tommytwoeyes's profile - activity

I'm an engineering student (sophomore / Calculus II level) with a plan to get a Master's or PhD in robotics, ultimately.

I got started in this after being introduced to Arduino and the maker scene, but my real passion is robotics. I'd like to begin learning ROS, but I don't understand the coordinate frames, transform frames, and other advanced stuff.

Is it still possible to learn ROS without an understanding of the more advanced mathematics and physics / kinematics concepts?

Hello all,

Please forgive my ignorance. This is my first time building a robot using ROS, and I'm really, really eager to get it going.

I'm using the Wild Thumper 6WD platform/chassis with differential drive. Another user has just informed me that ROS will not work—that I won't be able to proceed with my robot build (as is my understanding of the implications of his statements)—without wheel encoders?

Is this true? Is there any part of the build process I can work on without having wheel encoders? The Wild Thumper chassis does not come with encoders out of the box, although they are available from Pololu. However the chassis itself was pretty expensive, for my student budget, ($250 usd) and while I was already planning to get a pair of those encoders, it will be several weeks, if not months, before I can afford to do so.

Please, any advice about how to proceed with other parts of the build that aren't blocked by a lack of wheel encoders (if it is, in fact, true that one can't build a working ROS robot without them), or any advice regarding work-arounds for this issue (if it is an issue), would be very much appreciated. Thank you in advance!

Is it necessary to have the same ROS version installed on your workstation as is on your robot?

@ahendrix Based on all the comments and replies, it sounds like it's worth it for sure. I'm now beginning to think the reason I had so much trouble getting ROS working properly on my Pi 2 was because of the Class 4 card I was using. Thanks everyone.

Thanks man. I just read the IEEE Spectrum article from OSRF. How awesome is it that they've designed it to be both as affordable as possible (you can virtually print one, sounds like?!) & as versatile as possible? Anyway, it's nice to know where my tax refund is going. I was going to buy an iphone.

@gvdhoorn Not sure how much credibility to give this, but a couple posts on the RPi forum and Adafruit said the I/O rate designated by the SD card "Class" denotes only write speed, and that write speed is inversely proportional to read speed; e.g. Class 10 card may not read as quickly as Class 6

Really? Sweet. Sure sounds adequate, then. I'm almost more excited about the Turtlebot 3! This is the first I'm hearing about it. Any idea about a release date? (ballpark?) :)

@tfoote than following the tutorials, a book or two; DC motor controller & servo the only actuators; ultrasonic distance sensor or two, would you think an 8GB Class SD card would be adequate? I'll upgrade later; I'm just really eager to get started, but on a very tight student budget. Thanks again!

@tfoote Thanks! I'm barely past the beginner stage in my ROS experience, so any packages I build for my Wild Thumper-based, rover-type robot will be quite rudimentary for some time, I'm sure. I may log bag files for debugging. Can I ask: if you were in my shoes, doing nothing more intense with ROS

@ChriMo p.s. Why do you prefer the 12V NUCs?

Too bad I missed the special price on the RealSense kit. How much was it before Dec 30, 2016? It's $349 now. I was aiming for < $100. The UP Board is enticing, though. One has 2GB of RAM for $99. Thanks

Wow, yeah - those look really nice. A quad-core with about 4GB of RAM is what I'd think would be a decent single-board-computer platform for a ROS robot. Those are the specs that the netbook laptop that came with my Turtlebot had, anyway.

Thanks - that's awesome! I got a good deal on a TurtleBot v1 but haven't played with it because of this UGV project. It has an older model Create, but maybe the API won't be too different. Really cool. Maybe it's more feasible than I thought it would be.

toward the DIY PC and media center market. I didn't see a lot of Intel SBCs which looked like they were designed for hardware hacking (though that's changing, I think). Don't know much about the NUCs though. I'll check them out. Do you have any favorite models?

I tried getting up & running with a RPi 2, but it was painful. It clearly can be done, but I'd rather spend what time I have learning ROS. I looked for Intel-based solutions but didn't find one in my $ range. Also, how many GPIO pins do they have generally? The ones I found seemed to be geared

Thanks. I figured it must have some limitations in terms of processor-intensive tasks, or all the robots would have chucked their laptops over for SBCs. That'll be enough for me to get started and learn about for a while anyway.

I'm sure I/O speed is a factor, of course; but I'm trying to determine if the difference in Micro SD card Class (used to designate write speed only), say Class 4 vs Class 10, would have a significant impact (such as to be an impede robot operation) on ROS performance. I'm sure ROS does some writing, but from what (little) I know, it seems most I/O is conducted over a network. I'm not sure what a ballpark ratio of reading to writing during ROS runtime might be.

Does anyone know if saving a few bucks by buying a Class 4 or Class 6 MicroSD card would be a mistake? Some people say Class 4 is actually faster in applications with a high read-to-write ratio, because the faster write speed of the later Class 10 comes at the cost of slower read speeds. I don't know if that's true or accurate; sounds a bit unlikely to me.

Thanks

I'm a ROS beginner, and am also pretty n00b with respect to robotics in general. I'm really into it but with limited time for my hobby, I'm making slow progress.

I have a UGV-type robot project I'm working on. I have a platform/base (Wild Thumper), motor controller, ATmega and ATtiny MCUs for sensor/actuator interfacing, but haven't yet decided which computer will be best for central processing.

I'd really like to use a single-board computer, with the small size and cost, naturally. Clearly, lots of people have been using the Raspberry Pi for quite some time now, but with my limited experience, I'm having difficulty determining if it has sufficient resources.

I'm preparing to order a Pi 3, but just realized the Raspberry Pi foundation has announced all previous upgrade releases on Feb. 29. So maybe I should wait for a Pi 4, I thought. Some people speculate that the Raspberry Pi won't release as many significant updates this year, though, with upgrades aimed at industrial customers.

So, my bottom line: does the Pi 3 have sufficient resources for running ROS effectively? I realize some of the more resource-intensive components (rviz, maybe?) might not feasibly run on it, but otherwise is it a do-able? If so, what limitations can be expected? All OS compatibility/dependencies/etc issues aside, what limitations are imposed by running on a relatively meager platform (with respect to a laptop; no disrespect to the Pi!)?

I tried initially to use ROS on a Radxa Rock, which had a 1.8GHz quad-core ARM CPU and 2GB of RAM. Unfortunately, I had some difficulty installing and updating the OS, and support for it appears to have waned. I haven't found any other SBC more powerful than the Raspberry Pi within my budget, and the support available global Pi community is an enormous advantage.

Thanks!

I'm an amateur/hobby roboticist who loves robots but has only built a couple of simple, fairly dumb Arduino- and AVR-based robots that can avoid obstacles (most of the time), wander around aimlessly, etc.

I'd like to advance my skills, and learning to develop in ROS is an obvious place to start. I do have software development experience as a web developer (+12 years), and some experience with Python.

My robot will be an autonomous rover based on the 6WD Wild Thumper (34:1 gear ratio) chassis, using either a Radxa Rock Pro or Raspberry Pi 2 (haven't decided which yet), with an AVR ATmega1284 (possibly more than one) working as an intermediary to control the motors and possibly other peripherals, such as sensors, Bluetooth comms, etc.

My question is, I'm not sure of the best way to get started with ROS. There seem to be multiple good entry points to working with ROS; e.g. the ROS tutorials, of course. But while googling ROS tutorials today, I stumbled onto the MoveIt! project, which I had not heard of before.

The MoveIt! FAQ page said it MoveIt! is quickly becoming the preferred entry point to ROS development (for n00bs like me, presumably). However, some of the pictures and text on the home page cause me to wonder if the MoveIt! project isn't intended for mobile-manipulator-type robots, like stationary robotic arms and such. On the other hand, it did include robots like PR2 and other mobile robots in its list of supported robots. So, I guess, in short, I'm asking if the MoveIt! project would be appropriate and/or ideal for a complete ROS beginner.

Secondly, I wanted to ask whether diving into a somewhat complex project like the one I described above with the Wild Thumper chassis, with multiple devices involved, is the wisest approach, or whether it might be better to begin with a fairly simple project, like taking my new Raspberry Pi 2, installing ROS on it, and interfacing it with a single sensor, or something like a Bluetooth module. Is it reasonable to use a mobile rover-type robot project as your first intro to ROS (or MoveIt!), or would it be better to start small?

I've been wanting to learn ROS for a long time and just reached a long break in school (I'm a pre-engineering undergrad student, planning to major in electrical engineering), during which time I'll have almost 4 weeks of uninterrupted time in which to get comfortable with ROS programming.

Any advice, links, recommendations, etc., are quite welcome. Thank you!