Human trigger for automatic retraction

[Batienzaxcore]

Okay guys. Some of you may be familiar with this problem I'm having (I've been posting a lot about this topic). ANyway, to enlighten the others, I'mma go explain this one more time.

I am currently doing a portable robot manipulator for knee rehabilitation. The knee manipulator (Something like braces for a single leg) would do the flexion and extension for a certain patient that has knee injuries. Right now, I got the basic stuff using a 877a uC, a simple DC motor with a specialized 24V HBridge, a start button and a toggle switch (to determine the flex and extend mode of the manipulator) in a handheld controller thingy with an LCD display that would display its angle, measured by a potentiometer. The initial position of the manipulator is in a inverted 90 degrees shape, worn while a patient is sitting upright and elevated.

Now this comes the hard part. If ever that I use my robot manipulator to extend the leg of a patient, and the patient feels pain in his leg, some kind of human trigger would activate a sensor connected to the uC, and I want my robot manipulator to automatically go to the opposite direction until it reaches its initial position because of the fact that he cannot reach that certain level of . Trouble is, I don't know how would I do it. I mean, what kind of trigger from the patient could I use? Currently, I am trying to avoid a simple push button from the handheld controller, but if I had no choice...

Anyway, could you help me out? One of my professors advised me to do this. For example, the patient has his leg extended by the manipulator, then suddenly, he feels this sharp pain from his knee, he would have a tendency to resist the movement of the manipulator suddenly. That would be the trigger for the automatic retraction.

Any ideas? Questions will be entertained. I apologize for the bad English.

[mstacho]

Hm...you're facing a really fascinating (and very common but VERY hard to solve, in general) problem.  Think about what would happen if you used a "sudden increase" in resistance forces: since people have different leg sizes, weights, and (probably) muscle disorders, what is a "normal" resistance force for one person might be a very strongly resistive one for another.

I solved a similar problem once by using an initial probing input, then watching the derivative of the forces.  In theory (in VERY hand-wavy theory!) a few small, slower motions should give you a good idea of what the expected derivative of the forces would be during a normal run.  Record the max and the min derivatives encountered during this probe, then watch and make sure you aren't too far outside of those bounds during normal operation.

It's surprisingly effective, as long as you can ensure that your probing inputs are adequate! If you use too small a probe, then you'll always trigger your safety feature :-P

Since it's working with people, you'll probably always need emergency stops as well, but they should be clearly marked for emergencies only, and this should form an initial layer of safety...

Neat project!

MIKE

[Batienzaxcore]

Uhm, okay? What would be a very good sensor to use here? I'm thinking of an FSR (Force-sensing resistor) but then I do not know where to put it

It still confuses me mr. mstacho. Could you explain it to me in an easier, er, way? I'm not really good at comprehending stuff like this. What kind of probe? And what does that probe thing look like?

[newInRobotics]

The only way I can think of to measure pain level is to use brain activity monitor, which is definitely going to be difficult and expensive to implement, moreover it requires good neurological knowledge.

Pain is something not easy to measure, as it depends on ones perception and ones tolerance to pain, in my opinion, the best thing to do is to let patient to decide when tension starts feeling uncomfortable and click the button to relieve it. The relieve button can be fitted in a Hand Gripper like device.

[Batienzaxcore]

Fair point Mr. NewInRobotics. I even considered EMG and biofeedback, but then I was thinking of something much more easier and simpler to implement in my design project, so I decided that it was out of the question, given that I have less time and resources for that here in the Philippines.

How about a person's reflex? Maybe I could use something like that? Surely when he feels pain, he'd experience something like a reflex, and his leg would jerk away automatically.

[newInRobotics]

How about a person's reflex? Maybe I could use something like that? Surely when he feels pain, he'd experience something like a reflex, and his leg would jerk away automatically.

Your original idea of placing pressure sensor at heel support flange is something that might work very well. If tension device is going to be used with many different people, then, I assume, You'll have to do some data collection which would show how weight of the leg correlates to amount of pressure You get at normal operation (no discomfort is felt by patient); if this device is going to be used by one person only, then You can manually fine-tune threshold pressure value for that specific person.

Also, I think, You still have to implement safety cutoff switch which is going to be held by patient and pressed if any discomfort is felt; this is due to the fact that in rehabilitation process it is important not to cause any pain, as pain indicates that human tissue is still damaged and forces are applied to it, hence applying too much force can slow down, or reverse recovery process, therefore if patient feels that his/here limits are stressed more than it is personally acceptable, one should be able to stop that discomfort at command.

[mstacho]

I agree with NewInRobotics: if you're only using it for one person, then fine tuning the pressure should be perfectly fine.

The problem isn't so much with figuring out what to measure, but figuring out when that measurement represents "pain" compared to when it represents "normal".  My idea doesn't use a physical probe.  Instead, you make the robot move the leg slowly.  This is the "probe" input.  Since you're moving slowly, the operation should always be normal.  So from the probe input you get a sense of what is normal for a particular person, then you just keep checking to make sure you're inside that range of normal while the robot works.

You can do it with pressure sensors or whatever you need, as long as you can determine what normal is for a given person.

It's actually a really hard problem to solve in general: you're looking at all this sensor data. Somewhere in that data is a signal that represents "stop!".  But how do you tell the computer what to look for? :-P Welcome to the hardest problem of artificial intelligence.  However, solving it in your specific case might just require you to hook a whole bunch of people up and move them until they feel pain, then checking your sensor data and seeing if any obvious pattern emerges :-)

MIKE

[Batienzaxcore]

Hmmm... Attached to this are some of the pictures of the robot manipulator that I'm doing right now. Any thoughts about it? Anyway it's not really perfect, I still have... Rather, my group mate still has to improve it. Anyway, would using a force-sensing resistor work here, if I attach it to the heel part of the manipulator? The FSR would take the abrupt change in the force applied to it, and that would be a signal to the uC that it should retract.

Oh God, this is sooooooooo confusing. I have doubts about this. I DO really need help.

[mstacho]

Hm, it sounds like it should be a start at least.  This is engineering, right?  So the purpose of prototyping isn't just to test an idea, it's also to break an idea, to push it until it doesn't work anymore, and to learn lessons from that failure.

An FSR attached to the heel MIGHT register a large increase in resistance as the pain threshold is reached.  It's well worth testing out, and it represents a first idea.  What you might find is that some weird signals show up that can be used, with more sensors.  Or maybe the FSR is OK and it's worth using.  Or maybe it won't work :-P BUT you won't know until you build it and test it, then iterate over the design with the lessons learned in hand, right?

MIKE

[Batienzaxcore]

.

[mstacho]

  Can I suggest giving the patient a stop button anyway?  It's really unlikely that patients would be comfortable with the computer  telling them when they are in pain.  A patient would probably want to have that degree of control, plus it adds redundancy.  In our lab, we're doing upper limb rehab, and we have two main ways of stopping: if the patient's force exceeds a threshold (this is currently static and causes lots of problems :-P That's why I said it's hard!) and if the patient hits an emergency stop.

If we didn't have the E-stop, we wouldn't be able to get it approved by the university's ethics board, let alone any commercialization stuff...

MIKE

[Batienzaxcore]

Yep. Engineering... Maybe I ought to discard the whole automatic-retraction-triggered-by-patient-feeling-pain thing and use a simple E-Stop instead. God, this is so frustrating

Anyway, thanks for the help guys. I'mma reply here once I encounter more problems.

As of now, we got the motor and I am currently doing the gear parts. Once i get them all, we'd be doing the mechanical design of the manipulator before I apply the E-Stop thing.

[mstacho]

Well, the idea behind your automatic pain recognizer is very interesting and important.  In theory, if the computer could handle such decisions, it will add an extra layer of safety, plus will be more generally applicable (for example, we could use a similar idea in our lab...)

If I can make a suggestion: the automatic pain recognizer is a great add-on to a product that doesn't sound like it exists yet :-D  Perhaps testing the product first with an e-stop, then enhancing it with the automation, might be a better course of action.  Don't abandon it, by any means, just make sure that the basic idea (the lower limb rehabilitation) is possible before you start adding to it.

Does that make sense?

Good luck!

MIKE

[Batienzaxcore]

It does make sense Mr. mstacho. Thank you

For now, I will be trying to use the E-Stop. If ever that I find more information that could help me out, I will try to improve it, given the less time and resources that I have.

Thanks to you to Mr. NewInRobotics. You have been helpful as well

I will be posting here once I need your help again.

PS - Looks like I'll be mentioning a lot of people here in SoR once I document this project of mine

[Batienzaxcore]

Hmmm... What other kinds of reaction does pain (as stimuli) induce to the patient? reflexes seem to be unreliable. Resistance of movement as a response is unreliable as well. Do you guys get my question?

[newInRobotics]

Maybe it would be better to ask this question in some medicine forum? Also, why I are You so keen to detect pain? If patient feels pain, that means Your device went way too far and should have stopped 'just before pain'.

[Batienzaxcore]

Hmmm you've got a point there. Uhm, maybe I ought to do the automatic retraction thing before the pain kicks in. Anyway, a friend advised me that if I should use an FSR, I'll be needing a comparator for that. I'm not really sure what to compare, I think it's supposed to be one weight of something to one weight of another. I'm currently figuring that out.

[Batienzaxcore]

Thanks to the two of you, especially Mr. mstacho here who has been very helpful

Anyway, I am planning to continue my original idea of an FSR in the heel support flange (or maybe at that fat muscle located between the foot and the knee, I forgot its term). I recently realized something about resisting the movement of the robot. For example, suppose I lift my left arm with my right arm without any resistance, there is a force applied to the right arm which is equal to a part of the weight of the left arm, right? Now, if I would lift my left arm with my right arm, and my left arm would resist the lifting of my right arm, then the force applied to the right arm would be greater than without any resistance!

How could I have realized such things a little late >.<

Could you give me advice on this?

[newInRobotics]

Anyway, a friend advised me that if I should use an FSR, I'll be needing a comparator for that.

I don't know what value FSR You are going to use. If it's very low resistance (voltage drop across it is very small) and resistance change is small with pressure applied, then You can use op-amp (not comparator) to amplify the signal (scale it to be between 0V and 5V) so ADC port of microcontroller can detect it. If You supply microcontroller level voltage to FSR, proper output can be achieved by adding set value resistor in series to FSR, this would make voltage divider outputting voltage proportional to pressure applied to FSR.

Or Your friend could have thought of comparator being used as Schmitt Trigger, when certain voltage is reached comparator output goes HIGH (or low) causing Pin Level Shift interrupt in microcontroller. That is not very good solution as fine-tuning of pressure detection would have to be done by changing hardware, where if You use ADC port, it all can be done in software (faster, cheaper and more flexible).

[Batienzaxcore]

Yes, I have been planning to use a voltage divider. How I am going to do that to flowcode would be intriguing. It would require me to go back to pen and paper for a few minutes. Thanks for the help I'mma reply here once I meet more problems.

[vipulan12]

why not make it react to sound or specifically loud "ouch" sounds
most people do have a tendency to do that (unless they speak another language)
the technology should be cheaper than a sensor that detects pain from the nervous system

[Soeren]

Hi,

Or Your friend could have thought of comparator being used as Schmitt Trigger, when certain voltage is reached comparator output goes HIGH (or low) causing Pin Level Shift interrupt in microcontroller. That is not very good solution as fine-tuning of pressure detection would have to be done by changing hardware, [...]

There's this very high tech component called a trim-pot - perhaps it might help

If a system like this is to be worth the while, it should bypass the controller completely, to avoid it being neutralized by eg. a program landing in an endless loop as the result of electrical noise (like if a motor driven vehicle with an undampened ignition system drives by, or by fluorescense lamp starters, defect fridge thermostats etc.).


Batienzaxcore <-
Using a "brain wave" interface is out of the question. Ensuring a good low impedance fit and keeping it is impossible in daily life and you wouldn't be able to tell a pain response from eg. the P300 response, which happens when the brain detect a change (like if someone go bip-bip-bip-bip-bip-bip-bip-bop, where the change of the last sound will fire the P300 response with a fairly large amplitude).
The P300 response is named so, as the peak is Positive and it happens roughly 300ms from the stimuli.


EMG only gives a measure of muscle contraction, so isn't a reliable pain "detector" and the same problem with electrode placement and consistent contact applies.

Strain, whether measure with an FSR, a much more reliable and repeatable strain gauge, bridge configuration or not... Strain does not equate to level of pain and if you were to make a curve of strain vs. pain, it would be off in a short time.

Further, pain response is very variable, even within the same individual, depending on how well they slept last night, what they had for breakfast, ambient temperatures, how their perspective on life is and a host of other influences and it's impossible to know the "acceptable pain level'" at any given time, as you don't know all the determining parameters or their values - a prime example of Chaos (in it's real meaning).


As I see it, a better approach is sticking to a emergency switch, made so it can be easily actuated under severe conditions, without being actuated by accident all the time.
Any kind of rehab should be supervised by a therapist of sorts anyway, at least initially and this is where you get intelligence far beyond any microcontroller, as he/she "reads" the patient subcontiously and can stop a movement before it causes any real pain.

Just ponder this... To setup a trigger point (whatever method), would you deliberately cause the patient pain to do so?