Following the completion of the phase 1 design, a single leg module was manufactured to commence testing. Following this, using the data from our analysis, a set of redesigns were implemented to areas we found needed iteration. With these redesigns made, a full prototype was manufactured that, prior to the COVID-19 outbreak, was to be used for hollistic testing to assess the robot control system and inform further redesign. Unfortunately, hollistic testing was not possible before the shut down of university labs, however a full prototype was manufactured.
2. Single Leg
2.1 Single Leg Manufacture
To test the mechanical strength of our components, conduct temperature testing and visualise the control code in the robot for the first time, a single leg module was manufactured prior to the manufacture of the full robot.
The single leg was manufactured to meet the design as it was at the conclusion of the phase 1 design period.
2.2 Single Leg Testing
Using the test rig shown in the photo above, we ran a series of tests on the single leg. The first test measured both temperature and mechanical strength. Running the single leg through a squatting motion loop, we measured the temperature of the crucial electronics systems (motors, ODrive) to assess their temperature levels under load. After each cycle, a mass was added to the top of the module to increase the load of the robot. Through this testing, we saw a steady increase in the temperature of the electronics, but never as high as the upper threshold of 70 degrees we set prior to the test (the glass transition temperature of our PETG build material is 80 degrees).
At the higher loads, we found that the belts between the motors and concentric shafts were slipping slightly, a result of a lack of tension in the belts. This informed the first redesign point - the addition of belt tensioners. We also found that the design of our leg meant some bolts interefered at the bottom of the squat, the second redesign point was hence to alter the upper and lower leg designs to remove this interference.
Finally, when testing at higher speeds, we noticed some stuttering in the leg as well as latency between control commands. We suspected the origin of this to be the frequency of communcation between the python terminal on the laptop being used for control and the ODrive. Hence, we rewrote our control code into C++ and uploaded it to an Arduino Teensy, therein controlling the robot through this. After this change, there was no latency or stuttering and generally a smoother motion seen in the leg.
3. Phase 2
3.1 Phase 2 Redesign
Following the testing of the single leg, the following changes were made to the design of the quadruped, based on the obersvations made in testing.
The key change required after testing was the addition of tensioning to the belts, after we found there was slipping at higher loads. This was done in two parts, the first wsa tensioning to the motor belts via a new components fit onto the top of the motors.
This part added tension to the two belts, as well as adding an additional layer of rigidity through the axis of rotation of the motors. Upon its manufature and implementation into the single leg, we saw a clear improvement in the belt's tensioning, with no slipping occuring following repetitoin of the previous load testing.
The second belt tensioning addition coincided with the upper leg redesign. As well as adding tension to the femur belt, we needed to increase the range of motion of the lower leg. This was achieved by shortening the length of one side of the femur sandwich. Below you see a comparison of the leg before and after redesign; the femur plate has been shortened on one side, and a large bulge has been added to apply tension to the belt.
As you can see, the range of motion of the lower leg increased dramatically.
The final redesign point was an alteration to the lower leg design. This was done for both improvement to it's range of motion, for ease of manufacturing and a reduction in the wear of components. The aluminium tube arrangement designed in phase 1 was replaced by two carbon fibre panels, similar to the design of the upper leg. This change prompted a slight change in design to the foot to fit the new assembly.
This design change not only resulted in a sleeker design, it allows for the manufacture of the leg to be done entirely via computer controlled machinery (3D printing and water jet cutting) resulting in extremely precise parts versus the hand manufactured aluminium tubing from phase 1. To match this change, a small inset was designed into the foot such that the components still fit onto our existing knee pulley, and curvature was added to the top of the foot to further aid in the increase in range of motion (previously, the corner of the then square section of the foot interfered with the femur belt).
3.2 Full Prototype Assembly
Upon the completion of the previously described redesigns, we re-ran the tests conducted in phase 1 are were pleased to see that the changes made had solved the issues we were saw at the original testing stage. Hence, we were ready to assemble the full prototype.
Regretfully, we were unable to commence the hollistic testing we had planned due to the closure of the university labs, however, we were able to do some very basic testing of the prototype.
In the short time we had, we used a winch system to hang the robot and assess the full robot control code, visualising the different gait types we had designed as well as the robot's calibration sequence. We were unable to conduct a full walking test as we could not acquire a suitable power supply in time, however, we were able to do some low speed squat tests, which were succesful.
We opted to use aluminium tubing instead of carbon fibre as after analysis we found the physical properties of aluminium to be sufficient for our requirement, with the added benefit of recyclabilty over carbon fibre. Below is the FEA analysis we conducted on the assembly to confirm this change.
Upon the addition of a 50N force on the red shoulder module, aluminium tubing yielded a 0.9mm deflection at the corner, versus 0.6mm when using carbon fibre.
While the conclusion of this project was not as we had hoped, we are still proud of the design we have developed, and the prototype we manufactured. Our main goal this year was always to develop a strong foundation for future GDP groups to develop upon, and with the prototype platform we have finihsed this project with, next year's students have the opportunity to add more capabilities to the robot and continue its's development toward its eventual application - supporting astronauts on the moon.