Towards Embedded Computation with Building Materials

Area of Science:

• Materials Science
• Computational Science
• Information Processing

Background:

• The in materio computing paradigm leverages physical substrates for information processing.
• Reservoir Computing offers a model for understanding embedded in materio computation.
• Concrete's potential as a novel information processing substrate remains largely unexplored.

Purpose of the Study:

• To investigate the feasibility of using concrete-based materials as a substrate for signal classification within the in materio computing framework.
• To characterize the electrical and dynamical properties of concrete samples with varying additive concentrations.
• To explore the potential of these materials as a source for 'reservoirs of states' in computational models.

Main Methods:

• Electrical characterization of concrete samples with different additive concentrations.
• Dynamical analysis of selected specimens, including memfractive property assessment.
• Calculation of fractal dimensions, entropy parameters, Lyapunov exponents, and Detrended Fluctuation Analysis exponents.
• Signal waveform shape classification based on obtained parameters.

Main Results:

• Demonstrated the capability of concrete-based substrates for signal classification tasks.
• Identified memfractive properties in selected concrete specimens.
• Analyzed the richness of the reservoir configuration space through fractal and entropy parameters.
• Confirmed chaotic and self-affine signal characteristics using Lyapunov and DFA exponents.

Conclusions:

• Concrete-based materials show promise as a substrate for in materio computing and signal classification.
• The 'reservoir of states' generated by these materials can be effectively tuned for specific computational tasks.
• The complex dynamical behaviors observed in concrete support its application in advanced information processing scenarios.