The FB-V12 or “Featherless Biped (Version 12)” is a prototype walking robot I developed while doing research on robotics and control systems at Columbia University in New York. The development of this robot took 4 months to complete with over 800 hours spent on design, production, and testing. The design of this robot utilized artificial intelligence via evolutionary algorithms to achieve extreme weight reduction and structural support.
Gaining inspiration from “wind walking” wooden kinetic sculptures, I utilized a novel robotic linkage system which provided built-in stability from only 2 motors. I began the design of the FB-V12 with over 200 sketches, experimenting with 20 different walking design configurations. I also brainstormed 300 novel ideas to incorporate into this robot to provide unique stabilization techniques or other experimental robotic elements. I settled on a a particular system out of 3 I narrowed down. I created a specific 3D design with detailed traditional sketches, clay modeling, and computer aided design software.  I designed each element of this robot for extreme modularity and traditional manufacturability, meaning that different components had multiple uses and count be configured in many different ways and this base design could be manufactured using traditional mill/lathe/CNC machines. Next, I took each element I designed and optimized the morphology via the Altair CAD suite. This design phase took nearly 3 months. The assembly consisted of nearly 100 unique components. I rapidly prototyped each component using a FDM 3D printer, printing and testing each component as I designed the system. 
I utilized PLA plastic and ABS plastic with additional experimental 3D printing methods, such as utilizing steel rebar in my prints. Overall, printing and manufacturing all parts took over 400 hours. I assembled the system with COTS fastening and spring components. I also utilized 6 bus servo motors, with 4 for locomotion and 2 for stabilization. I soldered and integrated a converter, controller, and microprocessor. Using Python and libraries used to control the motors, I coded a GUI and walking patterns to control the robot. I experimented with multiple gait methods and walking frequencies. I also experimented with the robot simulation library PyBullet to teach the robot how to walk, algorithmically. 
Ultimately, this project was a success. The main improvements that could be integrated into this design are improved 3D printing parameters and improved materials. Some of the early components I printed failed structurally when tested, and lessons I learned throughout this production process could be used to improve this system. 
Back to Top