Owen Gillespie

Mudd Escapes

During Spring of 2018, my roommate and I decided we wanted to create our own escape room. We enlisted some friends and designed, built, wired, and programmed the room just in time to open it for the week before finals. Almost a quarter of Mudders got to play the room during that week. The room had 20 puzzles distributed through 4 areas. Teams of 6 were given 75 minutes to complete it with a 25% completion rate. The networked puzzles in the room ran on ESP8266's with a flask backend, and the other digital puzzles ran on a variety of Arduino boards. The game master could also monitor and interact with the game state using a react frontend. More information about this project and future escape rooms on the Mudd Escapes Website

Augmented Reality Labyrinth

At the 5C Hackathon 2016, my suitemates and I threw together an augmented reality version of the classic game labyrinth. Players can draw their own maze, and then control a marble in a virtual copy of their maze by holding the paper under a web cam and tilting the paper. The control works using OpenCV to compute a homography between the flat position and the current frame, and then decomposes the matrix to find the pitch and roll. Check out the code on GitHub, or a demo on YouTube

Twing

At MuddHacks 2016, my team set up an IMU on an ESP8266 inside of the tire swing at south dorm. We streamed the data over wifi to another ESP hosting a server, and used that data to control a set of neopixels on the tree according to the motion of the tire. Check out a quick explanation of the setup on YouTube.

HuntMaster

At LAHacks 2017, my team created HuntMaster, a DSL for scavenger or puzzle hunts that can be converted into a website with interactive modules to guide players through the hunt. Each hunt is composed of a set of stages, which either provide information to the players or require some type of input to advance. HuntMaster currently supports text input, scanning QR codes, GPS location, pictures of a certain type of object, matching a specific picture, and more as different inputs required of the player to advance. It is also easily extensible if hunt creators want to add new ways for the players to interact with the game. Check out the code on GitHub and an incredibly high production value demo of a hunt on YouTube.

ESP Tutorial

Cole and I ran a session to teach other students at Mudd how to use the control an ESP8266 with Lua for basic IoT projects. We were thrilled with the turnout and that we saw several cool projects based on the skills that we taught. Check out the presentation here and the code samples that go with it on GitHub.

Scorpion

I participated in FRC throughout highschool on team 246 Overclocked, but our 2015 robot, Scorpion, specifically was by far one of the coolest FRC robots I've ever seen, and one of my favorite projects. Featuring a 6 foot long 3 joint arm, a 3-wheel holonomic swerve drive, and a folding chassis to meet the size restrictions of the game, it stood out whenever it was on the playing field. I personally worked primarily on the motion control for the arm and the drive train, which was used to great effect in the autonomous sequence at the beginning of the game and in almost entirely automating the control of the arm. Check out the codebase on GitHub and a video of its gameplay on YouTube.

Label, Segment, Featurize

While I was working at FeatureLabs, I contributed to this paper that creates a framework for automated prediction engineering. Here is the fulltext of the paper.