Designing and building a quadruped robot has scratched every facet of engineering for me. From the endless list of pivotal mechanical design choices, the perception/mapping, and the optimal control theory that enables efficient and robust motion, this is the culmination of everything I have learned as an engineering student to this point in my life. I believe that control theory should lead mechanical design decisions wherever physics and cost allow, every design decision along the way has been scrutinized to this philosophy, with the end goal of obtaining a form that minimizes the mathematical model mismatch at every opportunity.
Stress Analysis in Ansys on the chassis
Water jetting the 16ga mild steel chassis parts
MIG welded the thin 16ga steel chassis
Chassis sand blasted and prepped for paint
High strength enamel paint
Chassis
I've chosen to go with mild steel for the chassis of this robot build. This allowed me to create a chassis that was extremely rigid and lightweight by introducing sheet metal bending at strategic points and weight relieving with a water jet CNC. Choosing to weld the chassis eliminated the need for any fasteners, and I was able to learn to MIG thin sheet metal (which was a totally eye-opening experience lol.) The result is a chassis that weighs only 800g, and still maintains a large factor of safety for my build requirements. It also allows for nearly infinite cable routing options, as well as excellent passive heat dissipation which is crucial for housing a high energy density custom battery pack as well as the abduct/adduct drive train.Â
Leg Sub-System
In quadruped control, it's common practice to assume the legs have negligible inertial affects on the dynamics of the body, which allows us to greatly simplify the model of the robot. For this assumption to be valid, the legs must have the absolute minimum mass possible while withstanding constant stresses and impacts with the ground. The motors have the highest mass density of all components, and without a significant rise in cost, this is not a variable I can alter. To mitigate the inertial effects of these components, I house them in the shoulder region. The legs are then constructed of 3mm water jet twill carbon fiber for critical stress absorption, and PLA 3D prints for structure wherever possible. The result is an extremely rigid leg that weighs only 716g, where the majority of the mass is located in the shoulder. I also embed IMUs in the feet instead of traditionally chosen force sensors. With this decision, I hope to detect foot slip conditions sooner, allowing for easier fall recovery control.