Spying on wildlife with biorobots: Interview with engineer Kamilo Melo

Kamilo Melo was building a robot shaped like a salamander for his Ph.D. in 2015 when an unexpected assignment came knocking on his door. Two producers at the BBC were keen to rope him in for a nature documentary.

The brief was simple and complicated at the same time: build robots shaped like a crocodile and a monitor lizard that could be left along the Nile River in Uganda with the aim of filming the real-life species and capturing their interactions with one another.

“It was one of the most challenging but, at the same time, the most fun things that I have done,” Melo, founder of KM-RoBoTa, a company that builds robots for scientific research and for industrial use, told Mongabay in a video interview.

For Melo, working with biorobots in the field in Uganda was an eye-opening experience. In a study published recently in the journal Science Robotics, Melo and his team described their experiences and takeaways while working on the BBC’s Spy in the Wild documentary series, which premiered in 2017.

Biorobotics combines biology and engineering to create robots that mimic biological systems. For instance, a biorobot built to look and move like a given animal, and equipped with a camera and audio recorder, can potentially document the animals behavior and interactions from up close.

Since the beginning of his career, Melo had been building robots that have been used for scientific research and study. “I use animals as a source of information to make robots that try to replicate certain things of those animals,” he said. “I do this in order to understand how animals work.”

Melo said there’s still a long way to go as far as using biorobots for biological research is concerned. Challenges abound in terms of the technology, logistics and funding. However, Melo emphasized the need for biologists and engineers to work together to get past the hurdles and make the field more efficient and applicable. “Biorobotics is more than building animals just for the sake of building animals,” he said. “It’s more like creating tools to try to understand how biology actually works.”

Kamilo Melo spoke with Mongabay’s Abhishyant Kidangoor about his experiences in Uganda, the scope of the field of biorobotics, and the challenges ahead. This interview has been edited lightly for length and clarity.

Mongabay: How did you get into biorobotics? Where did your interest stem from?

Kamilo Melo: I am actually not sure [laughs]. It started a long, long time ago. I was doing my bachelor’s project, and I decided to make a dragonfly. Back then, there was this professor in Berkeley, and he was working on a robofly. For me, it was very interesting to hear that they were replicating a little fly [at a scale of a] 25-centimeter [10-inch] wingspan or something. It was dynamic, where you would actually have everything moving.

It was very interesting to me to understand the physics behind it, and that’s why I started thinking about dragonflies. I was studying electrical engineering, and I realized that to achieve dynamic similarity, I need to build a machine that allows movements. For that reason, I started getting into robotics. I was interested in motion in nature, and animals give you all these possibilities. That was on one side. On the other hand, I think I’m also a frustrated biologist. I was always inclined to engineering, but I always admire the biology side and nature.

Mongabay: How would explain the scope and applications of biorobotics to someone who doesn’t know anything about it?

Kamilo Melo: From my perspective, I use animals as a source of information in order to make robots that try to replicate certain things of those animals. I do this in order to understand how the animals work. To break it down, I see the animals moving and know how their motion is very complex to understand. [So] we build machines that try to help us to explain it. If you make an animal replicate an experiment, it is very difficult for the animal to do it: you tell the animal, “Can you move your leg in the same way you moved it the last time?” and it will not be understood and it will be very difficult to do it. We just tried to replicate that leg with a robot. The robot can replicate the experiment many times and then you can have control in order to move the leg as you want and get all the measurements and statistical variability that you need.

For me, biorobotics is more than just building animals just for the sake of building animals. It’s more like creating tools to try to understand how biology actually works.

Mongabay: Could you give me a few examples of how this has worked in reality?

Kamilo Melo: A very recent one was a paper we published in 2021, which is about lampreys. We wanted to understand how it swims and how the swim is generated in the spinal cord. Lampreys are very primitive animals, and these animals are so simple that they’re just a spinal cord and some muscles that move. That means that they don’t have a very complex skeleton. They don’t have limbs. It’s just a spine that moves and oscillates. The question was how this swimming motion was generated. This is the biology question that is very important to understand because it sheds light on many of the things that we currently know about animals. So we built a robot in order to replicate the nervous system of that animal, and then tried to create a mathematical model that can control the robot. Having everything captured by this mathematical model, it was very easy to understand how these biological traits work.

I can give you another one. Three hundred million years ago, some of the first animals got out of the water and started walking on land. They were the amniotes. A group of scientists, they found the skeleton of one of these very primitive extinct animals and found some footprints of the animal. This is where robotics comes in. We cannot see and we cannot know how this animal walked because it is extinct. But we have a lot of modern animals like iguanas, crocodiles and salamanders that give us some idea about their locomotion. What we did was we took the skeleton of the animal and we reproduced that in a robot, and then we started testing different types of locomotion that are similar to these modern animals. We discovered that these animals that are supposed to be one of the first amniotes were actually quite athletic. That means they had a high posture and it gives us a new perspective of this animal which is different from the images we have in museums. That changes the perspective and it changes the ecology around these animals, and gives a little bit more veracity to what was happening long ago.

Mongabay: Could you walk me through your latest study based on your experience working with biorobots for the BBC documentary in Uganda?

Kamilo Melo: I have to say it was one of the most challenging but, at the same time, most fun things that I have done. At that time, in the laboratory where I was working, we were working with salamander robots. We build salamander robots in order to understand salamander locomotion. Then, these two people, producers from the BBC, came and said, “We want to disguise your salamander as a crocodile.” I told them that there is no way to do it. There is a big difference in the morphology of a salamander compared to the morphology of a crocodile. There are many changes in the limbs, in the size and the distribution of the mass and so on. So we had to build anew. Then my industrial and commercial salesman spirit came out and I said we needed to build two because it’s going to be very difficult to have repairs in the field and it will be very expensive. We got the budget to buy enough materials for building two robots. Then they said they also wanted a monitor lizard, which is the biggest enemy of the Nile crocodile. So we built that as well. We worked super hard and tirelessly for months.

Then we got it to the field, which was the other fun but also more stressful part. Because in the field there are many things that happen that you really don’t know will happen. For example, temperature was one of the biggest challenges. We knew that it was going to be hot. But we didn’t realize that inside the latex skin, it was going to be even hotter. So motors started shutting down and all the electronics were not made for that temperature. That was a big challenge.

Because of wild animals that can really kill you, we needed to build a robot that will be almost autonomous so you can control it from far away. This was also a little bit of a challenge. In a laboratory, when you work with a robot next to you, if something happens, you can immediately go and check. If there is a cable that is malfunctioning, you can change it quickly. But here, that wasn’t the case. We learned a lot about how to make things robust and how to make things easy to be changed in the field. Because if we damaged something, we would have just two hours of light to repair the whole thing. If we couldn’t repair things in those two hours of light, then the next day is basically lost and that’s a lot of money wasted.

Mongabay: What were your biggest learning experiences from that field trip?

Kamilo Melo: First of all, trust nature. When you design a robot, you take many decisions as a mechanical engineer or an electrical engineer. But you have never really figured it out till you put it to work in nature.

Also, I learned to be patient. When you are doing things in a laboratory, you probably do not get so frustrated as when you are in the field. When you’re in the field, you really need to be calm and try to outdo what’s going to happen in the future. This is why we put many of these things in the paper. Because we want to encourage people to do this, but also be aware of what can happen there.

The last thing is to keep everything as simple as possible. In the field, you’re not going to have anything. Probably just a screwdriver, and that’s it. If you make things very complicated and if you rely too much on very complicated techniques, it will get tough in the field. Like even 3D printing. It looks very simple to do things with 3D printing, but once you’re in the field, you don’t have your printer there. Instead of making very complex things, we should try to make it very simple and robust.

Mongabay: What were your takeaways from a biological perspective on how this could be better used for research purposes?

Kamilo Melo: A couple of people actually approached us later about how we can use these to study other animals. If you can bring a machine as close as possible to look like an animal, and not really invade their habitat, it will be a nice way to actually study behavior.

Right now, I am working with a person who studies monkeys in Africa. They want to bring a snake robot there, because they want to see the behavior of the monkeys and how they interact with snakes. A model of a snake that doesn’t move is not really good, especially for animals as smart as monkeys. When you bring something that is more realistic, you’re going to get better results and better data out of these studies. I guess this is one of the possible outcomes of using robots in the field: to use animal robots to study other animals.

Also, I had another request a couple of years ago from a professor in France. What he wanted to study was the calling of little animals to the parents, and how the parents react to this, and also the other way around. So we would have had to work with robots with speakers, microphones and cameras to get a first-person view of what is actually happening in the mind of the animals. It’s a more immersive experience than what you can do with just a camera overseeing everything.

Mongabay: How can this be done without intruding on the animals or causing harm to them?

Kamilo Melo: Our robots are not really harmful. I have a couple of robots which we brought into a space with babies and kids. It’s really very safe. We are not talking about industrial robots. We’re talking about the replication of motions that are very soft and cause no harm.

We do have to try to be as inoffensive as possible. But I suspect that if somebody from behavioral studies or from biology wants to use robots in order to bring them to the wild, I don’t think the impact is going to be bigger than what the real animal is doing, because all we try to do is to replicate. There will be a little bit of an impact because a human has to be close by. That human is going to now be the intruder into that environment. But it’s all for science.

Mongabay: What are the big challenges and gaps that continue to exist in the field of biorobotics?

Kamilo Melo: Oof, lots of them [laughs].

One is to do with actuation. The muscles animals have are a very complex machine that changes chemical energy into mechanical energy. So the losses between the amount of energy that you input as chemical and the output that you have are minimal. Even if there are a lot of losses, they’re minimal compared to what we have in an electrical system. That means that the force that you can produce [with a robot animal] is not really comparable to what a muscle can do. Not just in terms of the magnitude, but also the capability of the machine to carry its own weight. Muscles can carry a lot of weight and use less energy than what a motor will do because the motor is heavy by itself. We are far, far away from achieving the same behavior as muscles.

Muscles are also elastic and viscous. That means that you can cope with impacts and other disturbances in a better way. That’s something motors cannot do. In order to cope with that, you need to program the motors, add more complexity and add more engineering to try to get it as close to nature. But we are still far away from that.

In terms of control, there are now some machine-learning techniques that help robots behave and move better. But if you want to do all the motions of an animal, we still are far away from achieving that in a single robot. One piece of advice here is that you need to care about the hardware. It’s not just about software. Many roboticists work on simulation, but not many work on the actual hardware. There should also be a close link between the hardware and the system integration because everything is a single thing.

Mongabay: I am curious to know how artificial intelligence could play a role to make biorobotics more efficient.

Kamilo Melo: We know how to move the robots according to the laws of physics. With those, we can predict what’s going to happen. Now, we don’t have to have the robot make any decisions. This is where machine learning comes in, to make these decisions on a higher level. Once you build your robot and once you control all these low-level laws of physics, you can go to the higher level and make decisions about how to basically behave. In that case, artificial intelligence is more appropriate. It will be like a more developed brain.

At some point, when you start adding many legs to your animals, you have very complex behaviors that you need to replicate. For very simple things, like moving one leg, a dynamical mathematical model is enough. But at some point, when we start making more and more complex systems with more variables and more unknowns, it will become more complex. Then, it will be better to put everything in a black box and run one of these machine-learning algorithms, and let the robot learn. With that, you can achieve something close to animal behavior. Of course, you cannot unlink that from the hardware design. But there is a long way to go for all of this to happen.

Mongabay: How do you think robotics engineers and biologists can work better for wildlife research, monitoring and protection?

Kamilo Melo: The first thing is mutual understanding. Sometimes engineers are very rigid. They just want to know what it’s exactly like with the equations. Biologists can also be like that with their work. When we talk, there is a different language. So the important thing in this kind of collaboration is to have a person who serves as a bridge.

Secondly, from a commercial and industrial side, don’t cut funding. Robots are expensive. Good robots are more expensive. As a biologist, you shouldn’t see these as luxury machines, but look at them as laboratory equipment.

Abhishyant Kidangoor is a staff writer at Mongabay. Find him on 𝕏 @AbhishyantPK.

Citations:

Melo, K., Horvat, T., & Ijspeert, A. J. (2023). Animal robots in the African wilderness: Lessons learned and outlook for field robotics. Science Robotics, 8(85). doi:10.1126/scirobotics.add8662

Thandiackal, R., Melo, K., Paez, L., Herault, J., Kano, T., Akiyama, K., … Ijspeert, A. J. (2021). Emergence of robust self-organized undulatory swimming based on local hydrodynamic force sensing. Science Robotics, 6(57). doi:10.1126/scirobotics.abf6354