Complex neuromorphic adaptive core (neuracore) and physics enhanced neuromorphic adaptive controller

Patent metadata

Field	Value
Publication number	US-12499355-B1
Application number	US-202117369742-A
Country	US
Kind code	B1
Filing date	Jul 7, 2021
Priority date	Jul 13, 2020
Publication date	Dec 16, 2025
Grant date	Dec 16, 2025

Abstract

Described is a Neuromorphic Adaptive Core (NeurACore) cognitive signal processor (CSP). The NeurACore CSP includes a NeurACore local learning layer block that is operable for receiving as an input a mixture of in-phase and quadrature (I/Q) signals and mapping the I/Q signals onto a neural network to determine complex-valued output weights of neural states of the neural network. A global learning layer is included that is operable for adapting the complex-valued output weights to predict a most likely next value of the input I/Q signal. Further, a neural combiner is included that operable for combining a set of delayed neural state vectors with weights of the global learning layer to compute an output signal, the output signal being separate in-phase and quadrature signals.

First claim

What is claimed is:

1 . A Neuromorphic Adaptive Core (NeurACore) cognitive signal processor (CSP), comprising:
one or more processors and associated computer-readable medium having executable instructions encoded thereon, such that upon execution of the instructions by the one or more processors, the one or more processors implement:
a NeurACore local learning layer block, the NeurACore local learning layer block operable for receiving as an input a mixture of in-phase and quadrature (I/Q) signals and mapping the I/Q signals onto a neural network to determine complex-valued output weights of neural states of the neural network;
an adaption module, the adaptation module embedding a time-evolving physical model into the NeurACore local learning layer block;
a global learning layer, the global learning layer operable for adapting the complex-valued output weights to predict a most likely next value of the input I/Q signal; and
a neural combiner, the neural combiner operable for combining a set of delayed neural state vectors with weights of the global learning layer to compute and generate an output signal, the output signal being separate in-phase and quadrature signals.

2 . The NeurACore CSP as set forth in claim 1 , wherein the time-evolving physical model evolves the NeurACore local learning layer block based on at least one of a neural state vector, an input signal, embedded physical equations reflecting current dynamics of a physical system.

3 . The NeurACore CSP as set forth in claim 2 , wherein the NeurACore local learning layer block includes oscillators that represent an instantaneous spectrum of the input mixture of I/Q signals.

4 . The NeurACore CSP as set forth in claim 3 , wherein the input mixture I/Q signals are an input signal selected from a communication signal and a radar signal, with the output signal being denoised communication and radar signals.

5 . The NeurACore CSP as set forth in claim 4 , wherein the NeurACore local learning layer block is adaptable for frequency and quality factor.

6 . The NeurACore CSP as set forth in claim 5 , wherein the NeurACore local learning layer utilizes blind source separation to separate signals similar in frequency.

7 . The NeurACore CSP as set forth in claim 1 , further comprising a physics enhanced neuromorphic adaptive controller electronically coupled with both a physical system and with the NeurAcore CSP for controlling the NeurACore CSP.

8 . A physics enhanced neuromorphic adaptive controller, comprising:
one or more processors and associated computer-readable medium having executable instructions encoded thereon, such that upon execution of the instructions by the one or more processors, the one or more processors implement:
a physics-enhanced (PE) neuromorphic adaptive core (NeurACore), the PE NeurACore having an approximate physics-based system model that models operations of a physical system and a dynamic NeurACore with online learning, the PE NeuraCore being operable for receiving as an input a mixture of in-phase and quadrature (I/Q) signals and generating an output signal, the output signal being separate in-phase and I/Q signals; and
a controller coupled to the PE NEurACore, the controller operable for controlling the physical system based on the separate in-phase and I/Q signals.

9 . The physics enhanced neuromorphic adaptive controller as set forth in claim 8 , further comprising a physical system coupled to the physics enhanced neuromorphic adaptive controller.

10 . The physics enhanced neuromorphic adaptive controller as set forth in claim 9 , wherein the PE NeurACore includes system parameters, input layer weights, reservoir poles, and output layer weights, that are adjustable in real-time to capture behavior of the physical system.

11 . The physics enhanced neuromorphic adaptive controller as set forth in claim 8 , wherein the controller includes a compensator and inverse physical system model that uses learned parameters of the PE NeurACore to generate a pre-distorted input signal that, when fed to the physical system, produces the desired physical system output.

12 . A computer implemented method for signal processing using a Neuromorphic Adaptive Core (NeurACore) cognitive signal processor (CSP), comprising acts of:
using a NeurACore local learning layer block, receiving as an input a mixture of in-phase and quadrature (I/Q) signals and mapping the I/Q signals onto a neural network to determine complex-valued output weights of neural states of the neural network; using an adaption module to embed a time-evolving physical model into the NeurACore local learning layer block; using a global learning layer, adapting the complex-valued output weights to predict a most likely next value of the input I/Q signal; and using a neural combiner, combining a set of delayed neural state vectors with weights of the global learning layer to compute and generate an output signal, the output signal being separate in-phase and quadrature signals.

13 . The method as set forth in claim 12 , wherein the time-evolving physical model evolves the NeurACore local learning layer block based on at least one of a neural state vector, an input signal, embedded physical equations reflecting current dynamics of a physical system.

14 . The method as set forth in claim 13 , wherein the NeurACore local learning layer block includes oscillators that represent an instantaneous spectrum of the input mixture of I/Q signals.

15 . The method as set forth in claim 14 , wherein the input mixture I/Q signals are an input signal selected from a communication signal and a radar signal, with the output signal being denoised communication and radar signals.

16 . The method as set forth in claim 15 , wherein the NeurACore local learning layer block is adaptable for frequency and quality factor.

17 . The method as set forth in claim 16 , wherein the NeurACore local learning layer utilizes blind source separation to separate signals similar in frequency.

18 . A computer program product for signal processing using a Neuromorphic Adaptive Core (NeurACore) cognitive signal processor (CSP), the computer program product comprising:
a non-transitory computer-readable medium having executable instructions encoded thereon, such that upon execution of the instructions by one or more processors, the one or more processors perform operations of:
receiving as an input a mixture of in-phase and quadrature (I/Q) signals and mapping the I/Q signals onto a neural network to determine complex-valued output weights of neural states of the neural network;
embedding a time-evolving physical model into the NeurACore local learning layer block;
adapting the complex-valued output weights to predict a most likely next value of the input I/Q signal; and
combining a set of delayed neural state vectors with weights of the global learning layer to compute an output signal, the output signal being separate in-phase and quadrature signals.

Assignees

• Hrl Lab Llc

Inventors

• Adl Sanaz
• Fong Bryan H
• Petre Peter
• Martin Charles E
• Kabakian Adour V

Classifications

• G06N3/049

  Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs · CPC title

• G06N3/063 Primary

  using electronic means · CPC title

• G06N3/08 Primary

  Learning methods · CPC title