One of the challenges of using a belt drive system that is mostly FDM 3d printed is the fine line between too tight and too loose. The pulley sprockets have been iterated over and over again to provide maximum grip, and will likely become milled out of acetal copolymer, but for now they are more than sufficient. This leaves the tensioning method.
Since MIT Mini Cheetah is a good baseline, I basically just copied the adjustable roller arm idea, and 3d printed them out. This works for a short period of time, but as you can see in the photo they begin to bend after a few cycles of driving the belt. This would require constantly adjusting the set screws on the outside.
Since I didn’t have any suitable aluminum stock on hand, I decided to machine the same design out of some Delrin/acetal. This proved to work pretty well, and did not bend. Here’s the first test rollers out of the material:
Since this was just a test, I left cleanup for later, and installed them in the legs for a few days to see how tolerant they were to deflection.
Eventually the walls on the bearing-end of these rollers started to crack, I’m not 100% sure this was due to design or my hastily machined parts, so I decided to not take any chances with this extremely crucial part, and went ahead to redesign them with different bearings and thicker walls.
One of the challenges with any walking robot, is repeatability. If the belt tension is constantly changing then foot position placement will be very difficult, this is why I went through an iterative process for a seemingly simple mechanism. In fact, I’m still not totally happy with them but they should hold up well for now.
After doing all this disassembly and reassembly I decided for now to reprint the outer leg casing to include a removable ‘sight glass’ area to keep an eye on the rollers moving forward. If everything stays consistent I’ll probably reprint again to hide this stuff: