A borrowed microwave, a printed arm, and a great deal of orange tape.
A Pi Day hackathon sprint. Each cup is a pie assembled ahead of time — a crumb base of crushed crackers under pumpkin pie filling — so it only needs heating. Everything after that is the robot's job: the door, the cup, the start button, and the whipped cream on top.
The microwave is stock hardware with no electrical interface, so every input had to be made mechanically. Servos driven by an ESP32 work the door and the start button, and every bracket holding them — along with the fixture that clamps the whipped cream can upside down above the plate — is self-designed and 3D printed. The first door mechanism was a geared motor turning a printed spur gear against the door; it never produced the torque needed to break the latch, and that assembly is still taped to the top of the machine in the footage, no longer turning. What replaced it is the two-move open in step one: pull the handle until the latch pops, then push a clear acrylic bar taped upright along the door's top edge. The bar is a lever, and it buys back the reach and leverage a short arm does not have.
Plays each step in order.
1 Open two moves
The gripper pulls the handle until the latch pops, then reaches up and shoves the acrylic bar taped to the top of the door. The arm is too short to swing the door itself, so the door was given a longer handle.
Engineering
Two actuations, because one will not do. The latch needs a sharp pull right at the door's edge; the swing needs leverage from a point the arm cannot physically reach. The first attempt was a geared motor driving a printed spur gear, which never broke the latch, so the problem was moved from torque to geometry by fixing a rigid bar to the door and pushing that instead.
2 Load cup in
Cup off the top of the microwave and onto the turntable.
Engineering
A paper cup is an awkward payload: the gripper has to close hard enough to carry it and not hard enough to crumple it. The place is the harder half, since the cup has to go through a door opening and land flat on a turntable with the arm working near full extension.
3 Close door shut
The arm rises to near-vertical, reaches the top of the open door and pushes it until the latch catches.
Engineering
The same lever driven the other way. A latch only catches if the push lands near the door's free edge, so the arm has to reach past the hinge side rather than shove the middle of the door, where the force would just flex it.
detour Retry the tape gave up
The same close, twenty-three seconds earlier. The door swings almost shut, rebounds, and the tape holding the acrylic bar peels off. Someone re-tapes it and the next attempt holds.
Engineering
Worth showing because the failure is a fixturing problem, not a control one. The tape holding the lever peeled under repeated load and the bar slid down the door. The trajectory was fine; the mount was not, which is the failure mode you actually get on a hackathon timeline.
4 Start servo
A servo on a printed bracket presses the panel.
Engineering
The microwave has no electrical interface and was never going to be rewired, so pressing start stays a mechanical problem: an ESP32-driven servo on a printed bracket has to land on one membrane button, repeatably, from a fixed mount.
5 Unload cup out
Back out and onto the paper plate, which is parked under the nozzle for a reason.
Engineering
The pick is routine; the placement is the point. The cup has to come to rest under the dispenser nozzle, so this step's tolerance decides whether the next step can work at all.
6 Top whipped cream
The reason. The gripper tips the lever on an upside-down can and the cup fills from empty.
Engineering
Actuating a consumer aerosol valve with a robot gripper. The can is clamped inverted in a printed fixture so it will dispense at all, and the lever needs a steady push held for about a second rather than a tap.