Slashdot Mirror
High Accuracy Indoor Location Tracking?
aletterman asks: "I am looking for technology that can allow my company to track the position of a fork lift as it moves around a warehouse. This would allow us to factor out one problem situation - where the lift was when the driver dropped off the product. Based on the width of our warehouse locations, we need a resolution of about +-1ft. Standard GPS can't get that accurate or work well indoors. The fork lifts already have a VT220 terminal running 802.11b, so adding another device would not be difficult. I am currently looking at a product that can triangulate via the RSSI of the 802.11b network, but I am concerned the changes in the product mix and density of the corrugated boxes will change the RSSI and introduce a mis-positioning of the locations. I would prefer that the device transmit a position either serially or via our 802.11b network. Our warehouses are fairly large (300,000+ sqft) and have a large of amount of corrugated boxes. Any suggestions are greatly appreciated!"
Paint a grid on the floor, with squares of 1 sq. ft., possibly with magnetic paint, and install cameras or magnetic sensors to sense the lines passing underneath. Add a compass or rotational sensor to help determine the orientation. Add some maths, stir, and you might just have something that might work. (Or not...)
I suffer from attention surplus disorder.
If you could the speed and direction information along with known beacons placed in the warehouse, you should be able to have the forklift know where it's positioned at all times.
You are being MICROattacked, from various angles, in a SOFT manner.
Place optical encoders on the left and right front or rear wheels of the fork lift, so that it knows its delta position based on dead reconning. (about 1mm resolution)
Add 'bar codes' that run the width of the major isles, and point optical encoders downward on the fork lift so that as it drives over them, it can 'reset' its position and eliminate any errors that build up while using dead reconing.
Try to place them so that the fork lift is expected to run over a bar code every 5-15 minutes or so (depends on errors that you measure in encoder accuracy, and the resolution that you need).
Hook them up to a PIC that kicks out the counters for dead reconing as well as any known location events via serial to your computer on which the 802.11b is located. An AVR butterfly will do the job.
The world will not get better through technology. We must seek to be better people.
I dunno if the product you're looking at can handle this, but in principle, just because three points of reference theoretically allows a perfect calculation doesn't mean you can't add more to provide greater resolution in an error-prone environment.
I STILL can't find my car keys.
Don't blame Durga. I voted for Centauri.
Look into the technology used by the NHL on Fox for making the hockey puck more visible to the home viewers.
Here is an oversimplified diagram, but enough to get you pointed in the right direction.
Glonoinha the MebiByte Slayer
It's a bit more labor intensive, but i think it's the more sensible solution.
Each location (shelf/floor area) has a tag, RFID or iButton would both work i think. Each item/pallette has a similar tag. When dropping an item off he scans them both in with a mobile reader, uploaded to central, linking the two in a database.
Are you looking for the sexiest solution, or one that works?
Remember, it takes 42 muscles to frown and only 4 to pull the trigger of a sniper rifle.
Since you are concerned about RF absorption, you might look at Infrared. Some supermarkets are installing ceiling mounted sensors to track shopping carts and notify shoppers of "specials" near their location.
There was also this article published a while back:
http://www.ekahau.com/pdf/NYTimes_30Oct2003.pdf
I suppose you've ruled out a helium balloon on a string?
What about simply mapping each of the scanned RFID tags as they happen on a map of the warehouse?
Money is stored in cassettes in a fully automatic warehouse. The actual warehouse uses an automatic system, (if you have seen a storage robot, it isn't a whole lot different). However money is shipped from the loading/unloading dock to the procssing stations and then from the processing stations to the warehouse loading/unloading station using robot forklifts.
The forklift control system was German, but I can't remember whose. They used a pulsed transmission system and used the arrival time for navigation. The main control computer knew where the forklifts were to the centimetre and gave them orders.
See my journal, I write things there
Embed RFID tags in the floor. Put an RFID reader underneath the forklift. Have it report back over 802.11b with the RFID tag number(s) that it is closest to.
Let me know. I was looking for something similar for a cat locator project I was working on. This would plug into the cat locator portion of my website. My requirements are a little different since the locator module would have to be small enough to put around a cats neck.
So far, I have yet to see a solution presented in the comments that would work for my situation.
Things you think are in the Constitution, but are not.
This would allow us to factor out one problem situation - where the lift was when the driver dropped off the product.
Why not ask the driver?
Make yourself a grid of RF receivers on the ceiling of your wearhouse and stick an RF broadcaster on the forklift.
Have the RF broadcaster send of a 'ping' at a given frequency (your 'realtime resolution')
Since All the rf recievers know what time the 'ping' should have been sent, they can calculate how far away the forklift is from them by differencing the time they recieved it to the time that they got it.
Use three or more beacons (networked) and you have yourself a triangulation network. Im pretty sure this is how GPS works, except your recieves wouldn't be in space.
I know a lot of forklifts or other equipment sometimes have a flashing light on top. Could you do something similar, like have a bright light on top of the forklift (and if you have multiple forklifts, each could have a different color light), and have a few video cameras mounted on the ceiling. Every second or so a screencapture program could read a frame from each camera and scan it for the color of the light.
That might cause a problem with employees and their clothing, so you could even use two lights on a forklift, each a different color, so one lift is identified by red and green, or maybe several lights so it shows up as a large blob, bigger than any employee. You won't have to worry about echo or other problems that come from a radio signal in a large warehouse (which is probably made of metal).
Just need to keep an up-to-date database of what tags correspond to what tiles, so you can swap them out if they get damaged.
This strikes me as being much easier than magnetic paint or ultrasonic sensors, etc. SHould cost too much to retrofit the floor and the fork lifts. I don't know any specific numbers, however.
Step 1) Put a big round ball on a stick on top of each forklift, each a different color. (perhaps even lit up with infra-red colors). Use some colors you don't have a lot of around.
Step 2) Put cameras around the area frequently taking stills, and using edge detection (with color filter masks) to locate balls in the camera's 2-D space.
Step 3) Read a book about computer vision and how you can recover the 3-D posotion of something, given it's 2-D position from multiple cameras (5 cameras per ball is best).
You must DIE.
A little robotics might get you there. This may or may not be the best way, but it would work:
Careful odometry should get you within a foot or so, assuming you start from a known point. Yet, odometry coordinates inevitably degrade with time as errors build upon themselves.
There exists an algorithm, however, called the Extended Kalman Filter, that can help correct this. Using just a few sonar/iR sensors, corrections can be made to the coordinates.
Basically, as you move your forklift, its margin of error in position keeps increasing. Every once and awhile you fire off a handful of distance measurements from your sensors. Using the EKF, you can use this data along with a map of your warehouse to reduce your margin of error. The more measurements and the more accurate the readings, the closer you can properly position your forklift.
-dave
http://millionnumbers.com/ - own the number of your dreams
Put cameras on the ceiling and paint big orange numbers on top of the forklifts.
For every annoying gentoo user, are three even more annoying anti-gentoo crybabies. Take Yosh from #Gimp for example.
I assume that a warehouse as large as yours already has a system for numbering all the shelves.
In that case, why don't you just attach barcodes to all shelves and add a barcode reader with a long coilable cord to the forklifts?
When making a drop, the driver would scan the boxes to be dropped, and then the location they were dropped. To make a pickup, he could scan the box to be picked up, and then a special "pick up" barcode on a label somewhere on the forklift.
Ideally with a system such as this, a user interface requiring no user input other than scanning two barcodes.
This is much cheaper and reliable than any soft of positioning technology, and only relies on the tried-and-true technology of barcodes.
Sure, this is still prone to operator error, but no more so than a similar system using some kind of positioning. You're not going to gain any real benifits from positioning either way that you won't have when going with barcodes.
Check out a company called Dust Networks. They've been working on RFIDish localization for warehouse management based time of flight for years. I'm not sure if they've ever gotten it to work, but if anyone has it's these guys.
Otherwise, I think you're going to have to go with a bunch of cameras and some image processing. Maybe by taking pictures of the celing.
Paint a unique code on the roof of the forklift (something like a DataMatrix code or one of the other 2-D symbologies).
Put several cameras on the roof; fix them so they don't move; and calibrate them. You can probably use the WiFi cameras from Axis to not have to run cable.
This is what you need: the "Bat Ultrasonic Location System", developed at ATT Cambridge labs (former).
From this page:
Take heart---the problem you're contemplating has been fairly well studied.
I remember a professor here at CMU saying that you could do localization for forklifts by pointing a camera at the floor. Most warehouse floors have enough scratches and marks on them that as you wander around, you can get a pretty good idea of where you are by comparing them to a map (using techniques like Monte Carlo localization---google it!). Combined with encoders on the forklift wheels, you may be able to get the resolution you need.
Here is a paper describing technology like this. In the results they say they get accuracy down to a millimeter.
So, those talking about painting a grid on the floor have the right idea---but perhaps you don't even have to do that!
MAN SHOOTS ROVER!
This is basically how the Mattel Power Glove worked. The glove had 2 ultrasonic emitters on it, and then there was an array of 3 ultrasonic receivers that was placed on your TV.
The concept is similar to how GPS works, except that with ultrasound, your clocks don't need to be nearly as accurate and you can use MUCH lower frequencies.
retrorocket.o not found, launch anyway?
centimeter accuracy using indoor differential gps
e tail.jsp?id=3086
http://www.gpsworld.com/gpsworld/article/articleD
in most warehouses i've walked through there's often time product, grease, water, or any number of things on the floors. At times, the fork lift will slid through them in full brake. I've seen people play games with it it can become so slick (especially on polished concrete floors).
-Malakai
A Dragon Lives in my Garage
I assume you want something somewhat off-the-shelf and not ideas for designing one from scratch. Here's a link to a company that specializes in this type of stuff and might be able to provide a ready to use product. http://www.visonictech.com/info_page.asp?info_id=4 16
If you just look at the highlighted words you first get the requirement "track position forklift warehouse" So he wants to know where the forklifts are in a warehouse. Okay. But WHY?
Second set of words is "where driver dropped product. Ah. I was wondering. Who gives a toss where the forklift is? Oh I worked in enough warehouses to know they are sometimes a pain to find if people don't return them or use them as a portable chair to take their break outside BUT it would be cheaper just to buy another one then set up a tracking system.
So we just got a case of a user making it not very clear what he wants. All of the answers so far have focused on trying to keep track of the forklift. Some intresting solutions and some stupid ones (measuring wheel spin on a forklift forgets that these things slip OFTEN).
However none of you have so far questioned why he wants to do this NOR wondered if this was indeed the real requirement.
More likely he wants to keep tracks of the goods. In practice goods are often misplaced in warehouses. If something is in the wrong rack it can be a pain to find it. Or worse you only notice it when the wrong thing is sent to the customer because items were mixed up.
IF this is the real problem, tracking goods, then all the given solutions are at best incomplete and at worst totally and utterly wrong to a degree that explains why so many it projects are overtime and overbudget.
But surely knowing the location of the forklift whill help tracking goods? Nope. Why? Because there is no way to track were an item was dropped by the forklift. Why not? What if the forklift operator drops of his pallet for a moment to shift another pallet out of his way? What if he takes the contents of a pallet off to put them on another pallet?
Basically this has to do with warehouse procedures that aren't followed or don't match the real situation. Ideally every storage place in a warehouse should have a unique identifier. Each storage place is catogorized as to what it can contain. Every item coming in FITS inside the storage place (not 1 item requiring two spaces OR 2 items going into the same space). At receiving the computer assigns an empty storage space. The forklift operator then takes the item to the required location and stores it. Ideally you want some kind of system to verify this BUT in real live all you need is good employees and good managers who allow each other to do their jobs.
If this system is followed then no goods will go missing. Problem is that this often doesn't happen. Neither does it seem to happen in the posters warehouse. If the forklift operators kept track off where they dropped items off OR dropped them off at the pre-assigned location then things wouldn't go missing.
Sadly he seems to go for a tech solution. Nice but it won't work unless he shifts his attention from the forklifts to the items. They are the ones missing in action.
For the most obvious failure and the sure sign that this p
MMO Quests are like orgasms:
You may solo them, I prefer them in a group.
What you and others are perhaps comparing this with is truck tracking systems. BUT these systems are for a totally different need. They track trucks (and not what is in them) with gps because they already know what is in the truck from the loading bill AND because it helps them decide wich truck is closest to pick up an new freight.
Every good that comes in a warehouse is labelled. No matter how short it stays. That label is what is tracked. Item comes in, is labelled and assigned a location, moved to location. When it is time to deliver it is taken from location and the location is freed. No need for bloody forklift trucking. CHECK ACTUAL WAREHOUSES. The only time you need to track the forklift if your are automating the forklift itself. But this is very expensive as you need to add things like weight sensors to tell when the load comes on and off, scanners to scan the load in front, good positioning. Far easier just to do it the old fashioned way with good employees. It only fails because management puts to much stress on the employees and doesn't listen when they report that the system ain't working because of a shortage of location or because storage locations are to small for goods. Or simple time constraints.
MMO Quests are like orgasms:
You may solo them, I prefer them in a group.
Use an inertial navigation unit.
These are two companies I've heard of from Idealab where I used to work:
Newbury Network's Location Server products uses 802.11b signal monitoring to do location detection. They offer a virtual docent system that uses the technology to allow the virtual docents to provide location-appropriate information.
Evolution Robotic's VSLAM lets robots use odometry AND visual data to update its position information.
Depending on the size of the warehouse and the manner of "occlusion" that occurs, I'd say wiring up the warehouse with lots of cameras and triangulating to a beacon would be a pretty straightforward method... The hockey puck, indeed!
Mount the 802.11 wireless access points on the ceiling of the factory and put a long vertically mounted antenna on the forklift (to minimize or eliminate the effects of floor geometry).
Then use triangulation software that allows you to train it and keep collecting data points until your target precision is hit.
The reason that it can be true that 1+1 > 2 is that very peculiar nonzero value of the + operator
I implemented a similar system in a brewery warehouse about 12 years ago - we used ceiling mounted RF tags and a reader on the roof of the fork (with a 10-12 foot range), connected to the fork release/grab mechanism to match ceiling tag against location when the forkie picks up stock or puts it down - and there you go - he doesn't have to scan anything, but you know where that pallet is all the time. There are/were alternative solutions which involve mounting the reader under the body of the fork and burying some seriously small tags in the slab of the warehouse floor - this may be more practical depending on your individual needs. Out of doors, we used D-GPS (differential GPS) which was able to use standard GPS and a calibrated base station with a precisely-known position to iron out the fluctuations in GPS positioning - this allowed us to locate straddle forks at the local docks to the level of accuracy you describe, but as you say it's impractical indoors. I'd say any method attempting triangulation is doomed to being (i) expensive , (ii) overly complex, and (iii) unreliable in a normal warehouse environment where there's lots of lovely racking and stock to get in the way of RF transmissions. HTH Dan Shannon
I'm not sure if the resolution is tight enough for your needs, but this sort of problem is frequency used in large agri-businesses. Many of the larger operations use fixed position repeaters to increase the resolution of GPS and keep track of yields and soil conditions while planting, tilling, or harvesting.
Check out the MIT Cricket Indoor Location System. http://cricket.csail.mit.edu/.
? sid=116
It is commercially available from Crossbox Technologies http://www.xbow.com/Products/productsdetails.aspx
Ditch the driver and get yourself some Laser Guided Vehicles or Automated Guided Vehicles. These are in widespread use in industry and distribution in the US and Europe. If you go to the expense of tracking a forlift I can't help but believe that you could get a nice ROI by adding the hardware/software to replace the driver.
Simple people talk of people, better people talk of events, great people talk of ideas.
It sounds like a dirty word, but it's probably your solution: outfit your storage area with RFID, and put a sensor on the forklift so that when the forklift drops the item off, you can sense which RFID tag the forklift is close to.
The RFID tag may need to be in the floor, on the shelves, or whatever else: the beauty of this solution is that it's cheap: the tags are cheap and small and just need to be embedded around the floor: you then need to manually associate the tags with a location (shelf, etc).
What you need on the forklift is the sensor, which only needs to read the current tag and the current height of the forklift tray (not necessary if the tags are in the shelves themselves where height is implicit): the forklift then simply needs to transfer that information back to a base station, and 802.11 would probably be suitable (unless there's a ot of industrial noise). You could further automate the process by putting weight sensors onto the forklift to tell whether things have been loaded or unloaded and what weight is being transferred.
This would be a an alternative to using some sort of positioning system where you try to triangulate the position of the forklift / etc, which to me would seem to have greater probability of error.
Way better than 1ft resolution out of the box, but if you are serious, I bet they could dumb it down a bit.http://www.arcsecond.com/index.aspx
Seems to be sort of what you are looking for.
fiber optic gyro tracking stuff for warehouses
Just ran across the release and remembered this past ask slashdot. Hope it's useful info.
It's called Ranger, and uses radio to provide cm accuracy positions indoors or anywhere GPS isn't an option.r .htm
Check out the website:
http://www.ensco.com/products/homeland/rng/rng_ov
There's contact information at the top.
Because of the number of obstructions, no sonic or propogation delay mechanism would work. No stationary cameras would work. Electromagnetic mechanisms would only be accurate to feet. You could put barcodes in random locations and get a camera on the vehicle to automatically detect the barcodes.
Put a tall mast on the forklift and get two spools of fishing twine. Fasten the spools high on the walls in two adjacent corners and tie the twine to the mast. Measure how much twine is unspooled, and triangulate.
Tracking moving assets like fork lifts is tricky. Specially when you want it down to a foot of accuracy. I've tried with a dead reckoning FOG gyro, indoor gps and rfid. The expensive FOG gyro drifts over time and needs often recalibration (every minute). The indoor GPS has issues with obstacles like walls and shelves. RFID simply only reaches a few inches of rf range. Wherenet.com has an interesting solution, but is very expensive. A better solution is using a mesh network with both RSSI and Ultrasonic Technology. The ultrasonic sensors like Cricket from MIT are easy to setup and provide accuracy in inches. They do have an issue with walls and obstacles, so you want to use RSSI when the ultrasonic sensors are out of range. This combination would provide excellent location and a mesh network to communicate the status of assets back to your server. More details call me at net_mote@yahoo.com