Custom Planar Biped

Homebrewed Attempt at a Walking Robot:

After finishing my undergraduate degree at the University of Manitoba I began attempting to build another small biped. My intentions were to use this project as a means to learn more about legged locomotion, from design to controls. I was able to design, 3D print, assemble and program the robot to move around a bit before I started grad school where after it was put on indefinite hiatus.

At the time this was a major accomplishment for myself but looking back on this project it is very humbling to see just how little I knew at the time. I had a very limited budget, hence the small number of actuators. My guiding design principle was that I wanted the legs to be very light for a variety of reasons, the goal was to try and concentrate the mass of the robot up high so I googled around trying to find the coolest knee and hip mechanisms. Some take-aways of the whole project were:

• I was completely befuddled with how to have a light actuated ankle so I decided to make it entirely passive because I though springs were "really cool" and passive elements were somehow always good. I couldn't get enough tension on the springs so the feet ended up being very large which sort of contradicts the whole minimizing task space inertia goal which was the entire principle I was attempting to base the design around.

• I  was also completely unsure of how to couple the actuators together and ended up making these intersecting couplings that dramatically extended the length of the robot and were very hard to screw into place. This gave me a great appreciation for having to actually assemble something with your own two hands and how awful you can make that process if you don't think beforehand.

• Additionally because I was stumped on having an actuated ankle, for some reason I thought that the extra two actuators should be dedicated to hip ab/adduction. 

• Lastly I did not at all consider the packaging of where the electronics should go so I ended up designing a large tower that further extended the size of the robot.

In the end I was able to cobble something resembling a biped together and moved on to attempting to figure out its inverse kinematics (IK) so I could at least move the legs. At this point I had only seen IK in an undergrad course where we solve the equations by hand for simple planar manipulator arms, a closed loop system was something entirely different that I had to overcome. Having never even heard the term numerical programming or knowing what the field of optimization was, I instead chose to write out the elimination equations for each joint by hand and the typed everything into Wolfram-Alpha asking it to solve for the missing variables. What it spit out were unholy several hundred term long combination of trig functions, linear and various non-linear terms. Not being deterred that this was clearly the incorrect thing to do, I instead opted to painstaking type out the paragraph long functions, the relative size and impact of each individual term be damned. In the end however, I was able to move the legs through some very basic preprogrammed end effector trajectories that somewhat resembled what walking should look like while the robot was perched on a test stand.

This project really ignited my passion to better understand legged robotics and why everything I was doing was a complete disaster. While I am no designer I have always been very interested in how legged robots "should" be designed, in a way that maximizes both their dynamic capabilities as well as day-to-day usefulness. I am especially interested as a controls engineer in the tight coupling between design and controls, how certain robots are much more capable not just because they have better controls but superior design that lends itself well in conjunction to prevailing control techniques.

This project also cemented my love for controls in that for myself the most satisfying part of robotics is the act of getting something to move how you desire it to.

(The robot featured 6 Dynamixel Ax-12A robot actuators (3 per leg), and an Arbotix-M robot-controller)

Gesture Controlled Drawing Machine: Although my creations potential as a walking machine was something that would never be fulfilled, I was still able to get more usage out of the hardware. For my Introduction to Robotics (ROB514) final course project I used my biped to create a parallel 5-bar drawing machine. Due to no longer having access to a 3D printer I repurposed and modified some spare components to mixed success. The intention was to control the robot through simple finger gesture commands but the limitations of repurposing spare parts hindered the simple position control.