Employing Vector Field Techniques on the Analysis of Memristor Cellular Nonlinear Networks Cell Dynamics

Chandan Singh, Vasileios G. Ntinas, Dimitrios A. Prousalis, Yongmin Wang, Ahmet Samil Demirkol, Ioannis Messaris, Vikas Rana, Stephan Menzel, Alon Ascoli, Ronald Tetzlaff2023 · arXiv

Introduction

Cellular Nonlinear Networks (CNNs) represent a class of parallel processing architectures particularly effective for high-speed image processing tasks. The integration of memristors into these networks—forming Memristor Cellular Nonlinear Networks (M-CNNs)—introduces non-volatility and local memory capabilities. However, this integration significantly increases the complexity of the underlying dynamical systems. This paper presents a novel graphical analysis tool that utilizes vector fields and phase portraits to visualize and interpret the complex 2nd-order dynamics of M-CNN cells.

The Challenge of 2nd-Order Dynamics

Standard CNN cells are typically 1st-order systems, which researchers often analyze using the Dynamic Route Map (DRM). The addition of a memristor introduces a second state variable: the memristor's internal state (e.g., oxygen vacancy concentration, $N_d$). This transforms the cell into a 2nd-order dynamical system.

Existing analytical tools for 2nd-order systems, such as the DRM2, frequently only indicate the sign of state changes. They lack the ability to convey the magnitude or "speed" of the dynamics. This limitation makes it difficult for circuit designers to intuitively understand how individual components, such as the capacitor, influence the overall stability and transition behavior of the cell.

Circuit diagram of an M-CNN cell featuring a capacitor (C), resistor (R), memristor (M), and controlled sources.

Methodology: Enhanced Graphical Analysis

The proposed tool overcomes these limitations by combining traditional phase portraits with normalized vector fields and color-coded magnitudes.

Dynamical Modeling

The M-CNN cell is modeled using two coupled ordinary differential equations (ODEs). These equations describe the evolution of the capacitor voltage ($V_C$) and the memristor’s internal state ($N_d$). To ensure physical accuracy, the authors employ the Juelich-Aachen Resistive Switching Tool (JART) model, which specifically targets Valence Change Mechanism (VCM) memristors.

Vector Field Normalization and Color Encoding

A primary challenge in visualizing these systems is the disparate scales and units of $V_C$ (linear) and $N_d$ (often logarithmic). To address this, the authors normalize the vector lengths to a fixed ellipse, preserving only the direction of state evolution. The lost magnitude information—the Euclidean norm of the state derivatives—is then recovered by encoding it into the color of the arrows or the background heatmap.

Nullcline Mapping

By overlaying "nullclines" (lines where the derivative of a state variable is zero), the tool allows for the immediate identification of equilibrium points at their intersections. This provides a clear map of stable and unstable regions within the state space.

Key Insights and Results

The visualization tool reveals several critical aspects of M-CNN cell behavior:

  • Capacitance as a Dynamic Controller: The tool demonstrates that adjusting capacitance does more than just scale time; it alters the relative speed between the memristor and the capacitor. This can fundamentally change the system's trajectory toward equilibrium. For instance, increasing capacitance from $10nF$ to $1000nF$ can make the memristor the "fast" element, shifting the entire flow of the vector field.
  • Bistability Management: By tuning the parallel resistor ($R$), designers can transition the cell between "fully bistable" and "hybrid" modes. The vector field makes these transitions and the resulting stability regions immediately apparent.
  • Equilibrium Identification: The nullcline crossings accurately pinpoint stable versus unstable equilibria, which is essential for the design of reliable memory and logic circuits.
Visualization of the state derivative magnitude dVc/dt across the state space (Vc, Nd), with nullclines indicating equilibrium boundaries.

Conclusion and Future Directions

The developed graphical technique provides a powerful intuitive framework for designing and analyzing 2nd-order memristive systems. While the current focus is on isolated cells, future work will extend this approach to analyze the collective dynamics of interconnected M-CNN grids. Such advancements are crucial for developing efficient hardware for edge computing and autonomous systems where high-speed, local parallel processing is required.

Summary updated

Abstract

This paper introduces an innovative graphical anal-ysis tool for investigating the dynamics of Memristor Cellular Nonlinear Networks (M-CNNs) featuring 2<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">nd</sup>-order processing elements, known as M-CNN cells. In the era of specialized hardware catering to the demands of intelligent autonomous systems, the integration of memristors within Cellular Nonlinear Networks (CNNs) has emerged as a promising paradigm due to their exceptional characteristics. However, the standard Dynamic Route Map (DRM) analysis, applicable to 1<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">st</sup>-order systems, fails to address the intricacies of 2<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">nd</sup> -order M-CNN cell dynamics, as well the 2<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">nd</sup>-order DRM (DRM2) exhibits limitations on the graphical illustration of local dynamical properties of the M-CNN cells, e.g. state derivative's magnitude. To address this limitation, we propose a novel integration of M-CNN cell vector field into the cell's phase portrait, enhancing the analysis efficacy and enabling efficient M-CNN cell design. A comprehensive exploration of M- CNN cell dynamics is presented, showcasing the utility of the proposed graphical tool for various scenarios, including bistable and monostable behavior, and demonstrating its superior ability to reveal subtle variations in cell behavior. Through this work, we offer a refined perspective on the analysis and design of M-CNNs, paving the way for advanced applications in edge computing and specialized hardware.
BibTeX
@misc{singh2023employing,
  author = {Chandan Singh and Vasileios G. Ntinas and Dimitrios A. Prousalis and Yongmin Wang and Ahmet Samil Demirkol and Ioannis Messaris and Vikas Rana and Stephan Menzel and Alon Ascoli and Ronald Tetzlaff},
  title = {{Employing Vector Field Techniques on the Analysis of Memristor Cellular Nonlinear Networks Cell Dynamics}},
  year = {2023},
  eprint = {2408.03260},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2408.03260},
}
Figure