How Do Exoskeletons Integrate Sensors, Control Algorithms, and Actuators?

[Tom24]

Hi, I know this has been asked before but the answers haven't been in much detail. Iv previously browsed the web only to find stupid articles giving no information what so ever. Hydraulic pistons and servo motors working in conjunction together is the usual answer but to someone like me that's not at all helpful. If someone could spend the time to give a detailed description of how they work in conjunction together then that would be much appreciated.
The reason for this is my mother has had a neurological problem for years and its only getting worse, i don't have the money to buy an exoskeleton but i do have the money to make one, the muscle sensors is the only part i can think of that i would have a problem with other than that everything else seems well achievable. Please don't question my motives-thank you for your time,
Tom.

 

[JaredJames]

Do you understand what a servo motor is for and what pistons do?

If you can answer yes to both of those, it isn't that difficult to understand the functions they perform in an exo-skeleton.

The gist of what they do is aid the body by providing additional force for movement or absorption of impact.

The cost of construction will far outstrip the cost of buying one.

As a note, no one here will design it or tell you how to build one.

 

[Tom24]

i understand that servo motors activate when a pulse is sent from a wire to the motor, and that the length of the pulse will determine how far the motor turns. I think hydraulic pistons work by a force that is applied at one point is transmitted to another point using an incompressible fluid? iv read a few articles/howto's on them and get the general concept of them both. I know what they do but i don't get how they do it :L the HAL 3 suit is by far the best one iv seen and they are renting it at $2000 per month!

 

[JaredJames]

If you know what they do and how to get them to do it, then you can get on with building something. The "how they do it" serves no real purpose. The chances of you needing to take them apart is slim to none.

There's a reason the prices are that high. These suits aren't easy to build and take a lot of work. Tony Stark didn't do them justice in his 'rendition'.

 

[Baghdad_boy]

Hi Tom,

The complicated part of exoskeleton design is not the physical construction. The control alogrithms that go into converting the sensor data into actual movement is the complicated part.

I respect your determination and I know you are not expecting this to be an easy task so I will try to point you in the right direction.

The first thing I will say though is please be safe. Dealing with mechanical devices can be extremely dangerous especially when testing a new control method. The output is very rarely what you would expect. So please be safe :)

Ok, so with an exoskeleton, there are 3 levels; sensors, control, and actuators.
Between these levels there are interfaces. The interfaces are split into 2 sections; hardware and software.

So the final structure may be something like this
Sensors->Interface->Control->Interface->Actuators(eg motor)

The thing to know about sensors(especially EMG): The readings you get are very far from ideal. The readings must be filtered using hardware or software filtering techniques. This is usually done onboard for most modern sensors, so no need to worry too much about initial filtering stage. But you will have to filter the signal even more to be able to extract useful information from it. This is a complicated task to say the least. The signal may be non linear (meaning that doubling your muscle contraction may not double the signal).

Next, let's say you manage to extract useful information from the sensors (eg you know that the user has sent a signal to their muscles to contract 20%). Now you need to know what to do with this signal. How do you translate it to something useful? You have to identify in which direction the user wants to move. This is another very complex task. There is a lot of research into this type of prediction techniques.

Third, the controller must now translate this prediction into an actual actuation. This is another complex (yes everything is complex that's why there aren't many exoskeletons on the market) process. Let's say you put a 5V signal to control a DC motor. The motor spins. How fast is it spinning? If you half that voltage to 2.5V does this mean that the motor speed is halfed? What if an external force is applied to the shaft all of the sudden? Can the motor maintain its speed or position?

You can look into controller for actuators to learn more. One of the most popular types of controllers is PID controllers. This has 3 parts that manipulate the input signal to produce a desired output signal. They also track the error of the output over time an try to reduce this error as much as possible and as fast as possible.

The things I mentioned in this post are not even scratching the surface on the subject and I tried to keep the language as simple as possible.

I recommend you look up scholary publications on the subject by searching in databases such as the IEEE Xplore database. There is a wealth of knowledge there. However, don't expect this to be an easy task. If it was, many people would have already released exoskeletons. Even the HAL series of exoskeletons are not mature yet.

If you are able, I recommend taking a degree in Robotics Engineering at a reputable university. You may not be able to help your own mother now, but this may lead you to helping someone else's mother in the future.

I wish you and your mother all the best in health.

Salam,
Ibrahim

EDIT: I just noticed you mentioned that your mother has a neurological disorder. This makes the situation a magnitude more complex. When you mentioned "muscle sensors" I assume you are referring to Electromyography (EMG), which picks up signal sent through the nervos system from the brain or spinal cord to the muscles in order to cause actuation. The problem is that people with neurological disorders can't produce the correct signals in the first place and that is what causes the poor muscle control. So it is pointless to look into the natural signals from the body (which is what the HAL suit does) in order to predict the user's desired movement.

I have the same problem since my final year project is designing a rehabilitation device for stroke victims. The nervos system of stroke victims is damaged, and so predicting their desired movement can't be done by monitoring their nervos signals.