News & Events

ICDS Events

A drone flies over a riverbend surrounded by mountains.

A team of researchers at Penn State are working to make artificial intelligence (AI)-powered systems like unmanned autonomous vehicles (UAVs) more efficient for use in different environments. Credit: 9parusnikov/Adobe Stock.

AIMI director’s research team aims to make AI-powered systems more efficient

Posted on April 3, 2025

EDITOR’S NOTE: This story was originally published on Penn State News.

 

UNIVERSITY PARK, Pa. — A team of researchers at Penn State including Soundar Kumara, Penn State Center for Applications of Artificial Intelligence and Machine Learning to Industry director, are following a “selective learning” approach to make artificial intelligence (AI)-powered systems like unmanned autonomous vehicles (UAVs) more efficient.

For this project, Kumara, Allen E. and Allen M. Pearce Professor of Industrial Engineering, worked alongside Ankur Verma, who earned a doctorate from Penn State in 2024; Ayush Goyal, who graduated from Penn State in 2024 with a master’s degree in computer science; and Sanjay Sarma, Fred Fort Flowers and Daniel Fort Flowers Professor in Mechanical Engineering at the Massachusetts Institute of Technology.

Their approach collects only data needed for specific problems, instead of collecting all available data and sorting through to find what applies to a problem.

By combining a streamlined training approach, the team was able to reduce the amount of data, computing power and energy the systems need to maintain desired performance accuracies. The proposed approach, which was published in Scientific Records, resulted in a 435-fold reduction in computing resources not only in UAVs, but also wearable electronics.

According to Kumara, UAVs are aerial, on-land or underwater vehicles with autonomous navigation capabilities. This means that these vehicles use data collected by sensors to drive themselves and aren’t in need of human pilots. These vehicles can be used for search and rescue or resupply missions, infrastructure inspection, lunar and Martian exploration and navigating hazardous environments.

“Basically, we can put them in environments that may be dangerous for or inaccessible to humans to help achieve our goals,” Kumara said.

UAVs use different sensors such as cameras that see colors that humans can detect, as well as multi-spectral cameras, gyroscopes and accelerometers that can benefit from the team’s scientific approach.

“This approach has been tested by our team on sensor data from a variety of assets like industrial motors, pumps, gearboxes, gas turbines and more,” Kumara said. “Our approach exploits the structure or patterns of real-world sensor data, which is a universal property across different data modalities such as images, sounds or vibrations.”

The amount of sensor data to be collected for various sensing modalities and tasks is governed by the Shannon-Nyquist sampling theorem, which, according to Kumara, states that the sampling rate needs to be at least twice the highest frequency present in the signal to avoid information loss. For UAVs, this could be in kilohertz or megahertz, which is a massive amount of data in a very short period of time. An aircraft may also have up to 300 different sensors for measuring different physical properties which generate several gigabytes of data per flight, Kumara said.

UAVs also require charging and are typically battery powered. The team’s approach could extend the battery life and decrease the amount of computing power needed to analyze the sensor data. It could also help in reducing the weight of processors needed to fly.

To test their approach, Verma, who is the first author on the paper, developed novel neural network architectures that could jointly sample and train models on smaller amounts of data which resulted in fast training models and inference. Joint sampling, or collecting only the data needed for the project, and fast model training reduced the amount of data required by 10-fold and compute power by 435-fold.

Though the team has already applied their research to a few different use cases, they hope to expand their research through Lightscline, the company which Verma, Goyal and Kumara founded. They aim to commercialize this technology in various industries like space and defense.

The initial market study for this work is funded by the U.S. National Space Foundation I-Corps program.

 

 

 

Share

Related Posts