2D-locate moving objects in a room accurately

I have a room with lots of individually addressable LEDs on the walls. I want each LED to only light up, when a moving object is near them. (By near them I mean in terms of x/y coordinates. The z-axis (height of the room) should be ignored.
The room is about 5x7m. And 10cm² of accuracy would be nice.

A few ideas that came to my mind:

  • GPS
    how: using some GPS modules
    problem: requires every object to carry the modules and GPS will likely not work indoors

  • wifi/gsm triangulation
    how: maybe using 3 ESP8266 modules in sniffing mode
    problem: requires every moving object to carry a sort of smartphone with them

  • laser barrier grid array
    how: some ir lasers, some photo sensors
    problem: requires a lot of sensors to get precise coordinates => looks weird and ugly

  • distance measuring grid array
    how: ir or ultrasonic distance measurement sensors
    problem: requires a lot of sensors to get precise coordinates => very expensive and looks ugly
    most ir distance sensors only work up to 80cm; most ultrasonic sensors only work up to 450cm and have a 15° detection cone.

  • 2 axis motorized distance measurement
    how: a wall mounted construct that allows a distance measuring sensor to be moved via one or two motors
    problem: complicated algorithm required, building the motorized 2 axis arm will be very complicated, too.

  • real time camera output analysis
    how: using 2 cheap uart/spi camera modules mounted on different non opposing walls and a very sophisticated algorithm. and maybe some ir lights to detect movement in the dark
    problem: the algorithm would be extremely complicated. it also probably wouldn't cover the whole room and wide angle lenses would complicate the algorithm even more.

  • a hybrid out of the last two mentioned ones with only one camera
    how: one camera just checks for changes in pixels and the distance measurement sensor automatically gets directed onto that pixels position. So basically the camera will do the x-axis and the distance measurement sensor the y-axis.
    problem: the algorithm might be more simple than the one of the cameras-only idea, but it would still be extremely complicated to build the thing.

I just can't come up with a satisfying solution.
Any ideas?

Kinect?

IR emitters and receivers?

You mention "moving near them" which is very vague, but you also say
"requires a lot of sensors to get precise coordinates", which indicates you want
accuracy.

So I think you need to start putting numbers into this description to explain what
you really want clearly.

@AWOL That sounds very expensive and can probably not be controlled from an arduino.

@mwawrzyniec That's what I already mentioned under "laser barrier grid array".

@MarkT Sorry, you are right. I added two sentenced to the beginning of my first post.

granularity of measurement is key.

A simple solution would be ultrasound ranging sensors. May not be enough for 10cm2 granularity.

However, triangulation with a number of sensors may get close - taking care that only one sensor is active at the same time.

Of course with multiple targets moving within the space the problem/solution becomes more complex.

If the budget could stretch - thermal imaging could work for multiple targets.....possibly an array of small thermal sensors.

felic:
I have a room with lots of individually addressable LEDs on the walls. I want each LED to only light up, when a moving object is near them. (By near them I mean in terms of x/y coordinates. The z-axis (height of the room) should be ignored.
The room is about 5x7m. And 10cm² of accuracy would be nice.

A few ideas that came to my mind:

  • GPS
    how: using some GPS modules
    problem: requires every object to carry the modules and GPS will likely not work indoors

  • wifi/gsm triangulation
    how: maybe using 3 ESP8266 modules in sniffing mode
    problem: requires every moving object to carry a sort of smartphone with them

  • laser barrier grid array
    how: some ir lasers, some photo sensors
    problem: requires a lot of sensors to get precise coordinates => looks weird and ugly

  • distance measuring grid array
    how: ir or ultrasonic distance measurement sensors
    problem: requires a lot of sensors to get precise coordinates => very expensive and looks ugly
    most ir distance sensors only work up to 80cm; most ultrasonic sensors only work up to 450cm and have a 15° detection cone.

  • 2 axis motorized distance measurement
    how: a wall mounted construct that allows a distance measuring sensor to be moved via one or two motors
    problem: complicated algorithm required, building the motorized 2 axis arm will be very complicated, too.

  • real time camera output analysis
    how: using 2 cheap uart/spi camera modules mounted on different non opposing walls and a very sophisticated algorithm. and maybe some ir lights to detect movement in the dark
    problem: the algorithm would be extremely complicated. it also probably wouldn't cover the whole room and wide angle lenses would complicate the algorithm even more.

  • a hybrid out of the last two mentioned ones with only one camera
    how: one camera just checks for changes in pixels and the distance measurement sensor automatically gets directed onto that pixels position. So basically the camera will do the x-axis and the distance measurement sensor the y-axis.
    problem: the algorithm might be more simple than the one of the cameras-only idea, but it would still be extremely complicated to build the thing.

I just can't come up with a satisfying solution.
Any ideas?

Connect three strings to the object, rap the far end of each string around a motor with a know size pulley. use magnetic rotary encoders on the motor shafts to measure the 'distance' to the object. use math to triangulate the position from length of each string.

or Paint a grid on the floor, have the puck measure it's movement like a computer Mouse. Or better yet, use the mouse optic along with a magnetic gyro to measure relative distances.

Chuck.

Marvelmind sells a system for autonomous vehicles, claiming indoor localization accuracy of +/- 2 cm.

@AnalysIR As I already mentioned. The range and the angle of ultrasonic sensors isn't remotely good enough.
I also don't want to cover my walls with sensors.
And as I said, the budget is tight.

@chucktodd I really hope you are joking.

@jremington I bet this is pretty much the same thing as my second point "wifi/gsm triangulation". Besides that it is much more expensive and I can't use existing devices like smartphones.

How about Lidar?
"Accuracy: +/- 2.5cm at distances greater than 1m"

Have the thing panning around and determine its position, and send that out via cheap 433 MHz unit - master controller receives it and lights LEDs accordingly.

As you are discovering, this is a difficult problem, and not one you can expect to solve with a "tight budget".

How about Lidar?

I have the earlier version and it works very well. However, using lidar to locate an object in 2D is not at all trivial, and requires a map of the surrounding space. Look up SLAM (simultaneous location and mapping).

In a known rectangular room, just need the shortest distance from 2 perpendicular walls to determine location.

Maybe check these, see if any lower cost units have an output you can access.
http://www.homedepot.com/s/range%20finder?NCNI-5

In a known rectangular room, just need the shortest distance from 2 perpendicular walls to determine location.

Please elaborate! In an arbitrary rectangle, there are at least four independent solutions.

There have also been projects where the LEDs themselves were used as sensors, here's one example from Grumpy Mike's website:
www.thebox.myzen.co.uk/Workshop/LED_Sensing.html

With Lidar: