The collar you see there had a Zaurus PDA which ran the herding algorithm and managed the ad-hoc peer-to-peer wireless network between the cows. Each collar also had a GPS unit and a tiny guitar amplifier/speaker was used to amplify the sounds (WAV files played by the Zaurus).
We had only minor success in using sounds to herd them, mostly they would ignore the sounds after a while. CSIRO in Australia and a USDA researcher who patented the idea 20+ years ago extended the work further eventually using electric shocks from an ear tag to train the cattle to associate a sound with a shock, then later using just the sound to manage the cattle. They got very good results with this while minimizing the stress on the cattle due to shocks. So I'm somewhat dubious that this company has had much success with just a buzzing sound. Even shocks won't stop a cow if it is determined. They are herd animals however and feel uncomfortable if away from the herd, so if most of the herd obeys eventually the stragglers will come round.
We stopped working on it when we started getting inquiries from companies wanting to use the technology on prisoners. Herding people electronically is just wrong.
We also talked with an animal ethics researcher who was concerned that the cattle would not be visited by the ranchers as often and would suffer because of it. For example, sensors can tell you if a cow has a fever but not if it has a stone in its hoof. If you trust the sensors too much and don't visit the animals as much the animals can be harmed.
So I actually started a business back in 2012 specifically for animal tracking on farms (cows and sheep).
We tried BLE, image detection, loads of different approaches which worked at "prototype scale", the problems were:
1. Farms are geographically big, tend to have hills and metal stuff which bounces signals all over the place. LoRa is slow, BLE has a short area. If you're close enough to connect with the cow, you might as well get out the truck and do whatever it is yourself.
2. Animals, especially cows, can and will destroy anything they trample. That includes expensive dongles, wifi base stations, pipes, electric lines. This infrastructure is flimsy and hardening it costs money, especially against water ingress.
3. Cost. A $20 tech strapped to a $100 animal is a considerable percentage of your RoI. The margins are so low, it's just not worth doing this at scale.
The thing is, there is no tech solution that does cattle herding better than a farmer.
Did you look at drones recently? I ran into the same problems as you years ago on an outdoor monitoring PoC with stations spread over several dozen miles (it was cheaper to pay an intern than to deploy & upkeep the tech). For farming use cases, though, drones + CV at least seem to be a pretty decent solution to some problems.
Yeah, we've tried drones. Problem is identification. E-tags on animals require a proximity of below a metre and stock are jittery around flying robots.
Like you said, it's actually easier to get a quadbike and go look.
This might not work for traditional livestock operations but for certain boutique farms and ranches (think farms raising Wagyu and other sought after breeds - and even non-meat concerns) the proportional cost is much lower - and the benefits potentially much higher, given the potential economic losses if certain animals (such as a prized stud) were to encounter some kind of preventable incident.
Tags do not go on at birth. The mother cows have their calves and then much later when the rancher rounds them up the calves tend to stick by their mothers so we know what calf belongs to what cow.
Almarai, probably still the largest milk producer in the world with its main production facilities in Saudi Arabia, simply turned off the air conditioning in the right place at the right time, and the cattle herded freely to where it wanted to.
Coincidentally, exactly where the herders also wanted them to go.
Impressive.
Cows are just like any other people.
So for cow logistics rising global temperatures may come in handy!
Of course, in the long run you may have to tinker a bit with the heat resistance of these organisms.
I’ll just add that smart collars and sensors are pretty much the norm on cows, at least in Western Europe (which is generally at the forefront in terms of animal care and food safety regulations).
The novelty here is the remote herdkeeping functions. The limit to adapting such tech is usually collar cost and battery life (the industry usually aims for 5+ years of battery life).
Of course the environment is pretty hostile too. Neck collars get snagged on things, exposed to mud and rain, etc.
Anyway, it’s not like “technology finally makes it to farms”. Dairy cows are the worlds most monitored animals.
I did some video work for a Dutch company selling such tech into Australia. Quite amazing to see it upfront. One dairy family talked about how it took them from scraping by to rekindling their enthusiasm for their farm.
That particular farm had collars and auto-switching gates plus four milking robots.
Cattle destroy everything they can reach that isn't manufactured from steel. We have to build fences over our water tanks to prevent them from destroying automatic water valves. If they can get to e.g. a phone pedestal it is flattened within a week. Properly-built barbed wire fences deter them, but as soon as a post or wire loosens they're going through. Ear tags were a major innovation, in that they are made of plastic, in constant reach of cattle, and typically are only destroyed at a rate (I'm estimating here) of one per year per twenty cattle. I know that collars have been used to e.g. attach bells to cattle, and the pictures at the Halter site seem solidly-constructed, but I do wonder about reliability...
Pastures are most productive when cattle eat the grass in an individual section to a particular minimum height, and then are moved to the next section. Cattle are fairly easily controlled via trivial annoyances (which makes various barbaric methods like hotshots used by some ranchers particularly unethical), so it makes sense that noise/vibration could move them in a particular direction. There would probably be some leakage, since different individuals have different personalities and tolerances. It looks like you could see the animals who escape on the map in the app, so that would be nice.
Since we AI quite a bit the heat detection would be great. I wonder how it works? Maybe the collars communicate their relative positions to each other?
>Virtual fences keep the cows in designated areas,
While the idea of vibrating collars to steer cattle makes me uneasy (what comes next?), I am somewhat hopeful this can address the issue of barbed wire fences.
there is no way a vibrating collar will stop a cow in doing what they want to do. Cows are herd animal meaning they fight to establish rank. There is no way 2 cows close to a fence can be stopped from breaching the fence, no matter if it's barbed wire or if the fence carries 20K Volts. They will walk right through it - or better push each other right through it during a fight. Not even an electric shock would stop them and a vibration collar even less.
I can confirm, numerous times I've seen cows grazing or even running with barbed wire and fence posts dragging behind them. When they've decided they want to do something, it's pretty hard to stop them.
I don’t think the point of the collar is to stop a cow going somewhere stupid, but rather to influence them to move to feeding areas and milking stations.
If a cow automatically moves to the right area then they don’t have to be herded. This reduces time. Some cows will likely need encouragement but it sounds like this reduces not only time, but stress on the livestock.
They’re a real pain to move. The article barely covers the myriad benefits of being able to have a collar like this to do virtual herding/fences:
* more flexible and appropriate sizing of grazing areas based on current herd size and where the grass is literally greenest
* the above can also increase dairy yield per sq/m by a considerable amount (I can’t recall the exact numbers I was quoted)
* multiple startups I’ve spoken to in this area also think they’ll be able to use on device biometric sensors to detect when cows need to urinate. Before that happens herd them to dedicated area or drainage pans. Again expected increases in yields and less pasture damage.
* biometrics also provide early indicators of potential health issues.
Nothing wrong with wire fences though. They’ve clearly worked fine for over a hundred years.
Insurance companies have started requiring folks to remove them or pay extra for liability coverage from them. Someone told me the cost difference was worth swapping to electric fences and losing up to 6 cattle a year. I don't think they're accounting for who gets sued when their cows wander into the highway, tho.
This is a pretty awesome New Zealand company, been following them since they first got funding. I love farm tech! Another is OmniEye: https://www.irisdata.co.nz/omnieye I've seen this in person and it's fantastic stuff.
I note in the article they say "Holy cow!" as a play on words I guess, but here in Dunedin we have a milk company called Holy Cow: https://www.holycow.nz/
Always great to see a company from my city on here. This line on their page always cracks me up “creators of the world’s first sheep facial recognition model”
Cows already obey electric fences which are basically just a piece of string. (When really stressed/agitated they walk right through electric fences, apparently without much additional distress.)
This is just a wireless electric fence. Cue joke about the old lady and radio telegraphy[1]. :-)
I think the parent's premise is that the cows can't see a well defined fenceline with the smart collar, to visually know to stop, whereas they can with an electric fence (or other types of physical fencing). And that lack of visual cue would induce anxiety in the cow, perhaps not being as certain where it can go.
Re: visual cues, a de-powered electric fence almost works as well as a powered-on fence.
(It doesn't, because cows will reach through and around a fence to get bites of tall grass on the other side, and if it doesn't shock them at that point they'll start pushing through to reach more.)
What is 'far away' in Switzerland? I'm in Australia and have a small beef farm and there no way a bell could let you know where cows always are. They get into a gully on the other side of your property etc sound wont travel. I only have a small property I run as a side business vs many family farmers on thousands or 10's of thousands of acres. The really big properties get I to millions.
Also a bell doesn't stop them getting through a fence and into your neighbours property or heading bush.
A big part of this tech is pasture management. Ideally you want to keep a herd moving so they only eat the top third of pasture and move on. This can help automate that task.
Not least what they mention about real time feedback about cattle health.
I don't doubt bells are useful, but this tech is at another level of helpfulness. It's a smart phone vs yelling difference.
The big question is cost and measurable benifit. Having to mesh wifi a large property won't be cheap (maybe just water points for daily updates?), the cost per collar and how long they last and how this plays into grazing and health benifits will decide how viable this tech is.
I just want to add that mesh wifi is not necessary here. Given the amount of bandwidth needed (very little), LoRaWAN (1) with its ~3-10km range will be quite practical.
pasture management is indeed a big one, i heard that proper herding can increase drastically soil carbon intake in some cases. i do doubt however that climate management is the primary goal of most industrial farm with 1+ million cows. my comment was on the hi/low tech joke side originally, but i wonder if a technology that would allow even more industrial farming is a good thing.
depends what "training" means. it could be as simple as herding the cows as you normally would while also giving them the nagivational instructions from the collar at the same time, and then after 5 days you can start using the collar only.
It's aimed at dairy cows, so they'd be herded in for milking twice a day anyway.
Though having said that, the cows on the farms near where I grew up would line themselves up at the gate to be milked, and one farmer would just open the gate while another shooed the stragglers back towards the herd. If this thing is anything less than 100% effective it doesn't seem like it would save much work, since you'd still need someone on a quad bike to deal with the nuisances.
Cattle have individual personalities. Some like to stay centered within the herd, and some prefer to be hundreds of yards away. When left to find their own way in a large enclosure, my cattle typically form much smaller groups, like 15-20. This might not matter, if a particular pasture can be protected by controlling the herd's general behavior.
If we had this system the first thing to do would be to stop all this shitting on the road!
That was the theme of Quake2: captured humans were "automated" with implants. Today we start with cows, in 50 years we extend this to violent criminals and who knows, maybe in 100 years most of us will be automated for "safety reasons".
Suspiciously absent is info on the battery life of these collars. How does one go about changing them? Do you line up 100+ cows every few days and just sit there for a couple of hours changing batteries every time?
https://groups.csail.mit.edu/drl/animals_robots/cow_herding/...
The collar you see there had a Zaurus PDA which ran the herding algorithm and managed the ad-hoc peer-to-peer wireless network between the cows. Each collar also had a GPS unit and a tiny guitar amplifier/speaker was used to amplify the sounds (WAV files played by the Zaurus). We had only minor success in using sounds to herd them, mostly they would ignore the sounds after a while. CSIRO in Australia and a USDA researcher who patented the idea 20+ years ago extended the work further eventually using electric shocks from an ear tag to train the cattle to associate a sound with a shock, then later using just the sound to manage the cattle. They got very good results with this while minimizing the stress on the cattle due to shocks. So I'm somewhat dubious that this company has had much success with just a buzzing sound. Even shocks won't stop a cow if it is determined. They are herd animals however and feel uncomfortable if away from the herd, so if most of the herd obeys eventually the stragglers will come round.
We stopped working on it when we started getting inquiries from companies wanting to use the technology on prisoners. Herding people electronically is just wrong.
We also talked with an animal ethics researcher who was concerned that the cattle would not be visited by the ranchers as often and would suffer because of it. For example, sensors can tell you if a cow has a fever but not if it has a stone in its hoof. If you trust the sensors too much and don't visit the animals as much the animals can be harmed.