Monday, November 11, 2013

Robobear and a Note on Failure

My next post will be something interesting, probably more about my Daft Punk helmets. But for now, something a little less flashy.



I would like to talk about a recent project that ended up a failure. This is not the first failure I've had with projects related to this blog, but typically I just move on to a new project that has a better chance of succeeding. There's a lot that can be learned from failure, but often each lesson is small and not worth an entire written discussion. Collected together, these small lessons from failure provide an interesting look at personal development over time. Here are some of my own recent discoveries:

- Don't connect the terminals of a battery together to see if the battery is dead.
- When soldering, wait for the iron to reach the right temperature before you begin.
- Don't let a linear voltage regulator try to dissipate 10 Watts of power, they are not light bulbs.
- Know the structural limitations of your materials (duct tape, hot glue, epoxy, solder, nuts and bolts, wood, metal, acrylic, carbon fiber, fiberglass, etc.). You don't want them breaking at just the wrong time.
- Don't waste a microcontroller on something that can be done with a cheap IC (LED driver, motor driver, etc).
- Before you buy a bunch of electronic components, check eBay. Someone might be offering it for 10% the cost you would have payed elsewhere.
- If you need to make analog measurements (like a microphone), make sure the power source is stable and you've taken care to reduce electromagnetic interference.
- Prototype, collect data, and analyze your idea before committing time and money. This is like the advanced version of "measure twice, cut once".

Failure in the context of a hobby is a very productive kind of failure, since the risk is low. All you can waste is your own time and money. There isn't a risk of disappointing someone else (usually), and there isn't a risk of getting fired. The reason behind a failed project can be discovered and the lesson digested on your own time. Working on my own projects in my free time has allowed me the opportunity to stop and smell the roses and then truly understand why soaking roses in chloroform is a bad idea.

WARNING: The following is a long-winded story about the Robobear project and how it failed. If you are not interested in the details of how I attempted to motorize a wooden cart that would carry a bass player and a taxidermy bear, skip it. Alternately, here is a tldr summary: My job was to motorize a big wooden cart so it could move silently across a stage, the project failed despite solving many problems along the way.

A few weeks ago I was approached by my good friend Julie to come up with some ideas on how to get a wooden cart to move across a stage with minimal human interaction. The cart was 9 feet long, 4 feet wide, and only about 3 inches high. The purpose of the cart was to ferry a string bass player and a taxidermy bear on to and off of a stage for an artistic performance. I've heard of strange requirements from clients, but this was certainly unique.

I saw a few ways of moving the cart:
1 - Have someone push/pull the cart
2 - Have two ropes; one tied to the back for pulling, and one attached to a wind-up mechanism underneath the cart to make the cart move away when the second rope is pulled.
3 - Motorize some of the wheels and use wired, wireless, or fully automated controls.
After discussing the options, she chose to move forward with the motorized option and keep the other two as backup plans. And so, the idea of Robobear was born. I started planning, prototyping, and ordering parts. The budget was tight, so I had to plan carefully and keep cost in mind when designing the system.

The primary concern throughout this whole project was torque. The motors have to be capable of exerting enough torque to push the cart. If we pretend the cart is in a frictionless world (except the friction that allows the tires to push against the ground) and the ground is completely flat, then any amount of torque will do. Having more torque available would just increase the possible acceleration of the cart. Since we unfortunately don't live in such a world, we need the motors to be able to overcome friction and bumps in the ground before they can even start to move the cart. The first step in this project was estimating the necessary torque to get the cart rolling. By pulling the cart by hand with various amounts of weight on it (very scientific), I was able to estimate that it needed about 30 pounds of horizontal force.

To convert between force and torque, we need to know the lever arm distance (torque = force * distance). In this case that is the radius of the wheels that will be used. There wasn't a whole lot of space underneath the cart, so I had few options. Assuming 35mm radius wheels, the combined torque of the motors needed to be at least 4.7 Nm, or 660 oz-in. This is a fair amount of torque, I think. I decided to stick with DC gearmotors, and surfed around some websites I knew of that sold high powered ones. On Pololu, the most powerful motors they offer have around 200 oz-in of torque. Four of these would be necessary to get the cart moving. The speed matched up well enough, since a 70mm wheel rotating at 150rpm gives a max speed of 0.5m/s (1.8 fps). I went with plastic wheels found on the same site, and modified some mounting hubs to connect the large motors to such small wheels. Of course, I could have tried for heftier motors, like a golf cart motor or electric scooter motors. Unfortunately the size limitation was fairly strict, so I had to stick with motors that could fit underneath the cart.

For the motor driver, I went with a pre-built controller that could supply 24 Amps of current total, just enough to allow each of the four motors to stall (6A stall current) if they got stuck. Making a motor driver myself would have probably been cheaper, but most likely could not have supplied as much current without serious heat issues. Sometimes it's best to pay for peace of mind.


Two 3S, 2200mAh LiPo batteries were taken from an old hexacopter project to supply power. I connected them in series, providing a whopping 22V with a possible 40A of current. If you have put those figures together in your head and decided this was a bad idea, then I congratulate you both on your ability to calculate electric power and your good sense. The motors would end up drawing only 10 Amps at peak for only a few seconds, so 2200mAh was enough capacity for the job. The brains of the operation was an Arduino Pro Mini running a program that would interpret serial commands and run the motor driver. With all of the parts in place, I started mounting things to the cart.




After only a few days of work, I was ready to test the system on the floor of Julie's studio. It didn't budge. The floor of the studio was a slick, polished concrete surface that was impossible to get any traction on. It was also difficult to dance on, but that's another story. When I lowered the motors a little to press harder on the floor, the motor brackets just bent back under the weight of the cart. The motors seemed capable of exerting the planned amount of torque, but that torque was unable to make it to the ground to move the cart. Things were not looking good. To increase the traction, I added an extra wheel to each motor, and reinforced the motor brackets with stainless steel wire.



With 8 wheels making contact with the slick ground, the cart was finally able to move itself. However, it couldn't handle moving all of the weight of a human and taxidermy bear yet. The wheels were still slipping, but I accepted that the stage that the cart would need to perform on would not be so slick. The next step was to test the cart on the right floor and hope for the best. Unfortunately, the stage required special access and reservations, so by the time we were able to run the test, there were only 2 weeks left before the final rehearsals. We had to accept that if the cart could not move the required amount of weight by the end of the stage test, the motorization plan would have to be abandoned. We just couldn't keep hoping for a fix right up until the end and risk having an inoperable cart with no backup plan acted on.



We finally got the cart on to the stage and I began testing. At first, it didn't work. I tightened some screws that had gotten loose during transportation and it worked a little better. I tuned the motor response in software to prevent slipping (like traction control on a car) and it worked a little better again. I adjusted the height of each motor to allow more pressure to be applied evenly to each motor and it worked a little better. I could add around 200 pounds to the cart and it could move a few feet in each direction. Finally, the cart seemed to work.

I loaded up the cart with the human half of the intended payload and set it to move about 15 feet in one direction. It started up just fine, then stopped 5 feet short. I told it to reverse direction, thinking it was caught on a wire or bump in the floor, but all I got was a repetitive clicking sound from some of the motors. The motors had enough torque to move the cart, and the wheels had enough contact with the ground to exert enough horizontal force, but the rubber tires on the wheels were getting ripped from the plastic wheels after a few seconds of use.

I quickly came up with a few ideas that could fix this problem. One would be to fuse the tires to the wheels using heat or epoxy. Another way would be to create my own tires out of silicone rubber and mold them directly to the wheels. Unfortunately, every solution I came up with required another week or two of work and testing to confirm that it would fix the problem. We just didn't have the time to try something new. The project failed due to unforeseen complications and lack of time.



So the big question is: what did I gain from this failure? Some leftover motors and half broken wheels? An intimate hatred for slick floors? No, despite my reluctance I probably learned something about getting a job done on someone else's time-frame. Most of my projects are done at my own leisure, so unexpected setbacks can last as long as I like. It was a very different experience to have a looming deadline for something I consider to be a hobby. Not only was the idea of a deadline new, but the idea that there were smaller deadlines on the way to make sure the project as a whole was worth continuing. The biggest one here was making sure the wheels could get enough traction on the theater floor far enough in advance.

In all I came away with two big lessons. They are so big, they deserve a big font.
1) Given a large goal, set many smaller goals that lead up to it. This not only helps keep things on track and on time, but forces you to consider how realistic the main goal is.
2) Don't make your hobby into a job. If you do something for fun, keep it fun. Don't set deadlines or have other people depending on your for their job. I say this not because I regret doing this project, but because I've learned that failing at something is easier when no-one is watching.