Yiwen Wu
Engineer · Designer · Baker · Traveler
Project Homepage
Electrical System
Software System
Autonomous Robot Mechanical System
Mechanical Design
While designing the mechanical systems for our robot, we fully embraced the motto of “Fail early, fail often.” We went through four iterations of our shooter and three iterations of our main chassis. While assembling a new version of a component, we would immediately start a list of improvements for the next rebuild. Our chassis remained pretty similar from version one to version three, with additional mounting holes and assembly features added for ease of integration. We also tried to drop components lower and lower on each rebuild so that we would have enough space above the main deck of our robot to load balls in at 9 inches, load them into our shooter, and fire them.
Chassis and Drivetrain
We found ourselves lacking some space at times due to the circular construction of our robot, but we felt that the circular design definitely still paid off in the form of mobility. Our bot could never get itself stuck in a corner or against a flat wall, because it always had the ability to rotate its way out. One key change we made from chassis V1 to V2 was to make each of the two motor/drivetrain assemblies fully modular. We were able to remove the drivetrain to service it without having to deconstruct the entire upper portion of the robot. In our drivetrain, we used all metal pillow blocks, aluminum flex couplers, aluminum skate wheel adapters, and black rubber skate wheels in order to make the system as robust as possible. We did not want to run into issues with our drivetrain while simultaneously debugging electronics and software. We made sure to thread-lock all of our set screws and critical metal screw/nut interactions, so that they would not loosen up during the competition.
Shooter
Our shooting device started life as a flywheel shooter (pitching machine style). We wanted to standardize the process of loading balls as much as possible, so we designed a ramp loading system that would release one ball at a time down ramped rails into the flywheel shooter. We experimented with having the flywheel compress the balls against a flat surface, a continuation of the rails, as well as a v-shaped channel. We also added variable compression on the flywheel shooter in order to experiment with different amounts of compression. Despite these changes over the course of 3 iterations, we found that the flywheel shooter just was not accurate enough for this year’s 218b game. There were too many inconsistencies in ball shape/hardness and flywheel shape/hardness. Even with our plans to encode the flywheel to implement closed-loop speed control, we were not confident we could make the design sufficiently repeatable.
Initial prototypes were flywheel shooters with ramped ball-feeders
Luckily, we had begun prototyping shooters the first week of the project, so despite the fact that we made so many prototypes, the numerous iterations did not push the project schedule. After quickly prototyping with a plastic ruler and a rolled up piece of cardboard to cradle the ball, we found that we could make a very repeatable device by making our shooter much less sensitive to variations in preload distance. The basic idea in our design is to bend back a ruler with a servo arm until the arm loses contact and releases the now spring-loaded ruler. By using a plastic ruler, rather than a stiffer metal one, and extending the ruler out very far past the fulcrum, we were able to tune the system so that large variations in preload distance were required in order to change the landing zone of the projectile by a sizable amount. As long as our servo released the ruler in roughly the same spot each time, we knew the ball would land at almost the exact same distance each shot.
Beacon Alignment
The main part of our design that limited the accuracy of our shooter was our alignment system. The IR sensing worked great both during the grading session and during the competition, but we relied on a piece of heat-shrink tubing placed around the IR photodiode to narrow the window in which it was able to see the goal beacons. We could control the angular alignment of our shooter by slightly rotating the photodiode/heat-shrink assembly either direction. During the first round of the competition, our alignment was great, and we made two shots in a row, but in subsequent rounds, we kept misaligning and think the heat-shrink may have been knocked to the side. A more robust alignment system would have greatly improved our accuracy.