Carly's Portfolio

Welcome to the portfolio I’ve put together for my Ideo CoLab 2020 Makeathon application. Links to my project portfolio, my design fiction, and some more information about me can be accessed by the links on the left-hand sidebar.

I volunteer at the free public makerspace Hatch run by the Watertown Library. I mostly help people learn how to use our 3D printers, laser cutter, and to do basic Arduino code. I wanted to develop a project to help people learn and develop all 3 skills. I competed in a local sumo bot tournament and spent $80 on my base kit and another $30 on additional sensors and modifications. The experience made me think that a cheaper kit that used the 3d printer and laser cutter to make parts would be a good project for beginners.

The design has undergone several iterations to make it more user friendly. Below is version 2 (left) and version 3 (right) of the chassis. Making the wire hole pass throughs made it easier to feed wires through, and gave more options for routing wires around sensors. Notches in the side of the body make it easier to fit and attach the top half.

In addition to the custom body, I designed a custom pcb board (left image) to drive the motors with a small microcontroller. However, there were some mistakes such as tying a sensor input to the same pin as the microcontroller’s led pin which caused strange and inconsistent behavior. Also the motor connections were not in the best place for routing wires, and it was very hard to debug. Based on my feedback, another volunteer did a new iteration of the board design (right image) that changed the routing and added leds that can be controlled via the code to act like “print statements” to indicate the internal state of the code so that the user can debug code without tethering the robot to a computer.

Mimikyu is a type of Pokemon that is very small and wears a Pikachu (a very popular Pokemon) disguise as seen in the gif below.

I am currently working on building a Mimikyu animatronic that will move and tilt its head to indicate simple emotions such as happiness, sadness, curiosity, etc. Mimikyu is a good base for a simple animatronic because it has no visible feet and moves with a gliding motion which means a wheeled drivetrain can be used for a base. It also has no (visible) arms or hands and does not speak, which greatly simplify animating and controlling the animatronic.

In the video above and in the images below, you can see my first prototype of the “neck” joint. The foam head represents the stuffed Pikachu head of Mimikyu’s disguise, which means the neck is actually the of Mimikyu’s head.

A universal joint consists of 2 hinges mounted at 90 degrees from each other. A universal joint allows pivoting in 2 planes, and also transmits rotation even when the joint is pivoted (see gif from Wikipedia below). In my mechanism, this acts as connection of the base of the skull to the spine. Two four-bar linkages are made up of the servo horn, to its arm, to the neck plate, to the servo mount. This acts as the front of the neck. As seen in the video above, by controlling the servos with an Arduino, these linkages can be manipulated to create looking up or looking down motions, as well as tilting the head to the side.

The shaft that connects to the universal joint will be connected to a third servo that will rotate the shaft, providing a twisting motion, e.g. looking over your shoulder.

Right now all my plates and linkages are created out of wood. This allows me to quickly modify lengths and mounting points to test how variations in the geometry affect the motion of the head. It is possible to model the assembly in CAD, but programs like Solidworks have a very hard time solving for all the different degrees of freedom which can cause the model to break (see below). This would be more suited to a program like Maya, however, I don’t have access to such tools. So wood assemblies are a simple and cost effective way to solve the same problem by trial and error.

The concept of a circular economy requires us to not just recycle, but also repair and re-use more items. However, it can be difficult for small local repair shops to stay in business when they are located in small suburban or rural areas. Repair meet-ups ask individuals to on yet another time consuming to-do task. Shipping products to a repair service reduces the environment benefits. 

This is my beloved Cuisinart 4 Slice Belgian Waffle Maker - Square. A marvelous appliance that has one critical flaw: a shrill, piercing buzzer alarm that sounds when the waffles are done. Particularly as I am a morning person and particular family members are not. They did not appreciate being awoken by a loud shrill alarm on Saturdays that were supposed to be their chance to sleep in late.

I examined the waffle makers for the hole pattern in the casing that indicated where the noise maker lived. Removing some screws, I was able to find the buzzer assembly. I purchased a very small toggle switch that would neatly fit into the plastic wall nearest the wire that ran into the buzzer’s breakout board. I used my Dremel to make a small hole to house the switch, press fit the switch into place, and soldered it into the buzzer’s wire. I used some hot glue for insulation and to help secure the switch in place. It very neatly fits in with the existing design and use of the waffle maker.

Now when family members desire a quiet and late Saturday morning, I can turn the buzzer off and make waffles silently without disturbing anyone and when mornings are busy and hectic I can turn the alarm back on so I can be in other rooms and not risk overcooking or burning the waffles. 

This shelf was designed without any plans after seeing online photos of people using brackets to mount shelves above their doorways. We live in a rental with plaster and lathe walls so we are not allowed to put any holes in the wall and therefore cannot use brackets. I designed this shelf to be leveraged around the door frame and rest on the molding to stay in place, avoiding any sort of attachment to the wall, similar to doorway chin-up bars.

A horizontal bar made of 2″x1″ wood that is shorter than the doorframe sits on top of the back molding and pressed into the back wall. A horizontal 2″x1″ wooden cross bar longer than the doorframe spans the front. The ends of it are pressed into the doorframe by the weight of the plants above. Vertical supports and metal shelf brackets extend from the wide cross bar, to support a thin plywood shelf. The vertical supports have small L brackets that hook onto the front molding. The shelf is very thin to minimize weight. Two thin bars extend down along the door frame at the ends of the wide cross bar to help provide additional support and resist the desire of the shelf to rotate around the doorframe and dump the plants onto the floor.

For my Major Qualifying Project (senior capstone) at WPI, I built a calligraphy robot that used a series elastic actuator (SEA) for force control and compliance. I designed, built, and wired the robotic system as well as programmed the force control. Two grad students programmed the position control. 

This project was based on my hobby at the time of doing Spencerian calligraphy, which has very dramatic variations in line widths. However, very flexible pen nibs are very springy in a way that was challenging to model. Using an SEA meant that the sensor feedback inherently incorporated the springiness of the nib, and mechanically decoupled the motor from the load which protected the pen nib from damage if something in the code went wrong and the robot tried to drive the pen hard into the writing surface. The main technical calculation was the choice of the spring constant for the springs in the SEA, and you can see if the video presentation how I used a force plate to determine the force range needed, and the choice of springs was derived from there.

Design:

The main design requirements were as follows

Fit an SEA between the motor and the pen

Include a linear senor and rotary on the end-effector for input into the force control algorithm

Must have very rigid motion with minimal slop or backlash

Support rapid iteration of different designs

The CAD model was made in Solidworks

This was intended as a proof of concept of the design and control architecture that would never be used beyond our student team, so there were no UI/UX requirements

Build:

Most of the custom parts that made up the end effector were 3d printed to aid rapid iteration

Threaded inserts were pressed into the ABS with a soldering iron. These acted as captive nuts for bolts, which made assembly/disassembly quick and repeatable.

The aluminum extrusion frame was cut on a horizontal bandsaw and faced off on the milling machine

The sensors were mostly soldered with a few crimped connectors to ease rapid assembly/disassembly

Program:

I programmed a force control algorithm that ran on an Arduino Uno board

The potentiometers were connected to a low noise 12-bit sampling board to reduce noise readings

The Uno connected to the sampling board over SPI

The program read in the values from the linear and rotary potentiometers to determine how much force was being applied to the nib

It used a PD (Proportional Derivative) control loop to maintain the desired force

 The Uno received desired forces from the main program (the position control programmed by 2 grad students) and sent back the force error as a handshake over Serial.