Engineering’s Gourav Datta receives nearly $600K to further resilient and energy-efficient generative AI research
Gourav Datta, assistant professor of electrical, computer and systems engineering at Case School of Engineering, and a team of researchers were recently awarded a $592,000 National Science Foundation (NSF) grant to investigate resilient and energy-efficient generative AI for secure hospital robotics over NextG networks.
In addition to Datta, who is the principal investigator, the research team includes co-principal investigators:
- An Wang, assistant professor at the CWRU Department of Electrical, Computer and Systems Engineering;
- Zonghe Chua, PhD, assistant professor at the CWRU Department of Electrical, Computer and Systems Engineering; and
- Sandeep Chinchali, assistant professor of engineering at The University of Texas at Austin.
About the award
Hospitals are starting to adopt robots that assist surgeons, dispense medications and take care of patients. These robots must respond promptly and continue functioning efficiently even during low or disrupted internet connections without compromising patient confidentiality. Thus, this research seeks to design robots capable of operating independently, powered by their own lightweight AI brains that would interact with powerful cloud computers only when required to exchange crucial information efficiently and save energy. Moreover, the robots will be capable of learning and adjusting to the situation, as well as collaborating effectively within the robot team.
This project incorporates advanced wireless networks technology and artificial intelligence at the network edges to enable secure, real-time hospital robotics based on four integrated objectives. Specifically, the first objective is developing an energy-efficient method of synchronizing compact vision-language-action models and world models through the application of model distillation and event-driven model updates. The second objective involves developing adaptive coordination techniques to achieve privacy-awareness and resilience in the operations of multi-robot teams amid varying conditions of network connectivity. The third objective aims to develop a hierarchical federated learning algorithmic framework utilizing knowledge distillation and Byzantine-resilient aggregation and the fourth objective validates the developed algorithms and approaches on both physical robotic platforms and their high-fidelity simulations.
This research improves healthcare delivery by enabling secure, reliable, and energy-efficient robotic systems that support clinical staff and enhance patient care, with particular benefit to underserved and infrastructure-limited settings where dependable connectivity cannot be assumed. The algorithms and open-source testbeds will carry over to other safety-critical areas, among them eldercare, disaster response, and smart manufacturing. The work brings research outcomes into graduate and undergraduate courses, supports kindergarten through twelfth grade summer programs that expand participation in science and engineering and provides interdisciplinary training opportunities in artificial intelligence, networking, robotics and cybersecurity.
