Pseudo-Closed-Loop-Feedback for Sensorimotor Control

Rodney Brooks’s fashionable ‘the world is its own model’ idea obviates the traditional need for ‘representations’ within the mind. Rick Grush believes such representations are the basis of consciousness and reinforces ‘representation’ by drawing on ‘pseudo closed loop feedback’ (PCLF). PCLF is described by Andy Clark’s ‘Supersizing the Mind’, from pp.149-152. Here’s my understanding of it…

Pseudo Closed Loop Feedback

In an open-loop system, commands (desires) from the ‘mini-me’ inside my head are issued (1), go through my motor system (4) to drive parts of my body e.g. hand (5). There is no feedback as to whether my hand has moved in the right way.

In a closed-loop system, observations (7) of my hand pass back into me through my eyes (9) and visual processing. Errors are removed by adjusting (2) the commands to my motor system. There is a feedback loop from (2) out via (4) and back via (10).

Now, the main event: in a pseudo-closed-loop system, I call upon an internal emulation of what is actually happening outside of the ‘mind’. The emulation (11 and 12) models how the motor (4 and 5) and sensor (9 and 10) systems behave – the 3D+time physical world (7). I’ve only shown it as separate forward (11) and reverse (12) paths for comparison with the normal closed-loop system; in practice, there is no need to differentiate it – it’s just one lumped model. The switch (13) connects the emulator’s output back into the feedback loop.

The conjecture is that the emulation is the source of consciousness. And when I dream, my motor system (4) may be disengaged – switch (3) is open – but I can experience moving my hand and seeing the consequences in my dreams via my internal representations – the emulator model (11 and 12).

And why have this emulation model in the brain in the first place? Imagine trying to use a JCB digger for the first time. You are unaccustomed to the controls. You gingerly move the levers, see the JCB’s arm move and react accordingly. Progress is slow. An experienced operator won’t have to think twice about which lever does what but he will also build up experience of how far he has to move the lever to move the arm a certain amount. He would be so accustomed he could ‘do it with his eyes closed’ – or at least have a reasonable attempt at it with his eyes closed. Now imagine the JCB is being operated remotely – say, on Mars from Earth. It would take an experienced JCB operator a while to get used to the new dynamics of the situation – the time delay (both for commands to Mars and responses back to Earth). The pseudo closed loop system allows the brain (controller) to remember/learn the dynamics of the sensorimotor system and hence be able to control motor systems faster, without waiting for any slow external feedback response – teleassistance rather than teleoperation. When we want to pick something up, we can use our pseudo closed loop control to get our hand most of the way there quickly. Then we can operate in a closed-loop fashion with our eyes to complete the task.

(Review of PCLF in M. Kawato in Current Opinion in Neurobiology 9:718-727, 1999.)

This entry was posted in Uncategorized and tagged , , , , . Bookmark the permalink.

2 Responses to Pseudo-Closed-Loop-Feedback for Sensorimotor Control

  1. headbirths says:

    ‘Pseudo closed-loop feedback’ seems to me to be essentially the same as using observers in control theory. I was going to look this up in my old university lecture notes but can’t find them! So some googling around… A state observer can be used where the states of the system are not observable (which is commonly the case). It’s a ‘copy’ (model) of the system so that the model’s internal states can be tapped for making the feedback connections. See the ‘observer design’ section of this state space tutorial for clearer explanation and diagrams. With PCLF, the system itself is observable – indirectly, by moving limbs and seeing the result. But it is quicker and perhaps more energy efficient to use an ‘observer’ internal model. Example of a paper using observer design for robotics here – including passivity (energy efficiency).

  2. Pingback: Bubbles | Headbirths

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s