Turtlebot3 Adds 2 and 4GB RaspberryPi 4 Options

With all the discussions we have had about Pi4 vs Pi3 for GoPiGo3, it is significant that the Turtlebot3 ROS robot just added 2GB and 4GB RaspberryPi 4 options (+$10 or +$30) to the formerly RaspberryPi3B+ only configuration ($599) available for purchase.

They didn’t change the battery - even more seriously weak IMO at 11.1v 1800mAH, and they didn’t change the “About 2 hours 30 minutes” operating time specification shown on the website. The Pi4 power consumption is roughly 1.3 to 1.4 times that of the Pi3B+ so I will guess the operating time with the Pi4 options will be more like “About 1 hour 45 minutes”.

Dave isn’t smart enough to be jealous yet, and Carl is quietly acting smug with his “About 7 Hours 45 minute” playtime.


Wow. Those are pricey machines. Hard to understand the $900 difference between the Waffle and Burger. Granted, $400 of that is in the cost of the servos (and from the specs I’m not quite clear why there is a $200 difference in the per servo cost). But wouldn’t think the RPi camera and 16 extra plastic plates would add up to another $500. GoPiGo3 is clearly the winner here on the value side.


Double the torque, five times the cost per servo - what a deal.

There is also a laser range finder ($30 on GoPiGo), and a remote control with usb bluetooth interface that they want $80 for, and the $30 camera - let’s see that adds up to $140 - those are some really expensive plates.

I thought the price difference was because the Waffle came with an Intel RealSense camera, and different processor (TurtleBot3 Waffle used Intel Joule SBC with Intel Realsense R200 depth camera)but I don’t see it in the spec anymore- just the RPi camera. Maybe Intel quitting the RealSense camera business changed things.


That’s assuming you don’t use any high-current options, or turn on the motors, 'eh? :wink:

Seriously now, why would I want to spend almost a grand, (when all is said and done) for what the GoPiGo can do right out of the box?

What can the turtle bot do that the GoPiGo can’t? Is there any qualitative advantage?

Or. . . .

Should we be promoting the GoPiGo as a lower cost alternative to the Turtle Bot crowd?

Maybe it’s the “Mercedes/BMW” effect?
(i.e. “It’s more expensive so it MUST be better!”)


The reasons are many and valid.

If I had to make the choice again, I am almost certain I would pick the Turtlebot3 Burger $599 to learn ROS2 programming, and especially if I was looking for credibility in the ROS market.


I don’t get it.

What “credibility” would a Turtlebot give you on the “ROS market” that a GoPiGo won’t? Are you looking for a job?

AFAIK, ROS is ROS is ROS and it shouldn’t matter if you have one of Thomas Coyle’s beast 'bots or a GiggleBot. (Of course, if you can run ROS on a micro:bit, I want to see it!)

I don’t see how a GoPiGo is any less “credible” than a monster 'bot with six NVIDIA Xavier boards. Or is it that the snobs on the ROS forums laugh at the GoPiGo?

If that’s the case, 'naff 'em! They can [extremely gross and offensive physical suggestion censored!] if they’re that stuck up about what 'bot you’ve got.

I dumped Ubuntu and Fedora for just those reasons - I didn’t need the “holier-than-thou” attitude - and that’s why I like Mint’s forums and these forums as well: No B.S.

Like I said, I don’t get it.


Certainly a bigger community, which would be worth something. The Burger doesn’t seem too bad - it’s the Waffle that seems way overpriced. But I still think the GoPiGo3 is a good entry platform. From here I wouldn’t step up to a Turtlebot (no need) but probably something NVidia powered.

Would be nice to see a Raspberry Pi 4 version of the GoPiGo - maybe a GoPIGo4?? For schools that wanted to introduce ROS it would be way more cost effective. College students could buy and build their own even.




That already exists and Charlie is living proof!

All you need to do is attach a Pi-4 to your GoPiGo3 and you’re set.

You can attach a Seagate Expansion and/or Seagate One mini-SSD and you have 500+ gigs of whatever gawd-awful stuff you want to mess with. Multi-boot and you can run Raspbian for Robots, GoPiGo O/S, AND ROS all on the same box!

Just because it’s not a huge honkin’

Doesn’t mean it’s not going to get the job done!


Perhaps the least expensive ROS entry would be with micro-ROS on Arduino or Pico. Introduces Python, ROS, Programming, electronics, sensor interfacing.

1 Like

Yeah-ish. But it’s still not officially supported.

Again, yeah-ish. Granted that the Raspberry Pi still requires an external computer for any heavy lifting, it can still run a fair bit of the ROS stack. My impression was that micro-Ros was fairly limited, and mostly just a way to get data from microcontrollers. But I’ll confess I haven’t really explored this.



What about if a Pi-4 and a large SSD are installed?

1 Like

I suffer from prejudiced thinking of ROS robots as remote controlled sensor platforms. I hope to find out that some autonomous awareness and decision making are possible by removing real-time reaction constraints on higher-level functions.

Fast reactions require purpose-built mechanisms and communication paths. Last night I watched this YouTube video on what makes the dragonfly a four-times-more-successful-killer than raptors and big cats. In ROS’s defense, ROS2 adds the option of choosing tailored communication mechanisms, but @jimrh’s voicing of “Just what does ROS on GoPiGo3 enable?” (with my wife’s expectation that I should offer a cogent answer) haunts my efforts.

(My answer that “ROS offers using software I don’t have to write” was met with “Well then how come Dave still doesn’t do anything?”)

1 Like

For the GoPiGo3 your assessment is correct - you really need another computer to do some of the heavy lifting. You could try launching more nodes on Dave itself to see how it performs.

I have worked with robots where everything ran on the robot. The form factor was similar to the Turtlebot Waffle, but with a NUC-type computer rather than a Raspberry Pi. Needless to say they had a hefty appetite for power, and we used a decent sized lipo battery.



Still waiting for an answer to that one.

Just because you “don’t have to write it” doesn’t mean you understand it well enough to actually USE it! :wink:

That’s the challenge!

What functionality absolutely, positively, drop-dead requires real-time response and what can wait a few ms, or even seconds, without noticeable effects?

AFAIK, that’s application dependent.

Carl announcing verbally that he needs help finding his docking station is less time sensitive than the response to the key presses on an electric piano. One can wait milli-seconds or even seconds. The other would be dreadful with any delay whatsoever.

It’s up to you and what you expect.

Don’t be too hard on yourself OR Dave/Carl. Neither one is going to Mars and there isn’t a billion dollar grant riding on the outcome.

‘Ole Doc Harris’ prescription for robotics projects that are getting too serious is:

  1. Take the grandkids to Disney World.
  2. 4oz Jeigermeister, (or an acceptable substitute), orally, repeat as needed.
  3. Both, but delay #2 until you get back home and put the car away.

Remember: This is supposed to be FUN.

If it’s not fun anymore, you need to stop and do something else for a while.

What say ye?


While I find that statement presumptuous that anyone owes you an answer just because you posit a question (with clear prejudice), I did spend some time putting specifics to my general impressions derived from spending days in the Turtlebot3 ROS2 simulation environment following the Turtlebot ROS2 tutorials.

GoPiGo3 Advantages (as a ROS/ROS2 robot):

  • Half the cost
  • Platform support team’s English is impecable

Turtlebot3 Burger Advantages (as a ROS/ROS2 robot):

  • Standardized IMU and LIDAR units and mounting
  • Integrated LIDAR mounting won’t break plate drilling holes
  • Integrated LIDAR power supply with all necessary cabling provided
  • Smaller platform, and smaller turning circle
  • Stronger Construction
  • FPU on controller
  • 5.7 times more precise odometry
  • Motors have built-in, configurable acceleration profile
  • Install/Update via standard Linux packaging
  • OS images with ROS or ROS2 pre-configured are available
  • Active user community is large
  • Dynamixel motors are common in many robotic projects
  • Research community credibility
  • Actual specifications maintained on the website
  • Mature, thorough, and maintained ROS and ROS2 support
Feature GoPiGo3 + Pi3B, LIDAR, IMU Turtlebot3 Burger
ROS Market Penetration Book w/code “World’s Most Popular ROS Platform”
Cost Most affordable for education and hobbyists – approx $285 “most affordable platform for educations and prototype R&D” $599
Size L x W x H 210 x 140 x 150 mm 138 x 178 x 192 mm
Weight 750g 1000g
Controller ATMEGA328: 32-bit ARM Cortex-M0+ 32-bit ARM Cortex-M7 (with FPU and IMU)
Controller Software Proprietary Open Source (Update via USB)
Motor Control Modes 3: Velocity, Torque, Position - via controller features 4: Velocity, Torque, Position, Synchonized – via motor features (Cortex-M3 72MHz 32-bit built-in controller with configurable PID and Limits)
Encoders 720 “ticks” 0.5 deg/tick 4096 “ticks” 0.088 deg/tick
Max Translation Velocity 0.25 m/s 0.22 m/s
Max Rotational Velocity 4.8 rad/s (275 DPS) 2.84 rad/s (163 DPS)
IMU 9-axis w/Fusion Processor (flexible mounting options) 9-axis w/Digital Motion Processor (integrated, secure mounting)
Power Available 5v 3A 3.3v 800mA, 5V 4A, 12V / 1A
Supplied Power Adapter/Charger 12v 1A 12v 5A
Expansion Pins GPIO 12 pins GPIO 18 pins, Arduino 32 pins
Peripheral “Grove”[UART, 2xI2C, 2xADC/Digital/I2C], and 2x “3 pin Servo” 3xUART, CAN, SPI, 1xI2C, 5xADC,4xOLLO, 3xRS485 (Motors use 2 RS485), 3xTTL
Battery Li-Po 11.1V 3000mAh (33Wh 1C-3A) Li-Po 11.1V 1800mAh (20Wh 5C-9A)
Run Time 8 hours max – 4-5 hours typical “About 2 hours 30 minutes”
Charge Time About 3 hours About 2 hours 30 minutes
Status LEDs Power, WiFi Board, Arduino, Power
User LEDS 2xRed, 3xProgrammable color 1xRed,1xGreen,1xBlue
Built-in Audio Multi-frequency buzzer
Maximum Payload 15 kg
Turning Circle (radius) 140 mm 105 mm
Buttons/Switches On/Shutdown/Off switch 2x push buttons, reset, 2xDip Switch
Motor Feedback Velocity, position, stall, tick, Velocity, position, Load, Tick, Trajectory, Temperature, Input Voltage

No one should mis-interpret my intensions, here. I love my two GoPiGo3 robots. I am a GoPiGo3 “fan-boy” through and through.

With that established, after cracking a GoPiGo3 plate drilling mounting holes, smashing my head against the Internet to create a base OS that supports the GoPiGo3 software/hardware and ROS2, and having to create my own GoPiGo3 ROS2 node while learning ROS2, I should be allowed to question my decision to proceed via “sweat equity”, without having to be a Turtlebot3 Burger defender, or possibly seen as denigrating the GoPiGo3.

And further, if the question was specifically the “I” of you, you knew this discussion is totally about ROS for which you have no need, and therefore even if turtlebots walk on ROS water, it is not better for you!

This all sounds like I’m grumpy - I didn’t sleep well last night. I’m off to a nap now.



Isn’t that what the boards are for?

I’ve heard repeated statements about the supposed advantages and credibility of the TurtleBot for ROS. At something like 2x the cost, I was really interested in what the “where the rubber meets the road” advantages were.

(And yea, maybe I’m a bit protective.)

So, I am sorry if you found my request for information “presumptuous”. It wasn’t intended to be.

Of course you should. No one denies that.

You have the absolutely unquestioned right to wonder if putting a Pi-4 on a GoPiGo isn’t stretching sanity to its limits.

Others may disagree, but that’s OK too.

In my case, I absolutely did not understand, (make that read: wasn’t able to read between the lines of the previous postings), what all the shouting was about, so I asked.

When the answer didn’t happen, I gave it (what I thought was), a gentle nudge.

I am sorry that you thought I was climbing your derriere about this.

So, as I read this, the answer works out something like this:

GoPiGo vs TurtleBot:

GoPiGo - Inexpensive and can be modified to launch Saturn-V rockets, but YOU have to do the heavy lifting.

Real World Example:
Volkswagen Beetle with a blown V8 used for drag-racing.

TurtleBot - Costs like oil rights, but is the, (or one of the), standard reference implementations for ROS, therefore everything works right out of the box with minimal fiddling.

Real World Example:
The HP Laserjet 4 - the standard which all other printers are compared to - it’s a tank, and still working to this very day. (And I even have one!)

What say ye?


As a former HP engineer working with the then newly formed HP printer division, it warms my heart to hear that, but I also have a soft spot for my 11 year old Canon all-in-one that continues to print photos in 6 color ink that HP cannot touch … IMO of course.




With regard to ROS being applicable to me or not:

You are correct in assuming that it’s not applicable, (i.e. in scope), for me right now.

The key phrase is “right now”.

I am thoroughly enjoying the thread and watching both you and @KeithW fighting with this beastie. Even if I never touch ROS, I’m learning a lot just watching.

Once my plate empties a bit, I would like to try to duplicate some of your results on my 'bot without using ROS.

Depending on how much progress I make, I may take a crack at it myself - though I may well forgo the LIDAR.

So, in a way, this is applicable to me, if for no other reason than I get to watch two acknowledged masters battle with a worthy opponent.


To be precise, the latest GoPiGo3 encoders are capable of 1096 ticks/rev for 0.33 degree per tick precision but the current robot only reports 720 ticks per rev at 0.5 degrees per tick



  • Motors/encoders are dust/dirt/hair/water/bump protected