Physical Analytics - overview
Description: Physical Analytics is an emerging field at the intersection of big IoT (=Internet of Things) data, physical modeling and data analytics with the aim to provide the underlying intelligence for future and smarter IoT applications ("cognitive IoT").
One of the most remarkable developments in technology over the past 50 years has been the emergence of interconnected and ever becoming more powerful computing devices, which has created what we refer today as the internet (of computers). In recent years a similar trend has emerged with distributed and interconnected sensing devices, often integrated into appliances, and cell phones etc, which are measuring the physical world. This phenomenon is commonly referred to as the internet of things, which is expected to have profound implications not only for future technology and sciences.
The trend offers unique opportunities for very inter-disciplinary and highly relevant science. In the past most information on the internet has been originated by humans and computers. However now, vast amount of information is being created by devices, machines etc, which are placed in the real physical world. This convergence of the physical world with the digital domain allows applying physical principles to everyday problems in a much more effective and informed way than what was possible in the past. Very much like traditional applied physics and engineering has made enormous advances and changed our lives by making detailed measurements to understand the relevant physics of an engineered device, we can now apply the same rigor and principles to understand large-scale physical systems (to manage and control for example the flow of traffic in a city or the energy distribution within a complex grid system).
Physical Analytics aims at understanding macroscopic physical systems using physics-based models in combination with statistical and machine learning techniques. It leverages advances in communication and distributed sensing technologies. The research pursues the highest quality of science while making it matter (i.e., relevant to the most fundamental problems in our future). The Physical Analytics team at IBM is very interdisciplinary with researchers from physics, electrical engineering, mathematics, computer science, and other disciplines. The research includes both the application and development of a Physical Analytics platform, which requires the integration of several core technologies such as sensing and communication technologies, data management, numerical modeling methods, machine learning, control technologies etc. New science and insights are being generated from applying this platform to real-world systems.
News:
- Planned IEEE conference for Physical Analytics in Spring 2017
- Remarks at the Physics-inspired Machine Learning Conference at Santa Fe
- Keynote at CSTIC ("From Sensors to Smarter IoT Solutions")
Recent presentations:
Recent press:
- Harvard Business Review
- Data-ism by Steve Lohr (ISBN: 978-0-06-222681-5)
- Fortune Magazine
Key patents:
- Measurement and Management Technology Platform (US Patent 8,630,724)
- Multi-model blending (US Patent Application 20150347922)
Partners:
- Department of Energy: Sunshot, ARPA-E, EERE
- Gallo Wineries
- AT&T, PG&E, SCE, Georgia Tech
- New York Metropolitan Museum of Art
- University of Michigan
- Kansas State University
- Georgia Tech
- NREL, New England ISO, Green Mountain Power
- and many more
Other links: