Planar Impact Arena
My contribution to the project was in the experimental design, mechanical design and fabrication of the arena. There were a few interesting challenges in designing the arena. First, the object needed to be as close to free-fall as possible but remain in the plane. For this reason, we spaced the glass panes with precision cut aluminum bars and faced the objects with teflon to reduce friction when objects did touch.
Designing the magnetic latch was really interesting because as the robot attracts the object, normal force on the glass pane increases, thus increasing frictional force. This caused the object to drop off when the resulting frictional force (plus gravity) was greater than the upward component of the magnetic force. We added series elasticity to the end of our position-controlled arm so it could maintain contact with the surface, which increases magnetic force and frictional force equally but does not increase gravity. With this, we were able to lift objects and automatically reset them.
This was a lesson in system wear. When you impart energy on rigid bodies thousands of times, you are essentially fatigue testing them. Objects wore fairly quickly (3d printed PLA) against the aluminum base. However, if we made the objects out of aluminum, the base and the objects would wear each other, making them both require replacement. To continue rigid body experiments, I would re-design the contact area of the shape to distribute load more gracefully. As low as the friction is on this arena, there is still (almost imperceptible) frictional loss due to intermittent contact. To make it even more frictionless, we could tilt the setup off from vertical and levitate objects like air hockey pucks. Gravity would still apply but at a fraction of its normal constant. Lessons I learned from the project (improvements on the final product or the process). Skills I learned or flexed from doing it.
Devoted to the study of state-estimation, my lab-mate, Nima Fazeli, was interested in recovering inertia and coefficient of restitution from visually observing impact. This idea evolved into the evaluation of well-known rigid-body contact models by fitting experimental data collected by us. To enable this, we created the "planar impact arena" where objects are sandwiched between two glass sheets with small clearance, picked magnetically with a 6DoF robot arm, and thrown. A Gamma ATI force-torque sensor registers impact force data and the objects are tracked with the VICON IR parallax tracking system. Because the robot can automatically reset the experiment, we can gather tens of thousands of trials with minimal effort.