Rabble of Robots

Hello – long time no talk! That’s on me… I failed at rule one of putting together a blog which is really “just write the damn thing”. (really the only rule)

I’m two months into living here and officially two weeks into grad school. Part of the reason it has taken me so long to write this is because so much has happened! And every single thing that happened put another thing on the “add it to the next blog post” list, which just grew to a daunting size. Excuses, excuses.

So here’s the quick rundown of everything that I’ve been up to (for starting grad school). There’s also a soon-to-be-published “personal adventures” post!

Getting Started

Coming in, I had a plan – move in to my new house, get settled, and still have a few weeks to get a head start on my Ph.D. before classes began. The first part of my plan went off well enough – I was all moved in and (mostly) unpacked by early August. My bed was set up, clothes and kitchen unpacked, photos up on the walls, and restaurants, bars, and grocery stores figured out (it’s a pretty small town). Suddenly I was staring down the barrel of a big, amorphous, vague goal of “get started on my Ph.D.” with no clue how to do that. Thankfully, I had made a few contacts with professors in my prior visits, so I started emailing them and setting up meetings. The long-story-short of that was I had to start figuring out who I wanted to work with and what kind of projects interested me.

Most graduate students are tied to a specific professor (adviser) and research project from day 1. OSU robotics is fairly unique in that they allow students who don’t have an advisor to do a rotation their first year, which is what I’m doing. This means that I rotate through a 2-3 professors to figure out where I fit and what kind of research I really want to do.

Step one was to find a few people whose research was of interest to me. This is where the internet comes in handy – I basically had to professional-cyber-stalk every robotics faculty member to figure out what they have done in the last 5-10 years and what they plan on doing in the next 5-10. This means reading a lot of research papers and (sometimes out-of-date) websites to see what sparked my interest (reading quickly turns into a theme in grad school, I’ve learned). This process forced me to really sit down, consider, and define what’s important to me. Essentially, I want to…

• Create robots that have a positive impact
• Understand how people interact with robots
• Study the intersection of robotics and humanity

The last two are encompassed by a field called Human-Robotic Interaction (HRI). I am also interested in multi-agent systems, which is multiple robots that have different functionalities to working together as a group.

From that, I narrowed it down to four faculty, all of whom work on HRI in various capacities:

• Bill Smart: focuses on HRI. Specifically, improving interactions between people and roots, enabling robots to be self-sufficient for weeks / months, and determining how they can be used as personal assistants for people with severe motor disabilities
• Kagan Tumer: Works primarily on multi-agent systems. Also works on HRI and is interested in how we get robots to make ethical decisions.
• Geoff Hollinger: Mostly works with oceanographers on marine robots. This includes a touch of HRI and multi-agent systems.
• Julie Adams: works on swarms (many homogenous simple robots working together) and HRI

One piece of advice I have both received and read time and again is when choosing an adviser, the most important things are working style and temperament. It is far more important to find someone with whom you can work well than it is to find someone whose research exactly matches your interests. Research and projects can change over the course of a Ph.D., but a bad relationship is much more definitive.

After weeks of meetings, discussions, ponderings, and emails, I decided to work with Geoff, Bill, and Kagan. I’m working with Geoff first quarter on a project studying how to integrate an operator’s preferences into a robot’s decision making model. The goal is to allow the operator to be hands-off and still have the robot make the right cost / benefit decisions in an uncertain environment.

So that’s the research part of it. In spite of a Ph.D. being all about research, there’s a lot more that goes into it.

Classes

Generally the rule of thumb with Ph.D. students is you take two classes per quarter and fill the rest of your credits with thesis (up to 12 units). Ph.D. is for research. The classes you take should directly support whatever research you want to do or field you want to pursue. Nobody cares about your grades; they care that you learned what you were supposed to learn. This means that RESEARCH is the top priority, and classes come second to that. If you have a research and a class deadline on the same day, research always wins.

Classes started last week, and I was shocked that I’m actually interested and excited to go to class. I am engaged with the material and excited to work on the projects. Strange. Apparently that’s what happens when you’re passionate about the subject.

The two classes I’m taking this quarter are:

• Intro to Robotics – basically a broad-spectrum, very high-level view of robotics, the program at OSU, and skills we will need to succeed in the program. I really like how they structured this class, because it’s not all sensors-actuators-code (though there is that). Over the quarter, each faculty member comes in and does an hour talk on how they got here, what they struggled with, and how to succeed. The class also goes through ethics in robotics (though there are full classes focused solely on this), how to create a professional online presence, and other soft skills.
• Autonomous Agents and Multi-Agent Systems – (Multi-agent system: how to get multiple robots that have different functionalities to work together as a group) this class is SO COOL. It’s two lectures a week, a few homework assignments (though they take 15 – 20 hours apiece), and a quarter-long group project. Generally this class is a ton of work, but the material and format are fantastic. The very first lecture in this class really highlighted the difference between undergrad and grad school – the professors admit they don’t have all the right answers. As we were going through the class, we started debating the definitions / answers the professor was giving. After the class, he sent out updated slides because we had collectively decided that his answers were incorrect.

Everything else

The first two weeks have been a struggle for many reasons. One of the big reasons is start-up cost – there are so many things that I have to learn, and fast, outside of the material. I knew going in that I was going to be at a disadvantage at the start, given that I haven’t touched code in awhile or ever done research. But knowing something and experiencing it are different beasts.

In the first two weeks, I’ve had to:

• Figure out how to dual-boot Ubuntu, which crashed Windows and forced a factory reset. About a week after getting it stabilized, Windows completely crashed again due to automatic updates. I have yet to fix that problem… (If you don’t know what this means, basically I had to install a second operating system alongside Windows, and it did the software equivalent of exploding)
• Install and train myself to use the Robot Operating System (ROS) – a system that allows you to do a lot of robotic simulations and run code for robots. I’m still figuring out what it does. My goal for this weekend is to run through some tutorials.
• Learn Python (programming language… not the snake)
• Understand how to read research papers and start becoming a technical writer.
• How to use LaTeX (this is a technical document preparation system. Basically it’s like putting together a word doc, but instead of the built-in formatting, you use code)
• Do about 15 hours of online training for how to handle human trials. There are a lot of safety regulations in place and the dreaded IRB (Institutional Review Board) that put checks on any research that involves humans, including if you’re just going to ask a bunch of people a couple of questions. The purpose of this is to avoid things like the Stanford Prison Experiment.
• Figure out a system for cataloguing my references. I’ve been told that from the very start, I should be keeping track of every research paper I have read, including my own notes on the paper. Thankfully, we live in the age of amazing software that can do most of the heavy lifting for us. Now I just have to figure out how to use the software.
• Figure out, institutionally, what it means to be a grad student. We’re in this weird grey area between employee and student – it’s not straightforward figuring out which set of rules apply in what situations. It all depends on the day, department, set of rules, position of the moon, and number of cups of coffee collectively consumed on campus that day.
• Figure out seemingly straightforward logistical things, like how to get a desk. The program is still very new (only ~4 years old), so there aren’t many formal processes in place for assigning desks, getting a computer, etc.

How is it going? 

It has been a really tough couple of weeks. There is an overwhelming amount of work, projects, and expectations. I’ve already cried (which I almost never do) and panicked (which I do much more frequently). However, I have not ever, for one second, regretted or doubted my decision to pursue this. I was on the phone with my dad tonight talking about all the things that have made it tough, and he pointed out that in spite of everything, I had passion and enthusiasm in my voice. I was excited by all of it, even the overwhelming work and really hard problems. It’s true – I’m really excited to be here and I get little flutters every time I walk in the giant lab with various robots scattered everywhere in various states of moving and/or disrepair. This thing that I have the privilege of doing is so insanely cool.