Say you let loose a dozen robots into the ocean.
The launch is easy. Harder is figuring out how best to program this ‘search party’ to collect data, communicate with each other underwater, and send information home to the mother ship.
The Office of Naval Research has awarded a trio of Viterbi faculty members a three-year grant to find an effective way to use such a network of autonomous vehicles in dynamic environments, using advanced sensing technology.
Viterbi Professors Urbashi Mitra and Shri Narayanan.
“This project is a major undertaking,” says Mitra. “It requires an inter-disciplinary and collaborative approach to develop a holistic solution to the problems of joint sensing, classification, communication and control.”
Working on Mitra's team are two professors with critical areas of expertise. Gaurav Sukhatme, computer scientist who is a specialist in robotics and autonomous vehicles, has experience in underwater research while Shrikanth Narayanan, with appointments in electrical engineering, computer science, linguistics and psychology, is an expert in sensing and classification systems, and will also incorporate his research on the underlying language structure of human speech.
The three are collaborating with researchers from MIT, Northeastern University and The Woods Hole Oceanographic Institution and USC undergraduate and graduate students will also contribute.
There are a host of challenges central to communicating in an underwater domain. Radio-based technologies, such as those found in cell phones, are not effective. While sound can be used to transmit signals underwater, sending video is much more challenging due to bandwidth limitations. Using energy efficiently and selecting the best source of power for the autonomous vehicles are additional issues. “Where are you going to recharge your batteries when you’re 200 meters down in the ocean?” Mitra asks.
Another challenge? “We can’t put sensors everywhere,” says Narayanan. “There’s also the question of how to process information while requiring the least possiblr amount of human intervention.”
Ideally, the networked vehicles’ exploration will be guided by what they measure. Central to that task is programming them to make the right decisions about what to do next, over long ranges, with almost no error.
Viterbi Prof. Gaurav Sukhatme in his laboratory with a robot prototype.
“What’s the proper algorithm, the best sequence of instructions?” asks Sukhatme. “These are some of the issues we will be exploring.”
The team isn't developing the actual autonomous vehicles, at least not yet. Instead they are focusing on working out the theory behind the networked systems. However, Sukhatme’s lab has long specialized in autonomous seagoing robots and may later test the team’s work.
Graduate students from Mitra and Sukhatme's research groups are currently collaborating to instrument Sukhatme's robotic boats with Mitra's underwater acoustic transducers so that the boats can "talk" underwater.
Potential applications encompass civilian, military and industrial uses. Deploying a large number of networked, autonomous vehicles can help identify underwater mines, collect data for environmental impact assessments, and improve navigation systems for ships and submarines. Ultimately, such systems may one day help chart the deep, unknown depths of the ocean.
“Right now, we have a better map of Mars than the ocean bottom,” says Sukhatme.