The mechanical design and control systems of the robot were prototyped in CoppeliaSim, a robotics simulation software
This is an example of the early work done in CoppeliaSim to create a stable and adaptive ROS control system that lets the robot position its body at any height and angle, regardless of the terrain underneath it.
This simulation shows a much older version of the mechanical design, when the robot was much smaller. The sim work made me realize that the robot needed to be much lankier in proportion.
The first time my robot was able to climb stairs in simulation, with some of the control system visualized. It uses kinematics to calculate the center of mass, and uses it to shift its weight from the front to back legs during the stages of stair climbing.
I was experimenting with a lot of geometry changes, as I realized that the robot had to be much bigger to be able to climb the stairs. The 3D model did not catch up with the linkage size changes :).
This is the first truly usable attempt at stair climbing, after lots of work in iterating smaller linkage configurations (with the help of gradient descent), and improving the simulation to much more accurately simulate the movement of a mechanum wheel.
By positioning the wheels under the robot, it can move itself up and down.
The robot is able to produce almost 180 degrees of pitch.
This is the least favorable axis for this robot, as moving this axis too much moves the COM close to the side of the robot. However, it is helpful for leveling on rough terrain.
This version of the robot uses mecanum wheels to strafe on the ground and turn on the spot. It is a tradeoff of speed and rough terrain capabilities for more DOF.
(Yes I know half the wheels are backwards)
Three different sizes of robots were modeled using CAD to explore motor, controller, and battery pairings for various sizes of robots. The motors chosen for each design were selected based on calculations of motor specifications and robot weight/geometry.
While the differences in size may seem small, they have big implications on motor choice and the ability to climb stairs.
When fully sprawled, the robot is about two feet long. It fits on stairs by skipping a step. Because of its size, it can climb them quicker and easier than the other versions. However, because it is heavier and has longer arms, it needs bigger and more expensive electronics to drive.
The cooling system for this design is very interesting. The electronics box in the middle is pressurized by fresh air, which flows over the motor drivers and vents out through the motors themselves.
CAD:
The medium-sized model is shaped like its larger brother and uses many of the same components. it is a bit over a foot and a half fully spawled. By becoming smaller, this design is able to be lighter, and have smaller This design sits halfway between the small robot, which can fit on a single stair, and the larger robot, which can fit on two stairs at once.
CAD:
A foot and four inches at full spawl. This robot was made to be as small as it physically could while maintaining the ability to climb stairs. Based on dynamixels, this robot was designed to be easy to build as a first step in this project.
CAD: