| Polypods - Walking the Walk
Robots that rock and roll are fun but there's just something special about a robot that walks. The wheel is an almost exclusive human invention. Only a rare few examples of rotary motion exist in nature. You'd think that having served us so well we'd hold the wheel in higher esteem. But it seems that it's always what we can't do that we covet most dearly. And worst of all, nature makes it look so effortless.
Eventually I'd like to build a balancing biped but it's best for me to crawl before I try to run. To that end I'm trying to build a polypod with four to six legs.
There are some tremendous hexapods and quadrupeds out there already but they do tend to be expensive. One to two thousand isn't unreasonable for these beauties. My goal is to work the low end of the spectrum and come in under $200 for a quadruped.
The Journey Begins - May 29, 2011
To keep the price down I have to skimp on the servos because there are so many of them. Cheap servos don't typically have a lot of torque. Also as the legs get longer the mass of the robot gains leverage and applies more and more strain on the joints and the servos. This means keeping the size of the robot small and the weight low.
Because of the extremely low price ($2.77 as of May 2011) I've decided to use the TPro SG90 servos in the design. They claim to produce a torque in excess of 1Kg/cm. They weight about 9g each.
The body of the robot will be made from 1/8" MDF. It comes in at about 0.22g/sq.cm. The design I've come up with uses about 400 sq.cm of MDF and contributes about 88g.
For power, I'm planning on using 2 or 3 Li Ion batteries. I salvaged some US18650GR 2400 mAH Li Ion cells from a discarded SONY laptop. Although a few cells in the pack were damaged most were OK. I plan to use a 10A 6V buck regulator to deliver 6V to the servos. Each cell weights about 46g.
Inverse kinematics calculations can eat up a lot of processing power. Since I'm hoping to experiment with different schemes for generating gaits I don't want to be hobbled by lack of cycles. So I plan to give each leg its own DSPIC33FJ64GP802 40MIPS processor. These will in turn communicate via I2C with a dsPIC33FJ128GS804 40MIPS processor which will serve as the central brain. The leg dsPICs are about $5 in small quantities whereas the brain dsPIC is about $8.
Although the goal is for the robot to be complete autonomous I do plan to include a PIC18F2553 for USB support. This will be used during setup, calibration and testing. If all goes well I have some small 2.4GHz wireless cards I can add for remote monitoring or control.