One of the many interesting articles in the most recent Economist Technology Quarterly was a piece on virtual fencing for livestock. If pet containment systems at scale come to mind, you are on the right track.
The business problem:
"BUILDING and maintaining the fences needed to control livestock is an expensive and time-consuming business. The materials alone can cost more than $20,000 a kilometre. On top of that, there is the cost of repairing damage caused by wild animals and falling trees. And then there is the need to move some fences around, a bit at a time, so that grazing land can be used efficiently. Strange as it may seem, many ranchers would therefore like to get rid of fences—if they could."
The science problem:
"According to Dean Anderson, an animal scientist at America’s Department of Agriculture, and Daniela Rus, a computer scientist at the Massachusetts Institute of Technology, the answer is to move from real fencing to the virtual sort. The idea of virtual fencing is not entirely new. Pet “containment” systems, such as collars that give dogs a small electric shock if they roam outside a particular area, have been around since the early 1970s. But attempts to come up with a system for controlling free-ranging animals have failed."
The breakthrough idea (emphasis is mine):
"Dr Anderson and Dr Rus started from the observation that the job of a fence is merely to regulate an animal’s behaviour and asked if there was another way of achieving the same end. The result is a device dubbed the Ear-a-round, which acts both as a sensor of what an animal is up to and as a discipline on animals that are not behaving as their owner wishes.
The Ear-a-round consists of a small, light box that sits on top of a cow’s head, and a pair of earpieces made of fabric and plastic. The box contains a small computer, a GPS satellite-tracking device and a transceiver that enables it to be programmed remotely. The earpieces both keep the box upright and deliver commands—either sounds or electric shocks—to the animal wearing the device. The whole thing is powered by lithium-ion batteries topped up by solar cells."
The event processing part -- event detection, evaluation & response:
"The range that an animal is allowed to occupy is encoded using GPS co-ordinates. The GPS system determines the animal’s location, and an accelerometer and a compass housed inside the box track its rate and direction of travel. If an animal roams beyond the range specified, the device responds in a way determined by its wearer’s recent behaviour. The algorithms devised by Dr Rus are able to work out, based on past experience, how strong the message to turn back needs to be.
Minor transgressions lead to quiet sonic alerts or mild tingles; major ones to shouts or shocks. In both cases the cue is delivered to the ear opposite the direction that the animal is being nudged towards. Four years of research at a ranch in New Mexico have shown that cattle quickly cotton on to what they need to do."
The cow as downstream actor -- old fashioned stimuli:
"Not all the stimuli used to guide the animals are unpleasant—at least, not intentionally so. In April Dr Anderson set out to test whether a recording of him singing the “gathering songs” used during traditional round-ups would be as effective at herding cattle as irritating sounds such as barking dogs, or electric shocks. Four Brangus cows listened to recordings of him chirping “Come on, girls” at 30-second intervals. Almost immediately, he says, the herd began moving towards the corral. It is not clear whether they were encouraged by his singing, or were trying to get away from it."
The ROI -- cow leadership:
"Is all this really cheaper than fencing? At $600 a cow, not yet. But Dr Rus is working to get the price of the hardware down to the $100 that farmers will pay. Meanwhile Dr Anderson is about to start investigating how many cows actually need to be fitted with Ear-a-rounds to control an entire herd. He hopes that, by identifying a herd’s leaders and equipping them alone, this number can be reduced to a handful."
If this peaks your interest to read more about event processing, go here. On the other hand, if you've got "don't fence me in" on the brain, here's Bing Crosby and The Andrews Sisters.
Oh, give me land, lots of land under starry skies above,
Don't fence me in.
Let me ride through the wide open country that I love,
Don't fence me in.
Let me be by myself in the evenin' breeze,
And listen to the murmur of the cottonwood trees,
Send me off forever but I ask you please,
Don't fence me in...
This is creepingly close to our idea of where CEP makes most sense. We have a solution up and running which is rather close to this idea. Replace cow with truck and you can suddenly feed GPS positions, location update events in our world, to our CEP server (which is of course completely event driven) and produce immediate alerts when cows, eh, trucks are found in the wrong areas or off track. Stay tuned, it's public in a couple of months...
(No, we don't zap the drivers if ruleCore detects that they cross a virtual geofence, there are hopefully more civilized routines in place for this...)
Posted by: Marco Seiriö | October 24, 2008 at 08:17 AM
Marco - let me know when your solution is public. I imagine zapping the drivers could send them even further off-course! Perhaps you should consider the 'singing in the opposite ear' tactic ;-) -brenda
Posted by: brenda | October 24, 2008 at 12:05 PM
Sure thing Brenda, there's a public beta available in just a couple of weeks...
Posted by: Marco Seiriö | October 30, 2008 at 07:53 AM