View Full Version : Building an affordable Pressure Sensitive Floor...?

The Omen
12-17-2009, 08:05 AM

I'm new here and have a question regards to building my own pressure sensitive floor, akin to this one http://ame2.asu.edu/projects/floor/.

I hope you can help me out with it.

I'm an artist and am looking to build one as early as say now, for an exhibition in February of next year.

I need help from scratch. i have diddly squat for a budget so would need to build a very cheap version. So, I am open to all suggestions.

The exhibition will involve a large black square composed of soil (roughly 4 x 4m and 200mm in depth) that the general public will walk onto. Beneath the soil patch would lie the pressure sensitive panels. A live image of their movement on the soil will be projected onto a gallery wall. The plan is to have the projection resemble an aerial map of the soil patch, a simple monochrome that is black with white demarcation lines. The sensors would track the imprints and reconfigure the white lines.

I've attached an image to give you the general gist of the project.

Sincerely hope you can help.


The Omen

12-17-2009, 11:56 AM
Welcome to the forum!

You didn't give us a budget to stay within, so we don't have any kind of factor. Some budgets are just a physical impossibility - others aren't.

12-17-2009, 12:02 PM
Hey Omen,

Maybe a pressure sensative floor is not the best solution to the problem. I would think that having 20 cm of dirt between a sensor and the thing you are trying to sense would make your resolution so low that it would be sort of useless, or in the least would create a very general idea of what was going on, and would come at a relatively great expence.

Another idea would be to use video cameras mounted in the ceiling, looking down at the dirt. You could basically program a cheep webcam to look for differences in color and track peoples movement that way...heck you could probably get away with a sub $50 webcam and a borrowed computer.

Another idea would be to try to find some way to measure the thickness of the dirt in a given area...thus making the dirt itself into the sensor.

I don't know...I hope this is helpful.


The Omen
12-17-2009, 05:00 PM
Thanks for the response and thanks for the welcome.

Prior to the flooring, I thought of using GIS software and like you mentioned darkback, to have a camera installed overhead.

I'm still considering this an option. My main concern is how would the webcam articulate the differences between the imprints on the soil and the human figures. In the image projected, I just want a map of the soil and no human presence...is this possible? How can the webcam, connected to a program, only scan the surface of the soil alone?

Any thoughts you have, shoot away.

Thanks again.

12-17-2009, 05:18 PM
If you use an overhead camera, the sad truth is that you would have to fake it. You might see where people are, and estimate their heading, then just draw footprints/demarcation lines where you think their feet are. Lots of software exists to estimate the 2D location of a human from a camera view - check out sourceforge if you need a library.

I'm still a little vauge about what sort of resolution you want from the floor. The article you liked was able to discern quantitative pressure at the square millimetre level - do you need this kind of accuracy? Or do you just want to know where a person is standing, +- 30cm?

Two things are making this difficult - firstly, you want to spend as little as possible, which rules out a lot of options. Secondly, the dirt floor is going to cause a lot of problems with sensors - it will either clog them, or severely reduce their sensitivity.

Here is my solution - instead of using an overhead camera, why not use an underfloor camera? Take a glass floor and put a thin layer of dirt (mud might be better), and just aim the camera upwards. When a foot presses against the floor, it will part the dirt and show up as a white footprint shape. Assuming you want to project the footprints on the wall, very little processing is now required - just monochrome your image and perhaps increase the contrast.

The Omen
12-17-2009, 05:41 PM
Thanks badcommand,

I know it probably sounds a little too vague...apologies.

The underfloor camera is a great idea, but the amount of soil is a fairly important part of the work. I couldn't reduce it to a thin layer and the actual image of the footprints is not important. Moreover, I'd have to create a pretty high platform to have the cameras beneath the surface. Not something I want to do, as I want to keep the soil as low as possible.

The resolution is not important.

I'm basically creating a plot of land and a map of it and want to have that map manipulated by the imprints made by those that tread on the land. I don't want distinguishable footprints relayed back, just something that could track changes to the surface and have that reconfigure the map i've designed.

Thanks again

12-17-2009, 06:05 PM
In that case, If you don't think that the camera idea will work, I would head to Dick Smith Electronics/Jaycar/local electronics store and buy a bunch of cheapish push button switches (http://search.dse.com.au/electronics/Momentary%20Switch).

Then, use a hollow square of foam rubber as a spring, and put it on top of a square wooden box frame. Position the push switch just above the top of the wooden part, and then place a tile or square of chip board on top of that so that when a person is standing on the switch, it is in the ON position. Make mabye 36 (6x6) of these, and cover the whole thing in a sheet of plastic, which you can cover in dirt.

You now have 36 wires which represent individual grid cells, which need to be multiplexed (http://en.wikipedia.org/wiki/Multiplexer) together in some way for a computer to interpret them. I don't have much experience with good multiplexers, but maybe this (http://www.aztecmux.com/a1st1024.htm)? Anyway, you can connect your multiplexed signal to your parallel port and write some software which draws the view you want your projector to show.

That would be the cheapest and simplest option I could imagine (although I think the camera might offer better results if you can get it to work).

12-17-2009, 06:33 PM
I just had an idea...it would be sort of cool...but it isn't exactlly your idea...you could mount a camera in the ceiling...and have it track peoples movement...then you could have the video data create a sort of memory buffer of where people had been...So parts of the picture corrrelating to places people had been would change for a short period of time, and then fade...so you would have sort of trailes marking where people were.


12-17-2009, 07:17 PM
Dirt is heavy. You're looking at 3.2 cubic meters of dirt. That over 3,500 kg of material. Detecting the pressure from a foot print above all that material will be challenging. A mechanical switch as proposed by badcommandorfilename could work. You'll need to calculate the k value for your springs which will be proportional to your resolution (number of panels).

You said that resolution does not matter but it does. What resolution do you require for this project?

Besides posting your question, what research have you done to this point?

What technical skills you do have?

12-17-2009, 08:04 PM
Most importantly, what is your budget?

The Omen
12-17-2009, 08:46 PM
Thanks badcommand!!

Mike G and badcommand, posting my project on this forum is part of my research. I am just seeing what kind of options I have before I proceed and I'm relying on the good will of people like yourself for assistance. My technical skills are well below average in electriconics and programming, but fortunately I do have colleagues and friends who are tech savvy and may assist me once I have collated enough research material for production.

The push button panels sounds possible and inexpensive, thanks again badcommand. Each grid cell panel would represent a vector on the map and once pressed upon would change that vector. I could draw up several distortions for each individual vector. The vectors would be actual suburb demarcation lines and coastal lines taken from a map. (see attachment)

Sounds perfect.

I like this idea a lot as the soil that will cover the floor, has been collated from a whole entire city. I collected 100 bucket loads of soil from 100 properties on the real estate market, roughly covering most areas.

Now, comes the cluelessness. How the hell would I multiplex the signals to a program, and what program could achieve this? The program would need to identify the signals from the push buttons and move through the distorted vectors cell by cell as they are pressed on. Eeessh...too much time spent looking at art theory books, leaves me dumbfounded.

In terms of the physical cells themselves, is this something of what you had in mind badcommand?
Sorry just that I'm visual when it comes to stuff.


The Omen
12-17-2009, 09:10 PM
Apologies MikeG, i forgot to mention that I meant resolution, as in the detail of the imprint left, and the data of that imprint. I don't need it to be detailed.

But i guess the resolution wouldn't matter with the push button, as it would simply be an on/off trigger to change the vectors.

I guess the issue would be the k value of the springs in the pushbuttons. That's what you meant right? The concern being the amount of weight pressed onto them with the soil alone?

And my budget really is minimal. So please think efficient.

12-17-2009, 10:16 PM
That picture is helpful - I have a better idea of what you are trying to achieve.

And yes, that is what I was thinking of - the question is, is it what you had in mind?

I assume that the white lines on the diagram are what you will be drawing as people move around. If that's the case, maybe a regular grid like the one you've drawn is not the best strategy - you could cut the switches into the shape of the suburbs/counties and it might have a better result.

As for multiplexing (http://en.wikipedia.org/wiki/Multiplexer) - it is a hardware feature. What you will have is a large number of wires (logic bits) which represent where people are standing. However, your computer can only receive a finite number of input signals (ie, through a USB/serial/parallel port). Multiplexing is when you take many discrete signals and encode them into a single signal (number), which is unique to that particular state.

As a quick example, imagine a 2x2 grid. A multiplexer might read all of the states of the buttons, and convert it into a binary number:

1000 = top left pressed only
0011 = bottom two pressed
1111 = all pressed, etc.

It could then send this number to the computer's serial port, which can be decoded and converted to whatever image you need. I sense that this might be beyond your personal area of expertise, so talk to one of your friends and see if they can do it for you.

The Omen
12-17-2009, 11:07 PM
Hey badcommand,

sorry if i haven't made myself clear. i'll try again.

ok, so the image I just posted is a mock up of what the projected image would kind of look like, with each grid cell in the image corresponding to a single grid cell beneath the floor. if you were to step on say the top left grid cell, it would activate one of several already drawn up vectors/ images that are linked to that single cell. if that were the only cell stepped on, that would be the only one changing.

and like i said, this idea is perfect and is precisely what i was aiming for.

i've attached an image of the push button grid cell as i understood it from your reply. let me know if i'm on the right track with it.

thanking you ten fold.

12-18-2009, 06:24 AM
You are underestimating the gravity of this project. What is the minimum number of panels (resolution) you need to display a “good” picture; 2, 4, 8? How much earth will each panel hold up?

The electronics and programming has been done many times over. In the old days we called this a matrix keyboard; Google “matrix keyboard or keypad”.

The Omen
12-18-2009, 08:48 AM
Hi MikeG,

I've attached another image. It's got measurements, number of panels etc, minus weight of soil distribution per panel. also need to take into consideration weight per individual on panel. See if I can do that round up some numbers tomorrow. It's 2am here.

Thanks again

12-18-2009, 12:22 PM
This is not a robotics project, it is a construction project with some geometry and physics. Encoding a key press (panel) has been done many many times and is well documented.

I'm a little foggy on your idea of vector. Unless, you mean from a common origin like a corner but I think you mean a line like step1 to step2 or y=mx+b. Placing the lines end to end will produce a path through the grid over time. Plus it will help determine (estimate) which steps belong to which people if there are multiple folks walking on the grid.

12-18-2009, 03:08 PM
So you're looking at, what, about 1.6cubic meters of soil?

About 5700lbs? ~2600kg? (4m * 4m * 0.1m [stick with SI Units] = 1.6m3, right? 1 cubic meter = ~1.308 cubic yards. 1.6 * 1.31 = 2.096 If I were picking up a load of top soil, I would figure ~100lbs/cubic foot, to be safe, but it would probably be about 85lbs. If what you have is denser than clean/dry top soil by more than ten percent, those numbers would need to be adjusted.)

So each panel is supporting 160lbs (assuming even weight distribution) plus a potential 280lb human (Sure, they get heavier, but the deviation from norm is more than a couple sigma at that point) - maybe we should figure a 320lb human to be safe?

So your foam needs to support 500lbs per panel, but still deform enough to press a switch AND have longevity and recovery to not deform enough over the length of the work to depress said switch? That's a toughie.

12-18-2009, 03:17 PM
For my money - I'd fake it and use a soil analog.

As designed, you don't need a roboticist, you need a structural engineer and an architect.

IMHO: I wouldn't hazard an opinion beyond that, because we're talking safety and welfare of human life now. You need to find yourself a general contractor and a structural engineer that are licensed in your area, know the codes, and can design and construct the basic system. Then worry about how it gets translated into the projected image.

12-18-2009, 04:42 PM
Are you a university student, Omen?

Even if you aren't, I suggest you head down to your local engineering school and put up a few flyers asking for design consultations. Ask for a civil engineer for help with the floor (they know everything there is to know about soil too), and an electrical/mechatronic engineer for help with the switches and multiplexing. If software isn't your strong suit either (or you decide to try the webcam idea) try and get a computer science student to help you out too. Describe your project, and they can do the work you aren't skilled in. They will be able to do a much better job in half the time.

Offer around $10-$20 per hour. Most students will do it just for the reference on their CV.

Finally, don't forget to prototype what you come up with before doing a full-scale, and (like Adrenalynn said) get it approved by a licensed builder (you probably have a friend of a friend...) before putting it into production.