Approaches to a few thousand individually addressable widely dispersed LEDs

Question

I'm trying to come up with the best design for a project requiring between 1000 and 5000 individually-addressable LEDs, not in a strip, grid, ribbon, pegboard, etc. configuration. There is a lot of information about LED strips and arrays, but those don't seem directly applicable to my project.

What I want to do is similar to a "pick to light" system seen in warehouses, if you're familiar with that. I will have a room with many storage locations (various sizes) in it. I will have software that keeps track of what is in those locations, and when a user searches for one (either to remove or add one), the LED above the bin will light.

I need to support having an arbitrary set of the LEDs on at any given time -- it's not just one at a time (though realistically I'd expect well under 1/4 would ever be on at once). The bins will range from 3 inches to several feet apart, and the room will be on the order of 15 feet square, with storage locations spread throughout.

Reliability, a nice clean installation and ease of setup are more important than cost. Of course, all things being equal, lower cost is better -- it's just not the top priority.

Answer 1

This sounds like it might be a good application for a 1-Wire® bus, using something like a DS2413 to control every LED or two, at least if those parts were a bit cheaper. Otherwise, your best bet may be to attach each LED to a small microcontroller, and use a simple unidirectional communications scheme to send data to all the controllers. Using a unidirectional scheme would facilitate the construction of very simple repeaters (simple non-inverting buffers would probably suffice). Each microcontroller could use a small amount of flash or EEPROM to hold an address, so all devices could be individually addressed independently. The biggest difficulty might be configuring the network; that might be best facilitated by having each controller include a command to output its address by modulating its light output. An optical receiver attached to a portable computer could be used to visit each node, read its light pattern, and make note of its physical location.

Answer 2

If the cost of a 'standard' communication bus (Can, LIN, RS845-proprietary) and a uC at each location is no problem I would go for that. You'd probably need a two-stage bus, I can't think of any bus that supports 1k nodes, but 64 x 64 should be doable. For wiring I would select something that is available pre-assembled, ethernet patch cable might be a good choice.

A low-cost RF unit would be another option (RFM70? I just finished my C/C++ library), maybe with battery power (no cables!), the master could periodically scan all units and those who do not respond or detect a low battery are singled out for replacement. Or saturate the room with a broadcasted IR signal.

In the spring I will do a course on system architecture, this is a nice problem to illustrate the consequences of various approaches!

Answer 3

You could think of the whole room as one big 3d matrix of LEDs.

There is nothing to say that the LED has to be within a few inches of the chip driving it.

I assume you will be having, for example, a room with many racks. Each rack having many shelves. Each shelf having many bins.

You could have a controller per shelf - be that a microcontroller, or a shift register, or whatever, connected to the LEDs above each bin.

Those controllers/shift registers could be grouped into a per-rack controller. Finally, the per-rack controllers are grouped into the master room controller.

The room controller picks the rack and tells it the shelf and bin. The rack then tells the shelf which bin. The shelf lights the LED.

This could scale to multiple rooms by adding another layer to group the rooms together.

As an afterthought, you could have a button next to each LED which the operator presses when he has picked the item, thus turning off the LED and informing the stock management system that the item has been picked.

Answer 4

If you look through lapsed US [patents with my name on them (total of 1) you will find out how my client proposed to essentially exactly what you describe. Technically it worked well. Business wise he failed to sell it for whatever reason.

I don't have patent number to hand but can probably dig it out. Early 2000's AFAIR. Let me know if you find it.

There are many other ways, but I used an inductive loop that power fed all the electronics and addressed each bin digitally. Bidirectional data on an inductive powering loop is a reasonably good trick. I don't recall what was in the patent but you'd probably need more than was there for as full system as the patent may mainly have covered inductive powering.