“And we can now run image recognition on the drone itself,” Wolf says. So the drones could be programmed to distinguish a dog poo from, say, a rock? “We have databases on the drone where it can look up and compare images. It can differentiate between a human being, a bicycle, a car or a ship. So, if you go further, this is similar. This is a piece of paper or this is the rock or this is a dog poo. If it can look up a database and say, okay, this is usually what dog poo looks like, then this is all technology that can be used for that.”
Zack Jackowski, chief engineer for Boston Dynamics’ Spot robot, puts it more simply: “The way the machine learning works, if you can visually recognise it as a distinct thing, you can train a robot to recognise it. If you have an easy time picking it out, a robot can have an easy time picking it out.”
“Of course, there’s a lot of different forms of poo that can look very different,” Wolf says. “Form and sizes and consistencies can vary a lot, if it’s on grass and has sunk down or decomposed – but for sure it’s possible.” The really good news is that Wolf says the crap dangling from branches is the easiest to identify. “Something like a bag hanging in a tree would be very easy to detect, and flag, because it will have a very similar shape and colour.”
This is the sticking point. Drones would be ideal for flagging and tracking dog poo deposits, but not the actual cleanup. In 2017, a startup in the Netherlands claimed to have created two poop-scooping ‘Dogdrones’, but the idea never took off. Volunteers willing to help in the testing stages were, perhaps understandably, thin on the ground. Besides, the scooping drone of the pair was ground-based anyway.
“Picking up a bag might be something possible, I guess,” Wolf says. “Picking up the poo itself, with like a little shovel, that would be hard to implement. You need to increase the size of the drone, the utilities, then that will make everything bigger and more cumbersome.”
Robots to the rescue
Robots are frequently envisioned as fulfilling jobs involving the three Ds: “dirty, dangerous and dull”. Clearing up dog mess certainly ticks all these boxes. So, for reliable ground clearance, therefore, what we really need is a robot that can go wherever dogs can. This could be one of the best use cases for Spot yet. Indeed, the robot has already been fitted with its Spot Arm for clearing up trash outdoors.
Boston Dynamics itself says there is interest in a use case for “Spot + Spot Arm” to be used for cleaning of public spaces and along roadsides, and the operation is in essence similar to the “fetch” behaviour the BD engineers have already demonstrated.
It may not be as daring as, say, being used in military training, but Jackowski says devising ways that Boston Dynamics’ canine robot can help society, even if it’s just by cleaning up, is what actually gets his team fired up. “Our engineers love doing applications where robots can directly help people. There’s something that feels really good about having a robot cleaning the environment around us, and relieving people of this,” he says.
As for relieving people of picking up doggie doo-doo, Jackowski says Spot could handle this right now. “If it’s in a bag, that’s easy to recognise, and it’s actually probably pretty easy to pick up.” What about unbagged whoopsies? “You probably want to give the robot a plastic bag over its gripper. And, if you were really getting into it, you would have a mechanism that lets the robot have a plastic bag that it can change.