bookmark_borderArm exosuit part 2: Failure

After trying the fabric arm for a few minutes I felt lightheaded. Unfortunately the fabric arm needs to be very tight in order to be rigid enough to work. Being able to breath is more important than moving, therefore this option is not an option. My original plan, before plan fabric, was to use a “Steadicam” type device to balance my arm. Normally used to keep a camera perfectly balanced against gravity, it is also the mechanism used in many tool holding devices, lamps, microphone arms, and many existing mobile arm supports. Unable to afford one of these devices, I 3d printed my own. It may yet work but it has a few problems.

One, it only works when perfectly perpendicular to the ground. This will cause problems when I need to tilt my chair or drive on New England pavement. This could be fixed by attaching the whole assembly on a pivot that maintains its position with respect too gravity. This would not work when the chair is tilted in the Y plane, and adds bulk which leads to problem number two.

Two, this mechanism is very bulky. It adds a foot and a half of width to both sides of my chair. My chair is already 30 inches wide and barely fits through doors. Taking the devices off every time I encounter a door would negate any independence these devices give me.

Three, it may be too late. A passive exoskeleton still requires some strength to use, strength I had that is now gone. An active device is another story.

The easy part of an arm exoskeleton is the elbow, its only one degree of freedom and sensing my muscle movement should be easy. There are medical devices that do this already, however they are useless to me without the shoulder. Fortunately shoulder support exoskeletons are popping up everywhere. Intended for the factory environment to reduce workplace injuries from repetitive above head motions, they could be repurposed for my use instead. Not so fortunately, these products cannot be used while sitting down in a chair. These devices could be modified and placed on the back of my wheelchair however. The most compact design I have found so far is the Ottobock Paexo, and a close second is the ShoulderX from SuitX.

The ShoulderX could probably be attached to my wheelchair without any modification, however this design is not very flexible and must be placed in a particular way. This is disappointing as this design could easily be motorized without adding much bulk.

The Paexo design has given me an idea.

Here is a quick sketch of my next idea:

I’m nowhere near done but the basic idea is here. Here is a quick rundown:

  • E This is the brace that will be strapped to my arm, allowing the rest of the device to move my arm around.
  • D Just the rod connecting E to C, it has no joints except for a ball joint attaching to E
  • C This is a revolute joint between D and E
  • B The rod connecting C to A, not shown is B can twist, allowing side to side movement of E
  • A The attachment bracket for my wheelchair’s side rail. Also contains a revolute joints for B

This device will need to be motorized on every joint and that presents my largest hurdle. This device i’m now calling Plan C, is really a robot arm in disguise that uses my arm as it’s payload. I need to learn too many things before I can tackle this, advanced topics that need years of study to understand. I’m leaving the realm of screwing around and into Engineering. I can’t afford to hire an Engineer, so my only option is to become an Engineer myself. So I enrolled in college for Electrical Engineering and will start soon.

bookmark_borderAssuming direct control

Exoskeletons and other assistive robotics can never truly takeoff until they can be controlled properly. Even the best robots in the world can’t out-walk a toddler. This problem only gets worse when the robot is attached to a human. Humans and machines will always argue until the day they die over who is in charge.

Frustration occurs whenever a machine disobeys a direct order. Over-aggressive abs and traction control in a car, aim assist in video games, a wheelchair defeated by a half inch curb, all examples of machines gone wrong. All machines should aspire to be a microwave, you enter 30 seconds you get 30 seconds. If you tell a machine to jump it should not ask how high, it should just jump. If you wanted a certain height you would have specified.

There needs to be a way to control more outputs than current means. It takes two human arms to control one robot arm, see an excavator, and that does not include fingers. It would be much better if we had access to the spinal cord’s output to our limbs.

Reading intent from muscle movement is hard because we cant put sensors inside where they are needed.

Our body is controlled by opposing muscles, each pulling a joint up or down. Lets use the elbow as an example. The bicep pulls your forearm up, and the tricep pulls it down. However this is not all they do, your brain can also vary the stiffness of your limbs. It does this by pulling both the bicep and tricep at the same time, in order to lift more weight or have more precise control.

Muscles by themselves are not precision instruments but this effect makes them so. Now imagine if your brain had direct control over a electric motor.

Propose you had a device that could measure the intended power of every muscle in the human body. And you wanted to control the elbow motor of an exoskeleton. This motor is attached to a variable gearbox.

To gear the motor up or down we need simply to use the variable stiffness built in to our brain. When the bicep and tricep are both pulling hard, gear down, when they’re relaxed, gear up. Use the difference between the bicep and tricep for motor direction. Effectively the same mix for tank drive on a single joystick.

This would also save power because gearing down keeps joints in place without powering the motor. Also gearing down gives more torque, allowing for greater deadlifts and other feats of “strength”.

bookmark_borderArm exosuit part 1

My first attempt at an exosuit is going to be based on a cheap compression shoulder brace and elbow sleeve. These will be attached together with nylon webbing ensuring the load path travels past my shoulder. Actuation will consist of bowden cables attached to both the top and bottom of my elbow. Acting like a replacement for my bicep and tricep muscles.

This is based on the device in my second reference below. However they used a very complicated motor and pulley system that I believe is unnecessary. Instead I will use two springs, one for each bowden cable.

The A-gear, see below, also uses two springs. In fact it is my inspiration. From the A-gear I discovered two springs can statically balance 4 degrees of freedom.

These devices also share another similarity, they are the exact same device! Think about it, one is fabric, the other is aluminum but look closer and you will see. They both pull on the same points on the arm and forces are redirected in the same manner. Achieving the same goal with different materials.

The main problem with a passive spring based approach is it may be too late for me. A decade ago I would have had enough strength but now I have almost nothing. I tried the compression sleeves and the friction was noticeable. I have a backup plan for motor assistance just in case.

References:

1. A-gear dynamic arm support

2. Soft wearable assistive robotics: exosuits and
supernumerary limbs