Friday, June 27, 2008

Crash diet

I spent some time taking apart my Roomba to remove some extra weight. You can see the pictures of this process here.

In the pictures, you can see my laptop with the tear-down instructions found here (click on the Discovery Breakdown link on the right). They were quite simple to follow and worked great. In the end, I removed the brush deck, side brush, cliff sensors (broken anyway), and the front bumper. I'll probably eventually remove the wheels and wheel drop sensors in order to drop some more weight. I also have a little better understanding of where I can "plug in" to the sensors for later in the project.....

I stopped by Fry's and picked up an Antec Tri-Cool 120mm case fan (about $14 after taxes). And a 9v battery for my multimeter. I picked the Tri-Cool because it had decent CFM (79 at 12v), would work at lower voltages (as low as 5v), and wasn't overly expensive. After I got my meter working, it probably didn't matter about the low voltage, turns out that the dust bin connector is putting out between 14 and 15 volts.

I tried to get the bathroom scale to work (digital), but it just gave an error after replacing the battery. So I don't know the current weight after surgery.

Now I need to figure out best to attach the 120mm fan to my (currently) 1 inch hole. I'm a little leary of making the hole exactly match the fan and directly attaching it (main concerns are about whether the bag would get caught in the fan). I'm also considering two fans vs one (at 79CFM, I'm still short of the computed 93 to 115 I need based on this calculator ). Luckily, I still have enough MDF for another base if I screw it up (on purpose -- I plan on making a "good" one once I work out all of the kinks with the original, but if I need it before then, I'm good).

Monday, June 16, 2008

3, 2, 1....Blastoff?

So, I disassembled the dust bin to extract the vacuum fan. Then, using the all-pupose attachment implement (aka duct tape) I secured the fan to the hovercraft.

Next, I had to find a way to turn on just the fan motor. I tried the SCI Tester program that you can get from the RooTooth people, but it only ran in safe mode (which was a problem because it didn't override the cliff sensors). Next, I tried the Microsoft Robotics Studio, but I couldn't get it to connect (I'll have to dig more). Finally, I found a program by "Black 64" - that did the trick. I could send the 132 command (Full Mode) and then individually control each motor regardless of the cliff sensor status. The whirr of the motor sounded quite good at this point.

My first attempt was to just switch the leads and see if that switched the fan direction. Unfortunately, the circuit was smart enough to overcome that and continued to run in the "correct" direction which was the wrong direction for my intent. So my fan was sucking instead of blowing. So, I detached it and reoriented it so that it would then blow. Now set for a first flight.

So, it turns out that the Roomba fan motor isn't powerful enough to even lift the wooden platform off the ground, much less fully burdened with a 10lb Roomba. So I'll need to start investigating fan options and looking at stripping some weight off the Roomba. This is no longer going to be something that others could do without affecting the performance of their bot.

Images of the latest activity can be found here:

Monday, June 9, 2008

Construction begins

This weekend, I started constructing my HoverRoomba. After a trip to the orange home improvement store (about $4.50 for the 2'x4' sheet of MDF), here's what I did:

1. First, I cut the MDF down to 2'x2' (gives me spare wood to make another one if needed).
2. Next, I found the center of the 2'x2' sqaure using the corner technique.
3. I drove a small nail into the center of the circle and used some rope to draw a 13" circle (the Roomba is 13" in diameter.
4. Next, I sanded the edges to keep them from tearing the plastic sheeting.
5. After that, I drilled a small hole in what will be the rear of the craft. This is where I will be mounting the vacuum fan.
6. I had some heavy plastic sheeting left over from installing floating laminate flooring (the roll was about $18 originally, but I'm only using a small amount); this plastic was taped to the bottom with about 3/4" of "play" all around. This is the hovercraft skirt.
7. Next, I used some of the scrap MDF to finish the skirt and form the air pocket that the craft will hover on, I attached it with two wood screws.
8. I cut two vent holes in the skirt for the air to pass from the skirt to the coushin. When I first cut them, I made them too big (one of the pictures shows me testing the skirt by blowing through the hole). I covered them with tape and made smaller slits which seemed to work much better.

All of the pictures were taken by my 10-year old with her phone (2.0 megapixel), so composition is what it is.

My next steps are to fit it to the Roomba and fine tune the shape. Then, I will hook the vacuum fan to it and see if she'll go.

Sunday, June 1, 2008

Step 1: Hover - The research

Before I can make my Roomba (I need to name this guy to make it easier to talk about him -- suggestions taken in the comments) hover, I should do some research into what it will take and how to actually go about it.

The simplest plans I've seen on line are found here: (better pictures can be found here: These plans should scale down well to my Roomba since the plans call for a disk shape and the Roomba is already disk shaped. This is the design I will be basing my HoverRoomba on.

Obviously, I won't be using a leaf blower or shop-vac since the Roomba already provides this capability. The Roomba's vacuum motor is located in the dust bin and looks like this: My main concern is whether this will provide enough "oomph" to get the craft off the ground. If it does, phase one becomes a pure construction project that should take a weekend to complete. Otherwise, finding the right fan that will work (powerful enough without too much draw) could become a challenge. For this project, I will be building this in a manner where my Roomba actually rides the overcraft instead of rebuilding the entire undercarriage.

The Roomba's critical dimensions are 13 inches in diameter and about 10lbs in total weight (including battery, brush deck, etc.). If weight becomes problematic, I can shed some of it by removing non-essential pieces (i.e. the brush deck as was done in the Airsoft gun mod) but I would like to aviod that because it makes the mod more of an "accessory" than a "rebuild".
From this link: we see that the critical factors are surface area (40.85 sq. inches) and weight (10lbs). In order to get off the ground, I need at least 0.245 lbs/sq in. According to this page: I need to move about 1.8 cubic feet per second (184 cubic meters per hour).

I found a research paper ( that names the fan motor as the Mabuchi FK-260SA-12300 ( ). Using the picture of the motor listed earlier, we can see that the fan has 16 to 20 blades and should spin at around 7200 RPM and from the specs puts out about 0.002 HP.

So, what does all of this mean? Nothing really because I don't know enough about airflow, etc. to take it any further (and haven't had a lot of luck finding the formulas I need to see if it will work). In the end, I'm going to just try it. If the vacuum fan doesn't work, I'll have some numbers I can use to compare performancec stats against. I've looked at some PC fans and most of them are underpowered for what I'll need, but since the voltage draw of a PC fan is lower than what the Roomba motor will draw, it might be possile for me to use two PC fans to accomplish the task.

Next up, planning and materials.

Thursday, May 29, 2008

The victim....

I was the lucky recipient of a malfunctioning Roomba in a Bag of Crap from Woot in February of 2008. It's failure code indicated that the Cliff Sensors were not working properly (a quick check with my phone's camera confirmed that only one of the IR bulbs were lit). I could try fix it, but I thought it would be a lot more fun using it in some sort of hack project. I spent several months reading up on what other people were doing with their Roombas (which is why this blog didn't start until May and the Roomba was received in February). Most projects centered around what you could program a Roomba to do; I didn't want to do what everyone else was doing; I wanted something "GRAND".

With so many projects that focused on the software capabilities of the Robot and my respect for the modder community, I wanted to come up with a project that physically altered the attributes of my robot that left it functional but in a different way. I only found one other project that really altered the physical attributes of their Roomba [], so this will be somewhat uncharted territory. Along the way, I will have to learn some electronics and will have to build things that I have no expertise in (I'm a programmer, not an Electrical Engineer).

So, with the idea that I want to do something "GRAND", I brainstormed, I asked friends for input, I searched the web for "fun" projects. Eventually, I came up with the concept of a hovercraft. A Roomba is a vaccuum and in my youth, I remember projects that transformed a "normal household vaccuum into a hovercraft". While this isn't a "normal household vaccuum" it seemed like the perfect modern twist on this old garage project.

Any good project starts with a set of goals (for my programming projects, we'd call this the scope and this stage "Envisioning"). (On a side note, a lot of bad projects start this way, too.....I'm just hoping this isn't one of those projects.) So, here are my intended goals for this project and the order in which I plan on tackling them:

  1. Make my Roomba hover

  2. Give my Roomba some means of propulsion

  3. Use the existing Serial Command Interface (via a RooTooth) to control the "flight". iRobot provides a serial comminications port for the Roomba with a very detailed Serial Command Interface [warning: PDF].

  4. Remove the cliff sensors and replace them with some sort of sensor that is more useful to a hovercraft (first choice would be some sort of electronic compass -- there are four cliff sensors and four cardinal directions, seems logical to me)

  5. Write a full suite of Roomba software that allows the Roomba to navigate a simple obstacle course

Tuesday, May 27, 2008

And so it begins...

This blog will document the trials and tribulations that I experience as I attempt to convert a Roomba into an autonomous Hovercraft. Stay tuned for further posts.