Opensource Vacuum Robot

[assuncap]

Hi All,

We are all (well some of us are) passionated about our Roombas, Neatos or Trilobites but there is always something missing or something that we would have done differently... So that brings us slowly to opensource and in this case a Opensource Robot Vacuum Cleaner.
There are already opensource communities for 3d printers, UAVs, Tractors, etc. Maybe it is time for a ORVC.
I don't believe i have what it takes to start such a project on my own. So I thought i would share with all of you some of my ideas...
Here is a possible "shopping" list containing some of the areas that i think need to be developed in order to complete the vision.

Brain:
32bit Arm Micro Controller, netduino is a possible choice but there are others. The important thing is that it needs to be 32bits (arduino is only 8bit) and affordable.
Sensors:
IR, Sonar, Lidar, CCD, accelerometer, giroscope, compass, etc as long as it is openly available and affordable
Power:
Any battery tech available
Vacuum:
3D printed parts and a DC motor
Body Work:
3D printed parts, Mold casting or any other DIY method.
Firmware:
Battery Management, Sensor Management, Motor Management, SLAM, Behavior and Cleaning algorithms

So as you can see there is a lot to be done in order to achieve a functional ORVC.
I believe it can be done by publishing our ideas here (or somewhere else as long as people can access them and evolve them).

So please feel free to contribute with you comments.

[yvesdm3000]

I think it might be easier to start with a motherboard-replacement for a roomba, even putting a little embedded device like http://www.olimex.com/dev/sam9-L9260.html on top of an existing roomba and do some RS232/TTL communication would be a nice start. I'm still looking at a camera that I could connect to that olimex board to experiment with navigation like the Samsung bot does...

-Yves

[robotreviews]

This is a GREAT idea, and one that I have been thinking about for a very long time.

I would love to get an idea of how much interest there is for this. We have been working on a few things that are specifically designed to help make projects like this happen, and this would be the perfect project to start with!

So, it would be helpful if people could also post what sort of tools would be useful to make this sort of project easier to make a reality. (Wiki's, shared documents, dedicated forums, etc.)

Thanks!
Craig

[mfortuna]

I think a controller board with support for a "open" MCU could be developed without too much trouble. Using a roomba as a test vehicle seems like a good choice. Since all roombas use off-the-shelf MCUs I wonder if the existing firmware can be overwritten. I don't think the flash is locked since Irobot does issue updates from time to time. So if someone took a 500 series controller board, soldered a header to the JTAG connector on the board, it should be possible to completely overwrite the flash with a opensource image.

It's also possible to depop the MCU and wire in an Amega or other MCU. I remember someone doing this on a 400 series controller.

The next step would be to design a controller board with the various sensors and charging circuits. I think this would be a big effort for one person. A team approach seems better.

[assuncap]

@robotreviews
A dedicated wiki and forum for such a project would be ideal.

@mFortuna & Yves
Obviously both your suggestions are very interesting. But you can not limit your self to using a roomba as development platform. I see this as a several projects evolving hopefully into at least 1 complete functional Vacuum Robot. Along the way you have multiple paths, but you have common grounds as well.
I accept the idea the some people will want to play around with existing robots and develop not only new firmware or electronics but maybe also design different brushes or Vacuum Bins. I have looked several times at my Roomba next to my Dyson and thought on how to combine the two.
But there is other areas to consider, i can think of these:
SLAM - You need algorithms to navigate around the room
Sensors - You need effective ways to read the surroundings without having a big computational overhead and maintaining a low cost.
Cooperation - Developing ways for robots to work together.

You see the idea? If someone has a crazy idea just do some crazy drawing and publish it!

[yvesdm3000]

I think a controller board with support for a "open" MCU could be developed without too much trouble. Using a roomba as a test vehicle seems like a good choice. Since all roombas use off-the-shelf MCUs I wonder if the existing firmware can be overwritten. I don't think the flash is locked since Irobot does issue updates from time to time. So if someone took a 500 series controller board, soldered a header to the JTAG connector on the board, it should be possible to completely overwrite the flash with a opensource image.

I'm not really interested in changing current firmware of a roomba. I have no idea what I could do better than what iRobot created. To improve this bot, it needs a mapping system so something added to the bot like a top camera (with a wide angle lens) or a way where you can reliably have an indication of the angle the bot was in relation to its previous straight-line run.

-Yves

[vic7767]

Since the OS firmware on the iRobot robots is encrypted and proprietary a completely new CPU-MCU will be required in order to use the existing hardware sensors and motors as well as incorporate additional new cameras, sonar devices, laser devices etc. Some work has already been done on the Neato robot and the turrent mounted laser hardware and software. Possibly a merger of the two devices could be managed.

[TechGuy]

Since the OS firmware on the iRobot robots is encrypted and proprietary a completely new CPU-MCU will be required in order to use the existing hardware sensors and motors as well as incorporate additional new cameras, sonar devices, laser devices etc. Some work has already been done on the Neato robot and the turrent mounted laser hardware and software. Possibly a merger of the two devices could be managed.

I think the Roomba firware is stored on MCU flash memory. The opensource firmware will replace the Roomba firmware using JTAG interface. The MCU comes with source level on-chip debugging which is perfect for opensource development without buying expensive development tools. First step is to trace the Roomba circuit and find out which pins are used and not used.
Here is the STR735 (used in 531) overview --> http://microcontrollershop.com/CWSTR735.php

[sageman]

i possibly have access to a {mind block} pcb making machine {mind block} if one wants i'll ask my sister to ask this bloke at her school (i've left) about it

[assuncap]

I think it might be easier to start with a motherboard-replacement for a roomba, even putting a little embedded device like http://www.olimex.com/dev/sam9-L9260.html on top of an existing roomba and do some RS232/TTL communication would be a nice start. I'm still looking at a camera that I could connect to that olimex board to experiment with navigation like the Samsung bot does...

-Yves

I just remembered that because the MCU your refering actualy runs linux you could compile http://www.ros.org and run it on the actual robot. ROS has support for a lot of stuff so it would be a very good starting point for the robot firmware.

[bhylak]

i possibly have access to a {mind block} pcb making machine {mind block} if one wants i'll ask my sister to ask this bloke at her school (i've left) about it

For free?

This sounds awesome! I think the easiest approach would be to replace the MCU, like suggested. Develop a custom firmware, with mapping, like suggested. This could be awesome.

Count me in.

[sageman]

free or little charge (materials) (hopefully)

someone will have to make the design to go into the machine and what-not as that sort of stuff is about 5 miles over my head
anyway iil have my sister give him a letter about what machine it is and what program to use so the machine knows what to do etc after easter if everyone thinks its a good idea that soon

another thing is that from what ive seen the with of the pcb electric lines (if you get my drift) are quite thick i dont know weather its because he has the large bits for ease of soldering (teaching the year 7's or because of the method he uses it wont be effictant to make them smaller so will also ask the minimum width of those line things so pcb size can be umm known

its just a "CAM" machine that "scratches" 2 lines across a copper surface to make a electric channel thing (please tell me the correct name for these)

[bhylak]

You can actually do it with an ink printer.

[bhylak]

Guys, if you think about it, its not too hard to build a Roomba from scratch. Start with an arduino, hack some handheld vacuum, DC motors, maybe a 12v battery... I say ditch the bumper sensor, go with ultrasonic sensors. Mapping needs to be implemented. (Lasers, possibly on the bottom of the robot)

I say we call it Project Vacuum Freedom.

[assuncap]

I say we call it Project Vacuum Freedom.

Good name.
The trouble with arduino is that it is 8bit. So you don't have a lot of power available.
A netduino or similar are ARM based and therefore a lot more powerful.
One alternative i tought was to use the new IOIO (http://www.sparkfun.com/products/10585) board from sparkfun. Basically you can connect a Android device to it. So you could use all the processing power available on a Nexus one to control the robot vacuum, as well as the accelerometer, camera, giro, Wifi, etc.

My plan right now is to work on a low cost LIDER using a netduino, a camera and a line laser and no moving parts.

[bhylak]

Yes, some pda or something... I'll explore some fairly powerful yet cheap options... On ebay, they have those cheap arm powered netbooks for 100 dollars... that might be better. I could develop a small linux OS that could run on it...

[roboroomba]

take a look at my idea for a robot floor washer, it uses a spinning laser thing like the neato does. It would be cool if you could actually make one of those.

I would really like to see a carpet washing robot.

It would be like an old scooba with a lot of changes, but a similar cleaning system

the cleaning stages would go like this

1. Solution Injection
this would be where a bar like mister sprays down cleaning solution onto and into the carpet.

2. scrubbing brushes
this would be where the flat disk brushes spin around and scrub the solution into the carpet, and loosen up the dirt.

3. vacuum
this would be where a very strong vacuum sucks up all of the water and dirt through a squeegee thing like the roombas dirt bin does, except the rubber part will be larger, thicker rubber, and a lot further apart so that even large particles and hair can get sucked up.

4. drying
I am not sure if it needs this, but it could be were the exhaust from the vacuum would blow back down on the carpet and dry any remaining moisture.


It could have a soft water tank like the scooba 230

I also think that a navigation system like the Neato xv-11 would save a lot of cleaning solution and make it work better in general.


drawing



location of posts
http://www.robotreviews.com/chat/viewtopic.php?f=3&t=14577

[WebSpider]

I am very much interested in this open source vacuuming thread! (even registered especially for this )

I have a Roomba 535 PET. My Roomba frequently chooses illogical cleaning methods, or gets stuck in the lightbeam of a virtual wall, where it will circle until it panics.

Would I be able to influence the firmware myself, I would be able to make it a bit smarter (and feed that back to the community, offcourse ), instead of being annoyed at it

I've looked at the plastics that are used, and most of them can be printed. Gearbox, housing (possibly in pieces), motor cage, etc. Only thing that gets really difficult is printing brushes, sensors, etc

From what I'm seeing with my 3D printers now (I have two, and can get parts pretty easily since I run a webshop for them), the Arduino boards (or clones) have enough processing power to drive 4 independent steppers and run 5 sensors at the same time, with a polling frequency of 5ms. Really, that should be enough for a Roomba, no?

[bhylak]

I'm getting a Mini-Mill in a couple of days and adding CNC, so I can make some parts.

[Parker]

Question: Is this project about developing a controller that would fit in several existing robots?

I'm interested in open-source vacuuming robots. But an important criteria for me is that the end product has to be more than a toy. If it needs maintenance after one or two runs then forget it. If it can go a month without maintenance (e.g Electrolux trilobite, Karcher RC3000) then definitely count me in. The only way I can think of doing that is to make it replace the controller of an existing robot that has good build quality.