Advice for preparing for graduate school?

[ercus]

I'm new here but I know there have been several other posts asking what is the best grad school, etc., so I won't bore you with the same question.  Most topics have covered the desired experience and skills necessary to get into a grad school, but I'm more curious about the experience that I should have in order to succeed once I get there (assuming I get in). 

For some background, I'm entering my senior year as an electrical and computer engineering major, and somewhat recently became interested in robotics (mainly the AI/computer learning/control aspects).  I attend Loyola University Maryland, a good business school but with a small (but accredited) engineering program.  Unfortunately my curriculum hasn't offered a whole lot of useful experience in robotics.  The electrical concentration has focused more on signal processing, and the computer side has included some microprocessor/microcontrollers, but that's the only potentially useful thing that comes to mind.  The curriculum is kind of strange, so I don't even have much programming experience, although I have some experience in hardware description language, visual basic, and of course matlab, and some very limited experience on my own in java.  While I'm on the topic, is there any one language that I should spend my time learning/getting experience in?

I guess my overall question is, how much work do I have to do before considering a good graduate school in robotics?  I know no one can really tell me what I should do, but would I be crazy with this limited experience to apply to start grad school in 2012?  I realize I'm new to the field, but I'm willing to spend time learning on my own to make up for what my curriculum hasn't covered (but obviously I have a limited amount of time).  I don't mind going in with a little catching up to do, but I don't want to start if I clearly don't belong. 

I'm not too worried about qualifications for acceptance, I have a 3.92 gpa and two years of intern experience (although no published research), and I'm not looking at CMU or MIT.  However, I would like to go somewhere good, and I'm currently looking at Johns Hopkins, University of California - San Diego, University of Maryland, University of Pennsylvania.  I know I said this wasn't my focus, but does anyone have an idea of what my chances of getting into programs of this caliber would be?  Also, is it even a conceivable possibility to work full time while completing these programs?  I only ask because my (likely) employer offers tuition assistance for a masters degree, so that would be helpful, but obviously a good program is useless if I don't devote enough time to it.  Sorry for rambling about a topic that many others have asked about, but everyone here seems to know what they're talking about.  Any input would be appreciated.

Thanks.

[waltr]

The C programming language is the most common one used in embedded processors. So this one is a most know. Then it is always useful to know Assembler to some degree on a few different processors. This forces you to understand how a processor works at its lowest level.

As to filling in experience nothing beats actual job experience. Look for summer/part time electronic technician work. Be might be a bit boring at times but you will get to be hands on with real circuits and code. Try to find a company that makes equipment that uses embedded processors. They don't need to be 'robots' as any of the processor controlled systems share common basics with robot systems, the processor gets data from real-word sensors then an algorithm decides on output values. There are circuits to read those external sensors and circuits to control motors, solenoids, etc. All of this can be applied to robots. Automated assembly and processing equipment is really robotic in nature and there are many companies that build this type of equipment.

How you done any micro-processor and robot work on your own? If not then it is time you do. Even starting with the $50 robot project will teach you much and help bridge the gap between your theory course work and real circuits/code.

Do you know how to solder? Design, wire and code a simple micro-processor circuit that uses an AVR, PIC or ARM?
Can you read and understand a data sheets for logic chips, transistors, cap, resistors, ADCs, DACs etc?
These are all useful real-world skills.

It is possible to work full time while taking grad courses. Do expect to be busy and not have much of a social life (do you have a GF?). If working full time then you'll most likely be a part time student taking only a couple of courses per semester. Many people have done this.

Good luck and hope it all works out.

[mstacho]

I was in a similar situation when I started my descent into grad school.  My background is space engineering, so most of my stuff was in signal processing and communications, with a bunch of astrodynamics and control theory plugged in.  I didn't really find my undergrad education to offer much in the way of practical robotics experience, although of course we took the obligatory analog and digital electronics, I found that I could prove the stability of a PD controller but I couldn't tell you how to write one in C :-P

I started as an undergraduate research assistant in one of my professor's labs.  His was focused on control theory, so as an undergrad my job was to actually implement and test the algorithms.  I had no idea what i was doing, but after my four month coop, which was part time, I came out of there knowing how to do the basics.  Turned out I enjoyed it so much that I went back to work for him as a masters student, and did a really neat masters that was at the intersection of math, computer programming, and robotics.

When my masters was done, I still didn't feel that I had a great grasp of "robotics", but like you I knew I wanted to work in machine learning/control sort of stuff, so I found that the best way for me to approach that would be to do a PhD.  This is not the best choice for everyone, but it seemed to fit for me. 

I got into a PhD program in robotics, and my project involves machine learning.  But interestingly enough, my project is machine learning for robot grasping, which is essentially a glorified control scheme.  So I get to apply what I learned in my masters, see it in action, while also learning to build robots and work on the high and low level aspects of robotics.  I really love what I'm doing right now, by the by :-)

When I first arrived at my PhD, though, I didn't know half of what I know now.  Honestly, this forum has helped A LOT for building/coding robots, and my professors understood that my job was to learn -- I came in with knowledge in general, but my first year has been spent doing literature review, as well as building and testing a robotic hand.  I'm not yet even started into my research, but taking this first year to learn has made it so that when I do start my research in September, I'm going to hit the ground at full speed.

So don't worry about not knowing the exact stuff!  It's called grad SCHOOL for a reason.  But I'd also caution you to not jump into a field that you are woefully unprepared for.  I know for sure that I'm no mechanical engineer, so it'd be foolish for me to pursue grad studies in advanced robotic mechanical design!

MIKE