My Master of Engineering included a year-long capstone project. I spent the 2017-18 academic year working on developing a robust 6-bar sphere that could sustain a 10 meter fall from a drone, then walk away from it. The lab building these prototypes was quite large, and they had been doing it for awhile, including working with NASA to build large versions. They had built passive spheres that could survive high drops, and spheres that could walk, but not spheres that could do both. That was what I worked on.
My team of 3 included 2 other students who weren't ready to get started on the project so early in the school year, so much of the design and build I did myself. In order to familiarize myself with the dynamics of these spheres during a fall, (and to try out a build technique) I built a passive bot that I could progressively add springs to while dropping it from increasing heights. Here is my first passive prototype. It was approximately 500 mm in diameter, with a 1 kg payload in the center. The bars are aluminum tubing, and the tensile elements are UHMWPE string.
The way the tensile elements (spectra string in this bot) connect to the ends of the rods took some trial and error. There needed to be a solid place to connect the string, while providing space for a bumper to act as a foot and a damper in the fall scenario. The feet on the left are on the passive version, the feet on the right are what we ended up using for the mobile one.
Assembling the bot was extremely difficult due to the need to set up strings while the rods are in position, so I made these jigs that could hold the rods in position during assembly.Â
After testing I realized we would benefit from progressive springs; so we switched to latex tubes as their spring constant increases during elongation. This is the proto where we were testing different spring lengths.
When I landed on latex tubes as the springs, I had to devise a robust way to hold the end of the tube with a string without cutting the tube during a large extension, and while using as little distance as possible (to preserve range of motion for the strings.
The robot "walks" by pulling the strings sequentially in a specific pattern that enables the sphere to roll. I decided to go with some small geared DC motors that I mounted onto a little spool assembly. I waterjet some brackets out of steel sheetmetal, bent them up, pressed some bearings in, and assembed some bevel gears and rods in. These mounted to the aluminum tubes via a little aluminum block I machined.
The black shield acts as a guide for the string to spool in a consistent way.
This is a force tester we used to characterize the spring constant throughout the range of motion. This way we could calculate the approximate length of the tensile elements using only the rotation of the motors. This ended up being rather accurate and repeatable, but I imagine the degradation of the latex would lead to inaccuracies as the robot aged.
Motor control was performed by a teensy controlling a set of motor drivers on a custom board mounted in a lasercut wood enclosure that was suspended in the center of the sphere. It included batteries and a zigbee comms board so we could work on code dev remotely as it bumbled around the lab.
The final proto was dropped a few times from a building and by a drone from heights up to 10 meters. The robot survived the different drops with differing amounts of damage, including the control enclosure contacting the ground at the maximum drop height. The conclusion I came to was the spring stiffness required to keep the center enclosure from contacting the ground was too high for a reasonably sized motor to overcome while walking at a reasonable pace. This robot configuration might still have some promising characteristics for certain applications, but additional development is necessary.