This week I got the chance to observe and assist in a different project. A test subject came in to test a passive weightlifting ankle prosthesis, so I was able to see how the motion capture and force plate technology works. We tested the prosthesis with varying amounts of pressure while the tester walked on the force plates and did sets of ten squats. It was really interesting to see the motion capture technology. We placed small reflective dots on the tester's legs and hips and were able to see the dots move around the room on the screen. It was very cool.
TorsoBot has started showing successes. We began creating a series of code that would base the timing of the push-off leg on the while loops of our main program. After some trouble shooting and testing, we were able to get it functioning by itself. The next step was to integrate that program into the main one with the while loop. This took a long time. The counter in the program wasn't counting properly and our variables were not acting the way they should have. We eventually realized that we had defined our variables inside the while loop. By that I mean that every time the while loop repeated, the variables were redefined back to their original state. This tiny problem threw off the whole program and messed up the counters. After moving the variable definitions outside of the while loop, we had better luck.
However, new problems sprang up; the legs were uneven when they pushed off, the robot would lean backwards and fall, causing the wheels to spin rapidly in the air. Towards the beginning of the day, we tried to move the batteries to be attached to the bottom of the torso. We thought this would stop the tail from going under the torso and putting the whole robot off balance. After this, I taped a level to a meter stick and measured what angles would allow the legs to be even with each other. From this data, I found an equation that outputted the proper angle for one servo from the angle of another servo. After this, we were able to change the start and end angles of the push-off legs. This seemingly solved that problem. Yet, the new battery placement became an issue. As the robot moved forward and the push-off legs returned to start position, the tail piece pushed on the batteries. By pushing the batteries forward, the tail pushed the upper torso backward causing the robot to fall or go in the wrong direction as seen here:
I fixed this issue by moving the batteries to the front of the robot. This way, the batteries increase the weight on the front. In addition, I drew a little face on the front-facing foam piece to give it character and remind me which way the robot should be going.
When I went back to the lab the next day, I could not replicate the results of the previous day. So, I decided to test the robot at various desired speeds in order to find the perfect value that would get it to work. For some reason, in nearly every test, the robot fell over on the third step of the push-off leg. In order to fix this issue, I tried putting one battery on the front of the torso and one on the back. I then tested this new battery positions at different desired speeds. This battery position also failed. Then, we tested putting the batteries on either side near the wheels. This battery position seemed to work, but the robot began turning with uneven feet again, a glitch we thought we had already solved.
In order to fix this resurfacing problem, we tested different angles for the push-off legs. We had no success. It was time for my leg data to be put to use. We used the equation of the trend line to create around 94 data points ranging in between the start and end angles. Then, we took this data and created equations in our code for them. Next, we integrated these equations and variables in the parts of the code that moved the push-off leg. Our first test resulted in one push, then the push-off leg did not reset and the robot didn't move. Next week, we will work to fix this problem.
Comentários