Final Design
Our final design incorporates all of our lessons learned, including
- our final motor gearbox with 36.75:1 reduction,
- our original Chebyshev mechanism,
- strengthened legs,
- timing belts with manual tensioners, and
- our v4 counterweight.
|
|
|
|
Due to the extremely compact and feature-dense nature of the machine, drawing and analyzing every single piece of the robot in CAD was essential to the success of the project. Every single screw, washer, and belt was modeled and analyzed for interference and motion, allowing us to find and prevent many problems before fabrication.
Our final Solidworks CAD files are available here: Chebyshev Bipedal Walking Robot SLDWRKS.zip
and printable STL files are available here: Chebyshev Bipedal Walking Robot STL.zip
Final Construction
This combination of features resulted in a machine that can walk smoothly and reliably, typically traveling 10-20 steps between falls.
Improvements
At this point the limiting factor on walking stability and reliability appears to be the magnitude of dynamic disturbance caused by the counterweight movement. When the counterweight shifts from the lifting leg to the grounding leg, it accelerates continuously until it collides with the counterweight arm atop the grounding leg, and typically this quick transfer of inertia causes the robot to lift off the ground in an uncontrolled manner, rotating along the outer edge of the grounding foot. The impact force could be decreased by adding springs to the ends of the counterweight slider shaft so that the counterweight contacts said springs before colliding with the counterweight arm, though this could also produce underdamped second order dynamics that make balancing more complicated. A proper spring would have to be selected carefully.
It would not suffice to merely decrease the counterweight mass. The current weight is required to maintain balance with the current robot geometry. A lighter mass could be used if the counterweight shaft were longer, increasing the moment arm length about the edge of the support polygon, but this would mean that the counterweight travel time is longer, meaning the legs would have to slow down, and the legs are already fairly slow. The counterweight shaft angle also could not be reduced, as it is already near the minimum angle required to overcome static friction reliably and decreasing angle would also increase travel time. It might be possible to change the counterweight trajectory, to do something clever with a non-linear counterweight movement. Similar to how a brachistochrone is more efficient than a linear slope, it might be possible to find a curve that reduces our transit time and reduces impact force, but moving the counterweight along a specific curve instead of allowing it to slide along a straight shaft due to gravity is a very messy mechanical problem and we did not have time to address it.