A Penn State led research project has received a $12 million grant from the US Army Research Laboratory to develop a low power sensing system that could help autonomous drones and ground robots make faster, smarter decisions in challenging environments.
The proposed technology aims to give machines a form of digital intelligence inspired by the way insects and small animals process information. The researchers believe the approach could be particularly useful in remote or covert locations where reliable connections to cloud services are unavailable or undesirable.
Building intelligence at the edge
Many advanced AI systems rely on data centers, supercomputing infrastructure or substantial external power. That dependence creates challenges for drones and other edge devices that must operate independently in locations with limited connectivity and energy.
The Penn State team is working on an intelligent sensing node that could allow autonomous systems to process information onboard. Instead of sending large amounts of sensor data to a remote computer or cloud platform, the node would help machines interpret their surroundings locally.
This could enable drones and ground robots to respond more effectively in complex environments while reducing the need for bulky computing hardware and high power consumption, according to the university.
Learning from insects and animals
The project draws inspiration from the compact but highly effective brains of animals such as locusts and owls. Although these animals have relatively small brains, they can process visual and audio signals efficiently by filtering out irrelevant information and focusing on important cues.
Saptarshi Das, Ackley Professor of Engineering Science at Penn State and principal investigator on the project, said the research is focused on helping machines make intelligent decisions while using less energy.
The team also includes Wooram Lee, associate professor of electrical engineering and co principal investigator, along with other Penn State researchers.
Reworking how sensors process data
The planned sensing node will use sensors that can detect broadband electromagnetic signals emitted by electronic devices in the surrounding environment. The system will then process those signals using an architecture inspired by the cochlea, the part of the human ear that converts vibrations into signals the body can interpret as sound.
This design is intended to move some processing into the analog domain before the information is converted into digital data. The researchers say that approach could increase processing bandwidth and improve energy efficiency.
A major source of power consumption in many drone computing systems is the conversion of analog information from cameras and other sensors into digital data. By reducing or eliminating some of those conversion steps, the researchers aim to lower the energy needed for onboard computing.
Combining graphene and silicon
To build the system, the research team plans to combine graphene with conventional silicon semiconductor technology. Graphene is a one atom thick layer of carbon arranged in a honeycomb structure.
The materials will be used to transfer electrical signals into an array of memristors. These components can receive an electrical current and amplify the output, creating a potential pathway for more efficient sensor processing and computing.
The broader objective is to develop autonomous machines that can interpret information locally rather than depending on continuous communication with external infrastructure. That capability could prove valuable for drones and robots operating in areas where networks are unreliable, access is restricted or low visibility is a concern.
Implications for autonomous robotics
The project reflects a wider shift toward edge AI, where computing and decision making take place closer to the point at which data is collected. For autonomous drones and mobile robots, this can reduce latency, limit data transmission requirements and support operation in disconnected environments.
The US Army Research Laboratory is funding the project as part of its research efforts in advanced technologies for national security. Joshua Robinson, director of Penn State’s Materials Research Institute, said the work could contribute to more energy efficient systems for defense and other demanding applications.
If successful, the animal inspired sensing platform could offer a new way to design autonomous machines that make informed decisions with limited power and incomplete information. It also highlights how biological systems continue to influence the development of next generation sensors, processors and robotic platforms.
Source: Penn State University
