EUVIMEDEUVIMED entdecken
Crossref

Energy-Efficient Agentic AI: Long-Term Designs for Big Autonomous Systems

Rajesh Lomte, Shailaja Pede, Pallavi Nikumbh

International Journal of Drug Delivery Technology · 24.04.2026 · 16 (18s)

DOI: 10.25258/ijddt.16.18s.70 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Drug Delivery Technology, online erschienen am 24.04.2026. Verfasst von Rajesh Lomte und 2 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Autonomous AI agents are now operating on a scale which would have been considered utterly impossible even a few years ago. However, the computing requirements of these systems have increased in accordance with their capabilities and quite a lot more than when we examine the energy consumption patterns of large-scale agent systems, are the figures that it is difficult to ignore. This is to address this problem head-on. We have developed a framework hierarchy that enables autonomous agents to make informed decisions with a lot less energy. Our design contains a number of key ideas of novelty, including dynamic resource allocation, scheduling algorithms with regard to carbon content of energy sources, and design principles borrowed from the concept of neuromorphic computing. We demonstrate that our designs are 41% more energy efficient on average than traditional monolithic designs through use of industry standard benchmarks for evaluation. On average, our approach consumes 35% less energy than a set-up with more standard multi-agents. The performance of task completion is more than 94%. The system is scalable reasonably well. We have experimented with systems of any number of agents up to one thousand at a time and found sub-linear increases in energy with the size of the system. We have new protocols on coordinating energy-aware agents, algorithms on dynamic resource allocation in the characteristics of workloads and energy availability, and a framework of overall measurement of sustainability in agentic systems as our primary contributions. These advancements bring us into a closer step of scaling autonomous intelligence to non-unnecessary expenses on the environment.

Zitieren

APA

Lomte, R., Pede, S., & Nikumbh, P. (2026). Energy-Efficient Agentic AI: Long-Term Designs for Big Autonomous Systems. International Journal of Drug Delivery Technology, 16(18s). https://doi.org/10.25258/ijddt.16.18s.70

Vancouver

Lomte R, Pede S, Nikumbh P. Energy-Efficient Agentic AI: Long-Term Designs for Big Autonomous Systems. International Journal of Drug Delivery Technology. 2026;16(18s). doi:10.25258/ijddt.16.18s.70

BibTeX
@article{Lomte2026,
  author = {Lomte, Rajesh and Pede, Shailaja and Nikumbh, Pallavi},
  title = {Energy-Efficient Agentic AI: Long-Term Designs for Big Autonomous Systems},
  journal = {International Journal of Drug Delivery Technology},
  year = {2026},
  volume = {16},
  number = {18s},
  doi = {10.25258/ijddt.16.18s.70}
}
Kostenlos starten →