Intrinsic Information Processing and Energy Dissipation in Stochastic Input-Output Dynamical Systems
Abstract
Recent theoretical and experimental advances in nonequilibrium thermodynamics provide a new understanding of how ÒintelligentÓ control can convert information to energy. However, these approaches have yet to account for the diverse kinds of information that complex nonlinear systems are capable of producing and storing. This is particularly a concern regarding the diversity of differently structured irreversible processes, which play key role in limiting computation. Fortunately, computational mechanics accounts for this diversity, but in autonomous dynamical systems. Thermodynamic cycles that perform useful information processing are nonautonomous systems. To analyze information processing in nonautonomous systems, computational mechanics must be ex- tended to controlled dynamical systems. Additionally, it must be augmented to account for the energetics that support information generation and storage. The projectÕs goal was to synthesize the new nonequilibrium thermodynamics and computational mechanics into a single framework. The project successfully reached this goal and, leveraging that success, the research effort continues under the ARO-based MURI "Information Engines" (W911NF-13-1-0390), under the PI s leadership. The framework now provides techniques to analyze and predict the mechanisms by which energy flows support information processing. Using these, we are developing broadly applicable principles of emergent organization and hierarchy in complex physical and biological systems, including new bounds on the minimum required energy dissipation.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jun 25, 2021
- Source ID
- W911NF1210234
Entities
People
- James P. Crutchfield
Organizations
- Army Contracting Command
- United States Army
- University of California, Davis