Looking to nature for robot inspiration

The guiding principle of Kirstin Petersen’s Collective Embodied Intelligence Lab at Cornell Engineering is finding ways to create and deploy many simple robots that can accomplish more than one complex robot could get done alone. “And of course these simple robots can’t just succeed only in the lab,” says Petersen. “They need to be able to do what they do outside of the lab in the real world.”

As Petersen says this, she is seated at a table in her lab on the fourth floor of Rhodes Hall. Anywhere a visitor looks, there are robots and robot parts.

Which is why it is so surprising when Petersen starts to explain a current project she is very excited about. “We are working with Al Molnar’s group (associate professor in Cornell’s School of Electrical and Computer Engineering [ECE]) to affix sensors to honey bees so that we can use the bees to collect data for us.” Perhaps sensing the confusion of the interviewer, Petersen goes on, “Honey bees make the perfect biomonitors for the environment. If we can figure out what they know, then we don’t need to create bee-like robots to do the same job.”

Petersen is an assistant professor in ECE. She was first inspired to think seriously about robots when, as a high school student, she visited NASA’s Jet Propulsion Lab (JPL) in Pasadena, California. While there she saw the Limbed Excursion Mechanical Utility Robot (LEMUR) NASA was developing for the surface exploration of planets, moons, and asteroids.

Petersen grew up in Odense, Denmark. She attended the Odense University College of Engineering, where she earned her B.Sc. in electro-technical engineering. During her undergraduate studies, she went back to JPL, this time for an undergraduate internship. Petersen earned her Ph.D. in computer science from Harvard. She worked closely with Professor Radhika Nagpal at Harvard’s Wyss Institute for Biologically Inspired Engineering. The research that came out of Petersen’s time at Harvard was named one of the top ten scientific achievements of 2014 by Science magazine. The work involved a group of robots that were programmed to work together to build simple structures by sensing local progress and inferring the next steps.

Given this background it makes sense that Petersen would look to nature for inspiration for her robot swarms. Evolution has solved some very complex problems with some elegantly simple creatures. Termites, ants, and bees all build intricate structures without any single individual knowing the overall plan.

Petersen herself sounds a little surprised when she discusses her work with honey bees in more detail. “Al came to me and he said ‘I have these sensors—could these be useful for your robots?’ It wasn’t long before I realized we don’t need the robots at all—we already have the bees.” Petersen and Molnar want to use the bees to gather data about bee behavior, agricultural pollination patterns, and the role of plant species diversity to bee behavior. The hope is to be able to fit the bees with sensors and then allow them to go out and do their thing. As they return to the hive, a receiver will collect the gathered data wirelessly as the bee passes near it.

The sensors will use basic measurements—angle to the sun and time traveled—to give researchers a picture of where each bee has gone while it was out of the hive. Because the sensors are collecting such basic information, they can be very light and efficient. In fact, beekeepers already have sensors they use to gather some data and Molnar’s sensors are lighter and smaller than the ones currently in use.

This research excursion into apiology does not mean that Petersen has given up on robots. In fact, she has several other threads of research active in her lab. One is a soft robotic tool that can access and characterize grape clusters as they grow on the vine without damaging the fruit. Another is autonomous robotic construction. This project is in its early stages and involves soft robots that can put themselves in place and then become rigid and become part of a larger structure that is self-assembled.

As Petersen reels off the work happening in her lab, it becomes clear that she must have a lot of people working with her for all of these projects to be progressing. “I teach a class called Intelligent Physical Systems,” says Petersen, “and I have students from that class asking about research possibilities all the time. I had about 15 undergraduates in my lab this year, along with three Master’s students and seven M.Eng. students.” Petersen also has five Ph.D. students in her lab, and a technician.

“Much of the research really is driven by what the students want to spend time on,” says Petersen.  “At the same time, if you join my lab I will push you to do at least one thing that is out of your comfort zone—that’s often where we can learn the most.”