Tuning a multi-level control system

CSGnet Archive CSG2014
1

[From Bruce Abbott (2014.03.19.1150 EDT)]

Adam Matic – previously –

I can’t locate where you wrote it, but I recall your mentioning having trouble tuning a multi-level control system. It occurs to me that the problem may have to do with the relative speeds at which the different levels do their controlling. To work properly, the time required for a lower-level system to bring its controlled variable to reference must be shorter than the time required by a higher-level system – one setting the references of the lower-level systems – to do the same. This can be arranged via a slowing factor in a leaky-integrator output or by setting up longer loop delays in the higher-level system. In Bill’s Arm 2 demo, this is accomplished by executing the “evaluate” procedure that ties the physical properties of the arm to the muscle-level control systems four times in succession each time through the main iteration loop, where the higher-level system procedures are executed only once per iteration. (In my ArmDynamic test-bed program I commented out the extra calls to the “evaluate” procedure, because the demo lacks any higher-level control systems.)

If I recall correctly, Bill Powers said that the lower-level systems should be tuned first, then each level up in succession. However, it may be necessary to then go back and retune if, for example, the delays around the loop are too long in a lower-level system to work well with the system that sets its reference. Thus, tuning may require an iterative procedure.

Control engineers have developed mathematical procedures for finding optimal tuning; I don’t know whether these procedures work well with hierarchical control. In any case, I have no knowledge how to employ those procedures. For our demos, Bill and I worked by trial and error, e. coli fashion.

Bruce

2

[Adam Matic 2014.03.19 1700]

I’ve had some success in tuning reach, elevation and other coordination systems by slowing (filtering) the output, and in some cases the input too.

Adam

···

On Wed, Mar 19, 2014 at 4:50 PM, Bruce Abbott bbabbott@frontier.com wrote:

[From Bruce Abbott (2014.03.19.1150 EDT)]

Adam Matic – previously –

I can’t locate where you wrote it, but I recall your mentioning having trouble tuning a multi-level control system. It occurs to me that the problem may have to do with the relative speeds at which the different levels do their controlling. To work properly, the time required for a lower-level system to bring its controlled variable to reference must be shorter than the time required by a higher-level system – one setting the references of the lower-level systems – to do the same. This can be arranged via a slowing factor in a leaky-integrator output or by setting up longer loop delays in the higher-level system. In Bill’s Arm 2 demo, this is accomplished by executing the “evaluate” procedure that ties the physical properties of the arm to the muscle-level control systems four times in succession each time through the main iteration loop, where the higher-level system procedures are executed only once per iteration. (In my ArmDynamic test-bed program I commented out the extra calls to the “evaluate” procedure, because the demo lacks any higher-level control systems.)

If I recall correctly, Bill Powers said that the lower-level systems should be tuned first, then each level up in succession. However, it may be necessary to then go back and retune if, for example, the delays around the loop are too long in a lower-level system to work well with the system that sets its reference. Thus, tuning may require an iterative procedure.

Control engineers have developed mathematical procedures for finding optimal tuning; I don’t know whether these procedures work well with hierarchical control. In any case, I have no knowledge how to employ those procedures. For our demos, Bill and I worked by trial and error, e. coli fashion.

Bruce

3

[From Richard Kennaway (2014.03.19 1606 GMT)]

[From Bruce Abbott (2014.03.19.1150 EDT)]
If I recall correctly, Bill Powers said that the lower-level systems should be tuned first, then each level up in succession. �However, it may be necessary to then go back and retune if, for example, the delays around the loop are too long in a lower-level system to work well with the system that sets its reference. Thus, tuning may require an iterative procedure.

Control engineers have developed mathematical procedures for finding optimal tuning; I don�t know whether these procedures work well with hierarchical control. In any case, I have no knowledge how to employ those procedures. For our demos, Bill and I worked by trial and error, e. coli fashion.

For what it's worth, for my version of the inverted pendulum I did the mathematics, calculating the fourth-degree polynomial for the transfer function and choosing the gains of the four proportional controllers to put all the roots at the same place on the negative real axis. It's not sensitive to the exact values, but as a rule of thumb, the higher controllers generally have to be slower (i.e. have smaller gain) than the lower ones.

···

--
Richard Kennaway, R.Kennaway@uea.ac.uk, Richard Kennaway
School of Computing Sciences,
University of East Anglia, Norwich NR4 7TJ, U.K.

4

[From Bruce Abbott (2014.03.19.1225 EDT)]

Richard Kennaway (2014.03.19 1606 GMT)

[From Bruce Abbott (2014.03.19.1150 EDT)]

If I recall correctly, Bill
Powers said that the lower-level systems should be tuned first, then
each level up in succession. �However, it may be necessary to then go
back and retune if, for example, the delays around the loop are too
long in a lower-level system to work well with the system that sets its
reference. Thus, tuning may require an iterative procedure.

Control engineers have developed mathematical procedures for finding
optimal tuning; I don�t know whether these procedures work well with
hierarchical control. In any case, I have no knowledge how to employ
those procedures. For our demos, Bill and I worked by trial and error,
e. coli fashion.

RK: For what it's worth, for my version of the inverted pendulum I did the
mathematics, calculating the fourth-degree polynomial for the transfer
function and choosing the gains of the four proportional controllers to put
all the roots at the same place on the negative real axis. It's not
sensitive to the exact values, but as a rule of thumb, the higher
controllers generally have to be slower (i.e. have smaller gain) than the
lower ones.

Thanks, Richard; much appreciated! I just wish I had the mathematical
background to do that sort of thing.

Bruce

5

[From Adam Matic 2014.03.17 2140 cet]

Attached are some pictures of the robot arm I’m making and trying to tune. It is similar to the arm from Coordination Demo, but with less degrees of freedom.

Level one and two are angular velocity and angular position control for each joint, so each joint is like a servo motor system. There are five joints - rotation of the base, shoulder pitch, elbow pitch, wrist roll and wrist pitch. I’ve made these fairly low gain, but very small dt, around 1 ms, and large resolution of input angular position measurements - 13 bits.

IMAG0069-640x640.jpg

IMAG0070-640x640.jpg

IMAG0071-640x640.jpg

IMAG0072-640x640.jpg

···

Level three are reach, elevation, hand pitch, hand roll and lateral displacement systems. Perceptual signals for these are constructed like in Coordination Demo, using basic geometry formulas from arm segment lengths and joint angles. However, they are noisy, so they and their output are filtered.

I think they now work fairly well, considering how they worked before. There may still be room for improvement, but I’ll see about that as I go. At the moment, I’m working on adding touch/pressure sensing to the hand. There are three metal ‘nails’ on the hand that push a slider potentiometer up when pressed, and will be used to construct pressure perceptions. These pressure perceptions will be used to keep the hand holding a computer mouse. So, that will be a level above coordination of joints. I have no idea how I’ll tune these, but as Bill said, if it doesn’t work, just keep reorganizing.

When touch is finished, I plan to add a very simple visual system so I could do a tracking task. A camera will be pointed to the computer screen where I will have a target circle and a cursor circle, so that the Little Man Robot will control the distance between them.

Arm segments are cut from foam plastics and connected with screws, superglue and hot glue. Each joint has one or two micro geared DC motors and a potentiometer to measure joint angle. I’m using a Teensy 3.1 electronic board which is programmed like an Arduino, in C.

I have no idea when it will be finished. Soon, I hope. I’ll do some test and write a paper and this will be my M.A. in psychology in University of Zadar, Croatia. I have a B.A in psychology for which I did a standard tracking task in two dimensions. Before psychology I spent some time studying electrical engineering and computer science, but I wasn’t any good at electrical engineering so I dropped out and worked for I while, then decided to study psychology.

Adam

6

Hi Adam, this is really exciting to see! It will be very interesting to see some data and/or videos to show the control performance, and to see the specification in detail when it is complete. You must have such perseverence and focus to produce this and I wish you all the best in your continued work on it! Keep us all posted!
By the way, are you planning to equip it with the capacity for reorganisation?
Warren

···

On Wed, Mar 19, 2014 at 8:39 PM, Adam Matic adam.matic@gmail.com wrote:

[From Adam Matic 2014.03.17 2140 cet]

Attached are some pictures of the robot arm I’m making and trying to tune. It is similar to the arm from Coordination Demo, but with less degrees of freedom.

Level one and two are angular velocity and angular position control for each joint, so each joint is like a servo motor system. There are five joints - rotation of the base, shoulder pitch, elbow pitch, wrist roll and wrist pitch. I’ve made these fairly low gain, but very small dt, around 1 ms, and large resolution of input angular position measurements - 13 bits.

–
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences

Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

Level three are reach, elevation, hand pitch, hand roll and lateral displacement systems. Perceptual signals for these are constructed like in Coordination Demo, using basic geometry formulas from arm segment lengths and joint angles. However, they are noisy, so they and their output are filtered.

I think they now work fairly well, considering how they worked before. There may still be room for improvement, but I’ll see about that as I go. At the moment, I’m working on adding touch/pressure sensing to the hand. There are three metal ‘nails’ on the hand that push a slider potentiometer up when pressed, and will be used to construct pressure perceptions. These pressure perceptions will be used to keep the hand holding a computer mouse. So, that will be a level above coordination of joints. I have no idea how I’ll tune these, but as Bill said, if it doesn’t work, just keep reorganizing.

When touch is finished, I plan to add a very simple visual system so I could do a tracking task. A camera will be pointed to the computer screen where I will have a target circle and a cursor circle, so that the Little Man Robot will control the distance between them.

Arm segments are cut from foam plastics and connected with screws, superglue and hot glue. Each joint has one or two micro geared DC motors and a potentiometer to measure joint angle. I’m using a Teensy 3.1 electronic board which is programmed like an Arduino, in C.

I have no idea when it will be finished. Soon, I hope. I’ll do some test and write a paper and this will be my M.A. in psychology in University of Zadar, Croatia. I have a B.A in psychology for which I did a standard tracking task in two dimensions. Before psychology I spent some time studying electrical engineering and computer science, but I wasn’t any good at electrical engineering so I dropped out and worked for I while, then decided to study psychology.

Adam

7

Thanks Warren, I’ll make a video soon, I’ll post updates for any major steps.

I might try implementing reorganization in this arm after testing the tracking task, or perhaps in a different robotic system, I still don’t really know what I’ll do after the tracking task. There are many options.

For example, I could try going for more lower-level physiological realism and try making tension control systems with opposing flexor and extensor ‘muscles’ for each joint. I could try finding out more about higher levels of control, such as event and program control. Making more detailed version of the visual system would be very interesting. Reorganization is certainly one of the things I want to try out an learn about.

I plan to make detailed description of how the system was made and put source code online, so if anyone else would like to make similar projects, they could use it, and try out themselves what they find interesting.

Adam

···

On Thu, Mar 20, 2014 at 9:14 AM, Warren Mansell wmansell@gmail.com wrote:

Hi Adam, this is really exciting to see! It will be very interesting to see some data and/or videos to show the control performance, and to see the specification in detail when it is complete. You must have such perseverence and focus to produce this and I wish you all the best in your continued work on it! Keep us all posted!
By the way, are you planning to equip it with the capacity for reorganisation?
Warren

On Wed, Mar 19, 2014 at 8:39 PM, Adam Matic adam.matic@gmail.com wrote:

[From Adam Matic 2014.03.17 2140 cet]

Attached are some pictures of the robot arm I’m making and trying to tune. It is similar to the arm from Coordination Demo, but with less degrees of freedom.

Level one and two are angular velocity and angular position control for each joint, so each joint is like a servo motor system. There are five joints - rotation of the base, shoulder pitch, elbow pitch, wrist roll and wrist pitch. I’ve made these fairly low gain, but very small dt, around 1 ms, and large resolution of input angular position measurements - 13 bits.

–
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences

Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

Level three are reach, elevation, hand pitch, hand roll and lateral displacement systems. Perceptual signals for these are constructed like in Coordination Demo, using basic geometry formulas from arm segment lengths and joint angles. However, they are noisy, so they and their output are filtered.

I think they now work fairly well, considering how they worked before. There may still be room for improvement, but I’ll see about that as I go. At the moment, I’m working on adding touch/pressure sensing to the hand. There are three metal ‘nails’ on the hand that push a slider potentiometer up when pressed, and will be used to construct pressure perceptions. These pressure perceptions will be used to keep the hand holding a computer mouse. So, that will be a level above coordination of joints. I have no idea how I’ll tune these, but as Bill said, if it doesn’t work, just keep reorganizing.

When touch is finished, I plan to add a very simple visual system so I could do a tracking task. A camera will be pointed to the computer screen where I will have a target circle and a cursor circle, so that the Little Man Robot will control the distance between them.

Arm segments are cut from foam plastics and connected with screws, superglue and hot glue. Each joint has one or two micro geared DC motors and a potentiometer to measure joint angle. I’m using a Teensy 3.1 electronic board which is programmed like an Arduino, in C.

I have no idea when it will be finished. Soon, I hope. I’ll do some test and write a paper and this will be my M.A. in psychology in University of Zadar, Croatia. I have a B.A in psychology for which I did a standard tracking task in two dimensions. Before psychology I spent some time studying electrical engineering and computer science, but I wasn’t any good at electrical engineering so I dropped out and worked for I while, then decided to study psychology.

Adam

8

That’s wonderful. I’ll link to it from pctweb. I guess reorganisation is one way to optimise the parameters that otherwise you would have to do your own trial and error testing of, but when we have tried to implement reorganisation, it is no mean task as there are parameters within the reorganisation process that themselves need to be estimated (time windows, starting values for parameters & their rate of change, etc).
Warren

···

On Thu, Mar 20, 2014 at 10:39 AM, Adam Matic adam.matic@gmail.com wrote:

Thanks Warren, I’ll make a video soon, I’ll post updates for any major steps.

I might try implementing reorganization in this arm after testing the tracking task, or perhaps in a different robotic system, I still don’t really know what I’ll do after the tracking task. There are many options.

For example, I could try going for more lower-level physiological realism and try making tension control systems with opposing flexor and extensor ‘muscles’ for each joint. I could try finding out more about higher levels of control, such as event and program control. Making more detailed version of the visual system would be very interesting. Reorganization is certainly one of the things I want to try out an learn about.

I plan to make detailed description of how the system was made and put source code online, so if anyone else would like to make similar projects, they could use it, and try out themselves what they find interesting.

Adam

–
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences

Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

On Thu, Mar 20, 2014 at 9:14 AM, Warren Mansell wmansell@gmail.com wrote:

Hi Adam, this is really exciting to see! It will be very interesting to see some data and/or videos to show the control performance, and to see the specification in detail when it is complete. You must have such perseverence and focus to produce this and I wish you all the best in your continued work on it! Keep us all posted!
By the way, are you planning to equip it with the capacity for reorganisation?
Warren

On Wed, Mar 19, 2014 at 8:39 PM, Adam Matic adam.matic@gmail.com wrote:

[From Adam Matic 2014.03.17 2140 cet]

Attached are some pictures of the robot arm I’m making and trying to tune. It is similar to the arm from Coordination Demo, but with less degrees of freedom.

Level one and two are angular velocity and angular position control for each joint, so each joint is like a servo motor system. There are five joints - rotation of the base, shoulder pitch, elbow pitch, wrist roll and wrist pitch. I’ve made these fairly low gain, but very small dt, around 1 ms, and large resolution of input angular position measurements - 13 bits.

–
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences

Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

Level three are reach, elevation, hand pitch, hand roll and lateral displacement systems. Perceptual signals for these are constructed like in Coordination Demo, using basic geometry formulas from arm segment lengths and joint angles. However, they are noisy, so they and their output are filtered.

I think they now work fairly well, considering how they worked before. There may still be room for improvement, but I’ll see about that as I go. At the moment, I’m working on adding touch/pressure sensing to the hand. There are three metal ‘nails’ on the hand that push a slider potentiometer up when pressed, and will be used to construct pressure perceptions. These pressure perceptions will be used to keep the hand holding a computer mouse. So, that will be a level above coordination of joints. I have no idea how I’ll tune these, but as Bill said, if it doesn’t work, just keep reorganizing.

When touch is finished, I plan to add a very simple visual system so I could do a tracking task. A camera will be pointed to the computer screen where I will have a target circle and a cursor circle, so that the Little Man Robot will control the distance between them.

Arm segments are cut from foam plastics and connected with screws, superglue and hot glue. Each joint has one or two micro geared DC motors and a potentiometer to measure joint angle. I’m using a Teensy 3.1 electronic board which is programmed like an Arduino, in C.

I have no idea when it will be finished. Soon, I hope. I’ll do some test and write a paper and this will be my M.A. in psychology in University of Zadar, Croatia. I have a B.A in psychology for which I did a standard tracking task in two dimensions. Before psychology I spent some time studying electrical engineering and computer science, but I wasn’t any good at electrical engineering so I dropped out and worked for I while, then decided to study psychology.

Adam

9

Good point. I should probably use automated trial and error instead of doing it myself. I’m a bit afraid of this scenario: https://xkcd.com/1319/
:slight_smile:

Definitely need to try reorganization to improve parameters of control. Thanks for the input!

Adam

···

On Thu, Mar 20, 2014 at 12:15 PM, Warren Mansell wmansell@gmail.com wrote:

Warren

That’s wonderful. I’ll link to it from pctweb. I guess reorganisation is one way to optimise the parameters that otherwise you would have to do your own trial and error testing of, but when we have tried to implement reorganisation, it is no mean task as there are parameters within the reorganisation process that themselves need to be estimated (time windows, starting values for parameters & their rate of change, etc).

On Thu, Mar 20, 2014 at 10:39 AM, Adam Matic adam.matic@gmail.com wrote:

Thanks Warren, I’ll make a video soon, I’ll post updates for any major steps.

I might try implementing reorganization in this arm after testing the tracking task, or perhaps in a different robotic system, I still don’t really know what I’ll do after the tracking task. There are many options.

For example, I could try going for more lower-level physiological realism and try making tension control systems with opposing flexor and extensor ‘muscles’ for each joint. I could try finding out more about higher levels of control, such as event and program control. Making more detailed version of the visual system would be very interesting. Reorganization is certainly one of the things I want to try out an learn about.

I plan to make detailed description of how the system was made and put source code online, so if anyone else would like to make similar projects, they could use it, and try out themselves what they find interesting.

Adam

–
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences

Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

On Thu, Mar 20, 2014 at 9:14 AM, Warren Mansell wmansell@gmail.com wrote:

Hi Adam, this is really exciting to see! It will be very interesting to see some data and/or videos to show the control performance, and to see the specification in detail when it is complete. You must have such perseverence and focus to produce this and I wish you all the best in your continued work on it! Keep us all posted!
By the way, are you planning to equip it with the capacity for reorganisation?
Warren

On Wed, Mar 19, 2014 at 8:39 PM, Adam Matic adam.matic@gmail.com wrote:

[From Adam Matic 2014.03.17 2140 cet]

Attached are some pictures of the robot arm I’m making and trying to tune. It is similar to the arm from Coordination Demo, but with less degrees of freedom.

Level one and two are angular velocity and angular position control for each joint, so each joint is like a servo motor system. There are five joints - rotation of the base, shoulder pitch, elbow pitch, wrist roll and wrist pitch. I’ve made these fairly low gain, but very small dt, around 1 ms, and large resolution of input angular position measurements - 13 bits.

–
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences

Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

Level three are reach, elevation, hand pitch, hand roll and lateral displacement systems. Perceptual signals for these are constructed like in Coordination Demo, using basic geometry formulas from arm segment lengths and joint angles. However, they are noisy, so they and their output are filtered.

I think they now work fairly well, considering how they worked before. There may still be room for improvement, but I’ll see about that as I go. At the moment, I’m working on adding touch/pressure sensing to the hand. There are three metal ‘nails’ on the hand that push a slider potentiometer up when pressed, and will be used to construct pressure perceptions. These pressure perceptions will be used to keep the hand holding a computer mouse. So, that will be a level above coordination of joints. I have no idea how I’ll tune these, but as Bill said, if it doesn’t work, just keep reorganizing.

When touch is finished, I plan to add a very simple visual system so I could do a tracking task. A camera will be pointed to the computer screen where I will have a target circle and a cursor circle, so that the Little Man Robot will control the distance between them.

Arm segments are cut from foam plastics and connected with screws, superglue and hot glue. Each joint has one or two micro geared DC motors and a potentiometer to measure joint angle. I’m using a Teensy 3.1 electronic board which is programmed like an Arduino, in C.

I have no idea when it will be finished. Soon, I hope. I’ll do some test and write a paper and this will be my M.A. in psychology in University of Zadar, Croatia. I have a B.A in psychology for which I did a standard tracking task in two dimensions. Before psychology I spent some time studying electrical engineering and computer science, but I wasn’t any good at electrical engineering so I dropped out and worked for I while, then decided to study psychology.

Adam

10

[From Bruce Abbott (2014.03.20.0950 EDT)]

Adam Matic 2014.03.17 2140 cet –

Attached are some pictures of the robot arm I’m making and trying to tune. It is similar to the arm from Coordination Demo, but with less degrees of freedom.

Level one and two are angular velocity and angular position control for each joint, so each joint is like a servo motor system. There are five joints - rotation of the base, shoulder pitch, elbow pitch, wrist roll and wrist pitch. I’ve made these fairly low gain, but very small dt, around 1 ms, and large resolution of input angular position measurements - 13 bits.

Level three are reach, elevation, hand pitch, hand roll and lateral displacement systems. Perceptual signals for these are constructed like in Coordination Demo, using basic geometry formulas from arm segment lengths and joint angles. However, they are noisy, so they and their output are filtered.

I think they now work fairly well, considering how they worked before. There may still be room for improvement, but I’ll see about that as I go. At the moment, I’m working on adding touch/pressure sensing to the hand. There are three metal ‘nails’ on the hand that push a slider potentiometer up when pressed, and will be used to construct pressure perceptions. These pressure perceptions will be used to keep the hand holding a computer mouse. So, that will be a level above coordination of joints. I have no idea how I’ll tune these, but as Bill said, if it doesn’t work, just keep reorganizing.

When touch is finished, I plan to add a very simple visual system so I could do a tracking task. A camera will be pointed to the computer screen where I will have a target circle and a cursor circle, so that the Little Man Robot will control the distance between them.

Arm segments are cut from foam plastics and connected with screws, superglue and hot glue. Each joint has one or two micro geared DC motors and a potentiometer to measure joint angle. I’m using a Teensy 3.1 electronic board which is programmed like an Arduino, in C.

I have no idea when it will be finished. Soon, I hope. I’ll do some test and write a paper and this will be my M.A. in psychology in University of Zadar, Croatia. I have a B.A in psychology for which I did a standard tracking task in two dimensions. Before psychology I spent some time studying electrical engineering and computer science, but I wasn’t any good at electrical engineering so I dropped out and worked for I while, then decided to study psychology.

BA: Nice work, Adam. I’m looking forward to seeing the finished product in action!

Bruce