With current state of science/tech - can we replicate a tracking dog ?

My going in position is NO.

The question is around replicating a tracking dog that can follow a person’s scent on a flat field. With the current state of Science/Technology, can we have a robot that is capable of accomplishing this ? Here are some of the features of this robot to keep in mind :

1> Have only 2 sampling ports - spaced the same as a dog’s nose. Sampling to be done at same rate as a dog breathes
2> Be able to “pursue” the scent trail at the same speed a dog does.
3> Assume the terrain is flat and the robot can be on wheels (concentrate more on the detection and navigation abilities than locomotion abilities)

We’re not there yet, but we’re getting closer.

You will probably find this article interesting:

According to the article, they are currently fielding and testing robots designed to sniff out people in natural disasters. For example, you could have a building that is partially collapsed and is too dangerous to enter. So instead of sending a human into the building to check for signs of life, you send in a sniffer robot.

The article doesn’t give any details about how far along this testing is and how well the technology works at this point, but it’s definitely heading towards your robot dog.

I understand why you’re requiring the robot to move as fast as a dog, but why must it only have two nostrils? If we can do better than a dog in some ways, why shouldn’t we?

Why would we spend money developing such a thing, when evolution & breeding over thousands of years has produced bloodhounds (& other breeds) that are exceptionally good at this – and pretty cheap, and reproduce readily?

Robots don’t eat, defecate, or need medical care.

And the friendly ones don’t try to lick you to death. :slight_smile:

Robots ‘eat’ electricity & lubricating oil, defecate used oil & waste heat, and need routine maintenance care – pretty similar to a dog.

And you seem to think a friendly licking is a bad thing?

The amounts of “food” and volume/scheduling of maintenance for an animal versus a machine are very different. A co-worker of mine has a hunting dog he uses to flush pheasants, and described to me the thousand plus hours it took to train the dog to smell and seek out the birds. The dog needs to be fed and attended to every day, and he’s taken the dog out hunting one day on the past 8 months; and he complained that the dog didn’t do very well that day because it was out of practice.

One big advantage of machines is that when you don’t use them they require very little fuel or maintenance. I can park my quad with a dollar’s worth of fuel stabilizer for the winter and walk away… but an equivalent work animal like a horse would cost around $800 for food, general vet bills, hoof maintenance, bedding, etc… just to keep it alive and in reasonable condition, not to mention the burden of actually doing the daily chores.

You can’t switch a dog off and leave it in it’s box for months at a time though. Robo-pooch might also be more useful in noisy/hazardous environments. It could have capabilities exceeding those of a meat-dog, such as climbing walls, breaching doors, carrying equipment. You could even fit it with attachments like a net gun, a taser, signal flares etc.

Once you’ve got something with effectiveness comparable to a tracking dog, the technology can be refined and enhanced to the point where it is unquestionably superior.

Sorry Mmmiiikkkeee, you beat me to it.

Great Mikes think alike.

As for locomotion, you’d need to have human officers accompanying your tracker, anyway. So probably it’d take the form of a handheld tool that a human officer would carry, with the officer’s legs providing the locomotion.

You’ve evidently never been greeted by a large, VERY friendly, VERY exuberant dog. Too much of anything is apt to be unpleasant.

As there will already be technical limitations it wouldn’t make sense to precisely replicate the physical features of a dog in a robot unless there is a specific reason to do so. No reason to limit to one, two, three, or more nostrils; as well as the air flow. If one were to be building from the ground-up you go with what’s best.

Also, another thing about dogs is that like people, they get tired, are affected by environmental conditions, etc. I did a study on something related to this and it was some cool research. Like, the dog nostrils create a swirl of air to help capture particles and tumble into and out of the sinuses for greater perception. There was also studies on how well dogs performed in certain conditions which is why I had mentioned above about getting tired and the environment.

The advantage of a device that could analyze air-born and latent particles is that it could go on forever. You could have people in shifts work the device, or, automate it by some robot that just keeps going. Many reasons to use something like that over a dog.

Only thing is, nothing is that portable yet. But, for an example of timeline, look at alcohol breathalyzers. Those things used to be huge as in fire extinguisher huge, now, you can put one in your pocket.

My hypothesis is that the dog can tell the direction of the smell with the two nostrils - sort of like we can tell the direction of sound because of two ears. So yes, the Robot can have more more ports or nostrils but then it means that the “smell” sensors do not have enough resolution.

Not sure if that would work. IIRC the dog smells the ground closely - suggesting the smell is on the ground . Plus if using current sensors, they need a lot of time to analyze (and current sensors don’t have the resolution of a dog’s nose) - so the human officer will have a challenge deciding how much time to hold, which direction to go, etc. etc.

Besides what if you wanted the robot to detect say land mines.

OK, so you have a handheld device that’s close to the ground, like a metal detector. And maybe the sensors aren’t good enough yet, but that has nothing to do with the device being hand-held, because they wouldn’t be any better on a self-motive device.

The original question was about technological maturity. Smell sensors and locomotion equivalent to a dog are two separate technologies - agreed. However when you combine the two, usually you end up with a non-linear system that is difficult to solve. Even self driving cars cannot navigate through an unfamiliar terrain without roads. Sort of like guiding a missile to a target, except in this case, the target changes continuously and the machine itself has to detect the change and adjust.

There are several big advantages to an artificial sniffer.

  1. Training. As mentioned up-thread, it can take a long time to train a sniffer dog.
  2. Fatigue. As mentioned up-thread. I have heard that sniffer dogs can only actively work for short periods of time, before they need relatively long breaks. (Unfortunately I have no cite for this; I have not worked with dogs this way and I don’t even know if the specifics of my statement are true. Regardless, unless the technology is very finicky, it is likely to work many more hours a day than a dog.)
  3. Sensitivity (A). While dogs are very sensitive, it may be that an artificial sniffer may be more specifically sensitive to certain chemicals of interest. Additionally, it can “tell” the operator what it is perceiving. A dog may become distracted or follow a wrong scent (or maybe Timmy’s in the well).
  4. Sensitivity (B). Dogs are emotionally sensitive. This plays into environmental factors mentioned up-thread. Basically, the dog or the handler may be having a bad day, and this may affect performance. There are stories of dogs and handlers working earthquake sites, finding dead body after dead body becoming depressed by lack of success.
  5. Safety. Sniffer dogs lead with their noses. If looking for explosives or mines, that can be a bad thing. Dogs take a long time to train, and a skilled dog is a n expensive item. An artificial sniffer may represent an easier loss (depending on the cost of the sniffer). At any rate, if I was designing a sniffer for mine detection, I would use a very long intake tube, which might lower response time, but keep the expensive parts of my sniffer far from the point of action.

of course, this assumes an efficient, sensitive artificial nose.