Artificial intelligence based reconfigurable neuromorphic vision sensor fusion systems and methods thereof

Abstract

Embodiments herein provide system for an Artificial Intelligence (AI) based evolutionary and reconfigurable neuromorphic vision sensor fusion platform. The system comprises neuromorphic functional sensing units (NFSUs) to receive multiple sensory signals in flat or dome shaped arrays at the top. Further, the system comprises an underneath processing system comprising AI processing layers in a top-to-bottom bio-morphic hierarchical architecture. Further, each AI processing layers comprises of AI processing modules (AIP-modules) and AI processing interfaces (AIP-interfaces) capable of coupling or uncoupling neighboring AIP-modules in real-time during an operation. The AI processing layers with the thus dynamic AIP-modular structure in real-time allows to determine the type and relevance of the features of the incoming sensory signals with output tasks to be performed; autonomously configure or reconfigure the required subset of NFSUs with the needed AI modular processing functionalities; and generate the needed output during an operation.

Description

ARTIFICIAL INTELLIGENCE BASED RECONFIGURABLE NEUROMORPHIC VISION SENSOR FUSION SYSTEMS AND METHODS THEREOFTECHNICAL FIELD

[0001] Embodiments of the present disclosure generally relates Artificial Intelligence (AI)- based vision sensors, and more particularly to Al based reconfigurable neuromorphic vision sensor fusion systems and methods thereof.BACKGROUND

[0002] Typically, the increasing demand for autonomous systems, such as robotics, unmanned aerial vehicles (UAVs), autonomous vehicles, and remote sensing platforms, has made significant advancements in vision-based perception technologies. The autonomous systems rely on the integration of multiple vision sensors, including multiple vision sensory signals cameras, with UV, thermal, depth, and radiation imagers, to achieve accurate environmental understanding in diverse and dynamic conditions. However, traditional vision sensor fusion approaches face substantial challenges, including high computational complexity , significant power consumption, and substantial memory requirements, which are particularly prohibitive for edge and extremeedge devices operating with constrained resources.

[0003] Further, conventional sensor fusion techniques typically utilize centralized processing architectures that aggregate raw data from multiple sensors and apply computationally intensive Al algorithms, such as large-scale deep neural networks trained on large-scale global sensor data or Kalman filtering, to extract meaningful environmental features. While effective in controlled settings, the aforementioned methods often struggle to meet the stringent latency and energy constraints of real-time autonomous applications at the edge or extreme edge, especially in environments where resources are limited. Moreover, such methods are generally suitable for static configuration, lacking real-time adaptability to optimize sensor placement, configurations, or processing strategies in response to varying environmental conditions or requirements.

[0004] Furthermore, some existing systems have explored the use of neuromorphic computing to address the power and latency limitations of traditional architectures. Neuromorphic processors, inspired by biological neural systems, provides event-driven computation and asynchronous processing which enables significant reductions in energy consumption and processing delays for vision-based tasks. However, the integration of neuromorphic computing into multi-sensor fusion frameworks remains underexplored, particularly for optimizing the dynamic interplay between heterogeneous sensor modalities and the operational environments. Some existing neuromorphic implementations often focus on single-sensor applications whichlimits the ability to handle the complexity of multi-sensor (scalable to tens to thousands of sensors) data fusion required for robust autonomous operations of perception.

[0005] Furthermore, some existing vision sensor fusion systems often neglect the impact of spatial sensor geometry on perception accuracy which leads to less preferred performance in scenarios requiring precise localization or tracking, such as obstacle avoidance or target detection. The lack of real-time adaptive mechanisms to dynamically reconfigure sensor placements or processing priorities reduces the efficiency, particularly in three-dimensional (3D) operational volumes where environmental factors, such as occlusions or lighting variations affects the performance of the sensors. Also, the existing approaches fail to provide robust error bounds or performance metrics to quantify the reliability of fused sensor outputs, limiting the applicability in safety-critical autonomous systems.

[0006] Therefore, there is a need in the art for improved system and methods to address at least the aforementioned technical problems in the prior arts, by providing an Al based reconfigurable and scalable neuromorphic vision sensor fusion systems and methods thereof.SUMMARY

[0007] This summary is provided to introduce a selection of concepts, in a simple manner, which is further described in the detailed description of the disclosure. This summary is neither intended to identify key or essential inventive concepts of the subject matter nor to determine the scope of the disclosure.

[0008] An aspect of the present disclosure provides a system for an Artificial Intelligence (Al) based real-time reconfigurable neuromorphic vision sensor fusion for processing sensory’ signals using a layered and modular Al architecture. The system comprises a plurality of sensors configured to receive one or more sensory signals corresponding to a single or multi-modal data from one or more data sources. Further, the system comprises a processing system communicatively coupled to the plurality of sensors. Further, the processing system comprises a one or more Al processing layers arranged in a biomorphic hierarchical architecture. Further, the one or more Al processing layers are configured to determine one or more tasks to be performed based on real-time environmental conditions and operational requirements. Further, the one or more Al processing layers are configured to extract a plurality' of significant features from the received one or more sensory signals using a data-driven multiple modular Al in each Al processing layer. Further, the one or more Al processing layers are configured to determine a type and relevance of each of the extracted plurality of features with the determined one or more tasks to be performed. Further, the one or more Al processing layers configure each of the plurality' of sensors with at least one or more Al processing functions based on the determined type and therelevance. Further, the at least one or more Al processing functions is configured to process each of the plurality of significant features. Further, the one or more Al processing layers are configured to determine whether the assigned at least one or more Al processing functions match with the real-time environmental conditions and the operational requirements. Further, the one or more Al processing layers are configured to dynamically update the at least one or more Al processing functions to meet the real-time environmental conditions and the operational requirements based on the determination. Further, the one or more Al processing layers are configured to determine a subset of sensors and a subset of Al processing layers to be reconfigured based on the updated at least one or more Al processing functions. Further, the one or more Al processing layers are configured to autonomously reconfigure the determined subset of sensors and the subset of Al processing layers in real-time with the updated at least one or more Al processing functions. Further, the one or more Al processing layers are configured to perform the determined one or more tasks by executing the updated at least one or more Al processing functions at corresponding sensors and corresponding Al processing layers. Further, the updated at least one or more Al processing functions are configured to process each of the plurality of features. Furthermore, the one or more Al processing layers are configured to generate a processed output corresponding to the single or multi-modal data based on the performed one or more tasks.

[0009] Another aspect of the present disclosure provides a method for an Al based reconfigurable neuromorphic vision sensor fusion system. The method comprises determining, by a processor, one or more tasks to be performed based on real-time sensory signals, environmental conditions, and operational requirements. Further, the method comprises extracting, by the processor, a plurality of significant features from the received one or more sensory signals using a data-driven modular Al in each Al processing layer. Further, the method comprises determining, by the processor, a type and relevance of each of the extracted plurality of features with the determined one or more tasks to be performed. Further, the method comprises configuring, by the processor, each of the plurality of sensors with at least one or more Al processing functions based on the determined type and the relevance. Further, the at least one or more Al processing functions are configured to process each of the plurality of significant features. Further, the method comprises determining, by the processor, whether the assigned at least one or more Al processing functions match with the real-time environmental conditions and the operational requirements. Further, the method comprises dynamically updating, by the processor, the at least one or more Al processing functions to meet the real-time environmental conditions and the operational requirements based on the determination. Further, the method comprises determining, by the processor, a subset of sensors and a subset of Al processing layers to be reconfigured based on the updated at least one or more Al processing functions. Further, the method comprises autonomouslyreconfiguring, by the processor, the determined subset of sensors and the subset of Al processing layers in real-time with the updated at least one or more Al processing functions. Further, the method comprises performing, by the processor, the determined one or more tasks by executing the updated at least one or more Al processing functions at corresponding sensors and corresponding Al processing layers. Further, the updated at least one or more Al processing functions are configured to process each of the plurality of features. Furthermore, the method comprises generating, by the processor, a processed output corresponding to the single or multimodal data based on the performed one or more tasks.

[0010] To further clarify the advantages and features of the present disclosure, a more particular description of the disclosure will follow by reference to specific embodiments thereof, which are illustrated in the appended figures. It is to be appreciated that these figures depict only typical embodiments of the disclosure and are therefore not to be considered limiting in scope. The disclosure will be described and explained with additional specificity and detail with the appended figures.BRIEF DESCRIPTION OF ACCOMPANYING DRAWINGS

[0011] The disclosure will be described and explained with additional specificity and detail with the accompanying figures in which:

[0012] FIG. 1A illustrates a block diagram representation of an example Artificial- Intelligence (Al) based reconfigurable neuromorphic vision sensor fusion system capable of processing sensory signals using a layered and modular Al architecture, in accordance with an embodiment of the present disclosure;

[0013] FIG. IB illustrates a block diagram representation of an example electronics packaging, such as those show n in FIG. 1A, in accordance with an embodiment of the present disclosure;

[0014] FIG. 2A illustrates a schematic representation of an example Evolutionary Neuromorphic Vision fusion Sensor Platform (ENVSP) for processing one or more sensory signals received by a plurality7of sensors, in accordance with an embodiment of the present disclosure;

[0015] FIG. 2B illustrates a schematic representation of an example conceptual technology' of Al based reconfigurable neuromorphic vision sensor fusion system, in accordance with an embodiment of the present disclosure;

[0016] FIG. 2C illustrates a schematic representation side view of an example ENV SP w ith a Neuromorphic Functional Sensing Unit (NFSU) in an array layer coupled to underneath Al layers, in accordance with an embodiment of the present disclosure;

[0017] FIG. 3 illustrates a schematic representation of an example NFSU with an upper hardware receptor or lens like part and a lower Al software defined functionality part, in accordance with an embodiment of the present disclosure;

[0018] FIG. 4 illustrates a schematic representation of exemplary one or more NFSUs comprising one or more Al software defined multiple functionalities, in accordance with an embodiment of the present disclosure;

[0019] FIG. 5 illustrates a schematic representation of exemplary one or more Al processing layers configured with a plurality of Al software defined reconfigure-ability of a top layer (TL)-AI, in accordance with an embodiment of the present disclosure;

[0020] FIG. 6 illustrates a schematic representation of an example ENVSP with multiple NFSUs configured with multiple Al software defined functionalities processed through underlying Al layers with or without AlP-modules, in accordance with an embodiment of the present disclosure;

[0021] FIG. 7 illustrates a schematic representation of example ENVSP with multiple NFSUs comprising multiple Al defined functionality types configured in multiple different sequences where each Al software defined functionality is processed through Al processing layers with uncoupled AlP-modules, in accordance with an embodiment of the present disclosure;

[0022] FIG. 8 illustrates a schematic representation of an example ENVSP with multiple NFSUs comprising multiple Al defined functionality types configured in multiple different sequences, where each Al software defined functionality is processed through Al processing layers with coupled AlP-modules, in accordance with an embodiment of the present disclosure;

[0023] FIG. 9A illustrates a schematic representation of example top to bottom vertical view of biomorphic layered Al architecture of the ENVSP, in accordance with an embodiment of the present disclosure;

[0024] FIG. 9B illustrates a schematic representation of example side to side horizontal view of biomorphic layered Al architecture of the ENVSP, in accordance with an embodiment of the present disclosure;

[0025] FIG. 10 illustrates a schematic representation of plurality of Al based reconfigurable neuromorphic vision sensor fusion systems deployed within a plurality of network devices, in accordance with an embodiment of the present disclosure; and

[0026] FIG. 11 illustrates a flow chart representation of an example method for Al based reconfigurable neuromorphic vision sensor fusion system, in accordance with an embodiment of the present disclosure.

[0027] Further, those skilled in the art will appreciate that elements in the figures are illustrated for simplicity and may not have necessarily been drawn to scale. Furthermore, in termsof the construction of the device, one or more components of the device may have been represented in the figures by conventional symbols, and the figures may show only those specific details that are pertinent to understanding the embodiments of the present disclosure so as not to obscure the figures with details that will be readily apparent to those skilled in the art having the benefit of the description herein.DETAILED DESCRIPTION

[0028] For simplicity and illustrative purposes, the present disclosure is described by referring mainly to examples thereof. The examples of the present disclosure described herein may be used together in different combinations. In the following description, details are set forth in order to provide an understanding of the present disclosure. It will be readily apparent, however, that the present disclosure may be practiced without limitation to all these details. Also, throughout the present disclosure, the terms ”a" and “an” are intended to denote at least one of a particular element. The terms “a” and “an” may also denote more than one of a particular element. As used herein, the term “includes” means includes but not limited to, the term “including” means including but not limited to. The term “based on” means based at least in part on, the term “based upon” means based at least in part upon, and the term “such as” means such as but not limited to. The term “relevant” means closely connected or appropriate to what is being performed or considered.

[0029] For the purpose of promoting an understanding of the principles of the disclosure, reference will now be made to the embodiment illustrated in the figures and specific language will be used to describe them. It will nevertheless be understood that no limitation of the scope of the disclosure is thereby intended. Such alterations and further modifications in the illustrated system, and such further applications of the principles of the disclosure as would normally occur to those skilled in the art are to be construed as being within the scope of the present disclosure. It will be understood by those skilled in the art that the foregoing general description and the following detailed description are exemplary and explanatory of the disclosure and are not intended to be restrictive thereof.

[0030] In the present document, the word “exemplary” is used herein to mean “serving as an example, instance, or illustration”. Any embodiment or implementation of the present subject matter described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments. The terms “comprise”, “comprising”, or any other variations thereof, are intended to cover a non-exclusive inclusion, such that one or more devices or sub-systems or elements or structures or components preceded by “comprises... a” does not, without more constraints, preclude the existence of other devices, sub-systems, additional sub-modules. Appearances of the phrase “in an embodiment7’, “in another embodiment”, “in an exemplary embodiment” and similar language throughout this specification may, but not necessarily do, all refer to the same embodiment.

[0031] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this disclosure belongs. The system, methods, and examples provided herein are only illustrative and not intended to be limiting. A computer system (standalone, client, or server, or computer-implemented system) configured by an application may constitute a “module” (or “subsystem”) that is configured and operated to perform certain operations. In one embodiment, the “module” or “subsystem” may be implemented mechanically or electronically, so a module includes dedicated circuitry or logic that is permanently configured (within a special-purpose processor) to perform certain operations. In another embodiment, a “module” or a “subsystem” may also comprise programmable logic or circuitry (as encompassed within a general-purpose processor or other programmable processor) that is temporarily configured by software to perform certain operations. Accordingly, the term “module” or “subsystem” should be understood to encompass a tangible entity, be that an entity that is physically constructed permanently configured (hardwired) or temporarily configured (programmed) to operate in a certain manner and / or to perform certain operations described herein.

[0032] Neuromorphic computing refers to computational systems that emulate the structure and function of biological neural systems, utilizing analog, digital, or mixed-signal circuits to perform real time- or event-driven processing. Unlike traditional von Neumann architectures, neuromorphic systems process data asynchronously, responding to input events with low power consumption, making them suitable for real-time sensory applications. In the present disclosure, neuromorphic computing is implemented within neuromorphic functional sensing units (NFSUs), which are arranged in flat, dome, or faceted dome-shaped arrays. Each NFSU’s upper hardware receptor captures sensory signals (e.g., visual, infrared, or ultra-violet data), while the lower Al software part processes these signals using event-driven, low-power artificial intelligence algorithms inspired by neurolog ’ and compound eyes of insects and bees.

[0033] Artificial Intelligence (Al) may refer to computational systems that perform tasks requiring human-like intelligence, such as feature extraction, classification, or decision-making. In the present disclosure, Al encompasses ML, NN, and GA models within AlP-modules and AIP- interfaces. Software Function may denote a programmable operation executed by Al processing layers, such as detection, filtering, or reconfiguration of sensory data. Neuromorphic describes systems mimicking biological neural structures, focusing on time-dependent, event-driven, low- power processing. Biomorphic indicates a design inspired by biological systems, particularly hierarchical sensory' processing, without replicating biological components.

[0034] The following terms, as used in herein, are defined with general examples to clarify their scope. The definitions of terms provided herein, are provided for illustrative purposes to clarify their intended meaning within the context of the invention. These definitions are not intended to be exhaustive or to limit the scope of the terms to the specific examples or interpretations presented. The terms may encompass additional meanings, implementations, or applications consistent with the principles of the invention, as would be understood by one skilled in the art, and may be adapted to various embodiments, configurations, or technological contexts without departing from the scope of the invention as defined by the claims.

[0035] Neuromorphic Functional Sensing Units (NFSUs) may refer to sensor modules with an upper hardware receptor (e.g., lens) and a lower Al software part for processing signals. Flat, Dome, or faceted dome-shaped array may refer to configurations of sensor arrangements, where flat is planar, dome is continuously curved, and faceted dome comprises flat facets joined at angles to approximate a dome. Biomorphic hierarchical architecture may refer to a layered Al processing structure inspired by biological systems, with top, middle, and bottom layers performing sequential data reduction, knowledge enhancement, and task execution. Al processing layers may refer to software or hardware modules in the hierarchical architecture, including top Al (TL-AI), middle Al (ML-AI), and bottom Al (BL-AI) layers. Significant features, for example, may refer to attributes of sensory signals relevant to tasks, such as color, intensity, luminosity’, shape, pattern, feature, depth, polarization, frequency, amplitude, tone, pitch, or noise for vision sensory signals. Al processing functions may refer to operations performed by Al layers or Al modules forming the Al layers, including detection, filtering, classification, combination, or convolution of features. Al processing modules (AIP-Modules) may refer to modular Al software units with pre-trained ML, NN. or GA models for processing sensory data. Al processing interfaces (AlP-Interfaces) may refer to software modules coupling AIP-Modules using LA, WLA, SA, ML, NN, or GA, which could be pre-trained as well as further optimized during an operation for better efficiency and accuracy. This means that both AlP-modules and AlP-Interfaces come pre-trained and precomputed, and only the AlP-Interfaces could be further optimized during an operation for better efficiency and accuracy. Real-time environmental conditions may refer to dynamic external factors affecting sensor operation, such as surrounding lighting, weather, noise, and obstacles in the input sensor data. Operational requirements may refer to task-specific parameters, such as accuracy, speed, or energy constraints. Relevance score may refer to a numerical value indicating a feature's importance to a task, computed by an Al model. Coupled mode may refer to processing where sensory signals or AlP-modules interacting via AlP-interfaces during an operation. Uncoupled mode may refer to independent processing of sensory’ signals or AlP-modules. Tasks may refer to operations performed by the system, for example, but not limited to, image capture, patternrecognition, feature extraction, obstacle detection, depth estimation, location navigation, or event detection and classification.

[0036] Embodiments described herein provides a system for an Artificial Intelligence (Al) based reconfigurable neuromorphic vision sensor fusion. The system comprises a plurality of sensors configured to receive one or more sensory signals corresponding to a single or multi-modal data from one or more data sources. Further, the system comprises a processing system communicatively coupled to the plurality of sensors. Further, the processing system comprises a one or more Al processing layers arranged in a biomorphic hierarchical architecture. Further, the one or more Al processing layers are configured to determine one or more tasks to be performed based on real-time environmental conditions and operational requirements. Further, the one or more Al processing layers are configured to extract a plurality of significant features from the received one or more sensory signals using a data driven modular Al in each Al processing layer. Further, the one or more Al processing layers are configured to determine a t pe and relevance of each of the extracted plurality of features with the determined one or more tasks to be performed. Further, the one or more Al processing layers configure each of the plurality of sensors with at least one or more Al processing functions based on the determined type and the relevance. Further, the at least one or more Al processing functions is configured to process each of the plurality of significant features. Further, the one or more Al processing layers are configured to determine whether the assigned at least one or more Al processing functions match with the real-time environmental conditions and the operational requirements. Further, the one or more Al processing layers are configured to dynamically update the at least one or more Al processing functions to meet the real-time environmental conditions and the operational requirements based on the determination. Further, the one or more Al processing layers are configured to determine a subset of sensors and a subset of Al processing layers to be reconfigured based on the updated at least one or more Al processing functions. Further, the one or more Al processing layers are configured to autonomously reconfigure the determined subset of sensors and the subset of Al processing layers in real-time with the updated at least one or more Al processing functions. Further, the one or more Al processing layers are configured to perform the determined one or more tasks by executing the updated at least one or more Al processing functions at corresponding sensors and corresponding Al processing layers. Further, the updated at least one or more Al processing functions are configured to process each of the plurality of features. Furthermore, the one or more Al processing layers are configured to generate a processed output corresponding to the single or multi-modal data based on the performed one or more tasks.

[0037] Referring now to the drawings, and more particularly to FIG. 1 through FIG. 11, where reference characters denote corresponding features consistently throughout the figures,there are shown preferred embodiments, and these embodiments are described in the context of the following exemplary system and / or method.

[0038] FIG. 1A illustrates a block diagram representation 100A of an example Artificial- Intelligence (Al) based reconfigurable neuromorphic vision sensor fusion system capable of processing sensory signals using a layered and modular Al architecture, in accordance with an embodiment of the present disclosure. The system 102 (also referred herein as Evolutionary Neuromorphic Vision fusion Sensor Platform (ENVSP) 102) may include a plurality of sensors 104 (also individually referred herein as ‘sensor 104’ and collectively referred herein as ‘sensors 104’) configured to receive one or more sensory signals corresponding to a single or multi-modal data from one or more data sources.

[0039] Further, the sensors 104 may include, for example, but not limited to, a Neuromorphic Functional Sensing Units (NFSUs) 108. The NFSUs 108 further comprises an upper hardware receptor or lens like part 110 and a lower Al software-defined functionality part 112. The upper hardware receptor or lens like part 110 is configured to receive the one or more sensory signals. The lower Al software-defined functionality part 112 is configured to process the received one or more sensory signals. In one aspect of the present disclosure, the upper hardware receptor or lens like part 110 of each NFSUs 108 may refer to any or all, of for example, but not limited to, physical, chemical, materials, or structural composition of the upper hardware receptor or lens like part 110 of the NFSUs 108 under consideration. In another aspect, the Al software- defined functionality part 112 of each NFSUs 108 may include, for example, but not limited to, machine learning (ML), neural network (NN), or genetic algorithm (GA) based software or models or the like. The Al software-defined functionality part 112 may allow, for example, but not limited to, detection, filtering, classification, combination, or convolution of one or more sensory signals received by the NFSUs 108 within an input flat, dome, or faceted dome-shaped layer of the system 102.

[0040] The system 102 may include a processing system 130 communicatively coupled to the plurality of sensors 104. Further, the processing system 130 comprises one or more Al processing layers 1 16 arranged in a biomorphic hierarchical architecture. Further, the one or more Al processing layers 116 are configured to determine one or more tasks to be performed based on for example, but not limited to, real-time sensory signals, environmental conditions, and operational requirements. The one or more tasks may include, for example, but are not limited to, object detection, motion tracking, anomaly recognition, pattern classification, feature extraction or decision making or the like. The processing system 130, may be implemented on hardware such as application-specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), or Neuromorphic chips. The processing system 130, analyzes the one or more sensory signals,environmental data (e.g., light levels, weather and noise conditions), and predefined operational requirements (e g., task accuracy, energy constraints) using a machine learning (ML) classifier, such as, for example, but not limited to, a decision tree or neural network (NN). The classifier processes the one or more sensory signals, such as visual image pixel data or infrared thermal readings, alongside environmental metadata (e.g., ambient light intensity) to identify tasks such as for example, but not limited to, object and motion recognition, obstacle detection, mapping, navigation, or remote monitoring. In an example embodiment, a supervised ML model, pre-trained on labeled datasets of the one or more sensory signals and task outcomes, maps input patterns to task categories. For example, in an agricultural drone, the model analyzes visual images of grapevines and detects discolored leaves, determining a task of “disease detection7’ when low light and high humidity are detected. In another example, in industrial or manufacturing robotics with mixed human and humanoid-robots working in close proximity the model analyzes the surrounding environment to provide fast 360° situational awareness for collision or accident avoidance. In the agricultural drone example in a vineyard, the drone’s TL-AI layer 118 receives visual signals showing yellowing leaves and infrared data indicating elevated temperatures. The ML classifier, trained on crop health patterns, identifies “disease detection” and “navigation around trellises” as primary tasks based on cloudy conditions (environmental) and a requirement for precise monitoring (operational).

[0041] Further, the one or more Al processing layers 116 are configured to extract a plurality of significant features from the received one or more sensory signals using a data driven modular Al in each Al processing layer 116. The plurality of significant features may include, for example, but are not limited to, color, shape, size, texture, motion vectors, spectral frequency, edge contours, depth information, polarization, and spatial or temporal patterns or the like relevant to the received one or more sensory signals. The TL-AI layer 1 18, integrated with neuromorphic functional sensing units (NFSUs) 108, employs Al processing blocks 120 (also referred herein as AlP-modules or AlP-blocks) configured with convolutional neural networks (CNNs) to extract significant features, such as color, shape, or thermal signatures, from sensory signals. Each NFSU’s 108 lower Al software part 112 processes raw data (e.g., pixel intensities, infrared values) in parallel, reducing dimensionality while preserving task-relevant information. The ML- Al 124 and bottom Al 126 (BL-AI) layers further refine these features using additional AlP-modules 120 with unsupervised ML models, such as, for example, but not limited to, k- means clustering, to group similar features. In an example embodiment, CNNs in the TL-AI layer 118 may apply convolutional filters to visual data, extracting features such as edge patterns or color gradients. In the ML -Al layers 124, clustering algorithms may identify7significantfeature subsets (e.g., discolored leaf patterns) by grouping data points based on similarity metrics.

[0042] Further, the one or more Al processing layers 116 are configured to determine a type and relevance of each of the extracted plurality of one or more significant features with the determined one or more tasks to be performed. The TL-AI layer 118 may be. implemented as a neural network or weighted linear approximation (WLA) model, to classify extracted features into categories (e.g., color, shape, thermal) and compute their relevance to the identified tasks. Relevance scores may be calculated based on pre-trained weights correlating features with task outcomes. The TL-AI layer 118 may evaluate environmental conditions and operational requirements to priontize features critical to the task. In an example embodiment, a feedforward NN may assign relevance score by mapping features to task-specific objectives. For instance, a high score may be assigned to thermal anomalies for disease detection, while shape features may be prioritized for navigation.

[0043] Further, the one or more Al processing layers 116 configures each of the plurality of sensors 104 with at least one or more Al processing functions based on the determined ty pe and the relevance of the extracted plurality of one or more significant features. Further, the at least one or more Al processing functions is configured to process each of the plurality of significant features. In an example embodiment, the TL-AI layer 118 assigns Al processing functions, such as for example, but not limited to, detection, filtering, or classification, to each NFSU’s lower Al software part based on a feature type and a relevance. These functions may be implemented as AlP-modules 120 with pre-trained ML or NN models. For example, NFSUs 108 in the central array region may be assigned classification functions for high-relevance features, while peripheral NFSUs handle filtering for less cntical features. In an example embodiment, pre-trained AlP-modules 120, such as for example, but not limited to, a CNN for shape classification or a Gaussian filter for noise reduction, may be assigned to the NFSUs 108.

[0044] Further, the one or more Al processing layers 116 is configured to determine that whether the assigned at least one or more Al processing functions match with the real-time environmental conditions and the operational requirements. In an example embodiment, the ML-AI layers 124 may compare performance of assigned Al processing functions against the real-time environmental conditions and the operational requirements. The ML-AI layers 124 evaluates function outputs against expected performance metrics, such as classification accuracy or processing speed, stored in the BL-AI layer’s memory. In an example embodiment, a neuromorphic, analog or spike neural network (SNN) in the ML-AI layers analyzes time-series data from environmental sensors and function outputs to detect mismatches, such as reduced classification accuracy in low light or noisy conditions.

[0045] If the one or more Al processing functions mismatches with the real-time environmental conditions and the operational requirements, then the one or more Al processing layers 116 dynamically update the at least one or more Al processing functions in real-time to meet the real-time environmental conditions and the operational requirements based on the determination. The ML-AI layers 124 dynamically updates Al processing functions using, for example, a combination of genetic algorithms (GAs) and neural network-based selection of coupling parameters or weights within AIP -interfaces 122. Based on the mismatch detected previously, the ML-AI layers 124 modifies, adds, or switches functions (e.g., replacing a CNN with an infrared-focused NN model) to align with environmental changes and operational needs. Updates are propagated to NFSUs 108 via the TL-AI layer’s software-defined AIP Modules 120 and AIP interfaces 122. In an example embodiment, GAs optimizes coupling function locations and parameters by iteratively mutating and selecting configurations, guided by a fitness function evaluating performance against environmental and operational criteria.

[0046] Further, the one or more Al processing layers 116 determines a subset of sensors 104 and a subset of Al processing layers 116 to be reconfigured based on the updated at least one or more Al processing functions. Further, the process of reconfiguration may include for example, but not limited to, perform at least one or more of modification, addition, deletion, merger, and switching of the at least one or more Al processing functions between the plurality of one or more sensors 104 using one or more AlP-modules 120 and the one or more AlP-interfaces 122 to meet the real-time environmental conditions and the operational requirements. The one or more Al processing layers 116 may include, but not limited to, a Top Level-AI (TL-AI) layer 118, a Middle Level-AI (ML-AI) layer 124 and a Bottom Level-AI (BL-AI) layer 126.

[0047] In another example embodiment, the one or more Al processing layers 116 may include either a single Al layer or multiple Al layers comprising ‘N’ layers. Further, the one or more Al processing layers 116 may or may not be split into the top, bottom, or middle layers. In some implementations, the one or more Al processing layers 116 may act as a single layer, or multiple layers including many other intermediate layers, apart from the top, bottom, or middle layers.

[0048] In an example embodiment, the BL-AI layer 126, using, for example, a pre-trained decision tree, identifies which NFSUs 108 and Al processing layers require reconfiguration based on the updated functions. The BL-AI layer 126 maps the function requirements to sensor capabilities (e.g., infrared sensitivity) and layer roles (e.g., TL-AI for feature extraction), selecting a subset for reconfiguration in real-time to optimize task performance. In an example embodiment, the decision tree evaluates sensor and layer metadata (e.g., processing capacity, memory usage) against updated function requirements, selecting central NFSUs 108 andML- Al layers 124 for reconfiguration in real-time when infrared processing is prioritized. In another similar embodiment, for example, compound eyes of bees and insects with hundreds to thousands of lenses or NFSUs on dome shaped arrays prioritize visible and ultra-violet part of the spectrum over the red or infra-red parts when foraging for flowers in realistic noisy environments.

[0049] Further, the one or more Al processing layers 116 may autonomously reconfigure the determined subset of sensors 104 and the subset of Al processing layers 116 in real-time with the updated at least one or more Al processing functions. The BL-AI layer 126 initiates reconfiguration commands (e.g., reset, restart) that propagate upward through the ML- Al 124 to the TL-AI layer 118, adjusting the software-defined functionality 112 of the selected NFSUs and Al layers. Reconfiguration is executed using, for example, a software library of pre-trained AIP- modules 120 and their AlP-interfaces 122, optimized via simulated annealing (SA) to ensure fast and seamless integration with minimal memory adjustments. SA within AlP-interfaces 122 optimizes the coupling of updated AlP-modules 120, ensuring stable reconfiguration. For example, an SA algorithm adjusts weights in a WLA interface to connect an infrared-processing AIP- Module 120 to existing feature extraction modules.

[0050] In another example, a GA optimization under known local environmental and operations is used to not only refine and optimize the pre-computed weights of the relevant AIP- interfaces for higher accuracy results but also to replace the older pre-computed weights with the new optimized ones in the software library for future similar or across other applications uses.

[0051] Further, the one or more Al processing layers 116 may perform the determined one or more tasks by executing the updated at least one or more Al processing functions at corresponding sensors 104 and corresponding Al processing layers 116. Further, the updated at least one or more Al processing functions are configured to process each of the plurality of features. The processing system 130 executes tasks by applying the reconfigured Al processing functions across the TL-AI 118, ML- Al 124, and BL-AI 126 layers. The TL-AI layer 118 processes raw sensory data, the ML -Al layers 124 refine intermediate features, and the BL-AI layer 126 generates task-specific outputs using updated AlP-modules 120 and their AlP-interfaces within each layer. Tasks, such as, for example, image capture or obstacle detection, may be performed by coordinating sensor and layer operations in the BL-AI layer 126. Updated AIP- modules 120, such as, for example, an NN for classification or a GA-optimized filter, process features like thermal signatures or object shapes. The BL-AI layer 126 may integrate outputs using, for example, a decision-making NN to produce task results. Moreover, after reconfiguration of determined subset of sensors 104, the one or more Al processing layers 116 is configured to generate a processed output corresponding to the single or multi-modal sensory7signal based on the performed one or more tasks. The BL-AI layer 126, equipped with full memory, storage, andcommunication capabilities, aggregates task outputs from the hierarchical processing to generate a final processed output. This output, such as a spatial map or navigation command, is formatted for external communication (e.g., to a central system) using leaky communication protocols to minimize bandwidth. The output is stored in the BL-AI layer's memory for future reference or transmitted via a communication interface. In an example embodiment, the BL-AI layer 126 may be implemented as an NN, combines processed features (e.g., disease classifications, navigation paths) into a cohesive output, optimized for minimal data size using dimensionality7reduction techniques like principal component analysis (PCA).

[0052] Further, the system 102 may comprise the at least one or more Al processing functions configured to process each of the plurality of significant features received from the one or more sensory signals. The plurality of significant features may comprise for example, but not limited to, at least one of a color, an intensity, a luminosity, a shape, a pattern, a feature, a depth, a polarization, a frequency, an amplitude, a tone, a pitch, and a noise in any or all of the above.

[0053] Further, the system 102 may comprise the plurality of sensors 104 which may include, for example, but not limited to, a plurality of digital or neuromorphic functional sensing units (NFSUs) 108. The NFSUs 108 may be arranged in for example, but not limited to, at least one of a flat, dome, and faceted dome shaped array. Further, each NFSU 108 comprises an upper hardware receptor or lens like part 110 configured to receive the one or more sensory7signals and a lower Al software-defined functionality part 112 configured to process the received one or more sensory7signals. Further, the faceted dome shaped array is formed by an outermost dome-shaped array layer comprising one or more discontinuous faceted layers. Further, each discontinuous faceted layer may comprise a locally flat array connected to adjacent locally flat faceted arrays at specific angles to form a continuous curvature for the faceted dome shaped array.

[0054] Further, the one or more Al processing layers 1 1 may comprise one or more top Al processing layers 118 integrated with a lower Al software-defined functionality part 112 of the plurality7of sensors 104. Further, the one or more top Al processing layers 118 are configured to perform a first level of processing of the received one or more sensory signals to generate a first level processed data. The first level processed data represents a reduced-size dataset with enhanced details as compared to the received one or more sensory signals. The first level processed data may be achieved through for example, but not limited to, detection, filtering, classification, or convolution of the significant features.

[0055] Further, the one or more Al processing layers 116 may comprise one or more middle Al processing layers 124 communicatively coupled in between the one or more top Al processing layers 118 and one or more bottom Al processing layers 126. Further, the one or more middle Al processing layers 124 are configured to perform a second level of processing on the first levelprocessed data received from the one or more top Al processing layers 118 and generate a second level processed data. The second level processed data represent a further refined dataset with an even smaller data size and increased knowledge or inference. The second level processed data may be achieved through for example, but not limited to, additional detection, filtering, classification, or convolution operations.

[0056] Further, the one or more Al processing layers 116 may comprise one or more bottom Al processing layers 126 communicatively coupled to the one or more middle Al processing layers 124. Further, the one or more bottom Al processing layers 126 are configured to perform a third level of processing on the second level processed data received from the one or more middle Al processing layers 124. Further, generate an at least one of the third level and final processed data corresponding to the received one or more sensory signals. The at least one of the third level and final processed data corresponds to a processed output 128. The processed output 128 may comprise for example, but not limited to, navigational commands, obstacle avoidance signals, or event detection results, tailored to tasks for example, but not limited to, image capture, pattern recognition, or navigation. The processed output 128 may generate by full memory, storage, and communication capabilities in the one or more bottom Al processing layers 126. The processed output 128 may support for example, but not limited to, low-power, low-latency operations in dynamic edge or extreme edge environments.

[0057] Furthermore, the one or more Al processing layers 116 comprises one or more Al processing modules (AlP-modules) 120 coupled to the plurality of one or more neighboring AIP- modules 120 within the same Al processing layer 116. Further, the one or more AlP-modules 120 with at least one of the weights and parameters pre-determined and pre-trained for at least one or more Al processing functionalities of the said AlP-modules. Further, one or more Al processing interfaces (AlP-interfaces) 122, configured to at least one of the couples and connect at least two or more AlP-modules 120 with at least one of the weights and parameters predetermined, pretrained, and further optimized during an operation.

[0058] Further, the one or more Al processing layers 116 are configured to process the one or more sensory signals in an uncoupled mode. Further, one or more sensory signals and the corresponding AlP-module 120 operate uncoupled and independently through the one or more top Al processing layers 118, the one or more middle Al processing layers 124 and through the one or more bottom Al processing layers 126. Further, the one or more Al processing layers 116 is configured to process the one or more sensory signals in a coupled mode. Further, the plurality of sensory signals and the corresponding one or more AlP-modules 120 are at least one of the coupled and connected through the one or more AlP-interfaces 122 for at least one of the coupled and combined sensory signals processing through the one or more top Al processing layers 118, theone or more middle Al processing layers 124 and through the one or more bottom Al processing layers 126.

[0059] Further, the one or more Al processing layers 116 are configured to comprise one or more configurations of the AlP-modules 120 within the one or more top Al processing layers 118, the one or more middle Al processing layers 124, and the one or more bottom Al processing layers 126. Further, each AlP-module 120 configuration comprises of at least one of a machine learning (ML) mode, a neural network (NN) model, and a genetic algorithm (GA) model or like with pre-determined and pre-trained parameters. Further, one or more configurations of the AIP- interfaces 122 to either couple or connect two or more AlP-modules 120 within the one or more top Al processing layers 118, the one or more middle Al processing layers 124, and the one or more bottom Al processing layers 126. Further, each AlP-interface 122 configuration comprises of at least one of a linear approximation (LA), a weighted linear approximation (WLA), and a simulated annealing (SA), a machine learning (ML) model, a neural network (NN) model, and a genetic algorithm (GA) model or like with pre-determined and pre-trained parameters further optimized to enable one of a dynamic coupling and uncoupling of the one or more AlP-modules 120 in real-time. Further, the one or more Al processing layers 116 adjusts the modular configuration of the AlP-modules 120 and the AlP-interfaces 122 based on the type and relevance of the extracted plurality of significant features.

[0060] Further, to determine the type and relevance of each of the extracted plurality of significant features the one or more Al processing layers 116 are configured to classify each of the extracted plurality7of significant features into one or more categories using an Al model for example, but not limited to. the ML model, the NN model, the GA model or the like. The categories may comprise for example, but not limited to, object identification, environmental characterization, or event detection. For example, but not limited to, in a visual sensory context, categories may include object boundaries based on color and intensify, spatial attributes based on depth and shape, or surface properties based on polarization and pattern. In an audio context, categories may comprise for example, but not limited to, sound source types based on frequency and tone, or event indicators based on amplitude and noise. Further, the one or more Al processing layers 116 are configured to compute a relevance score for each classified features with the determined one or more tasks to be performed. Further, the relevance score indicates relevance level of each of the classified features with the determined one or more tasks to be performed.

[0061] Further, to dynamically update the at least one or more Al processing functions of one or more AlP-modules 120 to meet the real-time environmental conditions and the operational requirements based on the determination, the one or more Al processing layers 116 are configured to perform for example, but not limited to, at least one or more of modification, addition, deletion,merger, and switching of the at least one or more Al processing functions between the plurality of one or more sensors 104 using one or more AlP-modules 120 and the one or more AIP -interfaces 122 to meet the real-time environmental conditions and the operational requirements.

[0062] Further, to dynamically update the at least one or more Al processing functions of one or more AlP-modules 120 to meet the real-time environmental conditions and the operational requirements based on the determination, the one or more Al processing layers 116 are configured to assign significant Al processing functions of significant AlP-modules 120 to plurality of sensors 104 centrally located in significant region of an input array 202 with multiple NFSUs 108. Further, the one or more Al processing layers 116 are configured to assign non-significant Al processing functions of non-significant AlP-modules 120 to plurality of sensors 104 peripherally or edge located non-significant regions of the input array 202 with multiple NFSU 108 based on the realtime environmental conditions and the operational requirements.

[0063] Furthermore, to autonomously reconfigure the determined subset of sensors 104, the subset of Al processing layers 116 and the subset of AlP-modules 120 the one or more Al processing layers 1 16 are configured to determine the real-time environmental conditions and the operations requirements using a first set of sensors among the plurality of sensors 104. Further, the one or more Al processing layers 116 is configured to map operation parameters of each of the plurality of sensors 104 with the determined real-time environmental conditions and the operations requirements. Further, the one or more Al processing layers 116 is configured to determine a second subset of sensors 104 comprising the operations parameters deviating from the determined real-time environmental conditions and the operations requirements. Further, the one or more Al processing layers 116 is configured to determine the at least one or more Al processing functions of one or more AlP-modules 120 relevant to the determined second subset of sensors 104. Further, the one or more Al processing layers 1 16 is configured to autonomously reconfigure the determined second subset of sensors 104 to process the plurality of significant features based on the at least one or more Al processing functions of one or more AlP-modules 120.

[0064] Further, to autonomously reconfigure the determined subset of sensors 104 and the subset of Al processing layers, the one or more Al processing layers 116 are configured to initiate for example, but not limited to, at least one of a reset, a restart, and reconfiguration commands via the one or more bottom Al processing layers 126. Further, the one or more Al processing layers 116 propagates for example, but not limited to, at least one of the resets, the restart, and the reconfiguration commands upward to the one or more top Al processing layer 118 with minimal memory adjustments via the one or more middle Al processing layers 124. Furthermore, the one or more Al processing layers 116 autonomously reconfigure the determined subset of sensors 104 and the subset of Al processing layers using the one or more top Al processing layers 118.

[0065] Further, to autonomously reconfigure the determined subset of sensors 104 and the subset of Al processing layers, the one or more Al processing layers 116 are configured to selectively perform one of activation and deactivation of the determined subset of sensors 104 and the subset of Al processing layers based on the analysis of the real-time environmental conditions and the operational requirements.

[0066] Further, to perform the determined one or more tasks by executing the updated at least one or more Al processing functions of one or more AlP-modules 120 at corresponding plurality of sensors 104 and corresponding Al processing layers 116, the one or more Al processing layers 116 are configured to apply the updated at least one or more Al processing functions of one or more AlP-modules 120 onto a trained aforementioned Al model to determine optimal operational parameters for the determined subset of sensors 104. Further, the one or more Al processing layers 116 adjusts the current operational parameters of the determined subset of sensors 104 with the determined optimal operational parameters based on an output of the trained Al model. Furthermore, perform the determined one or more tasks on the determined subset of sensors 104 based on the adjusted operational parameters. The determined one or more tasks may comprise for example, but not limited to, an image capture, pattern, feature, obstacle and depth detection and traction, a location navigation, an event detection and classification, and many others without any limitations.

[0067] Example Use Case Scenario for the Evolutionary Neuromorphic Sensor Fusion Platform

[0068] In an exemplary embodiment, the Evolutionary Neuromorphic Sensor Fusion Platform (ENVSP) 102, as described herein, may be implemented in an autonomous agricultural drone configured for crop health monitoring and navigation within a vineyard under dynamic environmental conditions, including variable illumination and precipitation. The drone may comprise a plurality of neuromorphic functional sensing units (NFSUs) 108 arranged in a faceted dome-shaped array, each NFSU 108 comprising an upper hardware receptor or lens like part 110, such as a micro-lens assembly, for capturing multi-modal sensory signals including visual and infrared data, and a lower Al software-defined functionality part 112 for processing the multimodal sensory signals. The ENVSP 102 operates through a biomorphic hierarchical architecture comprising one or more top Al processing layers (TL-AI) 118, one or more middle Al processing layers (ML-AI) 124, and one or more bottom Al processing layers (BL-AI) 126. Upon deployment, the TL-AI layer 118, communicatively coupled to the NFSUs 108 and configured without memory, storage, or input-output capabilities, receives the one or more sensory signals comprising visual images of grapevines and infrared thermal data. The TL-AI layer 118 determines operational tasks, including identification of unhealthy vines and navigation around trellises, and extracts aplurality of significant features, such as leaf color, shape, and thermal signatures, using predetermined Al processing modules (AlP-modules) 120 implementing machine learning (ML) models. A neural network (NN) model within the TL-AI layer 118 may classify the plurality of significant features and computes relevance scores, prioritizing the plurality of significant features, indicative of disease, such as discolored leaves and elevated thermal signatures, thereby reducing data size while increasing task-relevant knowledge. The processed data is transmitted to the ML- AI layers 124, which, configured with minimal memory and no input-output capabilities, further process the data using additional AlP-modules 120 interconnected by Al processing interfaces (AlP-interfaces) 122 employing weighted linear approximation (WLA).

[0069] Upon detecting a change in environmental conditions, such as reduced illumination due to cloud cover, the ENVSP 102 determines a subset of NFSUs 108 and ML-AI layers 124 for reconfiguration. A genetic algorithm (GA) within the AlP-interfaces 122 may dynamically update the Al processing functions, reconfiguring centrally located NFSUs 108 to prioritize infrared signal processing for thermal detection and peripherally located NFSUs 108 to focus on shape detection for obstacle avoidance. The processed data, further reduced in size and enriched with inference, is passed to the BL-AI layer 126, which is configured with full memory, storage, inputoutput, reset, restart, reconfiguration, and leaky communication capabilities. The BL-AI layer 126, utilizing pre-trained AlP-modules 120 and simulated annealing (SA) interfaces, generates a final processed output comprising a spatial map of unhealthy vines and navigation commands to avoid trellises. In response to precipitation onset, the BL-AI layer 126 initiates a reconfiguration command, propagated upward through the ML-AI layers 124 to the TL-AI layer 118 with minimal memory’ adjustments, reconfiguring NFSUs 108 to optimize for wet conditions by activating functions resilient to moisture-induced noise. The drone may execute the determined tasks, including capturing images of affected vines, classifying disease severity, and navigating a collision-free path, while transmitting minimal data, such as coordinates of unhealthy vines, to an external system via leaky communication. This operation exemplifies the ENVSP’s 102 capability' to process multi-modal sensory signals, autonomously reconfigure sensors and Al processing layers functionalities in real-time based on environmental and operational requirements, and perform tasks with reduced power, memory, and latency at the edge.

[0070] The foregoing example use case scenario of the system 102, as described for an autonomous agricultural drone, is provided solely for illustrative purposes to demonstrate one potential application of the invention. The described scenario is not intended to limit the scope of the invention to the specific embodiment or application presented. The system 102 may be implemented in a wide variety7of other configurations, applications, and environments, including but not limited to autonomous vehicles, robotic systems, industrial automation, remote safety,securin’, and environmental monitoring, and other edge or extreme edge computing scenarios, utilizing various multi-modal sensory inputs and task requirements, without departing from the principles and scope of the invention as defined by the claims.

[0071] Those of ordinary skilled in the art will appreciate that the hardware depicted in FIG .1A may vary for particular implementations. For example, other peripheral devices such as an optical disk drive and the like. Local Area Network (LAN), Wide Area Network (WAN), Wireless (for example, Wi-Fi) adapter, graphics adapter, disk controller, input / output (1 / 0) adapter also may be used in addition or in place of the hardware depicted. The depicted example is provided for the purpose of explanation only and is not meant to imply architectural limitations with respect to the present disclosure.

[0072] Those skilled in the art will recognize that, for simplicity and clarity, the full structure and operation of all data processing systems suitable for use with the present disclosure is not being depicted or described herein. Instead, only so much of the system 102 as is unique to the present disclosure or necessary for an understanding of the present disclosure is depicted and described. The remainder of the construction and operation of the system 102 may conform to any of the various current implementation and practices know n in the art.

[0073] FIG. IB illustrates a block diagram representation 100B of example electronics packaging 132. such as those shown in FIG. 1A, in accordance with an embodiment of the present disclosure. In an example embodiment, the electronic packaging 132 may be configured to house and interconnect the hardware components via one or more electrical buses to support the hierarchical, biomorphic Al architecture of the system 102. The electronics packaging 132 may include for example, but not limited to, one or more processors 114 for executing Artificial Intelligence (Al) processing functions within Al processing modules (AlP-modules) 120, a memory / storage units 106 for storing pre-determined and pre-trained parameters and processed data, an input / output (I / O) components 134 for performing data communication between system components and external devices, a power supplies unit 138 to power the system 102 and to ensure low-power operation suitable for edge or extreme edge environments, secondary sensors 136 for capturing additional environmental data, and servos and motors 140 for enabling autonomous operations in applications for example, but not limited to, drones 1004, robots 1006. The components are housed within the packaging and are interconnected via one or more electrical buses and may be implemented on traditional ASIC or SOC hardware, FPGAs hardware, or advanced neuromorphic, analog-AI, or in-memory computing chips, ensuring efficient, low- latency, and low-power performance.

[0074] In one embodiment, the electronics packaging 132 may include the processor 114 configured to control and coordinate the operations of the layered Al architecture and execute theAl software-defined functionality part 112 processed through the one or more Al processing layers 116. The processor 114 may comprise, without limitation, a microcontroller, digital signal processor (DSP), field programable gate arrays (FPGA) processors, neuromorphic computing unit, or application-specific integrated circuit (ASIC), selected based on performance and power requirements for the target deployment environment. The processor 114 is operatively connected to a memory and storage unit 106, which provides volatile and non-volatile memory support for storing one or more sensory' signal, Al model parameters, intermediate feature maps, and final processed outputs 128.

[0075] Additionally, the electronics packaging 132 integrates one or more input / output (I / O) components 134 configured to allow communication between the system 102 and external devices, networks, or cloud systems. The I / O components 134 may include for example, but not limited to, communication ports, serial and parallel interfaces, wireless transceivers, and network modules. The electronics packaging 132 further comprises one or more secondary sensors 136 configured to capture additional environmental parameters for example, but not limited to, temperature, humidity, motion, pressure, light intensity, location, or radiation induced variations and noises.

[0076] FIG. 2A illustrates a schematic representation 200A of an example Evolutionary' Neuromorphic Vision fusion Sensor Platform (ENVSP) for processing one or more sensory signals received by a plurality of sensors 104, in accordance with an embodiment of the present disclosure. FIG. 2A shows a Top Level- Artificial Intelligence (TL-AI) layer 118 (also referred herein as ‘one or more top level Al processing layers’ 118) integrated with multiple NFSUs 108 arranged in a formation for example, but not limited to, groups or patterns. Further, the multiple NFSUs 108 are arranged on for example, but not limited to, a flat, dome or faceted dome shaped array for receiving one or more sensory signals from multiple channel visual sensing images or signals. The TL-AI layer 118, in conjunction with underlying Middle Level -Al (ML- Al) layer 124 (also referred herein as ‘one or more middle level Al processing layers' 124) and Bottom Level-AI (BL-AI) layer 126 (also referred herein as ‘one or more bottom level Al processing layers’ 126) , is able to configure or reconfigure the Al software defined functionality part 112. The significant or important features of the one or more sensory signals are arranged in a central region of an array, while the non-significant or less important features of the one or more sensory signals are positioned toward the edges of the array (also referred herein as ‘input NFSU array’ or ‘input array’). The TL-AI layer 118 processes the received one or more sensory signal which comprises of high storage data and may contain irrelevant data. The TL-AI layer 118 reduces the size of received one or more sensory' signal by performing for example, but not limited to, detection, filtering, classification, or convolution of the significant features. Further, the reduced data sizedinformation and knowledge, received from the TL-AI layer 118 is passed on to the one or more ML- Al processing layers 124 for further Al processing. Further, passing on the reduced data size and maximal knowledge size to the BL-AI layer 126. The BL-AI layer 126 is equipped with maximum memory, storage, input-output, reset, reconfiguration, and communication capabilities. The BL-AI layer 126 provides the final processed data which may comprise for example, but not limited to, navigational commands, obstacle avoidance signals, or event detection results, tailored to tasks for example, but not limited to, image capture, pattern recognition, or navigation.

[0077] FIG. 2B illustrates a schematic representation of an example conceptual technology' of Al based reconfigurable neuromorphic vision sensor fusion system 200B, in accordance with an embodiment of the present disclosure. FIG. 2B is divided into two sections, with the left section depicting an inspiration from a honeybee neurology 208. The right section illustrating a reconfigurable biomorphic vision sensor platform 212, highlighting the scalability and modularity' of the ENVSP 102.

[0078] The left section of FIG. 2B includes a depiction of the honeybee neurology 208. The honeybee neurology 208 is illustrated as a simplified cross-sectional diagram of the insect’s neural structure, emphasizing compound eyes and integrated vision-brain system, which sen e as biological inspiration for the ENVSP 102. Further, the ENVSP 102 depicts visual processing of the bee 206, particularly, a very efficient ability of a bee 206 to process multiple (5000-6000) visual sensory signals through multiple sensing units simultaneously, focusing on relevant data within noisy environments. Moreover, FIG. 2B illustrates example multiple micro sized analog- AI or Micro- Al, or Spike Neural Network (SNN) Models 204 adjacent to the honeybee neurology' 208. The Micro-AI Models 204 represent an example modular Al processing modules (AIP- modules) 120 (also referred herein as AlP-modules 120) and AIP interfaces 122 to couple two or more than two AIP-Modules as described above, and which make an analogous distributed neural network processing, inspired by the bee 206, possible as an embodiment of the present invention. The AlP-modules 120 are pre-determined, pre-computed, or pre-trained. The AlP-interfaces 122 are configured to at least connect or combine the AlP-modules 120, allowing real-time reconfiguration and processing of one or more sensory signals across a range of 10s to 1000s of micro-AI models 204.

[0079] In one aspect, the system 102 may include creating a library of one or more predetermined, pre-computed, and pre-trained AlP-modules 120 and one or more AlP-interfaces 122. The one or more AIP interfaces 122 are configured for joining, stitching, or combining one or more AIP modules 120 in one or more than one to all possible combinations. The combinations are made for use in real time for top-to-bottom biomorphic layered Al architecture, system, andmethod for processing one or more sensory signals either in series or in parallel during an operation.

[0080] The top right section of FIG. 2B illustrates a resulting Modular Evolutionary Reconfigurable Neuromorphic Artificial Intelligence (ERN-AI) Software 210 and a bio-morphic vision sensor platform 212 on the bottom right. The Modular ERN-AI software 210 and the bio- morphic vision sensor platform 212 are functionally and architecturally inspired by the compound eye structure and honeybee neurology' 208 of the bee 206. The modular ERN-AI software 210 comprises a configurable multi-layer structure, geometrically represented as a dome-shaped architecture, corresponding to the organization of compound vision systems. The modular ERN- AI software 210 includes dynamically reconfigurable software-defined logic blocks 116 (also referred herein as Al processing layers) capable of adapting the one or more Al functionality in real time based on the environmental conditions, task relevance, or the operational requirements, thereby mimicking context-aware adaptability of biological organisms. The bio-morphic vision sensor platform 212 further comprises a vertically stacked architecture including multiple Al processing layers configured with a plurality of NFSUs 108. Each Al processing layer 116 is configured to receive and preprocess the received one or more sensory signals. The bottom Al processing layer 126 comprises, for example, but not limited to, a memory' unit 106-2, communication interfaces, and reconfiguration logic modules. Collectively, the components enable the system 102 to perform real-time, parallel, and localized sensory processing in a low power, low memory, and low latency configuration suitable for edge or extreme-edge deployment environments. The architectural design of the system 102 allows for selective, task-specific sensor engagement and software-level reconfiguration in a modular and evolutionary manner. Hence, the system 102 supports scalable and autonomous vision processing across diverse applications for example, but not limited to, robotics, drones, autonomous vehicles, and remote sensing and monitoring systems.

[0081] FIG. 2C illustrates a schematic representation 200C of side view of an example ENVSP with a Neuromorphic Functional Sensing Unit (NFSU) in an array layer coupled to underneath Al layers, in accordance with an embodiment of the present disclosure. Further, the ENVSP 102 consists of one or more NFSUs 108 which may be arranged in groups or patterns on an outermost input array layer. The input array 202 with multiple NFSUs 108 may be, for example, but not limited to, a flat, a dome-shaped, or faceted dome-shaped, and a designed to receive one or more sensory signals either in parallel or sequentially. The individual or multiple NFSUs 108 are disposed on the outermost input array layer 202. The outermost input array layer 202 may be, for example, not limited to, flat, dome-shaped, or faceted dome-shaped. The NFSUs 108 are integrated with hierarchical one or more Al processing layers 116 comprising the TL-AI 118, theML- Al 124, and the BL-AI 126 layers. The integration allows receiving, processing, and analyzing the one or more sensory signals. The system 102 operates with a low memory usage, a low power consumption, a reduced input-output, and a leak}7data communication. This configuration allows a low-latency, a low-power autonomous operation at the edge or extreme edge computing of the input array. Further, the outermost dome shaped array layer with single or multiple NFSUs 108 may not be feasible in a single continuous dome shaped layer with a single continuous curvature of the dome. In such cases, the outermost dome shaped input array layer 202 may be considered as made of a series of discontinuous faceted dome shaped layer such that each locally flat faceted array is joined with other neighboring locally flat faceted arrays at very small angles forming the needed overall continuous curvature for an overall dome shaped layer. Moreover, the electronics packaging 132 comprises a unified structure incorporating a memon / storage unit 106, a processor 114, an input / output (I / O) component 134, a plurality' of secondary' sensors 136, a power supply unit 138, and servomechanical components including servos and motors 140.

[0082] FIG. 3 illustrates a schematic representation of an example NFSU with an upper hardware receptor or lens like part 110 and a lower Al software defined functionality7part 112, in accordance with an embodiment of the present disclosure. In one aspect, each NFSU 108 on an outermost flat, dome, or faceted dome-shaped array layer in the ENVSP 102 may include an individual upper hardware receptor or lens like part 110 or a combination of the upper hardware receptor or lens like part 110. The upper hardware receptor or lens like part 110 forms a top optical hardware sensing part for receiving the one or more sensory7signal by each NFSU 108. However, in another aspect, the bottom part of each NFSU 108 is lower Al software defined functionality7part 112. For example, the lower Al software defined functionality part 112 comprises one or more Al processing layer 116. The Al software defined functionality part 112 is configured to perform one or more basic functions. The one or more basic functions may include, for example, but are not limited to, detection, filtering, classification, combination, or convolution or the like. The one or more basic functions may be applied to one or more basic features of the received one or more sensory signals. The one or more basic features (also referred herein as ‘plurality of significant features’) may include, for example, but are not limited to, color, spectral frequency, luminosity7, shape, pattern, feature, depth, or polarization or the like.

[0083] In one aspect, the upper hardware receptor or lens like part 110, nature, or functionality of each NFSU 108 may refer to for example, but not limited to, any physical, chemical, material, structural, doping, or coating composition of the upper hardware receptor or lens like part 110 of the NFSU 108. The composition of the upper hardware receptor or lens like part 110 is fixed and does not change in response to the received one or more sensory7signals received by each NFSU 108 during real-time operation. In another aspect, a response behavior ofthe upper hardware receptor or lens like part 110 may vary based on changing significant features of the received one or more sensory signal and a surrounding environment or operating conditions of the ENVSP 102 during real-time operation. Furthermore, the lower Al software defined functionality part 112, nature, or functionality of each NFSU 108 may include, for example, but not limited to, machine learning (ML), neural network (NN), or genetic algorithm (GA) based software or the like. The lower Al software defined functionality part 1 12 may perform at least one, or all of detection, filtering, classification, combination, or convolution of one or more basic features of the received one or more sensory signal. The significant features may include, but not limited to, color, spectral frequency, luminosity, shape, pattern, feature, depth, or polarization, as defined by the lower Al software defined functionality part 112 of the NFSU 108.

[0084] FIG. 4 illustrates a schematic representation of exemplary one or more NFSUs comprising one or more Al software defined multiple functionalities, in accordance with an embodiment of the present disclosure. Parts (a)-(d) of FIG. 4 illustrates multiple NFSUs 108 capable of performing one or more basic functions. The one or more basic functions may include for example, but are not limited to, detection, filtering, classification, combination, or convolution or the like. The one or more basic functions may be applied to one or more basic features of the received one or more sensory signals. The significant features may include for example, but are not limited to, color, spectral frequency, luminosity, shape, pattern, feature, depth, polarization, or any combination thereof. FIG. 4(a) shows the combined processing of multiple significant features, while FIGs. 4(b)-4(d) shows individual processing of the significant features by the Al software-defined functionality7part 112 of NFSUs 108.

[0085] In one aspect of the present disclosure, the Al software-defined functionality part 112 of each individual NFSU 108 within the ENVSP 102 are controlled or configurable through underlying TL-AI 118, ML- Al 124, and BL-AI 126 layers in real time during the operation of the ENVSP 102. Further, the overall Al software-defined functionality part 112 of the ENVSP 102 are reconfigurable in real time. The reconfiguration of Al software-defined functionalities part of the NFSU 108 enables the ENVSP 102 to focus on the significant, relevant, or required functions among all available functionalities. The selection of relevant data depends on the received one or more sensory7signals, the environmental or operating conditions, and the specific application for which the ENVSP 102 is deployed.

[0086] FIG. 5 illustrates a schematic representation 500 of exemplary7one or more Al processing layers 1 16 configured with a plurality of Al software defined reconfigure-ability of a top layer (TL)-AI 118, in accordance with an embodiment of the present disclosure. Parts (a)-(d) of FIG. 5 depict top views of four example configurations or reconfigurations of the TL-AI 118 within the ENVSP 102. The example configurations of the TL-AI 118 represent possibleautonomous reconfigurations of the TL-AI 118, among many others, which may occur in real time during operation. The ability of autonomous reconfigurations of the TL-AI 118 allows the system 102 to use low power, low memory, and low latency one or more sensory signals. In one aspect, similar to compound eyes of the bees 206 or other insects, the system 102 allows receiving one or more sensory signals through multiple NFSUs 108, ranging from hundreds to thousands, either in parallel or in sequence. The system 102 is configured to focus only on the significant, relevant, or needed significant features of the available Al software-defined functionalities part 112. The ability' to focus on relevant significant features is particularly applicable when processing low- resolution or uncontrolled-resolution significant features received from the one or more sensory- signals within a close-proximity, congested, dark, or noisy environment dunng real-time operation.

[0087] In another aspect, a subset of the NFSUs 108 arranged in a group or pattern on the outermost flat, dome, or faceted dome-shaped array may be used to sense surrounding environmental significant features of the one or more sensory signals. The remaining NFSUs 108 in the same group or pattern may be autonomously reconfigured to focus on the significant, relevant, or required Al software-defined functionalities part 112 based on the focused or target tasks, in real time during operation.

[0088] FIG. 6 illustrates a schematic representation 600 of an example ENVSP 102 with multiple NFSUs configured with multiple Al software defined functionalities processed through underlying Al layers with or without AlP-modules 120, in accordance with an embodiment of the present disclosure. In one aspect, Al processing is performed through each of the TL-AI 118, ML- AI 124. and BL-AI 126 layers in a modular manner using the one or more AlP-modules 120. The one or more AlP-modules 120 are interconnected through one or more AlP-interfaces 122. which may be joined or separated in real time during operation. In another aspect, a library is constructed comprising one or more AlP-modules 120 and one or more AlP-interfaces 122. The library7includes rules for dynamically joining or separating AlP-modules 120 to enable autonomous reconfigurability through the TL-AI 118, ML-AI 124, and BL-AI 126 layers in real time during operation. Further, some AlP-modules 120 and AlP-interfaces 122 are pre-determined, precomputed, and pre-trained for reuse during operation or across multiple applications. The remaining AlP-modules 120 and AlP-interfaces 122 may be computed, trained, or optimized dynamically during operation or across applications. The micro-size modular nature of AIP- modules 120 and AlP-interfaces 122 allows some AlP-modules 120 and AlP-interfaces to be further computed, trained, or optimized dynamically during operation or across applications according to the local environmental and operational conditions.

[0089] Further, in part B of FIG 6 . the Al software-defined functionality part 1 12 of each NFSU 108, as well as the corresponding Al processing functionalities within the TL-AI layer 118, ML- Al layer 124, and BL- Al layer 126, are configured such that each NFSU 108 operates in an entirely uncoupled manner from the Al software-defined functionality part 112 of neighboring NFSUs 108 during operation. In another aspect, the response behavior of the upper hardware receptor or lens like part 110 of each NFSU 108 may vary based on the change in significant features of the received one or more sensory signals and the environmental conditions or operational requirements of the ENVSP 102 during real-time operation.

[0090] FIG. 7 illustrates a schematic representation 700 of example ENVSP 102 with multiple NFSUs 108 comprising multiple Al defined functionality types configured in multiple different sequences or configurations where each Al software defined functionality is processed through Al processing layers 116 with uncoupled AlP-modules 120, in accordance with an embodiment of the present disclosure. In one aspect, the Al software-defined functionality part 112 of each NFSU 108, along with the corresponding AlP-modules 120 and AlP-interfaces 122 within the TL-AI 118, the ML-AI 124, and the BL-AI 126 layers, remain uncoupled from one another layer. Each NFSU 108 may be processed independently, simultaneously, or in parallel during a top-to-bottom operational cycle of the ENVSP 102. In FIG. 7(a) NFSUs 108 are shown to possess assigned Al software-defined functionalities 702-1 for example, but not limited to. 1, 2, and 3 arranged in a specific sequence or configuration (1, 2, 3, 2, 1). Each NFSU 108 is independently configured to process the received one or more sensory signal’s significant features for example, but not limited to, color, shape, depth, or polarization. The output of each NFSU 108 is independently directed into corresponding uncoupled AlP-modules 120 within the one or more Al processing layers 116. The one or more Al processing layers 116 emphasizes modularity, where each AlP-module 120 operates autonomously and processes the received one or more sensory signal without being affected by adjacent AlP-modules 120. Similarly, the FIG. 7(b) depicts an alternate sequence or configuration of NFSUs 108 configured with the same Al software-defined functionalities 702-2 for example, but not limited to, 1, 2, and 3 but arranged in a different order (1, 3, 2, 3, 1). The Al functionalities of the NFSUs 108 are again processed through corresponding uncoupled AlP-modules 120 and associated one or more Al processing layers 116, with each processing the received one or more sensory' signals independently. Further, by using uncoupled configuration of AIP modules 120, each NFSU 108 to AIP- modules 120 path is insulated from the influence of other AIP- modules 120, ensuring individualized significant feature extraction and processing, which is particularly valuable in noisy, dynamic, or resource-constrained edge computing scenarios.

[0091] FIG. 8 illustrates a schematic representation 800 of an example ENVSP 102 with multiple NFSUs 108 comprising multiple Al defined functionality types configured in multiple different sequences, where each Al software defined functionality is processed through Al processing layers 116 with coupled AlP-modules 120, in accordance with an embodiment of the present disclosure. In one aspect, the Al software-defined functionality part 112 of each NFSU 108, along with the subsequent processing through the corresponding AlP-modules 120 and AIP- interfaces 122, remains logically and functionally coupled or dependent on the neighboring AIP- modules and AlP-interfaces. As such, in this case, the processing of each NFSU 108 may not occur in an uncoupled, individual manner during a top-to-bottom processing cycle of the ENVSP 102 during operation. The processing of each NFSU 108 and their neighboring NFSUs 108, in this case, may still occur in a simultaneous and parallel manner during top-to-bottom processing cycle of the ENV SP 102 during operation. The difference between the coupled and uncoupled cases is that the assigned functionality of an NFSU 108 is coupled or influenced by the assigned functionalities of at least first nearest-neighboring NFSUs 108 during top-to-bottom processing cycle of the ENVSP during an operation. This allows significant functionality passing or influencing between at least first nearest-neighboring NFSUs during top-to-bottom processing cycle of the ENVSP 102 during an operation. The bio-morphic layered and modular Al architecture of ENVSP 102 with AlP-blocks and AlP-interfaces thus allows simultaneous and parallel processing of each NFSU during top-to-bottom processing cycle of all the NFSUs during an operation.

[0092] In one aspect, the one or more AlP-modules 120 are an object-oriented Al software processing modules for example, but not limited to, machine learning (ML), neural network (NN), or genetic algorithm (GA) or the like with predetermined, pre-computed, or pre-trained model parameters. The one or more AlP-modules 120 are configured to process the received one or more sensory signals to generate processed output 128 and corresponding knowledge without altering any internal parameters, structure, or operational configuration of the one or more AlP-modules 120 during operation. In another aspect, the one or more AlP-interfaces 122 are further an object- oriented Al software processing interface for example, but not limited to, linear approximation (LA), weighted linear approximation (WLA), simulated annealing (SL), machine learning (ML), neural network (NN), and genetic algorithm (GA) or the like. The AlP-interface 122 may connect, join, stitch, or combine any two or more than two similar or different AlP-modules 120 with Al model which may be predetermined, pre-computed, or pre-trained or could be computed, simulated, trained, or optimized during a training or an operation.

[0093] In one aspect, a key difference between the Al processing modules (AlP-modules)120 and Al processing interfaces (AlP-interfaces) 122 is that generally the AlP-modules 120 arepre-determined, pre-computed, and pre-trained during an operation and application. The AIP- modules 120 may not change and may be used for a variety of other and new operations and applications as well. However, the AlP-interfaces 122 may be pre-set, pre-determined, precomputed, and pre-trained during an operation and application, and yet may be further optimized during an operation and application. The AlP-interfaces 122 may be changed and re-optimized for a variety of other operations and applications.

[0094] Moreover, FIG. 8 illustrate two different configurations of multiple or individual NFSUs 108 and the corresponding AlP-modules 120 and AlP-interfaces 122 within the TL-AI 118, ML-AI 124, and BL-AI 126 layers, during a top-to-bottom Al processing layers 116 architecture of an ENVSP 102 comprising multiple NFSUs 108. In one aspect, the ENVSP 102 operates with an Al software-defined configuration ordering of (1,2, 3, 2,1) 702-1 for top-to-bottom processing during the operation. In another aspect, the same ENVSP 102 operates with a different Al software-defined configuration ordering of (1,3, 2, 3,1) 702-2 for top-to-bottom processing during the same operation. Hence, the system 102 demonstrates software-defined reconfigurability during operation. In another aspect, Al software-defined reconfigurability of the ENVSP 102 comprising hundreds to thousands of hardware receptor or lens like part 110, NFSUs 108, and received one or more sensory signals enables tens of thousands to millions of configuration possibilities in real time during the operation. The reconfiguration is based on the nature of the one or more sensory signals received, the operational requirements, the environmental conditions of the ENVSP 102, and the target goals or objectives defined by the intended use-case scenarios. The reconfiguration is show n in FIG. 8, by the use of three different ty pes of NFSUs 108 and the corresponding AlP-modules 120. The AlP-modules 120 are connected, joined, stitched, or combined using multiple different types of AlP-interfaces 122. identified as (12, 23, 32, 21, 13, 31), among the multiple NFSUs and the AlP-modules 120 forming the ENVSP 102.

[0095] FIG. 9A illustrates a schematic representation of example top to bottom vertical view of biomorphic modular and layered Al architecture of the ENVSP 102 , in accordance with an embodiment of the present disclosure. FIG. 9B illustrates a schematic representation of example side to side horizontal view of biomorphic modular and layered Al architecture of the ENVSP 102, in accordance with an embodiment of the present disclosure. FIG. 9A illustrates a top-to-bottom biomorphic hierarchical architecture which enables maximum reception of one or more sensory signals through multiple NFSUs 108 without requiring memory. I / O, or storage in the TL-AI layer 118 at the top. At the bottom, the BL-AI layer 126 handles much lower or reduced sized processed data, inferred knowledge, and model parameters, with full memory storage 106, I / O 134, and communication support, including on-off, restart, reconfiguration, reset, and leaky communication capabilities. The system 102 supports top-to-bottom processing with fullinitialization, reset, and reconfiguration through bottom-to-top operation. The ML-AI layer 124 is configured to reduce the size of the first level processed data to relevant and necessary components during the downward flow and supports initialization, reset, and recovery during the upward flow of processed output 128 or also referred as third level processed data.

[0096] As shown in FIG. 9B, the described top-to-bottom biomorphic hierarchical architecture may also be implemented in a single-layered, side-by-side configuration. The horizontal configuration includes compatibility with, for example, but not limited to, current Application-Specific Integrated Circuits (ASIC), field programmable gate arrays (FPGAs), and System on a Chip (SoC)-based hardware chips and accelerators, as well as with existing and future neuromorphic analog- Al and in-memory computing chips and devices. The ENVSP 102 may be implemented on the ASIC designs of all current and future SoC hardware based traditional chips and or Al accelerator chips. On the other hand, the ENVSP 102 may also be implemented on upcoming neuromorphic. crossbar, analog- Al, in-memory computing hardware, and chips on SOC and new and novel Neuromorphic and materials and devices.

[0097] FIG. 10 illustrates a schematic representation 1000 of plurality of Al based reconfigurable neuromorphic vision sensor fusion systems also deployed within a plurality of network devices, in accordance with an embodiment of the present disclosure. The system 102 targets a range of use cases at the edge or extreme edge on a LAN or WAN network, addressing applications with extremely low Size, Weight, and Power (SWaP) requirements. The use cases may include, but are not limited to, GPS-free or visual navigation and mapping for robots 1006 and drones 1004, enabling autonomous operation in environments where traditional navigation systems are unavailable, for example, but not limited to, tunnels, mines, indoors of large buildings or structures, or remote locations, sites or the like. The system 102 also may support event-driven and selective cameras 1008-1 and 1008-2. The event driven and selective cameras 1008-1 and 1008-2 focus on significant occurrences to reduce bandwidth usage. Such selective cameras 1008- 1 and 1008-2 are suitable for applications in aerospace and defense, where real-time, low-power vision processing is critical. In manufacturing and industrial robotics, the modular ERN-AI software 210 facilitates real-time quality control and predictive maintenance, enhancing operational efficiency, and ENSP vision sensors 220 facilitate 360° situational awareness and accident or collision avoidance for humans and humanoid robots operating in close proximity and noisy environments. For smart cities and urban areas, the system 102 allows advanced infrastructure management through low-power, real-time vision sensing. In remote safety and security’ monitoring, the system 102 supports autonomous surveillance applications. In human healthcare, the system 102 aids in real-time diagnostics and vision repair systems for elderly users, and real-time vision sensing and responses using AR / VR / XR within immersive environments. Theapplications in such cases may leverage the system’s event-driven and low-latency capabilities. Within the local and wide area networked applications, the modular ERN-AI software (210) and bio-morphic ENSP (220) multiple sensor fusion platform maybe implemented on the routers and nodes of the network for further low latency synchronized operations.

[0098] FIG. 11 illustrates a flow chart representation 1100 of an example method for Al based reconfigurable neuromorphic vision sensor fusion system, in accordance with an embodiment of the present disclosure.

[0099] At step 1102, the method 1100 may include determining, by a processor 114, one or more tasks to be performed based on real-time sensory signals environmental conditions and operational requirements.

[0100] At step 1104, the method 1100 may include extracting, by the processor 114, a plurality of significant features from the received one or more sensory signals using a data-driven modular Al in each Al processing layer 116.

[0101] At step 1106, the method 1100 may include determining, by the processor 114, a type and relevance of each of the extracted plurality’ of features with the determined one or more tasks to be performed.

[0102] At step 1108, the method 1100 may include configuring, by the processor 114, each of the plurality of sensors 104 with at least one or more Al processing functions based on the determined type and the relevance. Further, the at least one or more Al processing functions are configured to process each of the plurality of significant features.

[0103] At step 1110, the method 1100 may include determining, by the processor 114, whether the assigned at least one or more Al processing functions match with the real-time environmental conditions and the operational requirements.

[0104] At step 1 112, the method 1 100 may include dynamically updating, by the processor 114, the at least one or more Al processing functions to meet the real-time environmental conditions and the operational requirements based on the determination.

[0105] At step 1114, the method 1100 may include determining, by the processor 114, a subset of sensors 104 and a subset of Al processing layers 116 to be reconfigured based on the updated at least one or more Al processing functions.

[0106] At step 1116, the method 1100 may include autonomously reconfiguring, by the processor 114, the determined subset of sensors 104 and the subset of Al processing layers 116 in real-time with the updated at least one or more Al processing functions.

[0107] At step 1118, the method 1100 may include performing, by the processor 114, the determined one or more tasks by executing the updated at least one or more Al processing functions at corresponding sensors 104 and corresponding Al processing layers 116. Further, theupdated at least one or more Al processing functions are configured to process each of the plurality of features.

[0108] At step 1120, the method 1100 may include generating, by the processor 114, a processed output corresponding to the single or multi-modal data based on the performed one or more tasks.

[0109] In an example, the method 1 100 may include defining, by the processor 114, a modular configuration of one or more AlP-modules 120 and the AlP-interfaces 122 within one or more top Al processing layers 118, one or more middle Al processing layers 124, and one or more bottom Al processing layers 126, each AlP-module 120 comprising at least one of a machine learning (ML) mode, a neural network (NN) model, and a genetic algorithm (GA) model or the like with pre-determined parameters. Further, the method 1100 may include connecting, by the processor 114, the AlP-modules 120 using the one or more AlP-interfaces 122 configured to perform at least one of a linear approximation (LA), a weighted linear approximation (WLA), and a simulated annealing (SA), a machine learning (ML) mode, a neural network (NN) model, and a Genetic Algorithm (GA) model to enable one of a dynamic coupling and uncoupling of the one or more AlP-modules 120 in real-time. Further, the method 1100 may include adjusting, by the processor 114, the modular configuration of the AlP-modules 120 and the one or more AIP- interfaces 122 based on the type and relevance of the extracted plurality of significant features.

[0110] In an example, the method 1 100 may include classifying, by the processor 114, each of the extracted plurality of features into one or more categories using an Al model. The categories may comprise for example, but not limited to, object identification, environmental characterization, or event detection. For example, but not limited to, in a visual sensory context, categories may include object boundaries based on color and intensity, spatial attributes based on depth and shape, or surface properties based on polarization and pattern. Moreover, the Al model may include for example, but not limited to, the ML model, the NN model, the GA model or like. In an audio context, categories may comprise for example, but not limited to, sound source types based on frequency and tone, or event indicators based on amplitude and noise. Further, the method 1100 may include computing, by the processor 114, a relevance score for each classified features with the determined one or more tasks to be performed. The relevance score indicates relevance level of each of the classified features with the determined one or more tasks to be performed. The computation of the relevance score is executed within the Al processing layers 116, comprising the Top Al processing layers (TL-AI) 118, the Middle Al processing layers (ML-AI) 124, and the Bottom Al processing layers (BL-AI) 126. The one or more Al processing layers 116 are hierarchically structured and comprise one or more Al Processing Modules (AlP-modules) 120 configured with machine learning (ML), neural network (NN), or genetic algorithm (GA) models.Each AlP-module 120 is pre-configured with pre-trained parameters and processing weights, which are stored in one or more memory units 106 and dynamically accessed during operation. Further, the processor 114 evaluates each significant feature using a weighted scoring function that reflects the feature’s relative importance to the operational objectives under prevailing environmental conditions. The real-time conditions may include for example, but are not limited to, low light, rain, fog, dust, or other event or context-specific variations or noises. The scoring function may involve a combination of for example, but not limited to, statistical correlation or mutual information between the significant features of received one or sensory signal, significant feature importance measures from trained ML models and fitness evaluations in evolutionary- models for example, but not limited to, genetic algorithm (GA) or the like.

[0111] In an example, the method 1100 may include performing, by the processor 114, at least one of modification, addition, deletion, merger, and sw itching of the at least one or more Al processing functions of one or more AlP-modules 120 between the plurality of sensors 104 using one or more AIP -interfaces 122 to meet the real-time environmental conditions and the operational requirements.

[0112] In an example, the method 1100 may include assigning, by the processor 114, significant Al processing functions of significant AlP-modules 120 to sensors centrally located in a significant region of an input array. Further, the method 1100 may include assigning, by the processor 114, non-significant Al processing functions of non-significant AlP-modules 120 to sensors 104 peripherally or edge located non-significant regions of the input array based on the real-time environmental conditions and the operational requirements. The significant features are critical to the determined tasks for example, but not limited to, image capture, pattern recognition, obstacle detection, navigation, or event classification for example, but not limited to, depth, shape, or intensity-, which directly contribute to objectives for example, but not limited to, obstacle avoidance or object identification. The significant features are assigned high relevance scores and processed by significant AlP-modules 120 centrally located for optimal data capture. However, the non-significant features have lower relevance to the task, for example, but not limited to, color or luminosity- in a navigation task where spatial attributes are prioritized, or noise in a visual context where clarity- is important. The non-significant features are processed by non-significant AlP-modules 120 which are generally at peripheral or edge regions, reducing computational load while maintaining low-power, low-latency performance in edge or extreme edge environments.

[0113] In an example, the method 1 100 may include determining, by the processor 1 14, the real-time environmental conditions and the operations requirements using a first set of sensors among the plurality- of sensors 104. Further, the method 1100 may include mapping, by the processor 114, operation parameters of each of the plurality of sensors 104 with the determinedreal-time environmental conditions and the operations requirements. The operation parameters refer to the configurable settings and significant features that controls the functionality of each NFSU 108 within the input array 202, enabling the sensors 104 to capture and process received one or more sensory signals effectively. The operation parameters may include for example but are not limited to, sensitivity thresholds for detecting intensity or polarization, sampling rates for frequency of one or more sensory signal capture, feature extraction priorities which may focus on shape or depth, power consumption levels, and Al processing function executed by the Al software-defined functionality part 112. Further, the method 1100 may include determining, by the processor 114, a second subset of sensors 104 comprising the operations parameters deviating from the determined real-time environmental conditions and the operations requirements. Further, the method 1100 may include determining, by the processor 114, the at least one or more Al processing functions of one or more AlP-modules 120 and the interfaces relevant to the determined second subset of sensors 104. Further, the method 1100 may include autonomously reconfiguring, by the processor 114, the determined second subset of sensors 104 to process the plurality of significant features based on the at least one or more Al processing functions of one or more AIP modules 120 and the interfaces.

[0114] In an example, the method 1100 may include initiating, by the processor 114, at least one of a reset, a restart, and reconfiguration commands via the one or more bottom Al processing layers 1 16. Further, the method 1 100 may include propagating, by the processor 1 14, the reset, the restart, and the reconfiguration commands upward to the one or more top Al processing layer 116 with minimal memory adjustments via the one or more middle Al processing layers 124. Further, the method 1100 may include autonomously reconfiguring, by the processor 114, the determined subset of sensors 104 and the subset of Al processing layers using the one or more top Al processing layers 1 16.

[0115] In an example, the method 1100 may include selectively performing, by the processor 114, one of activation and deactivation of the determined subset of sensors 104 and the subset of Al processing layers 116 based on analysis of the real-time environmental conditions and the operational requirements.

[0116] In an example, the method 1100 may include applying, by the processor 114, the updated at least one or more Al processing functions of one or more AlP-modules 120 and the AlP-interfaces 122 onto a trained Al model for example, but not limited to, the ML model, the NN model, the GA model or the like to determine optimal operational parameters for the determined subset of sensors 104. Further, the method 1100 may include adjusting, by the processor 114, current operational parameters of the determined subset of sensors 104 with the determined optimal operational parameters based on an output of the trained Al model. Further, the method1100 may include performing, by the processor 114, the determined one or more tasks on the determined subset of sensors 104 based on the adjusted operational parameters, wherein the determined one or more tasks comprise an image capture, pattern, feature, obstacle and depth detection and traction, a location navigation, an event detection and classification, and many others without any limitations. The system 102 provides several significant advantages, particularly for autonomous applications in edge or extreme edge environments. The system’s 102 biomorphic, hierarchical Al architecture, comprising Top Al Processing Layers (TL-AI, 118), Middle Al Processing Layers (ML-AI, 124), and Bottom Al Processing Layers (BL-AI, 126), integrated with Neuromorphic Functional Sensing Units (NFSUs, 108), enables low-power, low-latency, and low- memory processing of one or more sensory signals for example, but not limited to, visual, audio, or hybrid. The system 102 is inspired by the adaptability of bee 206 vision systems and supports autonomous reconfiguration of sensors 104 and Al processing functions allowing the system 102 to dynamically adapt to real-time environmental conditions for example, but not limited to, dusty, rainy, dark, or temperature and radiation fluctuation environments and operational requirements for example, but not limited to, navigation, mapping, obstacle detection, or event classification. The uncoupled operation of NFSUs 108 enhances modularity and scalability, enabling efficient feature extraction and prioritization through relevance scoring. The use of advanced hardware within the electronics packaging 132 further optimizes energy efficiency, making the ENVSP 102 ideal for resource-constrained devices like drones 1004, robots 1006, or event driven and selective CCTV cameras 1008. Additionally, the system's 102 ability to process only significant features and assign significant Al processing functions to centrally located sensors 104 minimizes computational overhead, ensuring robust performance in noisy or dynamic environments. The aforementioned advantages collectively enable the ENVSP 102 to deliver high-efficiency, context- aware, and autonomous sensory processing, significantly advancing the capabilities of edge-based autonomous systems.

[0117] The object of the present invention disclosure is to provide an Artificial Intelligence (Al) based architecture, system, and method of an Evolutionary and reconfigurable Neuromorphic Vision Sensor fusion Platform (ENVSP) for use in low power, memory, and latency operations of autonomous drones, robots, and other mobile and sensing devices and applications without any limitations.

[0118] The low power, memory and latency operations of robots, drones, and other mobile and sensing devices and applications are especially suitable for operations at the edge or extreme edge. The Al based autonomous operations in the edge devices, working in remote or hard to reach locations, the multiple input channel data are sensed locally and the resulting output data, knowledge or resulting operations are also employed locally. The amount of the data andknowledge to be send back and forth to an Al processing cloud, during each operation cycle, is significantly reduced, thereby requiring far less bandwidth, connectivity, and resources than otherwise may be necessary. Similarly for the extreme edge computing (EEC), mobile and sensing devices and applications, all the Al based processing and operations on the multiple input channel data, input and output, and the resulting operations are done locally at the end point device itself.

[0119] The present disclosure, without any limitation, consists of multiple Al processing layers 116 in a top-to-bottom biomorphic hierarchical architecture including: no memory, storage, or input-output in the top-level Al (TL-AI) processing layer 118; minimal or small memory, with no storage or input-output in one or more than one middle-level Al (ML-AI) processing layers 124; and full memory, storage, input-output, reset, restart, reconfiguration, and leaky communication capabilities in the core bottom-level Al (BL- Al) processing layer 126.

[0120] The top-level TL-AI layer 118, without any limitation, is directly integrated with one or more than one Neuromorphic Functional Sensing Units (NFSUs) 108 including one or more than one lenses arranged in software defined autonomously reconfigurable groups and patterns on flat, dome, or faceted dome shaped arrays. In another aspect, without any limitation, this is analogous to a biomorphic integrated brain visual-sensor system, in bees and insects, where the top TL-AI layer 118 is integrated directly above with the vision sensing layer consisting of one or more than one NFSUs 108 for incoming single or multiple channel input digital or analog visual sensory signals, in parallel or in-sequence, in stationary or in moving sensing platform or devices configurations.

[0121] In one aspect, without any limitation, each NFSU 108 on a flat, dome, or faceted dome shaped array layer in the ENVSP 108 for visual sensing may be an individual input lens or a combination of input lenses capable of performing one or more basic functions such as the detection, filtering, classification, combination, or convolution of one or more basic features like color, intensity, luminosity, shape, pattern, feature, depth or polarization and the like, of an input visual sensory signal by its hardware nature and or by its Al software defined functionality of that NFSU under consideration.

[0122] In one aspect the hardware functionality of each NFSU 108, without any limitation, may mean any physical, chemical, materials, or structural composition of the upper lens part 110 of the NFSU 108 under consideration. In another aspect, the Al software defined functionality 112 of each NFSU 108, without any limitation, may mean any machine learning (ML), Neural Network (NN), or Genetic Algorithm (GA) based software functionality for the detection, filtering, classification, combination, or convolution of one or more than one input visual sensory signals received by that NFSU 108 within the input flat, dome, or faceted dome shaped layer of the present ENVSP 102.

[0123] In one aspect, the Al software defined functionality 112 of each individual NFSU 108 within the ENVSP 102, without any limitation, is controlled, configurable, or reconfigurable through the underlying TL-AI 118, ML-AI 124, and BL-AI 126 layers in real time during an operation of the ENVSP 102. In another aspect, these groups and patterns of individual and multiple NFSUs 108, without any limitations, may be autonomously and dynamically configured or reconfigured in real time to decide the local or global Al softw are defined functionality 112 to focus only on the significant, relevant, or as needed among all the available functionalities of the ENV SP 102 during an operation.

[0124] In one aspect, similarly to the complex eyes of honeybees or other insect vision, without any limitation, the present system 102 towards receiving one or more than one vision sensing signals or images through multiple (100s to 1000s) of NFSUs 108, simultaneously in parallel or in sequence and focusing only on the significant, relevant, or the needed part of all the available software defined functionalities or reconfigurations possible in the low or uncontrolled resolution data received from the multiple vision sensing signals or images within an environment during an operation. In another aspect, this allows an autonomous "Chameleon like” Al software defined autonomous reconfigure-ability of the ENVSP 102, depending upon the changing environmental conditions, without any limitation, in real time during an operation.

[0125] Embodiments of the present invention, without any limitations, thus include low power, low-memory, low input-output, and low latency operations with Al software defined autonomously reconfigurable ENVSPs 102.

[0126] In one aspect, these are possible because of the top-to-bottom biomorphic hierarchical TL-AI 118. ML-AI 124, and BL-AI 126 layered architecture for Al processing without any limitations. In another aspect, the Al processing through each of the TL-AI 118, ML- AI 124, and BL-AI 126 is also kept massively modular and reconfigurable in real time during operation. In yet another aspect, the Al processing through each of the above layers, without any limitation, is performed through one or more than one Al processing modules 120 (AlP-blocks) joined or coupled together with one or more than one Al processing interfaces 122 (AIP- interfaces), as described later, able to join or separate in real time, without any limitation, during an operation.

[0127] In one aspect, the Al softw are processing defined processing and functionality 112 of each and every AlP-block 120 in the TL-AI 118, ML-AI 124, and BL-AI 126, without any limitation, is kept uncoupled from the Al software defined processing and functionality 112 of each and every nearest neighboring AlP-block 120, w ithin the same layer, during an operation. In another aspect, without any limitation, all the input vision sensory' signals received by each NFSU 108 as well as of all their neighboring NFSUs 108 are then processed uncoupled, individually,simultaneous or in parallel, during a top to bottom cycle through all the Al layers 116 of the ENVSP 108 during an operation. In another aspect, the overall Al software defined functionality 112 of such ENVSP 102, without any limitation, becomes a collection or seamless combination of the individual, uncoupled Al software defined functionalities of the uncoupled NFSUs and could be considered as one single Al software defined functionality 112 of the ENVSP 102 during an operation.

[0128] In another aspect, the Al software defined functionality 112, and processing of each NFSU 108, and subsequently through each and every of their corresponding AlP-blocks 120 and AlP-interfaces 122 in TL-AI 118, ML-AI 124, and BL-AI 126 software layers underneath of an ENVSP 102 remain coupled to each other, without any limitation, and may not be processed uncoupled or individually. However, the coupled processing of AlP-blocks 120 and AlP-interfaces 122 within each TL-AI 118, ML-AI 124, and BL-AI 126 software layers can still be simultaneous or in parallel, during a top to bottom cycle in real time during an operation. The difference between the coupled and uncoupled cases is that the assigned functionality of an NFSU 108 is coupled or influenced by the assigned functionalities of at least first nearest-neighboring NFSUs 108 during top-to-bottom processing cycle of the ENVSP during operation. This allows significant functionality passing or influencing between at least first nearest-neighboring NFSUs during top- to-bottom processing cycle of the ENVSP 102 during an operation. The bio-morphic layered and modular Al architecture of ENVSP 102 with AlP-blocks and AlP-interfaces thus allows simultaneous and parallel processing of each NFSU during top-to-bottom processing cycle of all the NFSUs during an operation.

[0129] In one aspect, examples of uncoupled or coupled one or more than one AlP-blocks 120 with one or more than one AlP-interfaces 122 show, without any limitation, Al software defined autonomous reconfigure-ability of an ENVSP 102 in real time during an operation. In another aspect, such Al software defined recon figure-ability of an ENVSP 102 including on 100s- 1000s of input lenses or NFSUs 108, like the complex eyes of bees and insects, may allow 10,000s - 1.000,000s of configurations in real time during an operation depending upon the visual sensory signals received and the environment in the which the ENVSP 102 operates, as per the target goals or objectives of the ENVSP 102 as desired and set by the use-case scenarios.

[0130] In one aspect, an AlP-block 120 is an object oriented Al software processing block, such as machine learning (ML), neural network (NN), or genetic algorithm (GA) or the like, with predetermined, pre-computed, or pre-trained model parameters, without any limitation, to act on the input-data to produce output-data and knowledge without changing anything else within or about the AlP-block 120 itself during an operation. In another aspect, without any limitation, an AlP-interface 122 is also an object oriented Al software processing interface, such as linearapproximation (LA), weighted linear approximation (WLA). simulated annealing (SL), machine learning (ML), neural network (NN), and genetic algorithm (GA) or the like, able to connect, join, stitch, or combine any two or more than two similar or different AlP-blocks 120 with model data or parameters predetermined, pre-computed, or pre-trained, without any limitation, or could be computed, simulated, trained or optimized during a training or an operation.

[0131] In one aspect, a key difference between the AlP-blocks 120 and the AlP-interfaces 122, without any limitation, is that generally the AlP-blocks 120 are pre-determined, precomputed, and pre-trained during an operation and application, they do not change and can be used for a variety' of other and new operations and applications as well; however the AlP-interfaces may- have been pre-set, pre-determined, pre-computed, and pre-trained during an operation and application, and yet can be further optimized during an operation and application, they can be changed and re-optimized for a variety- of other operations and applications.

[0132] Embodiments of the present invention thus include creating a library of one or more than one pre-determined, pre-computed, and pre-trained AlP-blocks 120 and one or more than one AlP-interfaces 122, for joining, stitching, or combining one or more than one AIP blocks 120 in one or more than one to all possible combinations, without any limitation, for use in real time for top-to-bottom biomorphic layered Al architecture, system, and method for processing one or more visual sensor signal processing in series or in parallel during an operation.

[0133] In one aspect, the present invention including a library of AlP-blocks 120 and AlP- interfaces 122, without any limitation, allows the seeding, building-up, growth, and maturing of such a library- and its content for a bottom-up evolutionary- ENVSP 102 and accompanying evolutionary Al software architecture, method, and system described in this present disclosure.

[0134] In one aspect, without any limitation, an initial or seed stage library may contain the AlP-blocks 120 and-AIP interfaces 122 joining them, for mainly the basic significant attributes or features of one or more sensory' signal processing for a “beginner” or “child-like” intelligent AL In another aspect, as the time progress, the new, larger, and derived AlP-blocks 120 and AlP- interfaces 122 for joining them are created, added, modified and improved continually, without any limitation, to grow the seed stage Al library into a mature, adult, and optimized Al for designated and new classes of operations and applications.

[0135] In one aspect, the initial creation, seeding, growth, and use of such a library of AIP- blocks 120 and AlP-interfaces 122 for one or more than vision sensor signal processing during an operation or application, without any limitation, are extendable to other sensor signal types such as audio, textual, video, chemical or gas, thermal, and mechanical, touch, or pressure sensor signal processing, and any combination of two or more than tw o, w ithout any limitation, as well.

[0136] Finally, as mentioned above, embodiments of the present invention include a top- to-bottom biomorphic hierarchical architecture, system, and method for multiple channel vision sensor signals processing through NFSUs 108, and TL-AI 118, ML-AI 124, and BL-AI 126 software processing layers without any limitations.

[0137] In one aspect, the TL-AI software layer 118 supports one or more than one NFSUs 108 joined together with one or more than one AlP-blocks 120 and AlP-interfaces 122 between them, without any limitation, processing one or more than one input visual sensory' signals forming the maximum input data size to reduce the processed data size, increase the knowledge or inference, and pass on the reduced data size and increased knowledge to the one or more than one ML-AI software processing layers 124 stack below. In another aspect, one or more than one ML- AI software processing layers 124 stack, consisting of one or more than one AlP-blocks 120 joined together with one or more than one AlP-interfaces 122 between them, without any limitation, further process the received data from the TL-AI software layer 118 above and pass on the further reduced data and further increased knowledge to the bottom most core brain-like BL-AI 126 software processing layer below. In yet another aspect, the core bottom-most BL-AI software processing layer 126, consisting of one or more than one AlP-blocks 120 joined together by one or more than one AlP-interfaces 122 between them, without any limitation, finally process the much reduced processed data sized and knowledge, from the layers stack above, with full memory and storage requirements for the final knowledge, data, all model parameters, input-output, on-off, re-start, reset, configurations and reconfigurations, and communication capabilities without any limitation.

[0138] In one aspect, the present invention of ENVSP 102 with top-to-bottom biomorphic hierarchical layered Al architecture allowing maximum input data through multiple incoming channels with no memory, I / O or storage requirement in the TL-AI layer 1 18 at the top, and lowest sized processed data, knowledge, model parameters with full memory', storage, I / O and communication requirements in the BL-AI layer 126 with on-off, restart, reconfiguration, reset and leaky communications capabilities at the bottom. In another aspect, this architecture allows top-to-bottom processing with full initialization, re-set, and reconfiguration in the bottom-to-top capabilities without any limitations. In yet another aspect, the above described top-to-bottom, biomorphic, hierarchical architecture can also be implemented in single layered side-by-side configurations for the current ASIC. FPGA. and SOC based hardware chips and accelerators, as well as the current and future Neuromorphic analog-AI, and In-memory computing chips and devices.

[0139] The above description is illustrative only and is not intended to be limiting in any way. The details of the one or more implementations of this invention are set forth in theaccompanying drawings and the related description above. Other features, objects, and advantages of the invention will be apparent from the description and drawings, obvious to the experts in the field, and are detailed in the form of claims in this submission.

[0140] In an embodiment, one of the ordinary skills in the art will appreciate that techniques consistent with the present disclosure are applicable in other contexts as well without departing from the scope of the disclosure.

[0141] What has been described and illustrated herein are examples of the present disclosure. The terms, descriptions, and figures used herein are set forth by way of illustration only and are not meant as limitations. Many variations are possible within the spirit and scope of the subject matter, which is intended to be defined by the following claims and their equivalents in which all terms are meant in their broadest reasonable sense unless otherwise indicated.

[0142] The written description describes the subject matter herein to enable any person skilled in the art to make and use the embodiments. The scope of the subject matter embodiments is defined by the claims and may include other modifications that occur to those skilled in the art. Such other modifications are intended to be within the scope of the claims if they have similar elements that do not differ from the literal language of the claims or if they include equivalent elements with insubstantial differences from the literal language of the claims.

[0143] The embodiments herein may include hardware and software elements. The embodiments that are implemented in software include but are not limited to, firmware, resident software, microcode, and the like. The functions performed by various modules described herein may be implemented in other modules or combinations of other modules.

[0144] A description of an embodiment with several components in communication with each other does not imply that all such components are required. On the contrary, a variety of optional components are described to illustrate the wide variety of possible embodiments of the invention. When a single device or article is described herein, it will be apparent that more than one device / article (whether or not they cooperate) may be used in place of a single device / article. Similarly, where more than one device or article is described herein (whether or not they cooperate), it will be apparent that a single device / article may be used in place of the more than one device or article, or a different number of devices / articles may be used instead of the show n number of devices or programs. The functionality and / or the features of a device may be alternatively embodied by one or more other devices which are not explicitly described as having such functionality / features. Thus, other embodiments of the invention need not include the device itself.

[0145] The illustrated steps are set out to explain the exemplary7embodiments shown, and it should be anticipated that ongoing technological development will change the manner in w hichparticular functions are performed. These examples are presented herein for purposes of illustration, and not limitation. Further, the boundaries of the functional building blocks have been arbitrarily defined herein for the convenience of the description. Alternative boundaries may be defined so long as the specified functions and relationships thereof are appropriately performed. Alternatives (including equivalents, extensions, variations, deviations, and the like., of those described herein) will be apparent to persons skilled in the relevant art(s) based on the teachings contained herein. Such alternatives fall within the scope and spirit of the disclosed embodiments. Also, the words "comprising," "having," "containing," and "including," and other similar forms are intended to be equivalent in meaning and be open-ended in that an item or items following anyone of these words is not meant to be an exhaustive listing of such item or items or meant to be limited to only the listed item or items. It must also be noted that as used herein and in the appended claims, the singular forms “a”, “an”, and “the” include plural references unless the context clearly dictates otherwise.

[0146] Finally, the language used in the specification has been principally selected for readability and instructional purposes, and it may not have been selected to delineate or circumscribe the inventive subject matter. It is therefore intended that the scope of the invention be limited not by this detailed description, but rather by any claims that issue on an application based here on. Accordingly, the embodiments of the present invention are intended to be illustrative, but not limited, of the scope of the invention, which is outlined in the following claims.

Claims

CLAIMSWe claim:

1. A system for processing sensory' signals using a layered and modular Artificial Intelligence (Al) architecture, comprising: a plurality' of sensors configured to receive one or more sensory signals corresponding to a single or multi-modal data from one or more data sources; and a processing system communicatively coupled to the plurality' of sensors, wherein the processing system comprises one or more Al processing layers arranged in a biomorphic hierarchical architecture, wherein the one or more Al processing layers are configured to: determine one or more tasks to be performed based on real-time sensory' signals, environmental conditions, and operational requirements; extract a plurality of significant features from the received one or more sensory signals using a data-driven modular Al in each Al processing layer; determine a type and relevance of each of the extracted plurality' of significant features with the determined one or more tasks to be performed; configure each of the plurality of sensors with at least one or more Al processing functions based on the determined type and the relevance, wherein the at least one or more Al processing functions are configured to process each of the plurality of significant features; determine whether the assigned at least one or more Al processing functions match with the real-time environmental conditions and the operational requirements; dynamically update the at least one or more Al processing functions to meet the realtime environmental conditions and the operational requirements based on the determination; determine a subset of sensors and a subset of Al processing layers to be reconfigured based on the updated at least one or more Al processing functions; autonomously reconfigure the determined subset of sensors and the subset of Al processing layers in real-time with the updated at least one or more Al processing functions;perform the determined one or more tasks by executing the updated at least one or more Al processing functions at corresponding sensors and corresponding Al processing layers, wherein the updated at least one or more Al processing functions are configured to process each of the plurality' of features; and generate a processed output corresponding to the single or multi-modal sensors data based on the performed one or more tasks.

2. The system of claim 1, wherein the plurality of significant features comprises at least one of a color, an intensity, a luminosity, a shape, a pattern, a feature, a depth, a polarization, a frequency, an amplitude, a tone, a pitch, and a noise.

3. The system of claim 1, wherein the plurality of sensors comprise a plurality of digital or neuromorphic functional sensing units (NFSUs) arranged in at least one of a flat, dome, and faceted dome shaped array, wherein each NFSU comprises an upper hardware receptor part configured to receive the one or more sensory signals and a lower Al software-defined functionality part configured to process the received sensory' signals.

4. The system of claim 3, wherein the faceted dome shaped array is formed by an outermost dome-shaped array layer comprising one or more discontinuous faceted layers, wherein each discontinuous faceted layer comprise a locally flat array connected to adjacent locally flat faceted arrays at specific angles to form a continuous curvature for the faceted dome shaped array.

5. The system of claim 1, wherein the one or more Al processing layers comprise: one or more top Al processing layers integrated with a lower Al software-defined functionality part of the plurality of sensors, wherein the one or more top Al processing layers are configured to perform a first level of processing of the received one or more sensory signals to generate a first level processed data; one or more middle Al processing layers communicatively coupled in between the one or more top Al processing layers and one or more bottom Al processing layers, wherein the one or more middle Al processing layers are configured to perform a second level of processing on the first level processed data received from the one or more top Al processing layers and generate a second level processed data; and the one or more bottom Al processing layers communicatively coupled to the one or more middle Al processing layers, wherein the one or more bottom Al processing layers are configured to perform a third level of processing on the second level processed data received from the one or more middle Al processing layers and generate an at least one of the third levelor final processed data corresponding to the received one or more sensory signals, wherein the at least one of the third level or final processed data corresponds to the processed output.

6. The system of claim 1, wherein the one or more Al processing layers comprise: one or more Al processing modules (AlP-modules) coupled to a plurality of one or more neighboring AlP-modules within the same Al processing layer, wherein the one or more AIP- modules, with at least one of the weights or parameters pre-determined and pre-trained for at least one or more Al processing functionalities of the said AlP-modules; and one or more Al processing interfaces (AlP-interfaces), configured to at least one of the couple or connect at least two or more AlP-modules with at least one of the weights or parameters predetermined, pre-trained, and further optimized during an operation.

7. The system of claim 6, wherein the one or more Al processing layers are configured to: process the one or more sensory signals in an uncoupled mode, wherein one or more sensory signals and the corresponding AlP-module operate uncoupled and independently through the one or more top Al processing layers, the one or more middle Al processing layers and through the one or more bottom Al processing layers; and process the one or more sensory’ signals in a coupled mode, wherein the plurality of sensory signals and the corresponding one or more AlP-modules are at least one of the coupled or connected through the one or more AlP-interfaces for at least one of the coupled or combined sensory' signals processing through the one or more top Al processing layers, the one or more middle Al processing layers and through the one or more bottom Al processing layers.

8. The system of claim 1 , wherein the one or more Al processing layers are configured to comprise: one or more configurations of Al processing modules (AlP-modules) within the one or more top Al processing layers, the one or more middle Al processing layers, and the one or more bottom Al processing layers, wherein each AlP-module configuration comprises of at least one of a machine learning (ML) mode, a neural network (NN) model, and a genetic algorithm (GA) model with pre-determined and pre-trained parameters or weights; one or more configurations of Al processing interfaces (AlP-interfaces) to at least one of the couple or connect two or more AlP-modules within the one or more top Al processing layers, the one or more middle Al processing layers, and the one or more bottom Al processing layers, wherein each AlP-interface configuration comprises of at least one of a linear approximation (LA), a weighted linear approximation (WLA), a simulated annealing (SA), amachine learning (ML) mode, a neural network (NN) model, and a genetic algorithm (GA) model with pre-determined and pre-trained parameters, which can be further optimized to enable one of a dynamic coupling and uncoupling of the one or more AlP-modules in realtime; and adjust the modular configuration of the AlP-modules and the AIP -interfaces based on the type and relevance of the extracted plurality of significant features.

9. The system of claim 1, wherein to determine the type and relevance of each of the extracted plurality of significant features, the one or more Al processing layers are configured to: classify each of the extracted plurality of significant features into one or more categories using one or more AlP-modules and the one or more AlP-interfaces; and compute a relevance score for each classified features with the determined one or more tasks to be performed, w herein the relevance score indicates relevance level of each of the classified features with the determined one or more tasks to be performed.

10. The system of claim 1, wherein to dynamically update the at least one or more Al processing functions of one or more AlP-modules to meet the real-time environmental conditions and the operational requirements based on the determination, the one or more Al processing layers are configured to: perform at least one or more of modification, addition, deletion, merger, and switching of the at least one or more Al processing functions betw een the plurality of one or more sensors using one or more AlP-modules and the one or more AlP-interfaces to meet the real-time environmental conditions and the operational requirements.

11. The system of claim 1, wherein to dynamically update the at least one or more Al processing functions of one or more AlP-modules to meet the real-time environmental conditions and the operational requirements based on the determination, the one or more Al processing layers are configured to: assign significant Al processing functions of significant AlP-modules to sensors centrally located significant region of an input array; and assign non-significant Al processing functions of non-significant AlP-modules to peripherally or edge located non-significant regions of the input array based on the real-time environmental conditions and the operational requirements.

12. The system of claim 1, wherein to autonomously reconfigure the determined subset of sensors, the subset of Al processing layers and the subset of AlP-modules the one or more Al processing layers are configured to: determine the real-time environmental conditions and the operations requirements using a first set of sensors among the plurality of sensors; map operation parameters of each of the plurality of sensors with the determined real-time environmental conditions and the operations requirements; determine a second subset of sensors comprising the operations parameters deviating from the determined real-time environmental conditions and the operations requirements; determine the at least one or more Al processing functions of one or more AlP-modules relevant to the determined second subset of sensors; and autonomously reconfigure the determined second subset of sensors to process the plurality7of significant features based on the at least one or more Al processing functions of one or more AlP-modules.

13. The system of claim 1, wherein to autonomously reconfigure the determined subset of sensors and the subset of Al processing layers, the one or more Al processing layers are configured to: initiate at least one of a reset, a restart, and reconfiguration commands via the one or more bottom Al processing layers; propagate the reset, the restart, and the reconfiguration commands upward to the one or more top Al processing layer with minimal memory adjustments via the one or more middle Al processing layers; and autonomously reconfigure the determined subset of sensors and the subset of Al processing layers using the one or more top Al processing layers.

14. The system of claim 1, wherein to autonomously reconfigure the determined subset of sensors and the subset of Al processing layers, the one or more Al processing layers are configured to: selectively perform one of activation and deactivation of the determined subset of sensors and the subset of Al processing layers based on the analysis of the real-time environmental conditions and the operational requirements.

15. The system of claim 1, wherein to perform the determined one or more tasks by executing the updated at least one or more Al processing functions of one or more AlP-modules at corresponding sensors and corresponding Al processing layers, the one or more Al processing layers are configured to:apply the updated at least one or more Al processing functions of one or more AlP-modules onto a trained Al model to determine optimal operational parameters for the determined subset of sensors; adjust current operational parameters of the determined subset of sensors with the determined optimal operational parameters based on an output of the trained Al model; and perform the determined one or more tasks on the determined subset of sensors based on the adjusted operational parameters, wherein the determined one or more tasks comprise an image capture, pattern, feature, obstacle and depth detection and traction, a location navigation, an event detection, and classification.

16. A method for processing sensory signals using a layered and modular Artificial Intelligence (Al) architecture comprising: determining, by a processor, one or more tasks to be performed based on real-time sensory signals, environmental conditions, and operational requirements; extracting, by the processor, a plurality of significant features from the received one or more sensory7signals using a data-driven modular Al in each Al processing layer; determining, by the processor, a ty pe and relevance of each of the extracted plurality of significant features with the determined one or more tasks to be performed; configuring, by the processor, each of the plurality of sensors with at least one or more Al processing functions based on the determined type and the relevance, wherein the at least one or more Al processing functions are configured to process each of the plurality of significant features; determining, by the processor, whether the assigned at least one or more Al processing functions match with the real-time environmental conditions and the operational requirements; dynamically updating, by the processor, the at least one or more Al processing functions to meet the real-time environmental conditions and the operational requirements based on the determination; determining, by the processor, a subset of sensors and a subset of Al processing layers to be reconfigured based on the updated at least one or more Al processing functions; autonomously reconfiguring, by the processor, the determined subset of sensors and the subset of Al processing layers in real-time with the updated at least one or more Al processing functions; performing, by the processor, the determined one or more tasks by executing the updated at least one or more Al processing functions at corresponding sensors and corresponding Alprocessing layers, wherein the updated at least one or more Al processing functions are configured to process each of the plurality of features; and generating, by the processor, a processed output corresponding to the single or multi-modal data based on the performed one or more tasks.

17. The method of claim 16, wherein autonomously reconfiguring the determined subset of sensors and the subset of Al processing layers in real-time with the updated at least one or more Al processing functions comprises: defining, by the processor, a modular configuration of one or more Al processing modules (AlP-modules) and the Al processing interfaces (AlP-interfaces) within one or more top Al processing layers, one or more middle Al processing layers, and one or more bottom Al processing layers, each AlP-module comprising at least one of a machine learning (ML) mode, a neural network (NN) model, and a genetic algorithm (GA) model with pre-determined parameters; connecting, by the processor, the AlP-modules using the one or more Al processing interfaces (AlP-interfaces) configured to perform at least one of a linear approximation (LA), a weighted linear approximation (WLA), a simulated annealing (SA), a machine learning (ML) mode, a neural network (NN) model, and a Genetic Algorithm (GA) model to enable one of a dynamic coupling and uncoupling of the one or more AlP-modules in real-time; and adjusting, by the processor, the modular configuration of the AlP-modules and the one or more AlP-interfaces based on the ty pe and relevance of the extracted plurality7of significant features.

18. The method of claim 16, wherein determining the ty pe and relevance of each of the extracted plurality7of significant features comprises: classifying, by the processor, each of the extracted plurality7of significant features into one or more categories using an Al model; and computing, by the processor, a relevance score for each classified features with the determined one or more tasks to be performed, wherein the relevance score indicates relevance level of each of the classified features with the determined one or more tasks to be performed.

19. The method of claim 16, wherein dynamically updating the at least one or more Al processing functions of one or more AlP-modules to meet the real-time environmental conditions and the operational requirements based on the determination comprises: performing, by the processor, at least one of modification, addition, deletion, merger, and switching of the at least one or more Al processing functions of one or more AlP-modulesbetween the plurality of sensors using one or more AlP-interfaces to meet the real-time environmental conditions and the operational requirements.

20. The method of claim 16, wherein dynamically updating the at least one or more Al processing functions of AlP-modules to meet the real-time environmental conditions and the operational requirements based on the determination comprises: assigning, by the processor, significant Al processing functions of significant AlP-modules to sensors centrally located in a significant region of an input array; and assigning, by the processor, non-significant Al processing functions of non-significant AlP-modules to sensors peripherally or edge located non-significant regions of the input array based on the real-time environmental conditions and the operational requirements.

21. The method of claim 16, wherein autonomously reconfiguring the determined subset of sensors and the subset of Al processing layers comprises: determining, by the processor, the real-time environmental conditions and the operations requirements using a first set of sensors among the plurality of sensors; mapping, by the processor, operation parameters of each of the plurality of sensors with the determined real-time environmental conditions and the operations requirements; determining, by the processor, a second subset of sensors comprising the operations parameters deviating from the determined real-time environmental conditions and the operations requirements; determining, by the processor, the at least one or more Al processing functions of one or more AlP-modules and AlP-interfaces relevant to the determined second subset of sensors; and autonomously reconfiguring, by the processor, the determined second subset of sensors to process the plurality of significant features based on the at least one or more Al processing functions of one or more AlP-modules and AlP-interfaces.

22. The method of claim 16, wherein autonomously reconfiguring the determined subset of sensors and the subset of Al processing layers comprises: initiating, by the processor, at least one of a reset, a restart, and reconfiguration commands via the one or more bottom Al processing layers; propagating, by the processor, the reset, the restart, and the reconfiguration commands upward to the one or more top Al processing layer with minimal memory' adjustments via the one or more middle Al processing layers; andautonomously reconfiguring, by the processor, the determined subset of sensors and the subset of Al processing layers using the one or more top Al processing layers.

23. The method of claim 16, wherein autonomously reconfiguring the determined subset of sensors and the subset of Al processing layers comprises: selectively performing, by the processor, one of activation and deactivation of the determined subset of sensors and the subset of Al processing layers based on analysis of the real-time environmental conditions and the operational requirements.

24. The method of claim 1 , wherein performing the determined one or more tasks by executing the updated at least one or more Al processing functions of one or more AlP-modules and the AIP -interfaces at corresponding sensors and corresponding Al processing layers comprises: applying, by the processor, the updated at least one or more Al processing functions of one or more AlP-modules and the AlP-interfaces onto a trained Al model to determine optimal operational parameters for the determined subset of sensors; adjusting, by the processor, current operational parameters of the determined subset of sensors with the determined optimal operational parameters based on an output of the trained Al model; and performing, by the processor, the determined one or more tasks on the determined subset of sensors based on the adjusted operational parameters, wherein the determined one or more tasks comprise an image capture, pattern, feature, obstacle and depth detection and traction, a location navigation, an event detection, and classification.

Images