ๐‘Œ(๐‘ ) + ๐‘ƒ(๐‘ ) โ‹… ๐ท(๐‘ )
18Matteo Mischiati Control theory primer
โ€ข Control theory framework
- linear time-invariant (LTI) case
โ€ข Properties of feedback
- internal model principle
โ€ข Common control schemes
- forward and inverse models, Smith predictor
- state feedback, observers, optimal control
โ€ข A model of human response in manual
tracking tasks (Kleinman et al. 1970, Gawthrop et al. 2011)
19Matteo Mischiati Control theory primer
Feedback + Feedforward
The feedback controller kicks in only if inverse model is incorrect.
๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ )
PLANT
++
๐‘ˆ ๐น๐ต(๐‘ )
๐ธ(๐‘ ) ๐ถ(๐‘ )๐‘Œ(๐‘ )
+
-
๐‘ƒโˆ’1
(๐‘ )
INVERSE MODEL
FEEDBACK
๐‘ˆ ๐น๐น(๐‘ )
20Matteo Mischiati Control theory primer
Feedback + Feedforward
The feedback controller kicks in only if inverse model is incorrect.
The corrective command from the feedback path can be also used as a
learning/adaptation signal by the inverse model.
๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ )
PLANT
++
๐‘ˆ ๐น๐ต(๐‘ )
๐ธ(๐‘ ) ๐ถ(๐‘ )๐‘Œ(๐‘ )
+
-
๐‘ƒโˆ’1
(๐‘ )
INVERSE MODEL
FEEDBACK
๐‘ˆ ๐น๐น(๐‘ )
21Matteo Mischiati Control theory primer
Feedback + Feedforward
The feedback controller kicks in only if inverse model is incorrect.
The corrective command from the feedback path can be also used as a
learning/adaptation signal by the inverse model.
Significant sensory delays are still a problem.
๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ )
PLANT
++
๐‘ˆ ๐น๐ต(๐‘ )
๐ธ(๐‘ ) ๐ถ(๐‘ )๐‘Œ(๐‘ )
+
-
๐‘ƒโˆ’1
(๐‘ )
INVERSE MODEL
FEEDBACK
๐‘ˆ ๐น๐น(๐‘ )
๐‘’โˆ’๐‘ ๐œ
22Matteo Mischiati Control theory primer
Forward model
The control signal is sent through a model of the plant (โ€œforward modelโ€) to
predict the sensory output.
๐‘ƒ(๐‘ ) ๐‘Œ(๐‘ )
PLANT
๐‘ˆ ๐ถ(๐‘ )๐ธ(๐‘ ) ๐ถ(๐‘ )๐‘Œ(๐‘ )
+
-
๐‘ƒ(๐‘ )
FORWARD MODEL
CONTROLLER
predicted sensory output
23Matteo Mischiati Control theory primer
Forward model
The control signal is sent through a model of the plant (โ€œforward modelโ€) to
predict the sensory output.
The (delayed) sensory output can be used as a learning/adaptation signal for
the forward model.
Direct use of the delayed sensory output in the control is problematic
because of time mismatch.
๐‘ƒ(๐‘ ) ๐‘Œ(๐‘ )
PLANT
๐‘ˆ ๐ถ(๐‘ )๐ธ(๐‘ ) ๐ถ(๐‘ )๐‘Œ(๐‘ )
+
-
๐‘ƒ(๐‘ )
FORWARD MODEL
๐‘’โˆ’๐‘ ๐œ
CONTROLLER
predicted sensory output
24Matteo Mischiati Control theory primer
Smith predictor
Assuming ๐‘ƒ ๐‘  โ‰ˆ ๐‘ƒ ๐‘  and ๐œ โ‰ˆ ๐œ:
๐‘Œ ๐‘  = ๐‘ƒ ๐‘  ๐‘ˆ๐‘ ๐‘  =
๐‘ƒ ๐‘  ๐ถ ๐‘ 
1 + ๐‘ƒ ๐‘  ๐ถ ๐‘†
๐‘Œ ๐‘ 
Delay has been moved outside the control loop.
PLANT
๐‘ƒ(๐‘ ) ๐‘Œ(๐‘ )๐‘ˆ ๐ถ(๐‘ )๐ธ(๐‘ ) ๐ถ(๐‘ )๐‘Œ(๐‘ )
+
- -
๐‘ƒ(๐‘ )
๐‘’โˆ’๐‘ ๐œ
๐‘’โˆ’๐‘  ๐œ
+ -
delay model
plant model
predicted sensory output
error in sensory output prediction
CONTROLLER
25Matteo Mischiati Control theory primer
Models of the cerebellum
1. Cerebellum as an inverse
model in a feedback+feedforward
motor control scheme
Wolpert, Miall & Kawato, 1998 โ€œInternal models in the cerebellumโ€
Not in the sense of my presentation !
26Matteo Mischiati Control theory primer
Models of the cerebellum
2. Cerebellum as a
forward model in a
Smith predictor
control scheme
Wolpert, Miall & Kawato, 1998 โ€œInternal models in the cerebellumโ€
27Matteo Mischiati Control theory primer
State feedback
๐’™ = ๐‘“ ๐’™, ๐‘ข
๐‘ฆ = ๐‘” ๐’™, ๐‘ข
๐‘ฆ
PLANT
๐‘ข๐‘ฆ ๐’› = โ„Ž ๐’›, ๐‘ฆ, ๐’™
๐‘ข = ๐‘– ๐’›, ๐‘ฆ, ๐’™
CONTROLLER
๐’™
๐‘ฆ
PLANT
๐‘ข๐‘ฆ
CONTROLLER
๐’™
๐’™ = ๐ด๐’™ + ๐ต๐‘ข
๐‘ฆ = ๐ถ๐’™ + ๐ท๐‘ข
๐’› = ๐ป๐’› + ๐บ ๐‘ฆ + ๐‘€๐’™
๐‘ข = ๐ผ๐’› + ๐ฟ ๐‘ฆ + ๐‘๐’™
Linear time-invariant case:
28Matteo Mischiati Control theory primer
State feedback
If the plant is reachable, it is possible to achieve any arbitrary choice of
closed-loop poles with an appropriate linear and memoryless controller:
u = โˆ’๐พ๐’™ + ๐พ๐‘Ÿ ๐‘ฆ
๐’™ = ๐‘“ ๐’™, ๐‘ข
๐‘ฆ = ๐‘” ๐’™, ๐‘ข
๐‘ฆ
PLANT
๐‘ข๐‘ฆ ๐’› = โ„Ž ๐’›, ๐‘ฆ, ๐’™
๐‘ข = ๐‘– ๐’›, ๐‘ฆ, ๐’™
CONTROLLER
๐’™
๐‘ฆ
PLANT
๐‘ข๐‘ฆ
CONTROLLER
๐’™
๐’™ = ๐ด๐’™ + ๐ต๐‘ข
๐‘ฆ = ๐ถ๐’™ + ๐ท๐‘ข
Linear time-invariant case:
๐พ
๐พ๐‘Ÿ +
-
29Matteo Mischiati Control theory primer
Observers
Observer: dynamical system designed to estimate the full state (when not
fully available)
If the plant is observable, it is possible to achieve lim
๐‘กโ†’โˆž
๐’™ ๐‘ก = ๐’™ (with right ๐ฟ)
๐‘ฆ
PLANT
๐‘ข
๐’™
๐’™ = ๐ด๐’™ + ๐ต๐‘ข
๐‘ฆ = ๐ถ๐’™
๐’™ = ๐ด ๐’™ + ๐ต๐‘ข + ๐ฟ(y โˆ’ C ๐’™)
OBSERVER
30Matteo Mischiati Control theory primer
Observers
Observer: dynamical system designed to estimate the full state (when not
fully available)
If the plant is observable, it is possible to achieve lim
๐‘กโ†’โˆž
๐’™ ๐‘ก = ๐’™ (with right ๐ฟ)
Separation principle: if the plant is reachable & observable, can replace ๐’™
with ๐’™ and design ๐พ independently of ๐ฟ (use observed state just as real one)
๐‘ฆ
PLANT
๐‘ข
๐’™
๐’™ = ๐ด๐’™ + ๐ต๐‘ข
๐‘ฆ = ๐ถ๐’™
๐’™ = ๐ด ๐’™ + ๐ต๐‘ข + ๐ฟ(y โˆ’ C ๐’™)
๐‘ฆ
๐พ
๐พ๐‘Ÿ +-
OBSERVER
31Matteo Mischiati Control theory primer
Optimal control
Linear Quadratic Gaussian (LQG) optimal output regulation: linear plant,
additive Gaussian white noise on state (with covariance ฮฃ ๐‘‘) and output (๐œŽ ๐‘›);
minimize quadratic cost :
๐”ผ lim
๐‘‡โ†’โˆž
1
๐‘‡ 0
๐‘‡
๐’™ ๐‘‡ ๐‘ก ๐‘„๐’™ ๐‘ก + ๐‘ข ๐‘ก 2 ๐‘‘๐‘ก
Solution is linear observer (Kalman filter) with linear memoryless controller:
๐ฟ = ๐‘ƒ๐ถ ๐‘‡ ๐œŽ ๐‘›
โˆ’1, ๐ด๐‘ƒ + ๐‘ƒ๐ด ๐‘‡ + ฮฃ ๐‘‘ โˆ’ ๐‘ƒ๐ถ ๐‘‡ ๐œŽ ๐‘›
โˆ’1 ๐ถ๐‘ƒ ๐‘‡ = 0
๐พ = ๐ต ๐‘‡ ๐‘†, ๐ด ๐‘‡ ๐‘† + ๐‘†๐ด + ๐‘„ โˆ’ ๐‘†๐ต๐ต ๐‘‡ ๐‘† = 0
๐‘ฆ
PLANT
๐‘ข ๐’™ = ๐ด๐’™ + ๐ต๐‘ข + ๐’…
๐‘ฆ = ๐ถ๐’™
๐’™
๐’™ = ๐ด ๐’™ + ๐ต๐‘ข + ๐ฟ(yn โˆ’ C ๐’™)
๐‘ฆ = 0
๐พ
+-
OBSERVER (KALMAN FILTER)
๐’…
++
๐‘›
๐‘ฆ ๐‘›
32Matteo Mischiati Control theory primer
Internal model principle
Internal model principle (state space): to achieve asymptotical tracking of a
reference signal (rejection of a disturbance signal) produced by an exosystem,
the controller must contain an โ€œinternal modelโ€ of the exosystem.
(Francis & Wonham, Automatica, 1970)
It is a necessary condition, not a sufficient condition (need also stability).
General principle with extensions to nonlinear systems.
๐‘ฆ
PLANT
๐’–๐‘ฆ
๐’™ = ๐ด๐’™ + ๐ต๐‘ข
๐‘ฆ = ๐ถ๐’™
๐œผ = ๐‘†๐œผ + ๐บ๐‘’๐’› = ๐‘† ๐’›
๐‘ฆ = ๐‘‡ ๐’›
+
-
๐‘’
STABILIZING
CONTROLLER
๐œผINT.MODELEXOSYSTEM
33Matteo Mischiati Control theory primer
โ€ข Control theory framework
- linear time-invariant (LTI) case
โ€ข Properties of feedback
- internal model principle
โ€ข Common control schemes
- forward and inverse models, Smith predictor
- state feedback, observers, optimal control
โ€ข A model of human response in manual
tracking tasks (Kleinman et al. 1970, Gawthrop et al. 2011)
34Matteo Mischiati Control theory primer
Human response model
D. L. Kleinman, S. Baron and W. H. Levison, โ€œAn Optimal Control Model of
Human Response. Part I: Theory and Validationโ€, Automatica, 1970
(revisited, more recently, by Gawthrop et al. 2011)
Task: by controlling a joystick (position, velocity or acceleration control),
subject is asked to keep a cursor on the screen as close as possible to a
target location, while unknown disturbances are applied by the computer.
Plant:
๐‘š๐‘œ๐‘›๐‘–๐‘ก๐‘œ๐‘Ÿ ๐‘Œ(๐‘ )
++
๐ท(๐‘ ) (computer)
๐‘ˆ(๐‘ )
๐‘—๐‘œ๐‘ฆ๐‘ ๐‘ก๐‘–๐‘๐‘˜ ๐‘ƒ๐ฝ๐‘€ ๐‘  โˆˆ ๐‘˜,
๐‘˜
๐‘ 
,
๐‘˜
๐‘ 2
๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ
๐‘“๐‘’๐‘’๐‘‘๐‘๐‘Ž๐‘๐‘˜
โ€œHuman controllerโ€:
๐‘›๐‘’๐‘ข๐‘Ÿ๐‘œ๐‘š๐‘œ๐‘ก๐‘œ๐‘Ÿ
dynamics
๐‘ˆ(๐‘ )
++
๐‘š๐‘œ๐‘ก๐‘œ๐‘Ÿ ๐‘›๐‘œ๐‘–๐‘ ๐‘’
๐‘›๐‘’๐‘ข๐‘Ÿ๐‘Ž๐‘™
computation
๐‘ƒ ๐‘ ๐‘ ๐ถ ๐‘ 
๐‘’โˆ’๐‘ ๐œ
๐‘ƒ ๐‘ ๐‘  โ‰ˆ
1
๐œ ๐‘› ๐‘ +1
๐œ ๐‘› โ‰ˆ 100๐‘š๐‘ 
๐œ โ‰ˆ 150 โˆ’ 250๐‘š๐‘ 
35Matteo Mischiati Control theory primer
Human response model
Task: Output regulation/disturbance rejection with linear time-invariant
plant and significant delay on any potential feedback line
๐‘’โˆ’๐‘ ๐œ
๐‘ƒ(๐‘ )
++
๐‘ˆ(๐‘ )
๐‘(๐‘ )
๐‘Œ(๐‘ )
++
๐ท(๐‘ )
๐‘ˆ ๐ถ(๐‘ ) ๐‘Œ๐‘›(๐‘ )๐ถ(๐‘ )
CONTROLLER
๐‘ƒ๐ฝ๐‘€ ๐‘  โˆ™ ๐‘ƒ ๐‘(๐‘ )
ANALYSIS problem: infer a model of the neural controller ๐ถ(๐‘ ) from the
observed performance of the subjects tested.
36Matteo Mischiati Control theory primer
Human response model
Task: Output regulation/disturbance rejection with linear time-invariant
plant and significant delay on any potential feedback line
๐‘’โˆ’๐‘ ๐œ
๐‘ƒ(๐‘ )
++
๐‘ˆ(๐‘ )
๐‘(๐‘ )
๐‘Œ(๐‘ )
++
๐ท(๐‘ )
๐‘ˆ ๐ถ(๐‘ ) ๐‘Œ๐‘›(๐‘ )๐ถ(๐‘ )
CONTROLLER
๐‘ƒ๐ฝ๐‘€ ๐‘  โˆ™ ๐‘ƒ ๐‘(๐‘ )
ANALYSIS problem: infer a model of the neural controller ๐ถ(๐‘ ) from the
observed performance of the subjects tested.
So what are the performances?
โ€ข Very good and robust to disturbances up to 2Hz (sum of sinusoids), for all
three types of joystick dynamics
โ€ข Apparently delay-free Must be some kind of
FEEDBACK + FORWARD model !
37Matteo Mischiati Control theory primer
Human response model
Hypothesis: optimal control to minimize average error & control effort
๐”ผ lim
๐‘‡โ†’โˆž
1
๐‘‡ 0
๐‘‡
๐’™ ๐‘ก 2
+ ๐›ผ ๐‘ข ๐‘ก 2
+ ๐›ฝ ๐‘ข ๐‘ก 2
๐‘‘๐‘ก
Theoretical solution * (with assumptions similar to LQG problem):
- Optimal observer (Kalman filter) to estimate delayed state (as in LQG)
- Optimal least mean-squared predictor to predict current state
- Optimal linear memoryless controller (as in LQG)
๐‘ฆ
PLANT
๐‘ข
๐’™(๐‘ก โˆ’ ๐œ)
๐’™ = ๐ด๐’™ + ๐ต๐‘ข + ๐’…
๐‘ฆ = ๐ถ๐’™
KALMAN FILTER
๐‘ฆ = 0
๐พ
+-
๐’…
++
๐‘›
๐‘ฆ ๐‘›
* D. Kleinman, โ€œOptimal control of linear systems with time-delay and observation noiseโ€, IEEE Trans. Autom. Control, 1969
๐‘’โˆ’๐‘ ๐œPREDICTOR
๐’™(๐‘ก) ๐‘ฆ๐‘›(๐‘ก โˆ’ ๐œ)
38Matteo Mischiati Control theory primer
Controller freq. response with plant
๐‘˜
๐‘ 
Controller freq. response with plant
๐‘˜
๐‘ 2
Matteo Mischiati Control theory primer 39
Human response model
Gawthrop et al. * (2011):
- Introduced, in both estimator and predictor, a copy of the exosystem
generating sinusoidal disturbances (internal model principle!)
- Show that intermittent control is also compatible with results
* P. Gawthrop et al., โ€œIntermittent control: a computational theory of human controlโ€, Biol. Cybern., 2011
Actual response to sinusoid Response without int.model
40Matteo Mischiati Control theory primer
โ€ข Crash course in control theory (for LTI systems)
- many concepts can be extended to more general settings
โ€ข An example of control-theoretic approach to
modeling sensorimotor loops
- need to iterate between modeling/experiments to discern
among alternatives and improve understanding of the system
Conclusions
THANK YOU FOR YOUR ATTENTION !
41Matteo Mischiati Control theory primer

More Related Content

Similar to Control-theoretic approach to sensorimotor loops

Lecture Notes: EEEC4340318 Instrumentation and Control Systems - Fundamental...
Lecture Notes:  EEEC4340318 Instrumentation and Control Systems - Fundamental...Lecture Notes:  EEEC4340318 Instrumentation and Control Systems - Fundamental...
Lecture Notes: EEEC4340318 Instrumentation and Control Systems - Fundamental...AIMST University
ย 
Vibration Isolation of a LEGOยฎ plate
Vibration Isolation of a LEGOยฎ plateVibration Isolation of a LEGOยฎ plate
Vibration Isolation of a LEGOยฎ plateOpen Adaptronik
ย 
Passivity-based control of rigid-body manipulator
Passivity-based control of rigid-body manipulatorPassivity-based control of rigid-body manipulator
Passivity-based control of rigid-body manipulatorHancheol Choi
ย 
Lecture 5 backpropagation
Lecture 5 backpropagationLecture 5 backpropagation
Lecture 5 backpropagationParveenMalik18
ย 
Non-linear control of a bipedal (Three-Linked) Walker using feedback Lineariz...
Non-linear control of a bipedal (Three-Linked) Walker using feedback Lineariz...Non-linear control of a bipedal (Three-Linked) Walker using feedback Lineariz...
Non-linear control of a bipedal (Three-Linked) Walker using feedback Lineariz...Mike Simon
ย 
Continuous control
Continuous controlContinuous control
Continuous controlReiji Hatsugai
ย 
Robust Presentation on h ifiinity based full state feedback controller design...
Robust Presentation on h ifiinity based full state feedback controller design...Robust Presentation on h ifiinity based full state feedback controller design...
Robust Presentation on h ifiinity based full state feedback controller design...VivekKumar265461
ย 
01-system_models_slides.pdf
01-system_models_slides.pdf01-system_models_slides.pdf
01-system_models_slides.pdfMammarSoulimane
ย 
Time Response in Control System
Time Response in Control SystemTime Response in Control System
Time Response in Control SystemAnshulShekhar3
ย 
Intro to Quantitative Investment (Lecture 1 of 6)
Intro to Quantitative Investment (Lecture 1 of 6)Intro to Quantitative Investment (Lecture 1 of 6)
Intro to Quantitative Investment (Lecture 1 of 6)Adrian Aley
ย 
14599404.ppt
14599404.ppt14599404.ppt
14599404.pptMonaIbrahim72
ย 
NIPS KANSAI Reading Group #5: State Aware Imitation Learning
NIPS KANSAI Reading Group #5: State Aware Imitation LearningNIPS KANSAI Reading Group #5: State Aware Imitation Learning
NIPS KANSAI Reading Group #5: State Aware Imitation LearningEiji Uchibe
ย 
Sampling method : MCMC
Sampling method : MCMCSampling method : MCMC
Sampling method : MCMCSEMINARGROOT
ย 
Stability analysis of impulsive fractional differential systems with delay
Stability analysis of impulsive fractional differential systems with delayStability analysis of impulsive fractional differential systems with delay
Stability analysis of impulsive fractional differential systems with delayMostafa Shokrian Zeini
ย 
Adaptive Type-2 Fuzzy Second Order Sliding Mode Control for Nonlinear Uncerta...
Adaptive Type-2 Fuzzy Second Order Sliding Mode Control for Nonlinear Uncerta...Adaptive Type-2 Fuzzy Second Order Sliding Mode Control for Nonlinear Uncerta...
Adaptive Type-2 Fuzzy Second Order Sliding Mode Control for Nonlinear Uncerta...rinzindorjej
ย 
Kalman filter for Beginners
Kalman filter for BeginnersKalman filter for Beginners
Kalman filter for Beginnerswinfred lu
ย 
Intro to Quant Trading Strategies (Lecture 6 of 10)
Intro to Quant Trading Strategies (Lecture 6 of 10)Intro to Quant Trading Strategies (Lecture 6 of 10)
Intro to Quant Trading Strategies (Lecture 6 of 10)Adrian Aley
ย 
Lecture 23 24-time_response
Lecture 23 24-time_responseLecture 23 24-time_response
Lecture 23 24-time_responseSyed Ali Raza Rizvi
ย 

Similar to Control-theoretic approach to sensorimotor loops (20)

Lecture Notes: EEEC4340318 Instrumentation and Control Systems - Fundamental...
Lecture Notes:  EEEC4340318 Instrumentation and Control Systems - Fundamental...Lecture Notes:  EEEC4340318 Instrumentation and Control Systems - Fundamental...
Lecture Notes: EEEC4340318 Instrumentation and Control Systems - Fundamental...
ย 
Vibration Isolation of a LEGOยฎ plate
Vibration Isolation of a LEGOยฎ plateVibration Isolation of a LEGOยฎ plate
Vibration Isolation of a LEGOยฎ plate
ย 
Passivity-based control of rigid-body manipulator
Passivity-based control of rigid-body manipulatorPassivity-based control of rigid-body manipulator
Passivity-based control of rigid-body manipulator
ย 
Lecture 5 backpropagation
Lecture 5 backpropagationLecture 5 backpropagation
Lecture 5 backpropagation
ย 
Fuzzy logic
Fuzzy logicFuzzy logic
Fuzzy logic
ย 
Non-linear control of a bipedal (Three-Linked) Walker using feedback Lineariz...
Non-linear control of a bipedal (Three-Linked) Walker using feedback Lineariz...Non-linear control of a bipedal (Three-Linked) Walker using feedback Lineariz...
Non-linear control of a bipedal (Three-Linked) Walker using feedback Lineariz...
ย 
Me314 week08-stability and steady state errors
Me314 week08-stability and steady state errorsMe314 week08-stability and steady state errors
Me314 week08-stability and steady state errors
ย 
Continuous control
Continuous controlContinuous control
Continuous control
ย 
Robust Presentation on h ifiinity based full state feedback controller design...
Robust Presentation on h ifiinity based full state feedback controller design...Robust Presentation on h ifiinity based full state feedback controller design...
Robust Presentation on h ifiinity based full state feedback controller design...
ย 
01-system_models_slides.pdf
01-system_models_slides.pdf01-system_models_slides.pdf
01-system_models_slides.pdf
ย 
Time Response in Control System
Time Response in Control SystemTime Response in Control System
Time Response in Control System
ย 
Intro to Quantitative Investment (Lecture 1 of 6)
Intro to Quantitative Investment (Lecture 1 of 6)Intro to Quantitative Investment (Lecture 1 of 6)
Intro to Quantitative Investment (Lecture 1 of 6)
ย 
14599404.ppt
14599404.ppt14599404.ppt
14599404.ppt
ย 
NIPS KANSAI Reading Group #5: State Aware Imitation Learning
NIPS KANSAI Reading Group #5: State Aware Imitation LearningNIPS KANSAI Reading Group #5: State Aware Imitation Learning
NIPS KANSAI Reading Group #5: State Aware Imitation Learning
ย 
Sampling method : MCMC
Sampling method : MCMCSampling method : MCMC
Sampling method : MCMC
ย 
Stability analysis of impulsive fractional differential systems with delay
Stability analysis of impulsive fractional differential systems with delayStability analysis of impulsive fractional differential systems with delay
Stability analysis of impulsive fractional differential systems with delay
ย 
Adaptive Type-2 Fuzzy Second Order Sliding Mode Control for Nonlinear Uncerta...
Adaptive Type-2 Fuzzy Second Order Sliding Mode Control for Nonlinear Uncerta...Adaptive Type-2 Fuzzy Second Order Sliding Mode Control for Nonlinear Uncerta...
Adaptive Type-2 Fuzzy Second Order Sliding Mode Control for Nonlinear Uncerta...
ย 
Kalman filter for Beginners
Kalman filter for BeginnersKalman filter for Beginners
Kalman filter for Beginners
ย 
Intro to Quant Trading Strategies (Lecture 6 of 10)
Intro to Quant Trading Strategies (Lecture 6 of 10)Intro to Quant Trading Strategies (Lecture 6 of 10)
Intro to Quant Trading Strategies (Lecture 6 of 10)
ย 
Lecture 23 24-time_response
Lecture 23 24-time_responseLecture 23 24-time_response
Lecture 23 24-time_response
ย 

Recently uploaded

1029 - Danh muc Sach Giao Khoa 10 . pdf
1029 -  Danh muc Sach Giao Khoa 10 . pdf1029 -  Danh muc Sach Giao Khoa 10 . pdf
1029 - Danh muc Sach Giao Khoa 10 . pdfQucHHunhnh
ย 
Call Girls in Dwarka Mor Delhi Contact Us 9654467111
Call Girls in Dwarka Mor Delhi Contact Us 9654467111Call Girls in Dwarka Mor Delhi Contact Us 9654467111
Call Girls in Dwarka Mor Delhi Contact Us 9654467111Sapana Sha
ย 
Measures of Dispersion and Variability: Range, QD, AD and SD
Measures of Dispersion and Variability: Range, QD, AD and SDMeasures of Dispersion and Variability: Range, QD, AD and SD
Measures of Dispersion and Variability: Range, QD, AD and SDThiyagu K
ย 
Mastering the Unannounced Regulatory Inspection
Mastering the Unannounced Regulatory InspectionMastering the Unannounced Regulatory Inspection
Mastering the Unannounced Regulatory InspectionSafetyChain Software
ย 
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...EduSkills OECD
ย 
Russian Call Girls in Andheri Airport Mumbai WhatsApp 9167673311 ๐Ÿ’ž Full Nigh...
Russian Call Girls in Andheri Airport Mumbai WhatsApp  9167673311 ๐Ÿ’ž Full Nigh...Russian Call Girls in Andheri Airport Mumbai WhatsApp  9167673311 ๐Ÿ’ž Full Nigh...
Russian Call Girls in Andheri Airport Mumbai WhatsApp 9167673311 ๐Ÿ’ž Full Nigh...Pooja Nehwal
ย 
BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...
BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...
BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...Sapna Thakur
ย 
Disha NEET Physics Guide for classes 11 and 12.pdf
Disha NEET Physics Guide for classes 11 and 12.pdfDisha NEET Physics Guide for classes 11 and 12.pdf
Disha NEET Physics Guide for classes 11 and 12.pdfchloefrazer622
ย 
Web & Social Media Analytics Previous Year Question Paper.pdf
Web & Social Media Analytics Previous Year Question Paper.pdfWeb & Social Media Analytics Previous Year Question Paper.pdf
Web & Social Media Analytics Previous Year Question Paper.pdfJayanti Pande
ย 
Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...
Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...
Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...fonyou31
ย 
Software Engineering Methodologies (overview)
Software Engineering Methodologies (overview)Software Engineering Methodologies (overview)
Software Engineering Methodologies (overview)eniolaolutunde
ย 
microwave assisted reaction. General introduction
microwave assisted reaction. General introductionmicrowave assisted reaction. General introduction
microwave assisted reaction. General introductionMaksud Ahmed
ย 
Z Score,T Score, Percential Rank and Box Plot Graph
Z Score,T Score, Percential Rank and Box Plot GraphZ Score,T Score, Percential Rank and Box Plot Graph
Z Score,T Score, Percential Rank and Box Plot GraphThiyagu K
ย 
Beyond the EU: DORA and NIS 2 Directive's Global Impact
Beyond the EU: DORA and NIS 2 Directive's Global ImpactBeyond the EU: DORA and NIS 2 Directive's Global Impact
Beyond the EU: DORA and NIS 2 Directive's Global ImpactPECB
ย 
1029-Danh muc Sach Giao Khoa khoi 6.pdf
1029-Danh muc Sach Giao Khoa khoi  6.pdf1029-Danh muc Sach Giao Khoa khoi  6.pdf
1029-Danh muc Sach Giao Khoa khoi 6.pdfQucHHunhnh
ย 
A Critique of the Proposed National Education Policy Reform
A Critique of the Proposed National Education Policy ReformA Critique of the Proposed National Education Policy Reform
A Critique of the Proposed National Education Policy ReformChameera Dedduwage
ย 
mini mental status format.docx
mini    mental       status     format.docxmini    mental       status     format.docx
mini mental status format.docxPoojaSen20
ย 

Recently uploaded (20)

Cรณdigo Creativo y Arte de Software | Unidad 1
Cรณdigo Creativo y Arte de Software | Unidad 1Cรณdigo Creativo y Arte de Software | Unidad 1
Cรณdigo Creativo y Arte de Software | Unidad 1
ย 
1029 - Danh muc Sach Giao Khoa 10 . pdf
1029 -  Danh muc Sach Giao Khoa 10 . pdf1029 -  Danh muc Sach Giao Khoa 10 . pdf
1029 - Danh muc Sach Giao Khoa 10 . pdf
ย 
Call Girls in Dwarka Mor Delhi Contact Us 9654467111
Call Girls in Dwarka Mor Delhi Contact Us 9654467111Call Girls in Dwarka Mor Delhi Contact Us 9654467111
Call Girls in Dwarka Mor Delhi Contact Us 9654467111
ย 
Measures of Dispersion and Variability: Range, QD, AD and SD
Measures of Dispersion and Variability: Range, QD, AD and SDMeasures of Dispersion and Variability: Range, QD, AD and SD
Measures of Dispersion and Variability: Range, QD, AD and SD
ย 
Mastering the Unannounced Regulatory Inspection
Mastering the Unannounced Regulatory InspectionMastering the Unannounced Regulatory Inspection
Mastering the Unannounced Regulatory Inspection
ย 
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
ย 
Russian Call Girls in Andheri Airport Mumbai WhatsApp 9167673311 ๐Ÿ’ž Full Nigh...
Russian Call Girls in Andheri Airport Mumbai WhatsApp  9167673311 ๐Ÿ’ž Full Nigh...Russian Call Girls in Andheri Airport Mumbai WhatsApp  9167673311 ๐Ÿ’ž Full Nigh...
Russian Call Girls in Andheri Airport Mumbai WhatsApp 9167673311 ๐Ÿ’ž Full Nigh...
ย 
BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...
BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...
BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...
ย 
Disha NEET Physics Guide for classes 11 and 12.pdf
Disha NEET Physics Guide for classes 11 and 12.pdfDisha NEET Physics Guide for classes 11 and 12.pdf
Disha NEET Physics Guide for classes 11 and 12.pdf
ย 
INDIA QUIZ 2024 RLAC DELHI UNIVERSITY.pptx
INDIA QUIZ 2024 RLAC DELHI UNIVERSITY.pptxINDIA QUIZ 2024 RLAC DELHI UNIVERSITY.pptx
INDIA QUIZ 2024 RLAC DELHI UNIVERSITY.pptx
ย 
Web & Social Media Analytics Previous Year Question Paper.pdf
Web & Social Media Analytics Previous Year Question Paper.pdfWeb & Social Media Analytics Previous Year Question Paper.pdf
Web & Social Media Analytics Previous Year Question Paper.pdf
ย 
Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...
Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...
Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...
ย 
Software Engineering Methodologies (overview)
Software Engineering Methodologies (overview)Software Engineering Methodologies (overview)
Software Engineering Methodologies (overview)
ย 
microwave assisted reaction. General introduction
microwave assisted reaction. General introductionmicrowave assisted reaction. General introduction
microwave assisted reaction. General introduction
ย 
Z Score,T Score, Percential Rank and Box Plot Graph
Z Score,T Score, Percential Rank and Box Plot GraphZ Score,T Score, Percential Rank and Box Plot Graph
Z Score,T Score, Percential Rank and Box Plot Graph
ย 
Mattingly "AI & Prompt Design: The Basics of Prompt Design"
Mattingly "AI & Prompt Design: The Basics of Prompt Design"Mattingly "AI & Prompt Design: The Basics of Prompt Design"
Mattingly "AI & Prompt Design: The Basics of Prompt Design"
ย 
Beyond the EU: DORA and NIS 2 Directive's Global Impact
Beyond the EU: DORA and NIS 2 Directive's Global ImpactBeyond the EU: DORA and NIS 2 Directive's Global Impact
Beyond the EU: DORA and NIS 2 Directive's Global Impact
ย 
1029-Danh muc Sach Giao Khoa khoi 6.pdf
1029-Danh muc Sach Giao Khoa khoi  6.pdf1029-Danh muc Sach Giao Khoa khoi  6.pdf
1029-Danh muc Sach Giao Khoa khoi 6.pdf
ย 
A Critique of the Proposed National Education Policy Reform
A Critique of the Proposed National Education Policy ReformA Critique of the Proposed National Education Policy Reform
A Critique of the Proposed National Education Policy Reform
ย 
mini mental status format.docx
mini    mental       status     format.docxmini    mental       status     format.docx
mini mental status format.docx
ย 

Control-theoretic approach to sensorimotor loops

  • 1. Control-theoretic approach to the analysis and synthesis of sensorimotor loops A few main principles and connections to neuroscience Neurotheory and Engineering seminar - 05/28/2013 Matteo Mischiati 1
  • 2. โ€ข Control theory framework - linear time-invariant (LTI) case โ€ข Properties of feedback - internal model principle โ€ข Common control schemes - forward and inverse models, Smith predictor - state feedback, observers, optimal control โ€ข A model of human response in manual tracking tasks (Kleinman et al. 1970, Gawthrop et al. 2011) 2Matteo Mischiati Control theory primer
  • 3. Control theory framework ๐‘ข = ๐‘–๐‘›๐‘๐‘ข๐‘ก (๐‘š๐‘œ๐‘ก๐‘œ๐‘Ÿ) ๐‘ฆ = ๐‘œ๐‘ข๐‘ก๐‘๐‘ข๐‘ก ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐’™ = ๐‘–๐‘›๐‘ก๐‘’๐‘Ÿ๐‘›๐‘Ž๐‘™ ๐‘ ๐‘ก๐‘Ž๐‘ก๐‘’ ๐‘‘ = ๐‘‘๐‘–๐‘ ๐‘ก๐‘ข๐‘Ÿ๐‘๐‘Ž๐‘›๐‘๐‘’ ๐‘› = ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐‘›๐‘œ๐‘–๐‘ ๐‘’ Assume you have a system (PLANT) for which you can control certain variables (inputs) and sense others (outputs). There may be disturbances on your inputs and your sensed outputs may be noisy. ๐’™ = ๐‘“ ๐’™, ๐‘ข ๐‘ฆ = ๐‘” ๐’™, ๐‘ข ๐‘ข ๐‘ฆ PLANT ๐‘ข ๐‘ ++ ๐‘› ++ ๐‘‘ ๐‘ฆ ๐‘› 3Matteo Mischiati Control theory primer
  • 4. Control theory framework ๐‘ข = ๐‘–๐‘›๐‘๐‘ข๐‘ก (๐‘š๐‘œ๐‘ก๐‘œ๐‘Ÿ) ๐‘ฆ = ๐‘œ๐‘ข๐‘ก๐‘๐‘ข๐‘ก ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐’™ = ๐‘–๐‘›๐‘ก๐‘’๐‘Ÿ๐‘›๐‘Ž๐‘™ ๐‘ ๐‘ก๐‘Ž๐‘ก๐‘’ ๐‘‘ = ๐‘‘๐‘–๐‘ ๐‘ก๐‘ข๐‘Ÿ๐‘๐‘Ž๐‘›๐‘๐‘’ ๐‘› = ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐‘›๐‘œ๐‘–๐‘ ๐‘’ SYNTHESIS problem: design a controller that applies the right inputs to the plant, based on the noisy outputs, to achieve a desired goal while satisfying one or more performance criteria ๐’™ = ๐‘“ ๐’™, ๐‘ข ๐‘ฆ = ๐‘” ๐’™, ๐‘ข ๐‘ข ๐‘ฆ PLANT ๐‘ข ๐‘ ++ ๐‘› ++ ๐‘‘ ๐‘ฆ ๐‘›๐‘ฆ ๐’› = โ„Ž ๐’›, ๐‘ฆ, ๐‘ฆ๐‘› ๐‘ข ๐‘ = ๐‘– ๐’›, ๐‘ฆ, ๐‘ฆ ๐‘› CONTROLLER 4Matteo Mischiati Control theory primer
  • 5. Control theory framework ๐‘ข = ๐‘–๐‘›๐‘๐‘ข๐‘ก (๐‘š๐‘œ๐‘ก๐‘œ๐‘Ÿ) ๐‘ฆ = ๐‘œ๐‘ข๐‘ก๐‘๐‘ข๐‘ก ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐’™ = ๐‘–๐‘›๐‘ก๐‘’๐‘Ÿ๐‘›๐‘Ž๐‘™ ๐‘ ๐‘ก๐‘Ž๐‘ก๐‘’ ๐‘‘ = ๐‘‘๐‘–๐‘ ๐‘ก๐‘ข๐‘Ÿ๐‘๐‘Ž๐‘›๐‘๐‘’ ๐‘› = ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐‘›๐‘œ๐‘–๐‘ ๐‘’ SYNTHESIS problem: design a controller that applies the right inputs to the plant, based on the noisy outputs, to achieve a desired goal while satisfying one or more performance criteria Possible Goals: โ€ข Output regulation (disturbance rejection, homeostasis) : keep ๐‘ฆ constant despite disturbance โ€ข Trajectory tracking : keep ๐‘ฆ ๐‘ก โ‰ˆ ๐‘ฆ ๐‘ก โˆ€ ๐‘ก Performance criteria: โ€ข Static performance (at steady state): e.g. lim ๐‘กโ†’โˆž ๐‘ฆ ๐‘ก โˆ’ ๐‘ฆ ๐‘ก โ€ข Dynamic performance: transient time, etcโ€ฆ โ€ข Stability: not blowing up! โ€ข Robustness: amount of disturbance that can be tolerated โ€ข Limited control effort ๐’™ = ๐‘“ ๐’™, ๐‘ข ๐‘ฆ = ๐‘” ๐’™, ๐‘ข ๐‘ข ๐‘ฆ PLANT ๐‘ข ๐‘ ++ ๐‘› ++ ๐‘‘ ๐‘ฆ ๐‘›๐‘ฆ ๐’› = โ„Ž ๐’›, ๐‘ฆ, ๐‘ฆ๐‘› ๐‘ข ๐‘ = ๐‘– ๐’›, ๐‘ฆ, ๐‘ฆ ๐‘› CONTROLLER 5Matteo Mischiati Control theory primer
  • 6. Control theory framework ๐‘ข = ๐‘–๐‘›๐‘๐‘ข๐‘ก (๐‘š๐‘œ๐‘ก๐‘œ๐‘Ÿ) ๐‘ฆ = ๐‘œ๐‘ข๐‘ก๐‘๐‘ข๐‘ก ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐’™ = ๐‘–๐‘›๐‘ก๐‘’๐‘Ÿ๐‘›๐‘Ž๐‘™ ๐‘ ๐‘ก๐‘Ž๐‘ก๐‘’ ๐‘‘ = ๐‘‘๐‘–๐‘ ๐‘ก๐‘ข๐‘Ÿ๐‘๐‘Ž๐‘›๐‘๐‘’ ๐‘› = ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐‘›๐‘œ๐‘–๐‘ ๐‘’ ANALYSIS problem: infer the functional structure of the controller, given the observed performance of the overall system in multiple tasks Goals: โ€ข Output regulation (disturbance rejection, homeostasis) : keep ๐‘ฆ constant despite disturbance โ€ข Trajectory tracking : keep ๐‘ฆ ๐‘ก โ‰ˆ ๐‘ฆ ๐‘ก โˆ€ ๐‘ก Performance criteria: โ€ข Static performance (at steady state): e.g. lim ๐‘กโ†’โˆž ๐‘ฆ ๐‘ก โˆ’ ๐‘ฆ ๐‘ก โ€ข Dynamic performance: transient time, etcโ€ฆ โ€ข Stability: not blowing up! โ€ข Robustness: amount of disturbance that can be tolerated โ€ข Limited control effort ๐’™ = ๐‘“ ๐’™, ๐‘ข ๐‘ฆ = ๐‘” ๐’™, ๐‘ข ++ ๐‘ข ๐‘› ๐‘ฆ PLANT ++ ๐‘‘ ๐‘ข ๐‘ ๐‘ฆ ๐‘›๐‘ฆ ๐’› = โ„Ž ๐’›, ๐‘ฆ, ๐‘ฆ๐‘› ๐‘ข ๐‘ = ๐‘– ๐’›, ๐‘ฆ, ๐‘ฆ ๐‘› CONTROLLER ? 6Matteo Mischiati Control theory primer
  • 7. Example of analysis problem: uncovering the dragonfly control system ๐’— ๐‘ซ๐‘ญ ๐’‰๐’†๐’‚๐’… ๐’“ โ€ข We want to precisely characterize the foraging behavior of the dragonfly (what it does) to gain insight on its neural circuitry (how it does it). ๐’— ๐‘ซ๐‘ญ ๐’‰๐’†๐’‚๐’… ๐’‡๐’๐’š ? ๐’“, relative to dragonfly, in ๐’‰๐’†๐’‚๐’… ref. frame dragonfly accel. head rotation dragonfly head, body & wing dynamics dragonfly visual system ๐‘๐„๐“๐ˆ๐๐€ ? ๐“๐’๐ƒ๐s ? ๐–๐ˆ๐๐† ๐Œ๐”๐’๐‚๐‹๐„๐’ ๐๐„๐‚๐Š ๐Œ๐”๐’๐‚๐‹๐„๐’ ๐’ƒ๐’๐’…๐’š ๐’ƒ๐’๐’…๐’š 7Matteo Mischiati Control theory primer
  • 8. Linear time-invariant systems ๐‘ข = ๐‘–๐‘›๐‘๐‘ข๐‘ก (๐‘š๐‘œ๐‘ก๐‘œ๐‘Ÿ) ๐‘ฆ = ๐‘œ๐‘ข๐‘ก๐‘๐‘ข๐‘ก ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐’™ = ๐‘–๐‘›๐‘ก๐‘’๐‘Ÿ๐‘›๐‘Ž๐‘™ ๐‘ ๐‘ก๐‘Ž๐‘ก๐‘’ ๐‘‘ = ๐‘‘๐‘–๐‘ ๐‘ก๐‘ข๐‘Ÿ๐‘๐‘Ž๐‘›๐‘๐‘’ ๐‘› = ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐‘›๐‘œ๐‘–๐‘ ๐‘’ ๐‘ƒ ๐‘  = ๐‘ โˆ’๐‘ง1 ๐‘ โˆ’๐‘ง2 โ€ฆ(๐‘ โˆ’๐‘ง ๐‘š) ๐‘ โˆ’๐‘1 ๐‘ โˆ’๐‘2 โ€ฆ(๐‘ โˆ’๐‘ ๐‘›) , ๐‘ง๐‘–โˆˆ โ„‚ = ๐‘ง๐‘’๐‘Ÿ๐‘œ๐‘ , ๐‘๐‘– โˆˆ โ„‚ = ๐‘๐‘œ๐‘™๐‘’๐‘  โ€ข Stability (of a system) โ‡” ๐‘1, โ€ฆ , ๐‘ ๐‘› have negative real part โ€ข Performance (of a system): depends on location of poles and zeros โ€ข Related to transfer function in frequency domain: ๐‘ข ๐‘ก = sin ๐œ”๐‘ก โ‡’ ๐‘ฆ ๐‘ก = ๐‘ƒ ๐‘—๐œ” โ‹… sin(๐œ”๐‘ก + ฮฆ ๐‘ƒ ๐‘—๐œ” ) ๐’™ = ๐ด๐’™ + ๐ต๐‘ข ๐‘ฆ = ๐ถ๐’™ + ๐ท๐‘ข ++ ๐‘ข ๐‘› ๐‘ฆ PLANT ++ ๐‘‘ ๐‘ข ๐‘ ๐‘ฆ ๐‘›๐‘ฆ ๐’› = ๐ป๐’› + ๐บ ๐‘ฆ ๐‘ข ๐‘ = ๐ผ๐’› + ๐ฟ ๐‘ฆ CONTROLLER ๐‘ƒ(๐‘ ) ++ ๐‘ˆ(๐‘ ) ๐‘(๐‘ ) ๐‘Œ(๐‘ ) PLANT ++ ๐ท(๐‘ ) ๐‘ˆ ๐ถ(๐‘ ) ๐‘Œ๐‘›(๐‘ )๐‘Œ(๐‘ ) ๐ถ(๐‘ ) CONTROLLER Laplace transform Y ๐‘  = ๐‘ƒ ๐‘  โˆ™ ๐‘ˆ(๐‘ ) 8Matteo Mischiati Control theory primer
  • 9. Linear time-invariant systems Laplace transform for signals: โ€ข Final value theorem : lim ๐‘กโ†’โˆž ๐‘ฆ ๐‘ก = lim ๐‘ โ†’0 ๐‘  โˆ™ ๐‘Œ ๐‘  (if limit exists) โ€ข If u ๐‘ก โŸท ๐‘ˆ ๐‘  , then u ๐‘ก โŸท ๐‘ˆ ๐‘  ๐‘  Typical reference/disturbance signals: - Step ๐‘Œ ๐‘  = ๐‘Ž 1 ๐‘  - Ramp ๐‘Œ ๐‘  = ๐‘Ž 1 ๐‘ 2 - Sinusoid ๐‘Œ ๐‘  = ๐œ” ๐‘ 2+๐œ”2 ๐‘ƒ(๐‘ ) ++ ๐‘ˆ(๐‘ ) ๐‘(๐‘ ) ๐‘Œ(๐‘ ) PLANT ++ ๐ท(๐‘ ) ๐‘ˆ ๐ถ(๐‘ ) ๐‘Œ๐‘›(๐‘ )๐‘Œ(๐‘ ) ๐ถ(๐‘ ) CONTROLLER Y ๐‘  = ๐‘ƒ ๐‘  โˆ™ ๐‘ˆ(๐‘ ) ๐‘ก ๐‘ฆ(๐‘ก) ๐‘Ž ๐‘ก ๐‘ฆ(๐‘ก) ๐‘Ž๐‘ก ๐‘ก ๐‘ฆ(๐‘ก) sin(๐œ”๐‘ก) 9Matteo Mischiati Control theory primer
  • 10. โ€ข Control theory framework - linear time-invariant (LTI) case โ€ข Properties of feedback - internal model principle โ€ข Common control schemes - forward and inverse models, Smith predictor - state feedback, observers, optimal control โ€ข A model of human response in manual tracking tasks (Kleinman et al. 1970, Gawthrop et al. 2011) 10Matteo Mischiati Control theory primer
  • 11. Feedforward (inverse model) Stability Depends on poles (and zeros!) of ๐‘ƒ(๐‘ ) Performance (static and dynamic) Arbitrarily good if ๐‘ƒ ๐‘  โ‰ˆ ๐‘ƒ ๐‘  and its inverse exists and is stable: Y ๐‘  โ‰ˆ ๐‘ƒ ๐‘  โˆ™ ๐‘ƒโˆ’1 ๐‘  โ‹… ๐‘Œ(๐‘ ) โ‰ˆ ๐‘Œ(๐‘ ) Robustness to disturbance (disturbance rejection) None : Y ๐‘  = ๐‘ƒ ๐‘  โˆ™ ( ๐‘ƒโˆ’1 ๐‘  โ‹… ๐‘Œ ๐‘  + ๐ท(๐‘ )) โ‰ˆ ๐‘Œ(๐‘ ) + ๐‘ƒ(๐‘ ) โ‹… ๐ท(๐‘ ) ๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ ) PLANT ++ ๐ท(๐‘ ) ๐‘ˆ ๐ถ(๐‘ )๐‘Œ(๐‘ ) ๐‘ƒโˆ’1 (๐‘ ) CONTROLLER Y ๐‘  = ๐‘ƒ ๐‘  โˆ™ ๐‘ˆ(๐‘ )Uc ๐‘  = ๐‘ƒโˆ’1 ๐‘  โˆ™ ๐‘Œ(๐‘ ) 11Matteo Mischiati Control theory primer
  • 12. Properties of Feedback ๐‘Œ ๐‘  = ๐‘ƒ ๐‘  ๐ท ๐‘  + ๐ถ ๐‘  ๐ธ ๐‘  = ๐‘ƒ ๐‘  ๐ท ๐‘  + ๐ถ ๐‘  ( ๐‘Œ ๐‘  โˆ’ ๐‘Œ ๐‘  ) โ‡’ ๐‘Œ ๐‘  = ๐‘ƒ ๐‘  ๐ถ ๐‘  1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† ๐‘Œ ๐‘  + ๐‘ƒ ๐‘  1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† ๐ท(๐‘ ) Stability Depends on 1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† . Can potentially stabilize unstable plants. Disturbance rejection Depends on 1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† . Can potentially attenuate/cancel effect of ๐ท ๐‘  . ๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ ) PLANT ++ ๐ท(๐‘ ) ๐‘ˆ ๐ถ(๐‘ )๐ธ(๐‘ ) ๐ถ(๐‘ ) CONTROLLER ๐‘Œ(๐‘ ) Y ๐‘  = ๐‘ƒ ๐‘  โˆ™ ๐‘ˆ(๐‘ )Uc ๐‘  = ๐ถ ๐‘  โˆ™ ๐ธ(๐‘ ) + - 12Matteo Mischiati Control theory primer
  • 13. Properties of Feedback Static performance lim ๐‘กโ†’โˆž ๐‘’ ๐‘ก = lim ๐‘ โ†’0 ๐‘  โˆ™ ๐ธ ๐‘  , ๐ธ ๐‘  = 1 1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† ๐‘Œ ๐‘  โˆ’ ๐‘ƒ ๐‘  1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† ๐ท(๐‘ ) Letโ€™s see how different controllers perform: ๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ ) PLANT ++ ๐ท(๐‘ ) ๐‘ˆ ๐ถ(๐‘ )๐ธ(๐‘ ) ๐ถ(๐‘ ) CONTROLLER ๐‘Œ(๐‘ ) + - ๐‘’. ๐‘”. 1 1 + ๐œ๐‘  ๐‘˜ 13Matteo Mischiati Control theory primer
  • 14. Properties of Feedback Static performance lim ๐‘กโ†’โˆž ๐‘’ ๐‘ก = lim ๐‘ โ†’0 ๐‘  โˆ™ ๐ธ ๐‘  , ๐ธ ๐‘  = 1 1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† ๐‘Œ ๐‘  โˆ’ ๐‘ƒ ๐‘  1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† ๐ท(๐‘ ) Proportional controller: ๐ถ ๐‘  = ๐‘˜ lim ๐‘กโ†’โˆž ๐‘’ ๐‘ก = lim ๐‘ โ†’0 ๐‘  โˆ™ 1 1+๐‘˜ ๐‘ƒ ๐‘  ๐‘Ž ๐‘  = ๐‘Ž 1+๐‘˜ โ‰  0 errors in tracking a step (but small if ๐‘˜ is large) lim ๐‘กโ†’โˆž ๐‘’ ๐‘ก = lim ๐‘ โ†’0 ๐‘  โˆ™ 1 1+๐‘˜ ๐‘ƒ ๐‘  ๐‘Ž ๐‘ 2 โ†’ โˆž cannot track a ramp at all ๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ ) PLANT ++ ๐ท(๐‘ ) ๐‘ˆ ๐ถ(๐‘ )๐ธ(๐‘ ) ๐ถ(๐‘ ) CONTROLLER ๐‘Œ(๐‘ ) + - ๐‘Ž ๐‘  ๐‘Ž ๐‘ 2 ๐‘’. ๐‘”. 1 1 + ๐œ๐‘  ๐‘˜ 14Matteo Mischiati Control theory primer
  • 15. Properties of Feedback Static performance lim ๐‘กโ†’โˆž ๐‘’ ๐‘ก = lim ๐‘ โ†’0 ๐‘  โˆ™ ๐ธ ๐‘  , ๐ธ ๐‘  = 1 1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† ๐‘Œ ๐‘  โˆ’ ๐‘ƒ ๐‘  1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† ๐ท(๐‘ ) Proportional+Integral (PI) controller: ๐ถ ๐‘  = ๐‘˜ ๐‘ + ๐‘˜ ๐‘– ๐‘  = ๐‘˜ ๐‘ ๐‘ +๐‘˜ ๐‘– ๐‘  lim ๐‘กโ†’โˆž ๐‘’ ๐‘ก = lim ๐‘ โ†’0 ๐‘  โˆ™ 1 1+ ๐‘˜ ๐‘ ๐‘ +๐‘˜ ๐‘– ๐‘  ๐‘ƒ ๐‘  ๐‘Ž ๐‘  = 0 perfect in tracking a step lim ๐‘กโ†’โˆž ๐‘’ ๐‘ก = lim ๐‘ โ†’0 ๐‘  โˆ™ 1 1+ ๐‘˜ ๐‘ ๐‘ +๐‘˜ ๐‘– ๐‘  ๐‘ƒ ๐‘  ๐‘Ž ๐‘ 2 = ๐‘Ž ๐‘˜ ๐‘– โ‰  0 errors in tracking a ramp ๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ ) PLANT ++ ๐ท(๐‘ ) ๐‘ˆ ๐ถ(๐‘ )๐ธ(๐‘ ) ๐ถ(๐‘ ) CONTROLLER ๐‘Œ(๐‘ ) + - ๐‘Ž ๐‘  ๐‘Ž ๐‘ 2 1 1 + ๐œ๐‘  ๐‘˜ ๐‘ + ๐‘˜๐‘– ๐‘  15Matteo Mischiati Control theory primer (but small if ๐‘˜๐‘– is large)
  • 16. Properties of Feedback Static performance lim ๐‘กโ†’โˆž ๐‘’ ๐‘ก = lim ๐‘ โ†’0 ๐‘  โˆ™ ๐ธ ๐‘  , ๐ธ ๐‘  = 1 1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† ๐‘Œ ๐‘  โˆ’ ๐‘ƒ ๐‘  1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† ๐ท(๐‘ ) To perfectly track: We need: Step Ramp How about sinusoid? We need: ๐‘Ž ๐‘  ๐‘Ž ๐‘ 2 ๐œ” ๐‘ 2 + ๐œ”2 ๐‘ƒ(๐‘ )๐ถ ๐‘  = 1 ๐‘  โ‹… (๐‘ƒ ๐‘  ๐ถ ๐‘  )โ€ฒ ๐‘ƒ(๐‘ )๐ถ ๐‘  = 1 ๐‘ 2 โ‹… (๐‘ƒ ๐‘  ๐ถ ๐‘  )โ€ฒ ๐‘ฆ ๐‘ก = sin ๐œ”๐‘ก โ‡“ ๐‘’ ๐‘ก = ๐บ ๐‘—๐œ” โ‹… sin(๐œ”๐‘ก + ฮฆ ๐บ ๐‘—๐œ” ) ๐บ ๐‘  ๐‘ƒ(๐‘ )๐ถ ๐‘  = 1 ๐‘ 2 + ๐œ”2 โ‹… (๐‘ƒ ๐‘  ๐ถ ๐‘  )โ€ฒ โ‡“ ๐บ ๐‘—๐œ” = 1 + ๐‘ƒ ๐‘—๐œ” ๐ถ ๐‘—๐œ” โˆ’1 โ†’ 0 16Matteo Mischiati Control theory primer
  • 17. Internal model principle To perfectly track: We need: Step Ramp Sinusoid Internal model principle: To achieve asymptotical tracking of a reference signal (rejection of a disturbance signal) via feedback, the controller (or the plant) must contain an โ€œinternal modelโ€ of the signal. It is a necessary condition, not a sufficient condition (need also stability) ๐‘Ž ๐‘  ๐‘Ž ๐‘ 2 ๐œ” ๐‘ 2 + ๐œ”2 ๐‘ƒ(๐‘ )๐ถ ๐‘  = 1 ๐‘  โ‹… (๐‘ƒ ๐‘  ๐ถ ๐‘  )โ€ฒ ๐‘ƒ(๐‘ )๐ถ ๐‘  = 1 ๐‘ 2 โ‹… (๐‘ƒ ๐‘  ๐ถ ๐‘  )โ€ฒ ๐‘ƒ(๐‘ )๐ถ ๐‘  = 1 ๐‘ 2 + ๐œ”2 โ‹… (๐‘ƒ ๐‘  ๐ถ ๐‘  )โ€ฒ 17Matteo Mischiati Control theory primer
  • 18. Feedback vs. Feedforward Feedback โ€ข is needed if plant is unstable or for disturbance rejection โ€ข does not require full knowledge of the plant โ€ข incorporating the knowledge of possible reference and disturbance signals is very useful (internal model principle) Feedforward โ€ข if plant is known, and no disturbance, its performance canโ€™t be beat โ€ข no sensory delays ๐‘Œ ๐‘  = ๐‘ƒ ๐‘  ๐ถ ๐‘  1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† ๐‘Œ ๐‘  + ๐‘ƒ ๐‘  1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† ๐ท(๐‘ ) ๐‘Œ ๐‘  = ๐‘ƒ ๐‘  โˆ™ ( ๐‘ƒโˆ’1 ๐‘  โ‹… ๐‘Œ ๐‘  + ๐ท(๐‘ )) โ‰ˆ ๐‘Œ(๐‘ ) + ๐‘ƒ(๐‘ ) โ‹… ๐ท(๐‘ ) 18Matteo Mischiati Control theory primer
  • 19. โ€ข Control theory framework - linear time-invariant (LTI) case โ€ข Properties of feedback - internal model principle โ€ข Common control schemes - forward and inverse models, Smith predictor - state feedback, observers, optimal control โ€ข A model of human response in manual tracking tasks (Kleinman et al. 1970, Gawthrop et al. 2011) 19Matteo Mischiati Control theory primer
  • 20. Feedback + Feedforward The feedback controller kicks in only if inverse model is incorrect. ๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ ) PLANT ++ ๐‘ˆ ๐น๐ต(๐‘ ) ๐ธ(๐‘ ) ๐ถ(๐‘ )๐‘Œ(๐‘ ) + - ๐‘ƒโˆ’1 (๐‘ ) INVERSE MODEL FEEDBACK ๐‘ˆ ๐น๐น(๐‘ ) 20Matteo Mischiati Control theory primer
  • 21. Feedback + Feedforward The feedback controller kicks in only if inverse model is incorrect. The corrective command from the feedback path can be also used as a learning/adaptation signal by the inverse model. ๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ ) PLANT ++ ๐‘ˆ ๐น๐ต(๐‘ ) ๐ธ(๐‘ ) ๐ถ(๐‘ )๐‘Œ(๐‘ ) + - ๐‘ƒโˆ’1 (๐‘ ) INVERSE MODEL FEEDBACK ๐‘ˆ ๐น๐น(๐‘ ) 21Matteo Mischiati Control theory primer
  • 22. Feedback + Feedforward The feedback controller kicks in only if inverse model is incorrect. The corrective command from the feedback path can be also used as a learning/adaptation signal by the inverse model. Significant sensory delays are still a problem. ๐‘ƒ(๐‘ )๐‘ˆ(๐‘ ) ๐‘Œ(๐‘ ) PLANT ++ ๐‘ˆ ๐น๐ต(๐‘ ) ๐ธ(๐‘ ) ๐ถ(๐‘ )๐‘Œ(๐‘ ) + - ๐‘ƒโˆ’1 (๐‘ ) INVERSE MODEL FEEDBACK ๐‘ˆ ๐น๐น(๐‘ ) ๐‘’โˆ’๐‘ ๐œ 22Matteo Mischiati Control theory primer
  • 23. Forward model The control signal is sent through a model of the plant (โ€œforward modelโ€) to predict the sensory output. ๐‘ƒ(๐‘ ) ๐‘Œ(๐‘ ) PLANT ๐‘ˆ ๐ถ(๐‘ )๐ธ(๐‘ ) ๐ถ(๐‘ )๐‘Œ(๐‘ ) + - ๐‘ƒ(๐‘ ) FORWARD MODEL CONTROLLER predicted sensory output 23Matteo Mischiati Control theory primer
  • 24. Forward model The control signal is sent through a model of the plant (โ€œforward modelโ€) to predict the sensory output. The (delayed) sensory output can be used as a learning/adaptation signal for the forward model. Direct use of the delayed sensory output in the control is problematic because of time mismatch. ๐‘ƒ(๐‘ ) ๐‘Œ(๐‘ ) PLANT ๐‘ˆ ๐ถ(๐‘ )๐ธ(๐‘ ) ๐ถ(๐‘ )๐‘Œ(๐‘ ) + - ๐‘ƒ(๐‘ ) FORWARD MODEL ๐‘’โˆ’๐‘ ๐œ CONTROLLER predicted sensory output 24Matteo Mischiati Control theory primer
  • 25. Smith predictor Assuming ๐‘ƒ ๐‘  โ‰ˆ ๐‘ƒ ๐‘  and ๐œ โ‰ˆ ๐œ: ๐‘Œ ๐‘  = ๐‘ƒ ๐‘  ๐‘ˆ๐‘ ๐‘  = ๐‘ƒ ๐‘  ๐ถ ๐‘  1 + ๐‘ƒ ๐‘  ๐ถ ๐‘† ๐‘Œ ๐‘  Delay has been moved outside the control loop. PLANT ๐‘ƒ(๐‘ ) ๐‘Œ(๐‘ )๐‘ˆ ๐ถ(๐‘ )๐ธ(๐‘ ) ๐ถ(๐‘ )๐‘Œ(๐‘ ) + - - ๐‘ƒ(๐‘ ) ๐‘’โˆ’๐‘ ๐œ ๐‘’โˆ’๐‘  ๐œ + - delay model plant model predicted sensory output error in sensory output prediction CONTROLLER 25Matteo Mischiati Control theory primer
  • 26. Models of the cerebellum 1. Cerebellum as an inverse model in a feedback+feedforward motor control scheme Wolpert, Miall & Kawato, 1998 โ€œInternal models in the cerebellumโ€ Not in the sense of my presentation ! 26Matteo Mischiati Control theory primer
  • 27. Models of the cerebellum 2. Cerebellum as a forward model in a Smith predictor control scheme Wolpert, Miall & Kawato, 1998 โ€œInternal models in the cerebellumโ€ 27Matteo Mischiati Control theory primer
  • 28. State feedback ๐’™ = ๐‘“ ๐’™, ๐‘ข ๐‘ฆ = ๐‘” ๐’™, ๐‘ข ๐‘ฆ PLANT ๐‘ข๐‘ฆ ๐’› = โ„Ž ๐’›, ๐‘ฆ, ๐’™ ๐‘ข = ๐‘– ๐’›, ๐‘ฆ, ๐’™ CONTROLLER ๐’™ ๐‘ฆ PLANT ๐‘ข๐‘ฆ CONTROLLER ๐’™ ๐’™ = ๐ด๐’™ + ๐ต๐‘ข ๐‘ฆ = ๐ถ๐’™ + ๐ท๐‘ข ๐’› = ๐ป๐’› + ๐บ ๐‘ฆ + ๐‘€๐’™ ๐‘ข = ๐ผ๐’› + ๐ฟ ๐‘ฆ + ๐‘๐’™ Linear time-invariant case: 28Matteo Mischiati Control theory primer
  • 29. State feedback If the plant is reachable, it is possible to achieve any arbitrary choice of closed-loop poles with an appropriate linear and memoryless controller: u = โˆ’๐พ๐’™ + ๐พ๐‘Ÿ ๐‘ฆ ๐’™ = ๐‘“ ๐’™, ๐‘ข ๐‘ฆ = ๐‘” ๐’™, ๐‘ข ๐‘ฆ PLANT ๐‘ข๐‘ฆ ๐’› = โ„Ž ๐’›, ๐‘ฆ, ๐’™ ๐‘ข = ๐‘– ๐’›, ๐‘ฆ, ๐’™ CONTROLLER ๐’™ ๐‘ฆ PLANT ๐‘ข๐‘ฆ CONTROLLER ๐’™ ๐’™ = ๐ด๐’™ + ๐ต๐‘ข ๐‘ฆ = ๐ถ๐’™ + ๐ท๐‘ข Linear time-invariant case: ๐พ ๐พ๐‘Ÿ + - 29Matteo Mischiati Control theory primer
  • 30. Observers Observer: dynamical system designed to estimate the full state (when not fully available) If the plant is observable, it is possible to achieve lim ๐‘กโ†’โˆž ๐’™ ๐‘ก = ๐’™ (with right ๐ฟ) ๐‘ฆ PLANT ๐‘ข ๐’™ ๐’™ = ๐ด๐’™ + ๐ต๐‘ข ๐‘ฆ = ๐ถ๐’™ ๐’™ = ๐ด ๐’™ + ๐ต๐‘ข + ๐ฟ(y โˆ’ C ๐’™) OBSERVER 30Matteo Mischiati Control theory primer
  • 31. Observers Observer: dynamical system designed to estimate the full state (when not fully available) If the plant is observable, it is possible to achieve lim ๐‘กโ†’โˆž ๐’™ ๐‘ก = ๐’™ (with right ๐ฟ) Separation principle: if the plant is reachable & observable, can replace ๐’™ with ๐’™ and design ๐พ independently of ๐ฟ (use observed state just as real one) ๐‘ฆ PLANT ๐‘ข ๐’™ ๐’™ = ๐ด๐’™ + ๐ต๐‘ข ๐‘ฆ = ๐ถ๐’™ ๐’™ = ๐ด ๐’™ + ๐ต๐‘ข + ๐ฟ(y โˆ’ C ๐’™) ๐‘ฆ ๐พ ๐พ๐‘Ÿ +- OBSERVER 31Matteo Mischiati Control theory primer
  • 32. Optimal control Linear Quadratic Gaussian (LQG) optimal output regulation: linear plant, additive Gaussian white noise on state (with covariance ฮฃ ๐‘‘) and output (๐œŽ ๐‘›); minimize quadratic cost : ๐”ผ lim ๐‘‡โ†’โˆž 1 ๐‘‡ 0 ๐‘‡ ๐’™ ๐‘‡ ๐‘ก ๐‘„๐’™ ๐‘ก + ๐‘ข ๐‘ก 2 ๐‘‘๐‘ก Solution is linear observer (Kalman filter) with linear memoryless controller: ๐ฟ = ๐‘ƒ๐ถ ๐‘‡ ๐œŽ ๐‘› โˆ’1, ๐ด๐‘ƒ + ๐‘ƒ๐ด ๐‘‡ + ฮฃ ๐‘‘ โˆ’ ๐‘ƒ๐ถ ๐‘‡ ๐œŽ ๐‘› โˆ’1 ๐ถ๐‘ƒ ๐‘‡ = 0 ๐พ = ๐ต ๐‘‡ ๐‘†, ๐ด ๐‘‡ ๐‘† + ๐‘†๐ด + ๐‘„ โˆ’ ๐‘†๐ต๐ต ๐‘‡ ๐‘† = 0 ๐‘ฆ PLANT ๐‘ข ๐’™ = ๐ด๐’™ + ๐ต๐‘ข + ๐’… ๐‘ฆ = ๐ถ๐’™ ๐’™ ๐’™ = ๐ด ๐’™ + ๐ต๐‘ข + ๐ฟ(yn โˆ’ C ๐’™) ๐‘ฆ = 0 ๐พ +- OBSERVER (KALMAN FILTER) ๐’… ++ ๐‘› ๐‘ฆ ๐‘› 32Matteo Mischiati Control theory primer
  • 33. Internal model principle Internal model principle (state space): to achieve asymptotical tracking of a reference signal (rejection of a disturbance signal) produced by an exosystem, the controller must contain an โ€œinternal modelโ€ of the exosystem. (Francis & Wonham, Automatica, 1970) It is a necessary condition, not a sufficient condition (need also stability). General principle with extensions to nonlinear systems. ๐‘ฆ PLANT ๐’–๐‘ฆ ๐’™ = ๐ด๐’™ + ๐ต๐‘ข ๐‘ฆ = ๐ถ๐’™ ๐œผ = ๐‘†๐œผ + ๐บ๐‘’๐’› = ๐‘† ๐’› ๐‘ฆ = ๐‘‡ ๐’› + - ๐‘’ STABILIZING CONTROLLER ๐œผINT.MODELEXOSYSTEM 33Matteo Mischiati Control theory primer
  • 34. โ€ข Control theory framework - linear time-invariant (LTI) case โ€ข Properties of feedback - internal model principle โ€ข Common control schemes - forward and inverse models, Smith predictor - state feedback, observers, optimal control โ€ข A model of human response in manual tracking tasks (Kleinman et al. 1970, Gawthrop et al. 2011) 34Matteo Mischiati Control theory primer
  • 35. Human response model D. L. Kleinman, S. Baron and W. H. Levison, โ€œAn Optimal Control Model of Human Response. Part I: Theory and Validationโ€, Automatica, 1970 (revisited, more recently, by Gawthrop et al. 2011) Task: by controlling a joystick (position, velocity or acceleration control), subject is asked to keep a cursor on the screen as close as possible to a target location, while unknown disturbances are applied by the computer. Plant: ๐‘š๐‘œ๐‘›๐‘–๐‘ก๐‘œ๐‘Ÿ ๐‘Œ(๐‘ ) ++ ๐ท(๐‘ ) (computer) ๐‘ˆ(๐‘ ) ๐‘—๐‘œ๐‘ฆ๐‘ ๐‘ก๐‘–๐‘๐‘˜ ๐‘ƒ๐ฝ๐‘€ ๐‘  โˆˆ ๐‘˜, ๐‘˜ ๐‘  , ๐‘˜ ๐‘ 2 ๐‘ ๐‘’๐‘›๐‘ ๐‘œ๐‘Ÿ๐‘ฆ ๐‘“๐‘’๐‘’๐‘‘๐‘๐‘Ž๐‘๐‘˜ โ€œHuman controllerโ€: ๐‘›๐‘’๐‘ข๐‘Ÿ๐‘œ๐‘š๐‘œ๐‘ก๐‘œ๐‘Ÿ dynamics ๐‘ˆ(๐‘ ) ++ ๐‘š๐‘œ๐‘ก๐‘œ๐‘Ÿ ๐‘›๐‘œ๐‘–๐‘ ๐‘’ ๐‘›๐‘’๐‘ข๐‘Ÿ๐‘Ž๐‘™ computation ๐‘ƒ ๐‘ ๐‘ ๐ถ ๐‘  ๐‘’โˆ’๐‘ ๐œ ๐‘ƒ ๐‘ ๐‘  โ‰ˆ 1 ๐œ ๐‘› ๐‘ +1 ๐œ ๐‘› โ‰ˆ 100๐‘š๐‘  ๐œ โ‰ˆ 150 โˆ’ 250๐‘š๐‘  35Matteo Mischiati Control theory primer
  • 36. Human response model Task: Output regulation/disturbance rejection with linear time-invariant plant and significant delay on any potential feedback line ๐‘’โˆ’๐‘ ๐œ ๐‘ƒ(๐‘ ) ++ ๐‘ˆ(๐‘ ) ๐‘(๐‘ ) ๐‘Œ(๐‘ ) ++ ๐ท(๐‘ ) ๐‘ˆ ๐ถ(๐‘ ) ๐‘Œ๐‘›(๐‘ )๐ถ(๐‘ ) CONTROLLER ๐‘ƒ๐ฝ๐‘€ ๐‘  โˆ™ ๐‘ƒ ๐‘(๐‘ ) ANALYSIS problem: infer a model of the neural controller ๐ถ(๐‘ ) from the observed performance of the subjects tested. 36Matteo Mischiati Control theory primer
  • 37. Human response model Task: Output regulation/disturbance rejection with linear time-invariant plant and significant delay on any potential feedback line ๐‘’โˆ’๐‘ ๐œ ๐‘ƒ(๐‘ ) ++ ๐‘ˆ(๐‘ ) ๐‘(๐‘ ) ๐‘Œ(๐‘ ) ++ ๐ท(๐‘ ) ๐‘ˆ ๐ถ(๐‘ ) ๐‘Œ๐‘›(๐‘ )๐ถ(๐‘ ) CONTROLLER ๐‘ƒ๐ฝ๐‘€ ๐‘  โˆ™ ๐‘ƒ ๐‘(๐‘ ) ANALYSIS problem: infer a model of the neural controller ๐ถ(๐‘ ) from the observed performance of the subjects tested. So what are the performances? โ€ข Very good and robust to disturbances up to 2Hz (sum of sinusoids), for all three types of joystick dynamics โ€ข Apparently delay-free Must be some kind of FEEDBACK + FORWARD model ! 37Matteo Mischiati Control theory primer
  • 38. Human response model Hypothesis: optimal control to minimize average error & control effort ๐”ผ lim ๐‘‡โ†’โˆž 1 ๐‘‡ 0 ๐‘‡ ๐’™ ๐‘ก 2 + ๐›ผ ๐‘ข ๐‘ก 2 + ๐›ฝ ๐‘ข ๐‘ก 2 ๐‘‘๐‘ก Theoretical solution * (with assumptions similar to LQG problem): - Optimal observer (Kalman filter) to estimate delayed state (as in LQG) - Optimal least mean-squared predictor to predict current state - Optimal linear memoryless controller (as in LQG) ๐‘ฆ PLANT ๐‘ข ๐’™(๐‘ก โˆ’ ๐œ) ๐’™ = ๐ด๐’™ + ๐ต๐‘ข + ๐’… ๐‘ฆ = ๐ถ๐’™ KALMAN FILTER ๐‘ฆ = 0 ๐พ +- ๐’… ++ ๐‘› ๐‘ฆ ๐‘› * D. Kleinman, โ€œOptimal control of linear systems with time-delay and observation noiseโ€, IEEE Trans. Autom. Control, 1969 ๐‘’โˆ’๐‘ ๐œPREDICTOR ๐’™(๐‘ก) ๐‘ฆ๐‘›(๐‘ก โˆ’ ๐œ) 38Matteo Mischiati Control theory primer
  • 39. Controller freq. response with plant ๐‘˜ ๐‘  Controller freq. response with plant ๐‘˜ ๐‘ 2 Matteo Mischiati Control theory primer 39
  • 40. Human response model Gawthrop et al. * (2011): - Introduced, in both estimator and predictor, a copy of the exosystem generating sinusoidal disturbances (internal model principle!) - Show that intermittent control is also compatible with results * P. Gawthrop et al., โ€œIntermittent control: a computational theory of human controlโ€, Biol. Cybern., 2011 Actual response to sinusoid Response without int.model 40Matteo Mischiati Control theory primer
  • 41. โ€ข Crash course in control theory (for LTI systems) - many concepts can be extended to more general settings โ€ข An example of control-theoretic approach to modeling sensorimotor loops - need to iterate between modeling/experiments to discern among alternatives and improve understanding of the system Conclusions THANK YOU FOR YOUR ATTENTION ! 41Matteo Mischiati Control theory primer

Editor's Notes

  1. Fast feedback loop involving the forward model can be seen as playing the role of an inverse model.
  2. Fast feedback loop involving the forward model can be seen as playing the role of an inverse model.
  3. Here we consider an output regulation problem.
  4. Example: S=[0 โ€“omega; omega 0] produces oscillatory signals.