Ani Hsieh’s Robot Teams are Exploring New Frontiers

by Lida Tunesi

A heterogeneous team of robots equipped with different sensing and measurement capabilities coordinating to monitor distinct regions in the environment. Credit: Prorok et al., ICRA 2016

As a child, Ani Hsieh wanted to become an astronaut, but she soon realized there were a few obstacles that no amount of studying or training could overcome.

“I’m short, I’m hopelessly nearsighted, and I get really bad motion sickness,” says Hsieh, research associate professor in the Department of Mechanical Engineering and Applied Mechanics. Thankfully, she saw an alternative. “I thought the next best thing would be to be an engineer. If nothing else, I could work on space robots.”

Ani Hsieh

Hsieh did find her way into robotics, but rather than the cosmos, her work primarily focuses on the similarly unknown and isolated world of the oceans.

“Robots make perfect sense for ocean exploration because humans can’t operate there without a lot of support infrastructure,” Hsieh says. “Same with space exploration. The way I like to think about robots is as tools that extend our reach and give us a richer view into this fascinating world.”

Hsieh’s research group programs teams of robots, from small groups up to large-scale swarms, for marine studies and exploration.

“We are interested in coordinating teams of robots to do useful tasks,” Hsieh says.

Though individual robots might be suited to some tasks, there are often benefits to using teams.

“The reason you want to work with more than one is because of the naturally built-in redundancy,” Hsieh says. “The hope is that if you’re clever about how they coordinate and cooperate as an entity, they can act as more than the sum of their parts.”

A team of robots cooperatively transport an object or a larger vessel in an aquatic environment. Credit: Sartoretti, et al. ICRA 2016

Groups of robots can be useful in many situations where multiple spaces need to be monitored or explored at once, such as surveying the safety of rooms in a building after a fire before first responders go in, or in the ocean, where robot teams can help answer environmental questions.

“It’s such a big space that you can’t take one point measurement and say something about the entire ocean,” Hsieh says, “so it makes sense to use a large team of robots. For example, people are interested in studying coral bleaching. To do this, you have to monitor the health of the reefs, which means you have to look at the entire reef, not just certain spots.”

Hsieh uses tools from fluid mechanics and dynamical systems theory to model and program how teams of robots can work together to achieve a goal. Her work considers the many complications and constraints that arise in trying to get a team to work together — in or out of the seas.

“A good example is Amazon warehouses,” Hsieh says. “At the high level, the robots are all doing the same thing, so you really get the benefit of having more than one. But at the same time, there are new challenges. You can’t let them run into each other, and you have to make sure the workload is spread out so no one is doing the robot equivalent of twiddling their thumbs.”

Each different environment has its own considerations. Things would change, for instance, if the robots were searching the woods rather than working in a warehouse. “Natural environments introduce a new layer of complexity. Engineered things are predictable because they tend to have a set of rules everyone abides, but if you’re out in the ocean or in a forest, things are more challenging because there’s a lack of structure,” Hsieh says.

Marine environments present their own set of challenges. For one, the environment affects the robots. This wasn’t always an issue, says Hsieh.

“Historically, robots were massive things,” Hsieh says. “It was important that the robot didn’t punch a hole in the wall or send a human flying. It was all about making sure the robot did what it was supposed to do without causing any harm. Now, as robots get smaller, the way they move is more impacted by forces in the environment. When you’re big, the wind can blow on you and nothing happens. When you’re small, a wind gust can blow you off course. I’m interested in that connection — how does the environment interact with the robot?”

Fortunately, these forces aren’t necessarily a problem. In fact, one of Hsieh’s main interests lies in using those forces to the robots’ advantage.

“In one of our projects, we are thinking about how to plan energy-efficient trajectories for marine vehicles that leverage the flow of the water,” Hsieh says. If you have mapped out the locations of ocean features, such as sinks or high-current regions, you could plan the robots’ trajectory to use that naturally occurring flow. “How do you leverage flow to make sure you stay in one region, or go between regions to get the measurements you want?”

Conversely, in the water, the presence of the robots also changes the way the water moves.

“For ocean monitoring, we also need to consider how the vehicles deform the water around them as they take measurements, and how that impacts the measurements,” Hsieh says. “One of our projects delves more deeply into that interaction. The idea is that if we understand it, we can do a better job of getting vehicles to navigate and cooperate effectively.”

A robot tracking a moving ocean front by planning and navigating in a pattern that ensures the robot captures measurements on both sizes of the front. Credit: Kularatne et. al., ICRA 2015

Right now, Hsieh’s lab is predominantly focused on programming the robots, but they are hoping to get more into the design of the vehicles, exploring the relationship between the shape of the robot and its ability to move in water and air.

To study these questions, Hsieh’s group uses a mix of experiments. They employ mathematical models to understand how things should work in a simplified setting, and download ocean current data for a more realistic simulation.

To simulate the messier conditions of the natural world, Hsieh uses the enormous water tank in her lab. “We do things in the water tank partly because ocean data is sparse. And while it’s very hard to create realistic flows perfectly in a tank, it’s a nice way to introduce real-world conditions in a controlled setting.”

The variety of ideas and methods involved in robotics holds a lot of appeal for Hsieh. “My undergraduate degree was in general engineering,” she says. “That very general background fits with my interdisciplinary approach. When I decided to go to graduate school I thought robotics seemed to have the right mix of application versus theory.”

Hsieh received her Ph.D. at Penn Engineering, studying how to coordinate swarms of vehicles, and found many reasons to come back.

“There is a can-do attitude, and a very collaborative environment here. Nobody says ‘no,’ they say ‘we’ll figure out how to do it.’ It’s very nice to be in that situation when my research is so interdisciplinary,” Hsieh says. “I rely a lot on my collaborators — oceanographers, applied mathematicians, physicists — and we enjoy working with each other.”

Hsieh plans to keep making the most of these opportunities. “We’re still in that fundamental stage of research, but I hope that our work will inspire new ways for humans to interact with the ocean through the use of robots.”