An award-winning concept imagines robotic rovers roaming alien planets while learning from their past computations and each other.
The research by University of Southern California Viterbi School of Engineering students shows a group of autonomous robots working together to process and analyze the environment without plugging into a central server.
The team made this possible by using edge computing, an emerging technology behind the internet of things – an ecosystem of interrelated devices that can exchange data even when it is not connected to the internet.
The study had four small rover-like robots, called Turtlebots, linked similarly. It allowed the robots to “learn” from each other while processing data through a purposefully developed task-scheduling software and analyzing the information faster.
“As these robots complete their tasks, they collaborate with each other to do whatever computation is needed,” Bhaskar Krishnamachari, the team’s faculty advisor, said in a statement.
The software is based on a graph convolutional neural network (GCN), a deep learning method used to equip machines with the ability to analyze visual imagery by representing them as graphs.
“The scheduler gets as much as possible out of graph structures, like in the case of image detection,” Lillian Clark, a Ph.D. student and team member said. “This is really useful if you’re doing something like looking for signs of life or water melting on another planet.”
Aside from space exploration, such autonomous robots present an array of potential applications closer to home, including in areas as diverse as facial recognition, baggage handling, and military reconnaissance.
The Viterbi team won first place for their work at the 2nd Student Design Competition on Networking Computing on the Edge. An international contest focusing on edge computing recognized it for demonstrating how the use of GCN could increase robot computing efficiency without central servers.
Other examples of graph neural networks used to improve existing technology include a tool to more accurately predict flashover events, the leading cause of firefighter deaths.
More from Cybernews:
Subscribe to our newsletter