Building a walking robot is not a trivial task, but luckily I do not have to start from scratch. For starters, the detailed project image sketch is based off of the project image for #Robotics Chat by @Radomir Dopieralski. I will also use the knowledge I gained from my 2014 Hackaday Prize Entry #ANDY: A Multi-Purpose "Humanoid" Robot
After browsing the internet for quite some time, I came to the conclusion that either you had to pay thousands of dollars to get a advanced humanoid robot (almost always using proprietary technology) or hack together your own. There are a few options in between, but still two expensive with too little support for my liking.
Of course I had to start out by building my own BOB the biped (the image on the left). I 3D printed out the files in white PLA on my Lulzbot Mini and assembled him with machine screws and SG90 servos purchased off of ebay. since pretty much all the code for BOB is based off Arduino, I used a Adafruit Pro Trinket 5V Hackaday Edition as the MCU.
I flashed the first BOB arduino sketch I could find, and lets just say things didn't go smoothly. First, the servos tried moving to a position that didn't exist, and then the robot made jerky movements before tipping over. I played with the code for an hour or two before finally giving up and looking for more code to try. Next I some code that used some heavy duty mathematics to figure out how BOB should position his steps. This code partially worked, but the robot would still tip over every 2 or three steps, or stall the servos. When the robot did take the steps successfully, it was not in a straight line at all. After modifying the foot size writing my own code from scratch, I finally got decent results.
After some more research I discovered the Zowi project
from BQ Labs. Their code
is heavily optimized for their Arduino derivative with Futaba 3003 servos. This is proof that you can get great results with just a 4 DOF biped, so that is what I am going to start out with on TyroBot.
Although Zowi has the walking part down, the project mainly relies on the app limiting its practical use.
In conclusion, TyroBot will work just fine with just four degrees of freedom. It will just require a sturdy structure (which BOB lacks) a good codebase (like Zowi) and a Accelerometer to help TyroBot make accurate steps. I included a list of project goals in the project description.