Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
DETAILED ACTION
This Office Action is in response to the application filed on 08/05/2024. Claims 1 and 11 are independent claims. Claims 1-20 have been examined and are pending. This Action is made non-FINAL.
Drawings
The drawings were received on 08/05/2024. These drawings are reviewed and accepted by the Examiner.
Claim Objections
Claims 1-2, 11, and 15 are objected to because of the following informalities:
Regarding claim 1; the acronym ARF should be spelled out in full as its first occurrence. Appropriate correction is required.
Regarding claim 2; the acronym CNN should be spelled out in full as its first occurrence. Appropriate correction is required.
Regarding claim 11; the acronym AI should be spelled out in full as its first occurrence. Appropriate correction is required.
Regarding claim 14; Claim 14 recites the limitation “LoRa (Long Range) wireless communication protocol.” This format presents the abbreviation before the full term, which is contrary to standard patent drafting practice. The conventional format is to present the full term first, followed by the abbreviation in parentheses (e.g.: "Long Range (LoRa)) wireless communication protocol". Appropriate correction is required.
Regarding claim 14; Claim 14 recites the limitation “the US 902-923 MHZ frequency band”. It is suggest for the clarity that the limitation be further amended as “the U.S. 902-923 MHZ frequency band” or “the 902-923 MHZ frequency band allocated in the United States” Appropriate correction is required.
Regarding claim 15; the acronym MQTT should be spelled out in full as its first occurrence. Appropriate correction is required.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1 and 5 are rejected under 35 U.S.C. 103 as being unpatentable over Mahishi et al. (“Mahishi,” US 2025/0307694) in view of Liu et al. (“Liu,” US 2024/0378263), and further in view of (Wang et al. (“Wang,” US 2021/0279618)
Regarding claim 1, Mahishi teaches a system for enhancing reliability and trustworthiness in cyber-physical systems, the system comprising:
a plurality of sensors configured to collect real-time data from an environment or system (Mahishi: par. 0106, An IoT platform may receive data from multiple sensors that report a variety of metrics (e.g., temperature, speed, position, pressure, noise, image data, vibration, motion, personnel detection, mode of operation, task being performed, etc.”) , “The registered sensors may report to the platform periodically, sporadically, or otherwise be pushed into the platform or pulled in by the platform”; par. 00108, IoT sensors embedded in machinery and equipment can continuously collect data on various parameters like temperature, vibration, and noise levels.), wherein the plurality of sensors includes at least two sensor types selected from the group consisting of: vibration sensors, thermal sensors, acoustic sensors, and environmental sensors (Mahishi: par. 0106, An IoT platform may receive data from multiple sensors that report a variety of metrics (e.g., temperature, speed, position, pressure, noise, image data, vibration, motion, personnel detection, mode of operation, task being performed, etc.” ); par. 00108, IoT sensors embedded in machinery and equipment can continuously collect data on various parameters like temperature, vibration, and noise levels);
a data processing unit comprising
one or more processors and a memory, the memory storing instructions that (Mahishi: par. 0106, fig. 5, pars. 0113, 0123, 0125, 0136), when executed by the one or more processors, cause the data processing unit to: receive, from the plurality of sensors, collected real-time data (Mahishi: par. 0106, An IoT platform may receive data from multiple sensors that report a variety of metrics; par. 0108);
Mahishi discloses using "a machine learning model" but does not explicitly disclose "plurality of machine learning models." Additionally, while Mahishi teaches detecting anomalies, it does not explicitly recite "identify patterns" as a separate function alongside anomaly detection. Mahishi does not explicitly disclose using a "plurality of machine learning models", "identify patterns and detect anomalies" , wherein the plurality of machine learning models comprises an ARF model and a probabilistic circuit model; cross-validate the analysis results across the collected real-time data from the plurality of sensors to verify the accuracy and reliability of the analysis; generate, based on the cross-validated analysis, one or more recommendations for improving the operation of the cyber-physical system; a communication interface configured to transmit the one or more recommendations to a user interface for presentation to an end-user.
However, in an analogous art, Liu discloses
analyze the collected real-time data using a plurality of machine learning models to identify patterns and detect anomalies, (Liu: par. 0042: "advanced detection algorithms can be applied to the selected features from block 210 to identify instances of non-synchronization. This block can utilize a combination of machine learning models, such as neural networks, decision trees, and ensemble methods like random forests or boosted trees, to analyze patterns and anomalies that suggest synchronization issues.") wherein the plurality of machine learning models comprises an ARF model (Liu: par. 0042, teaches using "ensemble methods likes" as part of the combination of machine learning models. Applicant's specification describes the "ARF model" as using random forest techniques, specifically training "a plurality of decision trees using different subsets of the collected real-time data" and computing "anomaly scores" based on the "proportion of decision trees that classify the data point as an anomaly"; par. 0042, Liu's teaching of using "random forests" and "decision trees" for detecting anomalies);
cross-validate the analysis results across the collected real-time data from the plurality of sensors to verify the accuracy and reliability of the analysis (Liu: par. 0076, "incorporate fail-safes and redundancy checks to ensure that the corrective measures it enacts are both appropriate for the detected events."); and
generate, based on the cross-validated analysis, one or more recommendations for improving the operation of the cyber-physical system (Liu: par. 0076, "alert generation and response activation system...Upon detection, this system can autonomously initiate a series of predefined corrective actions, ranging from system parameter adjustments to more complex remedial protocols"; "corrective measures... do not inadvertently introduce further system instabilities"; "reducing system downtime."); and
a communication interface configured to transmit the one or more recommendations to a user interface for presentation to an end-user (Liu: par. [0080], teaches a system with a "user interface" that facilitates "proactive monitoring and automated rectification of non-synchronization events in real-time"; par. 0081, teaches a "user interaction and control device" that "serve[s] as a primary interface for user interaction with the overall system...can be equipped with specialized software that allows for the real-time visualization of system status, alert notifications, and detailed reports on synchronization metrics.").
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Liu with the method and system of Mahishi to include analyze the collected real-time data using a plurality of machine learning models to identify patterns and detect anomalies, wherein the plurality of machine learning models comprises an ARF model; cross-validate the analysis results across the collected real-time data from the plurality of sensors to verify the accuracy and reliability of the analysis; and generate, based on the cross-validated analysis, one or more recommendations for improving the operation of the cyber-physical system; and a communication interface configured to transmit the one or more recommendations to a user interface for presentation to an end-user. One would have been motivated to provide to implement Liu's teaching of using multiple machine learning models (combination approach) in Mahishi's IoT anomaly detection system to improve detection accuracy and robustness. The motivation to combine is well-established in the art: using multiple diverse machine learning models in an ensemble approach provides better predictive performance than a single model by leveraging the complementary strengths of different modeling techniques. Liu explicitly teaches this principle by describing the use of a "combination of machine learning models" rather than a single model. Applying Liu's multiple-model approach to Mahishi's IoT anomaly detection system would have been an obvious improvement to increase the system's ability to accurately detect anomalies while reducing false positive. Furthermore, both references operate in similar technical contexts (industrial monitoring and anomalydetection), both use machine learning for automated detection, and both aim to improve system reliability. One of ordinary skill would recognize that Liu's combination approach could enhance Mahishi's IoT platform's anomaly detection capabilities. The combination merely applies a known technique (ensemble machine learning) to a known system (IoT anomaly detection) to achieve a predictable result (improved detection accuracy).
Liu does not explicitly teach a probabilistic circuit model as one of the plurality of machine learning models.
However, in an analogous art, Wang discloses wherein the plurality of machine learning models comprises a probabilistic circuit model (Wang: par. 0003, "Examples of learning machines include, but are not limited to, kernel machines, decision trees, decision forests, random forests, Sum-product networks, Bayesian networks, Boltzmann machines, and neural networks"; par. 0007, "The reference learning machines in the system may include but are not limited to Sum-product networks, Bayesian networks, Boltzmann machines, and neural networks"; par. 0055, "The reference learning machine 101 may include but are not limited to Sum-product networks, Bayesian networks, Boltzmann machines, and neural networks." Sum-product networks are a well-known type of probabilistic circuit in the machine learning art. Sum-product networks consist of a graph structure with sum nodes that compute weighted sums and product nodes that compute products of their inputs, used for computing probability distributions and generative modeling.).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Wang with the method and system of Manishi and Liu to include wherein the plurality of machine learning models comprises a probabilistic circuit model. One would have been motivated to use diverse model types (random forests + probabilistic circuits + neural networks, etc.) provides complementary approaches that improve overall detection capability. This is a well-known principle in ensemble learning and multi-model machine learning systems: different model architectures capture different patterns and relationships in data, and combining their predictions yields more robust and accurate results than any single model. All three references operate in the machine learning field, all address using learning machines to analyze data and make predictions, and one of ordinary skill would recognize that including probabilistic circuits alongside random forests and other models in an ensemble would provide the benefits of diverse model combination. The result improved anomaly detection through multi-model analysis is predictable based on known principles of ensemble machine learning.
Regarding claim 5, the combination of Mahishi, Liu, and Wang teaches the system of claim 1. The combination of Mahishi, Liu, and Wang further teaches, wherein the environmental sensors include at least one of: a temperature sensor; a humidity sensor; and a water leak sensor (Mahishi: par. 0106, An IoT platform may receive data from multiple sensors that report a variety of metrics (e.g., temperature, speed, position, pressure, noise, image data, vibration, motion, personnel detection, mode of operation, task being performed, etc.”) , “The registered sensors may report to the platform periodically, sporadically, or otherwise be pushed into the platform or pulled in by the platform”; par. 00108, IoT sensors embedded in machinery and equipment can continuously collect data on various parameters like temperature, vibration, and noise levels.).
Claim 2 is rejected under 35 U.S.C. 103 as being unpatentable over Mahishi et al. (“Mahishi,” US 2025/0307694) in view of Liu et al. (“Liu,” US 2024/0378263), and Wang et al. (“Wang,” US 2021/0279618), and further in view of Delmonico et al. (“Delmonico,” US 2025/0342386)
Regarding claim 2, the combination of Mahishi, Liu, and Wang teaches the system of claim 1. The combination of Mahishi, Liu, and Wang teaches, wherein the plurality of machine learning models but does not explicitly disclose further includes a CNN model trained to identify patterns in data collected by the sensors.
However, in an analogous art, Delmonico discloses CNN model trained to identify patterns in data collected by the sensors (Delmonico: par. 0062, the edge device 108A can use one or more AI and/or machine learning models that are trained to identify a type of emergency or other type of event from different types of sensor signals, correlated sensor signals, changes/deviations in the sensor signals, etc. The model(s) can receive the generated sensor signals as model inputs, process the model inputs, and generate output such as event/emergency classification information. The edge device 108A can categorize the type of emergency based on patterns of the generated sensor signals across different edge devices 108A-N and/or sensor devices 112. The models can be trained using deep learning (DL) neural networks, convolutional neural networks (CNNs), and/or one or more other types of AI, machine learning techniques, methods, and/or algorithms).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Delmonico with the method and system of Mahishi, Liu, and Wang to include CNN model trained to identify patterns in data collected by the sensors. One would have been motivated to use the AI and machine learning techniques can be deployed on the edge to provide for real-time or near real-time monitoring of conditions and emergency response, thereby decreasing time needed to detect and respond to an emergency while ensuring the safety and wellbeing of relevant users (Delmonico: par. 0005).
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Mahishi et al. (“Mahishi,” US 2025/0307694) in view of Liu et al. (“Liu,” US 2024/0378263), and Wang et al. (“Wang,” US 2021/0279618), and further in view of Formicola et al. (“Formicola,” US 2025/0047702)
Regarding claim 3, the combination of Mahishi, Liu, and Wang teaches the system of claim 1. The combination of Mahishi, Liu, and Wang teaches do not explicitly disclose, wherein the instructions further cause the data processing unit to: determine a trust score indicating a level of reliability and trustworthiness of the cyber-physical system based on the cross-validated analysis; and transmit the trust score to the user interface for presentation to the end-user.
However, in an analogous art, Formicola discloses
determine a trust score indicating a level of reliability and trustworthiness of the cyber-physical system based on the cross-validated analysis (Formicola: abstract, an aggregation of multiple risk scores produces a trust score for the overall system; par. 0007, The computer processor may aggregate a plurality of risk scores to generate a trust score for the system. Using the trust score the computer processor may reallocate an updated portion of the plurality of security measures based on the generated trust score. Periodically, the computer processor may recalculate risk scores for a plurality of security threats to compute an updated trust score and update the selected security measures based on the updated trust score.); and
transmit the trust score to the user interface for presentation to the end-user (As previously applied, Liu: pars. [0080]-[0081] teaches transmitting information to user interface for display to users. Liu teaches a system with a "user interface" that facilitates operations [0080] and a "user interaction and control device" that "serve[s] as a primary interface for user interaction with the overall system...can be equipped with specialized software that allows for the real-time visualization of system status, alert notifications, and detailed reports" [0081]. It would be obvious to transmit Formicola's trust score to users via user interface as taught by Liu, so that system operators can monitor overall system trustworthiness. Displaying system status metrics to users is fundamental in monitoring systems).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Formicola with the method and system of Mahishi, Liu, and Wang to include determine a trust score indicating a level of reliability and trustworthiness of the cyber-physical system based on the cross-validated analysis; and transmit the trust score to the user interface for presentation to the end-user. One would have been motivated to One would have been motivated to provide such a combination because both Mahishi and Formicola address monitoring and analyzing systems to assess reliability. Mahishi teaches using machine learning to analyze IoT sensor data for anomaly detection, while Formicola teaches generating a "trust score" by aggregating multiple analysis results to represent overall system trustworthiness. It would have been obvious to apply Formicola's technique of generating a trust score to Mahishi's IoT anomaly detection system - i.e., aggregate the cross-validated analysis results from multiple sensors to generate a single trust score representing overall cyber-physical system reliability. This provides users with a simple, intuitive metric for assessing system health rather than requiring them to interpret multiple individual sensor analysis results. Generating aggregate metrics from multiple data sources to simplify user understanding is a known technique in monitoring systems. The combination merely applies a known technique (aggregating analysis results into a single metric) to a known system (multi-sensor IoT anomaly detection) to achieve a predictable result (simplified system health assessment).
Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over Mahishi et al. (“Mahishi,” US 2025/0307694) in view of Liu et al. (“Liu,” US 2024/0378263), and Wang et al. (“Wang,” US 2021/0279618), and Formicola et al. (“Formicola,” US 2025/0047702), and further in view of Menon et al. (“Menon,” US 2022/0270183)
Regarding claim 4, the combination of Mahishi, Liu, Wang, and Formicola teaches the system of claim 3. The combination of Mahishi, Liu, Wang, and Formicola further disclose, wherein the instructions further cause the data processing unit to:
track the trust score over time (Formicola: par. 0007, The computer processor may aggregate a plurality of risk scores to generate a trust score for the system. Using the trust score the computer processor may reallocate an updated portion of the plurality of security measures based on the generated trust score. Periodically, the computer processor may recalculate risk scores for a plurality of security threats to compute an updated trust score and update the selected security measures based on the updated trust score).
Mahishi, Liu, Wang, and Formicola do not explicitly teach generating a visualization of the trust score over a user-selected time period.
However, in an analogous art, Menon teaches generating visualizations of data over user-selected time periods (Menon: abstract, A graph representatively illustrates the data retrieved by the database query and displays business revenue-related information on computer interfaces; relevant to the selected business area; par. 0193, Referring to FIG. 18, historical dashboards and corresponding data for a context or insight may be displayed by clicking on the history icon (revolving arrow). A popup 1800 is shown in FIG. 18 which asks the user to select an execution time and then either to apply or to cancel. For example, the user may select an execution that was performed a week ago to see the corresponding data/graph for the context or insight from a week ago. Referring to FIG. 19, the user may also select the calendar icon for a context or insight and then select a date and time on a calendar interface 1900, and then select “set” to see results based on the selected date/time, or “cancel” to exit the calendar).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Menon with the method and system of Mahishi, Liu, Wang, and Formicola to include generate a visualization of the trust score over a selected time for display on the user interface. It would have been obvious to one of ordinary skill in the art to apply Menon's visualization technique to Formicola's trust score data. Formicola teaches generating trust scores over time through periodic recalculation creating temporal trust score data. Menon teaches visualizing temporal data with user-selectable time ranges. One would have been motivated to combine these teachings because both address displaying system monitoring data to users. When you have temporal data (Formicola's trust scores calculated over time), it is obvious to visualize that data using time-based graphing techniques with user-selectable time ranges (as taught by Menon) to enable users to analyze trends and patterns. Providing graphical visualization of metrics over selectable time periods is a fundamental feature in monitoring systems, allowing users to quickly understand temporal trends rather than examining raw numerical data. The combination applies a known visualization technique (time-selectable graphs from Menon) to known monitoring data (trust scores from Formicola) to achieve a predictable result (trend visualization for system health monitoring).
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Mahishi et al. (“Mahishi,” US 2025/0307694) in view of Liu et al. (“Liu,” US 2024/0378263), and Wang et al. (“Wang,” US 20220221/0279618), and further in view of Gorman et al. (“Gorman,” US 2025/0333931)
Regarding claim 6, the combination of Mahishi, Liu, and Wang teaches the system of claim 1. Mahishi, Liu, and Wang do not explicitly disclose, wherein the instructions further cause the data processing unit to: generate a sensor fusion model that combines the collected real-time data from the plurality of sensors; and use the sensor fusion model in the analysis to improve the accuracy and reliability of anomaly detection.
However, in an analogous art, Gorman discloses
generate a sensor fusion model that combines the collected real-time data from the plurality of sensors (Gorman: par. 0017, The integration of diverse sensor technologies into a cohesive system, analyzed through advanced algorithms, provides a comprehensive loading technique that surpasses the capabilities of any single sensor type. Sensor fusion further leverages the strengths and mitigates the weaknesses of individual sensors, offering several advantages: 1) combining visual, thermal, and distance data ensures a well-rounded perception of the operational environment, enabling accurate decision-making under various conditions; 2) the redundancy offered by multiple sensor types enhances system reliability, allowing for cross-verification of data and reducing the likelihood of false positives or missed detections; 3) the ability to adapt to different environmental conditions and operational requirements significantly improves, ensuring consistent performance regardless of visibility, weather, or terrain challenges; and 4) machine learning algorithms can integrate and analyze data from all sensors, offering predictive insights that improve safety and operational efficiency, such as anticipating maintenance needs or identifying potential hazards before they become critical.);
and use the sensor fusion model in the analysis to improve the accuracy and reliability of anomaly detection (Gorman: par. 0017, The integration of diverse sensor technologies into a cohesive system, analyzed through advanced algorithms, provides a comprehensive loading technique that surpasses the capabilities of any single sensor type. Sensor fusion further leverages the strengths and mitigates the weaknesses of individual sensors, offering several advantages: 1) combining visual, thermal, and distance data ensures a well-rounded perception of the operational environment, enabling accurate decision-making under various conditions; 2) the redundancy offered by multiple sensor types enhances system reliability, allowing for cross-verification of data and reducing the likelihood of false positives or missed detections; 3) the ability to adapt to different environmental conditions and operational requirements significantly improves, ensuring consistent performance regardless of visibility, weather, or terrain challenges; and 4) machine learning algorithms can integrate and analyze data from all sensors, offering predictive insights that improve safety and operational efficiency, such as anticipating maintenance needs or identifying potential hazards before they become critical.).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Gorman with the method and system of Mahishi, Liu, and Wang to include generate a sensor fusion model that combines the collected real-time data from the plurality of sensors; and use the sensor fusion model in the analysis to improve the accuracy and reliability of anomaly detection. One would have been motivated to such a combination because both Mahishi and Gorman address using multiple sensors and machine learning for analyzing systems. Mahishi teaches using multiple sensor types (vibration, temperature, noise) with machine learning for IoT anomaly detection, while Gorman teaches using sensor fusion to combine data from multiple sensor types (visual, thermal, distance) with machine learning algorithms to improve accuracy and reliability. It would have been obvious to apply Gorman's sensor fusion technique to Mahishi's multi-sensor IoT anomaly detection system to achieve the same benefits of improved accuracy and reliability. Sensor fusion is a well-known technique in the art for improving detection performance when multiple sensors are available. Gorman explicitly teaches that sensor fusion "leverages the strengths and mitigates the weaknesses of individual sensors", which is a general principle applicable to any multi-sensor system. One of ordinary skill would recognize that applying sensor fusion to Mahishi's system would predictably improve anomaly detection accuracy and reliability by combining complementary sensor data, providing cross-verification, and reducing false positives and missed detections. The combination merely applies a known technique (sensor fusion) to a known system (IoT multi-sensor anomaly detection) to achieve a predictable result (improved detection accuracy and reliability).
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Mahishi et al. (“Mahishi,” US 2025/0307694) in view of Liu et al. (“Liu,” US 2024/0378263), and Wang et al. (“Wang,” US 20220221/0279618), and further in view of Woolf (“Woolf,” US 2021/0342743)
Regarding claim 7, the combination of Mahishi, Liu, and Wang teaches the system of claim 1. The combination of Mahishi, Liu, and Wang teaches, wherein the communication interface but does explicitly discloses further configured to: receive feedback from the end-user regarding the one or more recommendations; and transmit the feedback to the data processing unit, wherein the instructions further cause the data processing unit to refine the plurality of machine learning models based on the feedback.
However, in an analogous art, Woolf discloses
receive feedback from the end-user regarding the one or more recommendations (Woolf: par. 0026, Allows users to provide feedback to refine the shared models and improve their capabilities for all members of the model collective"; par. 0025. "Allows users to download the shared encapsulated models and validate the results against their own data to approve/reject the updated results".); and
transmit the feedback to the data processing unit, wherein the instructions further cause the data processing unit to refine the plurality of machine learning models based on the feedback (Woolf: par. 0026, "Allows users to provide feedback to refine the shared models and improve their capabilities...by encapsulating the aggregated model and uploading to the model sharing collective, i.e., centralized repository"; par. 0020, "The user may further iteratively configure and refine the machine learning model to generate the desired model outcomes"; par. 0024, upload the new user-directed iterative (UDI) machine learning models...into a centralize repository".).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Woolf with the method and system of Mahishi, Liu, and Wang to include receive feedback from the end-user regarding the one or more recommendations; and transmit the feedback to the data processing unit, wherein the instructions further cause the data processing unit to refine the plurality of machine learning models based on the feedback. One would have been motivated to provide methods and tools that not only afford machine learning models that are easily created and configured without the necessity of hard coding by the user, but also to afford multiple users with the ability to share their “know-how” derived from these models to collectively improve the models while maintaining privacy by obscuring the original training data, so that no confidential or proprietary information is shared between users of this collective model. Users may thereby rapidly teach the machine learning models to interpret their data without programming, personalizing the system's analysis and filtering capabilities, and then encapsulate their domain expertise in machine learning models that can be leveraged at scale and shared throughout a single or across multiple enterprises (Woolf: par. 0006).
Claim 8 is rejected under 35 U.S.C. 103 as being unpatentable over Mahishi et al. (“Mahishi,” US 2025/0307694) in view of Liu et al. (“Liu,” US 2024/0378263), and Wang et al. (“Wang,” US 20220221/0279618), and further in view of Kini et al. (“Kini,” US 2023/0316370)
Regarding claim 8, the combination of Mahishi, Liu, and Wang teaches the system of claim 1. Mahishi, Liu, and Wang do not explicitly, wherein the instructions further cause the data processing unit to: identify, based on the cross-validated analysis, one or more systems being monitored by of the plurality of sensors that are not functioning normally; and generate a notification recommending maintenance or replacement of the identified one or more systems.
However, in an analogous art, Kini discloses
identify, based on the cross-validated analysis, one or more systems being monitored by of the plurality of sensors that are not functioning normally (Kini: par. 0040, The personalized recommendation(s) … enable the particular user to identify and/or make changes to one or more underutilized or overutilized resource devices 108”; par. 0025, provider systems may generate various recommendations based on utilization data of a resource device to enable a user to identify whether the resource device is underutilize; par. 0032, "machine learning models" including "KNN (K-Nearest Neighbor) model" are used for analysis.); and
generate a notification recommending maintenance or replacement of the identified one or more systems (Kini: par. 0025, Some provider systems may generate various recommendations based on utilization data of a resource device … such that the user may appropriately shutdown, reconfigure, replace, or otherwise manipulate the resource device based on the recommendations; par. 0040, personalized recommendation(s) may include .. a shutdown recommendation, a rightsize recommendation, a burstable recommendation, and/or the like. A shutdown
recommendation is a recommendation to shut down the relevant resource device 108 or put the resource device 108 in a low power/sleep state, for example, if the relevant resource device 108 has not been used for a predetermined period of time (e.g., 7 days). A rightsize recommendation is a recommendation to identify and/or use a more suitable resource device that may be used instead of (e.g., that may replace) the relevant resource device 108. A burstable recommendation is a recommendation to modify a size or a capacity of the relevant resource device 108.).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Kini with the method and system of Mahishi, Liu, and Wang to include identify, based on the cross-validated analysis, one or more systems being monitored by of the plurality of sensors that are not functioning normally; and generate a notification recommending maintenance or replacement of the identified one or more systems. One would have been motivated to provide such a combination because both Mahishi and Kini address monitoring IoT and computing systems for abnormal peration to enable proactive management. Mahishi teaches using IoT sensors and machine learning to detect anomalies in physical systems, while Kini teaches using machine learning (KNN model) to identify specific resource devices (including IoT devices) that are underutilized or overutilized and generating actionable recommendations for maintenance or replacement. It would have been obvious to apply Kini's approach of identifying specific problematic systems and generating targeted maintenance/replacement recommendations to Mahishi's IoT anomaly detection system. Both systems use machine learning to analyze monitored systems, detect abnormal operation, and provide actionable recommendations to improve system performance. The combination addresses the known problem of efficiently managing multiple monitored systems by identifying which specific systems require attention and recommending appropriate corrective actions, yielding the predictable result of improved system management and reduced downtime.
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over Mahishi et al. (“Mahishi,” US 2025/0307694) in view of Liu et al. (“Liu,” US 2024/0378263), and Wang et al. (“Wang,” US 20220221/0279618), and further in view of Yankovskiy et al. (“Yankovskiy,” US 2020/0304304)
Regarding claim 9, the combination of Mahishi, Liu, and Wang teaches the system of claim 1. Mahishi, Liu, and Wang do not explicit discloses wherein the instructions further cause the data processing unit to: encrypt the collected real-time data using a cryptographic key before transmitting the data to the data processing unit; and decrypt the encrypted data using the cryptographic key before analyzing the data using the plurality of machine learning models.
However, in an analogous art, Yankovskiy discloses
encrypt the collected real-time data using a cryptographic key before transmitting the data to the data processing unit (Yankovskiy: par. 0074, As already noted, the preferred embodiment uses symmetric encryption to encrypt data at-rest on gateway devices 104 of FIG. 1. Such an encryption regime may be represented by the following equation: ENC(K.sub.i,P.sub.i)=C.sub.i Eq. 1; par. 0075, Here K.sub.i is the symmetric cryptographic key of the above discussion that is used by encryption operation ENC for encrypting plaintext IoT data P.sub.i to produce corresponding ciphertext C.sub.i; par. 0076, Since key K.sub.i is symmetric, the converse process of decryption is performed by inverting the above operation. The decryption process is typically carried at a target backend server or computer in data center 116 where plaintext IoT data P.sub.i is to be used and is conveniently represented by the following equation: DEC(K.sub.i,C.sub.i)=P.sub.i Eq. 2; par. 0078, Enc operation 200 of FIG. 2 takes as inputs plaintext data 202 P.sub.i, P.sub.i, P.sub.k from respective IoT device i, j, k (not shown in FIG. 2) and encrypts them with respective symmetric encryption keys 204 K.sub.i, K.sub.j, K.sub.k respectively. Enc operation 200 produces ciphertext data 210 C.sub.i, C.sub.1, C.sub.k to respective plaintext data 202 P.sub.i, P.sub.j, P.sub.k as produced by applying Eq. 1 above); and
decrypt the encrypted data using the cryptographic key before analyzing the data using the plurality of machine learning models (Yankovskiy: par. 0076, Since key K.sub.i is symmetric, the converse process of decryption is performed by inverting the above operation. The decryption process is typically carried at a target backend server or computer in data center 116 where plaintext IoT data P.sub.i is to be used and is conveniently represented by the following equation: DEC(K.sub.i,C.sub.i)=P.sub.i Eq. 2; par. 0079, In a similar manner, FIG. 3 shows decryption operation 250 of Eq. 2 above that takes as inputs symmetric encryption keys 204 K.sub.i, K.sub.j, K.sub.k and ciphertext data 210 C.sub.i, C.sub.j, C.sub.k to produce respective plaintext data 202 P.sub.i, P.sub.j, P.sub.k.).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Yankovskiy with the method and system of Mahishi, Liu, and Wang to include encrypt the collected real-time data using a cryptographic key before transmitting the data to the data processing unit; and decrypt the encrypted data using the cryptographic key before analyzing the data using the plurality of machine learning models. One would have been motivated to provide such a combination because both Mahishi and Yanko address IoT sensor systems, and securing IoT data transmission is a well-known concern in IoT deployments. Yanko teaches that IoT data should be encrypted at gateway devices before transmission to backend servers to provide data security. It would have been obvious to apply Yanko's encryption/decryption scheme to Mahishi's IoT sensor system to protect sensor data during transmission over networks. This combination addresses the known problem of securing IoT communications using the well-known solution of symmetric encryption, yielding the predictable result of secure data transmission. The combination merely applies a known security technique (encryption/decryption) to a known IoT system (Mahishi's anomaly detection platform) to achieve the expected benefit of protecting data in transit.
Claim 10 is rejected under 35 U.S.C. 103 as being unpatentable over Mahishi et al. (“Mahishi,” US 2025/0307694) in view of Liu et al. (“Liu,” US 2024/0378263), and Wang et al. (“Wang,” US 20220221/0279618), and further in view of Kadota (“Kadota,” US 2022/0141811)
Regarding claim 10, the combination of Mahishi, Liu, and Wang teaches the system of claim 1. The combination of Mahishi, Liu, and Wang teaches, wherein the plurality of sensors are configured to communicate with the data processing unit but does not explicitly disclose using a low-power wide-area network (LPWAN) communication protocol.
However, in an analogous art, Kadota discloses that the communicate between sensor information obtainment device 2 and IoT sensors device using low-power wide-area network (LPWAN) communication protocol (Kadota: par. 0032, communication performed between the sensor information obtainment device 2 and the IoT sensors 4 may be wired or wireless.. As communication standards based on LPWA, for example, SIGFOX (sub-GHz band (866 MHz band, 915 MHz band, and 920 MHz band), with a maximum transmission speed of about 100 bps and a transmission distance of about several tens of kilometers), LoRa (sub-GHz band, with a maximum transmission speed of about 250 kbps and a transmission distance of about a maximum of 10 km), and the like can be given)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Kadota with the method and system of Mahishi, Liu, and Wang to include wherein the plurality of sensors are configured to communicate with the data processing unit using a low-power wide-area network (LPWAN) communication protocol. One would have been motivated to provide an information processing device, an information processing method, and a program capable of efficiently obtaining data from a plurality of devices (Kadota: par. 0006).
Claims 11-12 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Mahishi et al. (“Mahishi,” US 2025/0307694) in view of Nam et al. (“Nam,” US 2024/0394269), and further in view Liu et al. (“Liu,” US 2024/0378263)
Regarding claim 11, Mahishi teaches a computer-implemented method for enhancing decision-making in an AI system, the method comprising:
collecting, using a plurality of sensors, real-time data from an environment or system (Mahishi: par. 0106, An IoT platform may receive data from multiple sensors that report a variety of metrics (e.g., temperature, speed, position, pressure, noise, image data, vibration, motion, personnel detection, mode of operation, task being performed, etc.”) , “The registered sensors may report to the platform periodically, sporadically, or otherwise be pushed into the platform or pulled in by the platform”);
receiving, by a data processing unit, the collected real-time data from the plurality of sensors (Mahishi: par. 0106, fig. 5, pars. 0113, 0123, 0125, 0136);
Mahishi discloses receiving, by a data processing unit, the collected real-time data from the plurality of sensors but does not explicitly disclose analyzing, by the data processing unit, the collected real-time data using a machine learning model to identify patterns and detect anomalies via consensus building machine learning models to increase overall detection accuracy, wherein analyzing the collected real-time data comprises: generating, using a probabilistic circuit model, a generative model of the environment based on the collected real-time data;
However, in an analogous art, Nam discloses
analyzing, by the data processing unit, the collected real-time data using a machine learning model to identify patterns and detect anomalies via consensus building machine learning models to increase overall detection accuracy, wherein analyzing the collected real-time data comprises (Nam:par. [0007]: "a model management unit configured to generate a query response model by training a probabilistic circuit-based model using training data collected by devices included in an approximate query processing network." Nam teaches analyzing collected data using a probabilistic circuit-based machine learning model. Abstract: "in a distributed network environment where various terminals (sensors, mobile computing and communication devices, and the like), network devices, and cloud infrastructures autonomously participate and continuously collect data." par. [0008]: "The model management unit may map the devices included in the network to leaf nodes of the probabilistic circuit-based model and generate the probabilistic circuit-based model by configuring sum nodes for binding the same type of data and product nodes for binding different types of data on the basis of information on data types provided by the devices included in the network.").
generating, using a probabilistic circuit model, a generative model of the environment based on the collected real-time data (Nam: abstract; par. [0007]: "a model management unit configured to generate a query response model by training a probabilistic circuit-based model using training data collected by devices included in an approximate query processing network." ; par.: "The model management unit may...generate the probabilistic circuit-based model by configuring sum nodes for binding the same type of data and product nodes for binding different types of data on the basis of information on data types provided by the devices included in the network.".; par. 0008, aches that the model "map[s] the devices included in the network to leaf nodes" and uses "information on data types provided by the devices included in the network.").
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Nam with the method and system of Mahishi to include analyzing, by the data processing unit, the collected real-time data using a machine learning model …; “ generating, using a probabilistic circuit model, a generative model of the environment based on the collected real-time data;” One would have been motivated to improve scalability and efficiency for approximate queries about operations (aggregation, statistics, and the like) mainly used in exploratory data analysis on the basis of a tractable probabilistic circuit (TPC) in a distributed network environment where various terminals (sensors, mobile computing and communication devices, and the like), network devices, and cloud infrastructures autonomously participate and continuously collect data (Nam: par. 0005).
Nam does not teach identifying, using an ARF model, anomalies in the collected real-time data based on the generative model; cross-validating, by the data processing unit, the analysis results …; generating, by the data processing unit and based on the cross-validated analysis …; transmitting, via a communication interface, the one or more recommendations to a user interface; and displaying, on the user interface, the one or more recommendations to an end-user.
However, in an analogous art, Liu discloses identifying, using an ARF model, anomalies in the collected real-time data (Liu: par. 0042, "advanced detection algorithms can be applied to the selected features from block 210 to identify instances of non-synchronization. This block can utilize a combination of machine learning models, such as neural networks, decision trees, and ensemble methods like random forests or boosted trees, to analyze patterns and anomalies that suggest synchronization issues." Note that Liu explicitly teaches using "ensemble methods like random forests" to "analyze patterns and anomalies").
cross-validating, by the data processing unit, the analysis results by comparing the identified patterns and detected anomalies across the collected real-time data from the plurality of sensors (Liu: par. [0042]: "This block can utilize a combination of machine learning models, such as neural networks, decision trees, and ensemble methods like random forests or boosted trees, to analyze patterns and anomalies." Liu [0076]: "incorporate fail-safes and redundancy checks to ensure that the corrective measures it enacts are both appropriate for the detected events. Liu's use of a "combination of machine learning models" [0042] inherently involves cross-validation by comparing model outputs. When multiple diverse models (neural networks, decision trees, random forests) independently analyze the same data to detect "patterns and anomalies", cross-validation is achieved by comparing their outputs).
generating, by the data processing unit and based on the cross-validated analysis, one or more recommendations for improving the system (Liu: par. 0076, "alert generation and response activation system...Upon detection, this system can autonomously initiate a series of predefined corrective actions, ranging from system parameter adjustments to more complex remedial protocols"; "corrective measures... do not inadvertently introduce further system instabilities"; "reducing system downtime.");
transmitting, via a communication interface, the one or more recommendations to a user interface (Liu: par. [0080], teaches a system with a "user interface" that facilitates "proactive monitoring and automated rectification of non-