Autonomous Networking for Autonomous AI

Tools powered by artificial intelligence are getting better at talking with people. Experts at Florida Atlantic University have their eyes on a future where AI devices are better at talking with each other.

The generative AI boom of the 2020s has been unleashed by chatbots and text-to-image tools that respond to natural language and produce something that people find useful in their everyday lives. Tech companies are building AI into daily life as people increasingly trust these tools to write their emails, manage their grocery lists or even generate their computer code.

But the future of AI will go beyond text and images. Robotic AI platforms will be connecting with each other, training together and carrying out their tasks collaboratively, said Dimitris Pados, Ph.D., director of the Center for Connected Autonomy and Artificial Intelligence at the College of Engineering and Computer Science at Florida Atlantic.

“This center is building towards a future where AI-powered computing and sensing platforms will not only operate on their own, but network autonomously,” said Pados, professor in the college and the Charles E. Schmidt Eminent Scholar in Engineering. “Through distributed learning, they will share their training and perform together based on a distilled understanding of their environment and mission.”

The College of Engineering and Computer Science officially unveiled the Center for Connected Autonomy and Artificial Intelligence on its Boca Raton campus in 2021. The center has since accumulated millions of dollars in grant support from institutions including the U.S. Department of Defense and National Science Foundation. CA-AI has also attracted new high-level faculty experts in AI, mechanical engineering, robotic navigation, big data analytics computer and other areas.

While those are common fields to bring together for any AI robotics research, CA-AI also features the skills of internationally renowned scholars in networking and communications, said Stella Batalama, Ph.D., dean of the College of Engineering and Computer Science.

“AI and wireless communications are very much a joint problem, and the Center for Connected Autonomy and Artificial Intelligence has special expertise in wireless connectivity foundations,” Batalama said. “That’s the unique characteristic that goes beyond what other AI centers might be doing.”

The center isn’t just working on networks for AI devices: They’re building the technology that will allow AI platforms to network themselves.

“It goes both ways,” Pados said. “We develop connected systems of AI-powered, autonomous devices, and this connectivity is itself autonomous and based on AI principles.”

In 10 or 20 years, this could look like a fleet of robotic submersibles, aerial drones and ground vehicles all traveling in concert with true autonomy. Not following a programmed route along GPS waypoints or piloted remotely by people, but finding their own way based on their training and assessment of current conditions.

And each unit won’t just act on the input from its own sensors, but as a node in a network formed over whatever available connection best suits them. They will share data and pursue a single mission based on their collective intelligence.

“The future is autonomous AI systems across all modalities, from the bottom of the oceans to the sea surface, from the ground to air and space,” Pados said. “The vessels will talk with each other as needed and maintain that connectivity on their own. If something goes wrong, they will find each other. They do whatever is possible to maintain communication.”

The center’s researchers are already producing the foundational breakthroughs that will make all that possible.

One recent example uses machine learning concepts to overcome a common obstacle in wireless communications: electromagnetic interference. This kind of signal disruption can come from other traffic over a busy frequency or even a malicious jamming effect. But researchers at CA-AI, collaborating with the U.S. Air Force Research Laboratory, have “cracked the code” that allows wirelessly connected devices to automatically overcome interference on the fly, Pados said.

This new technology takes advantage of software-programmable radio chips that send and receive information over multiple antennas—a bedrock method for increasing wireless signal capacity known as multiple-input multiple-output, or MIMO.

The breakthrough is a set of algorithms that enable the machines to intelligently use the power of multiple antennas to precisely shape the wave carrying their signal to avoid interference.

“The waveforms are shaped by the nodes themselves using advanced machine learning concepts that our center has developed from scratch,” Pados said. “If the link is disrupted by interference, the two nodes can figure it out, changing dynamically at a millisecond scale.”

This work by CA-AI was recently featured in May 2024 by Nature Reviews as a research highlight. The first open, live demonstration by CA-AI was recognized with the best demo award at the Institute of Electrical and Electronic Engineers Consumer Communications and Networking Conference in January this year.

Fully realized AI networks for autonomous AI devices won’t arrive without the next generation of scientists. CA-AI supports opportunities for students from high school to post-doctoral researchers in computer science, mechanical and electrical engineering, and more.

“It’s really not one discipline,” Pados said. “You want people with different backgrounds and interests to come together and work towards this objective. This is really the future of AI”