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 .
Status of Claims
Claims 1, 3-4, 9-11, 13-14 and 19-20 are amended. Claims 6-8 and 16-18 are cancelled. Claims 1-5, 9-15 and 19-20 are pending in the application.
Claims 1-5, 9-15 and 19-20 are rejected under 35 U.S.C. 103.
Response to Amendment
The amendment filed April 15th, 2026 has been entered. Claims 1, 3-4, 9-11, 13-14 and 19-20 are amended. Claims 6-8 and 16-18 are cancelled. Claims 1-5, 9-15 and 19-20 are pending in the application. Applicant’s amendments to the claims and specification have overcome each and every objection and 112(b) rejection previously set forth in the Non-Final Rejection Office Action mailed January 15th, 2026.
Response to Arguments
Applicant’s arguments, see page 16 - 22, filed April 15th, 2026, with respect to 35 U.S.C. § 101 rejection have been fully considered but are moot in view of the amendments. The 35 U.S.C. 101 rejection for claims 1-5, 9-15 and 19-20 has been withdrawn.
Applicant’s arguments, see page 22 – 25, filed April 15th, 2026, with respect to 35 U.S.C. § 103 rejection have been fully considered but are not persuasive. Applicant argues that Mozo and Bonati do not teach modifying the one or more network scenarios based on feedback. The rejection has been modified to further rely on Xia (NPL: ”Genet: Automatic Curriculum Generation for Learning Adaptation in Networking”) for this amended feature. Mozo is still relied upon for teaching a network digital twin used for AI/ML training and testing in generated network scenarios. Bonati is used for teaching O-RAN closed loop analytics and control. Xia further introduces automatic curriculum generation in a networking environment, where network environments used for training are selected and updated based on the current model’s performance. The combination of references teaches using the digital twin O-RAN training environment of Mozo and Bonati with Xia’s technique of adapting network training environments based on model performance feedback. Accordingly, the 103 rejection is maintained.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1 – 5, 9, 11 – 15, 19 – 20 are rejected under 35 U.S.C. 103 as being unpatentable over Mozo et al. (NPL: “B5GEMINI: AI-Driven Network Digital Twin“) in view of Bonati et al. (NPL: ”Intelligence and Learning in O-RAN for Data-driven NextG Cellular Networks”), further in view of Xia et al. (NPL: “Genet: Automatic Curriculum Generation for Learning Adaptation in Networking”).
Regarding Claim 1, Mozo teaches
receiving network data from a network, the network data comprising performance metrics data; (Pg. 11 of Mozo states “This module is responsible for monitoring the complete operation of all information exchanged within the NDT. Once all the DTs are provisioned and configured, the monitoring module will control the activation of port-mirroring functionalities within each one of the subnets present in the resulting NDT, with the aim of being able to use this information traffic for the generation of datasets that allow for the improvement of the of the complete emulation of the NDT.“)
generating a digital twin of the network, wherein the digital twin is calibrated based on the network data, wherein the generating the digital twin of the network comprises; (Pg 3. of Mozo states “B5GEMINI goes one step beyond and also focuses on the use of NDT for training and validating AI/ML components in a controlled scenario when potentially harmful situations could occur if trained or tested in a real network. B5GEMINI will provide a virtualized representation of the 5G/6G network meant to analyze, diagnose, emulate, and control the physical network… With the support of AI, complex data sources and telemetry (e.g., IoT sensors, network data) will be fed into sophisticated data interpretation processes to facilitate the replica of 5G/6G network components.” Pg. 7 of Mozo states “From this point of view, using a bidirectional data flow between both worlds, the DT is able to continuously adapt to operational changes based on real-time data and information coming from the physical twin, being able, among other things, to monitor and even predict the future state of the physical twin. In addition, the DT can also be manipulated, and changes made to it can be automatically transferred to the physical twin.“ Pg. 8 of Mozo states “One of the crucial elements that differentiates B5GEMINI from Mouseworld is the two-way communication capability with the real network, which allows real‐time synchronization between the real and virtual networks. That is, the configuration of the real network can be replicated in the virtual twin in real‐time and the optimizations applied to the virtual twin can be seamlessly deployed in the real network”)
analyzing a performance of an artificial intelligence (AI) model or a machine learning (ML) model; (Pg. 3 of Mozo states “It is worth noting that current NDT proposals tend to apply AI to add intelligence in different ways to network orchestration processes. B5GEMINI goes one step beyond and also focuses on the use of NDT for training and validating AI/ML components in a controlled scenario when potentially harmful situations could occur if trained or tested in a real network.” Pg. 17 of Mozo states " In addition, for some cases, such as Reinforcement Learning (RL), the ML model proposes changes to the NDT and once completed the result can be evaluated." Pg. 25 of Mozo states “To demonstrate the advantages provided by the architecture proposed in Section 4.3, we present a use case for energy optimization that builds on the DMap concept to optimize the energy consumption of an NDT system by managing its resources.” B5GEMINI’s training, validation, and energy optimization use cases require monitoring ML performance in the NDT.)
generating one or more network scenarios via a radio access network (RAN) scenario generator based on the performance of the AI model or the ML model, and (Pg. 6 of Mozo states “One of the corner stones of B5GEMINI is the Mouseworld Lab [22], a controlled environment set up in the Telefónica I+D premises for running experiments that allow deploying complex network scenarios in a controlled way and generate realistic labeled data sets for training supervised ML components and validate supervised and unsupervised solutions.” Pg. 7 of Mozo states “Therefore, the Machine Learning Orchestration is part of the Digital Twin Entity Management and is composed of a Topology Generator and an Experiment Launcher. The first uses predefined templates, interacts with OSM (ETSI NFVSOL-005 interface) and OpenStack (Glance Image service API) provisioning NSD (Network Service Descriptors) and VNFD (VNF descriptors) compatible with OSM, detailing the network functions, the links, and their day-1 configurations. The Experiment Launcher makes day-2 configurations and triggers the emulations functions for dataset generation, and it is described with OSM’s ProxyCharms or with dedicated scripts. The current version uses a configuration file to define the statistical distribution of the traffic, the number of intervals in which the experiment is divided and the duration, and the type of service that it is emulating.” The topology generator and experiment launcher are like a scenario generator. They define scenarios then trigger emulation runs that feed the ML training pipeline.)
modeling the network based on the network data and the generated one or more network scenarios. (Pg. 8 of Mozo states “the development of smart agents that can be deployed in the target network, which are in charge of collecting all the information necessary for the DT generation (topological information, hardware and software information, states, etc.). This information is introduced to the deployment module in the JSON or XML format” Pg. 2 of Mozo states “By replicating different environments in a lab and running multiple scenarios, NDTs offer a cost-effective way to assess performance, predict the impact of environmental changes (such as cyber threats), and optimize network processes and decision making accordingly.” Pg. 3 of Mozo states “With the support of AI, complex data sources and telemetry (e.g., IoT sensors, network data) will be fed into sophisticated data interpretation processes to facilitate the replica of 5G/6G network components.” Mozo teaches modeling the network from both real network data and generated network scenarios.)
training the AI model or the ML model based on data generated from the digital twin; (Pg. 6 of Mozo states “One of the corner stones of B5GEMINI is the Mouseworld Lab [22], a controlled environment set up in the Telefónica I+D premises for running experiments that allow deploying complex network scenarios in a controlled way and generate realistic labeled data sets for training supervised ML components and validate supervised and unsupervised solutions.” Pg. 3 of Mozo states “B5GEMINI goes one step beyond and also focuses on the use of NDT for training and validating AI/ML components in a controlled scenario when potentially harmful situations could occur if trained or tested in a real network… B5GEMINI will enable an extensive use of advanced AI mechanisms to realize several valuable AI applications, such as (i) the training and testing of ML components to deploy smart applications such as cybersecurity or network management in real-time environments; … (iii) the use of the NDT as a platform to perform distributed training and inference processes for ML and DL models using on-demand GPU virtualization;”)
monitoring a performance of the modeled network (Pg. 11 of Mozo states “This module is responsible for monitoring the complete operation of all information exchanged within the NDT. Once all the DTs are provisioned and configured, the monitoring module will control the activation of port-mirroring functionalities within each one of the subnets present in the resulting NDT, with the aim of being able to use this information traffic for the generation of datasets that allow for the improvement of the complete emulation of the NDT.“)
optimizing the modeled network based on the provided feedback to the RAN scenario generator (Pg. 2 of Mozo states “Additionally, by creating an NDT, the real time optimization of its corresponding physical equivalent is also possible. Furthermore, NDTs can help secure traditional networks by enabling fast identification and isolation of network failures to quickly respond to security threats.” Pg. 17 of Mozo states “In contrast, B5GEMINI can provide a real-time assessment of the 5G network with little overhead, allowing for faster and more accurate assessment. In addition, with this approach, AI methods can be more easily integrated to gain feedback from the network configuration, provide suggestions to optimize it and enforce the configuration automatically in the live environment. In this way, B5GEMINI can serve as a key tool for network operators to manage the deployment of 5G networks in a closed loop fashion to ensure timely and efficient network operations.”)
Mozo does not explicitly teach that
operating the trained AI model or the trained ML model within the network.
providing feedback to the RAN scenario generator based on the monitored performance of the modeled network;
modifying the one or more network scenarios based on the feedback;
However, Bonati teaches that
and operating the trained machine learning model within the network. (Pg. 5 of Bonati states “We embedded the DRL agents into xApps running in the near real-time RIC (right of Fig. 3), for a total of 12 DRL agents running in parallel and making decisions with a time granularity of 500 ms. Agents connect with the network BSs through the ORAN E2 interface.” Pg. 6 of Bonati states “To train our DRL agents we generated some 7 GB of training data of various performance metrics” Trained DRL agents deployed as xApps in the near RT RIC.)
Xia teaches that
providing feedback to the RAN scenario generator based on the monitored performance of the modeled network; (Pg. 1 of Xia states “Genet automatically searches for such environments and iteratively promotes them to training. Three case studies—adaptive video streaming, congestion control, and load balancing—demonstrate that Genet produces RL policies that outperform both regularly trained RL policies and traditional baselines.” Pg. 6 of Xia states “Each iteration consists of three steps (which will be detailed shortly): 1. First, we update the current RL model for a fixed number of iterations over the current training environment distribution; 2. Second, we select the environments where the current RL model has a large gap-to-baseline; and 3. Third, we promote these selected environments in the training environments distribution used by the RL training process in the next iteration… After a certain number of iterations, the current RL model and a pre-determined rule-based baseline are given to a sequencing module to search for the environments where the current RL model has a large gap-to-baseline.” Xia teaches using monitored model performance as feedback to determine which network environment configurations should be selected and promoted for later training. Even though Xia does not use the term “RAN scenario generator”, it teaches the same relevant feedback function in a networking scenario generation context. It would have been obvious to combine with Mozo’s RAN scenario generator in the O-RAN digital twin system of Mozo and Bonati.)
modifying the one or more network scenarios based on the feedback; (Pg. 6 – 7 of Xia states “Once a new configuration is selected, the environments generated by this configuration are then added to the training distribution as follows. When the RL training process samples a new training environment, it will choose the new configuration with 𝑤 probability (30% by default) or uniformly sample a configuration from the old distribution with 1 − 𝑤 probability (70% by default), and then create an environment based on the selected configuration. Next, training is resumed over the new environment distribution... Genet restarts the BO search every time the RL model is updated. The reason is that the rewarding environments can change once the RL model changes.” Xia modifies the distribution of generated network environment configurations by adding configurations identified through performance feedback. A PHOSITA would recognize as modifying the network scenarios generated by the scenario generator for training iterations.)
It would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to combine the teachings of Mozo, Bonati, and Xia. Bonati teaches an O-RAN architecture where ML based agents are trained using data generated on a large RF emulation and then deployed in the non-RT and near-RT RICs to optimize RAN performance. Mozo teaches using a network digital twin to deploy complex network scenarios in a controlled environment and generate realistic labeled datasets for training and validating ML components, and emphasize two way coupling between the digital twin and the real network. Xia teaches a networking domain curriculum generation technique in which monitored RL model performance is used to select and promote network environment configurations for subsequent training. One with ordinary skill in the art would be motivated to incorporate the teachings of Bonati, Xia into the teachings of Mozo in order to improve the safety, flexibility, and efficiency of training RIC ML agents, to reuse synthetic and real data across both environments, to provide manageable environment for testing before deployment, and to make sure that generated network scenarios are updated based on monitored model/network performance. It would have been predictable combination to improve robustness and performance of the trained AI/ML model by exposing the model to network scenarios where its current performance was weak, before deployment in the physical network.
Regarding claim 2, the rejection of claim 1 is incorporated herein. Furthermore, the combination of Mozo, Bonati, and Xia teaches that
wherein the network is based on an open radio access network (O- RAN). (Pg. 1 of Bonati states “This article explores the disaggregated network architecture proposed by the O-RAN Alliance as a key enabler of NextG networks… The second core innovation—which is likely to be even more impactful—is the Radio Access Network (RAN) Intelligent Controller (RIC), a new architectural component that provides a centralized abstraction of the network, allowing operators to implement and deploy custom control plane functions. In both its non and near real-time versions, the RIC facilitates RAN optimization through closed-control loops, i.e., autonomous action and feedback loops between RAN components and their controllers.”)
Regarding claim 3, the rejection of claim 2 is incorporated herein. Furthermore, the combination of Mozo, Bonati, and Xia teaches that
operating the digital twin within a non-real-time (Non-RT) radio access network intelli- gent controller (RIC) framework as part of an AI or ML workflow for training and testing the AI model or the ML model. (Pg. 3 of Bonati states “Indeed, since AI/ML techniques usually rely upon a randomized initialization, O-RAN requires all ML models to be trained and validated offline before their deployment [10] ... Online AI/ML techniques could still be used in O-RAN compliant architectures by allowing models to be trained with offline data in the non real-time RIC, and then perform online learning in the near real-time RIC… The O-RAN Alliance defines non real-time any control loop that operates on a timescale of at least one second. As shown in Fig. 2, this involves the coordination between the non real-time and near real-time RIC through the A1 interface. This control loop manages the orchestration of resources at the infrastructure level, making decisions and applying policies that impact thousands of devices. These actions can be performed using data-driven optimization algorithms processing data from multiple sources, and inference models deployed on the non real-time RIC itself.” Pg. 3 of Mozo states “It is worth noting that current NDT proposals tend to apply AI to add intelligence in different ways to network orchestration processes. B5GEMINI goes one step beyond and also focuses on the use of NDT for training and validating AI/ML components in a controlled scenario when potentially harmful situations could occur if trained or tested in a real network.”)
Regarding claim 4, the rejection of claim 2 is incorporated herein. Furthermore, the combination of Mozo, Bonati, and Xia teaches that
operating the digital twin within a near-real-time (Near-RT) radio access network intelli- gent controller (RIC) framework as part of an AI or ML workflow for training and testing the AI model or the ML model. (Pg. 4 of Bonati states “Near real-time control loops operate on a timescale between 10 ms and 1 s… Because one near real-time RIC is associated to multiple gNBs, these control loops can make decisions affecting up to thousands of UEs, using user-session aggregated data and Medium Access Control (MAC)/PHY layer KPIs. ML-based algorithms are implemented as external applications, i.e., xApps, and are deployed on the near real-time RIC to deliver specific services such as inference, classification, and prediction pipelines to optimize the per-user quality of experience, controlling load balancing and handover processes, or the scheduling and beamforming design.” Pg. 5 of Bonati states “We embedded the DRL agents into xApps running in the near real-time RIC (right of Fig. 3), for a total of 12 DRL agents running in parallel and making decisions with a time granularity of 500 ms. Agents connect with the network BSs through the ORAN E2 interface.” Pg. 3 of Mozo states “It is worth noting that current NDT proposals tend to apply AI to add intelligence in different ways to network orchestration processes. B5GEMINI goes one step beyond and also focuses on the use of NDT for training and validating AI/ML components in a controlled scenario when potentially harmful situations could occur if trained or tested in a real network.” Combining Mozo and Bonati, we get a DT whose data and behavior feed into the near RT RIC ML workflow.)
Regarding claim 5, the rejection of claim 2 is incorporated herein. Furthermore, the combination of Mozo, Bonati, and Xia teaches that
operating the digital twin in parallel with the O-RAN, wherein the digital twin is further operated outside of a non-real-time (Non-RT) radio access network intelligent controller (RIC) framework and a near-real-time (Near-RT) radio access network intelligent controller (RIC) framework. (Pg. 8 of Mozo states “One of the crucial elements that differentiates B5GEMINI from Mouseworld is the two‐way communication capability with the real network, which allows real‐time synchronization between the real and virtual networks… That is, the configuration of the real network can be replicated in the virtual twin in real-time and the optimizations applied to the virtual twin can be seamlessly deployed in the real network” Pg. 3 of Bonati states “This control loop manages the orchestration of resources at the infrastructure level, making decisions and applying policies that impact thousands of devices. These actions can be performed using data-driven optimization algorithms processing data from multiple sources, and inference models deployed on the non real-time RIC itself.“ Pg. 4 of Bonati states “Moreover, since the non real-time RIC is endowed with service management and orchestration capabilities, this control loop can also handle the association between the near real-time RIC and the DUs/CUs. This is particularly useful in virtualized systems where DUs and CUs are dynamically instantiated on-demand to match the requests and load of the RAN.” B5GEMINI describes a virtual twin running alongside the real network, synchronized. Bonati’s Non-RT RIC is the ML framework with the DT/NDT as component. )
Regarding claim 9, the rejection of claim 1 is incorporated herein. Furthermore, the combination of Mozo, Bonati, and Xia teaches that
wherein the digital twin comprises an offline simulation module and a runtime simulation module. (Pg. 3 of Mozo states “It is worth noting that current NDT proposals tend to apply AI to add intelligence in different ways to network orchestration processes. B5GEMINI goes one step beyond and also focuses on the use of NDT for training and validating AI/ML components in a controlled scenario when potentially harmful situations could occur if trained or tested in a real network… (iii) the use of the NDT as a platform to perform distributed training and inference processes for ML and DL models using on-demand GPU virtualization;” Pg. 8 of Mozo states “One of the crucial elements that differentiates B5GEMINI from Mouseworld is the two-way communication capability with the real network, which allows real-time synchronization between the real and virtual networks. That is, the configuration of the real network can be replicated in the virtual twin in real-time and the optimizations applied to the virtual twin can be seamlessly deployed in the real network.” Pg. 3 of Bonati states “Second, data-driven solutions must be trained and validated offline to avoid causing inefficiencies—or even outages—to the RAN. Indeed, since AI/ML techniques usually rely upon a randomized initialization, O-RAN requires all ML models to be trained and validated offline before their deployment [10]… Online AI/ML techniques could still be used in O-RAN compliant architectures by allowing models to be trained with offline data in the non real-time RIC, and then perform online learning in the near real-time RIC..” The NDT/.DT environment used for training and testing ML components, and O-RAN’s requirement that models be trained or validated offline maps to offline simulation module. The part of the twin that is synchronized with live data and participates in real or near real time control loops from Bonati maps to runtime simulation module.)
Claims 11 – 15, 19 recite substantially similar subject matter as claims 1 – 5, 9 respectively, and are rejected with the same rationale, mutatis mutandis.
Claim 20 recites substantially similar subject matter as claim 1 respectively, and is rejected with the same rationale, mutatis mutandis.
Claim 10 is rejected under 35 U.S.C. 103 as being unpatentable over Mozo et al. (NPL: “B5GEMINI: AI-Driven Network Digital Twin“) in view of Bonati et al. (NPL: ”Intelligence and Learning in O-RAN
for Data-driven NextG Cellular Networks”), Xia et al. (NPL: “Genet: Automatic Curriculum Generation for Learning Adaptation in Networking”), further in view of Yilmaz et al. (NPL: “Radio Environment Map as Enabler for Practical Cognitive Radio Networks”).
Regarding claim 10, the rejection of claim 9 is incorporated herein. Furthermore, the combination of Mozo, Bonati, and Xia teaches that
generating user equipment (UE) mobility pattern within the offline simulation module; (Pg. 5 of Bonati states “We have emulated a 5G network with 4 BSs and 40 UEs (Fig. 3, left) in the dense urban scenario of Rome, Italy.” Pg. 6 of Bonati states “Agents have been trained under network configurations obtained by varying the distance between BSs and UEs and the mobility of the UEs.” Pg. 7 of Mozo states “The Experiment Launcher makes day-2 configurations and triggers the emulations functions for dataset generation, and it is described with OSM’s ProxyCharms or with dedicated scripts. The current version uses a configuration file to define the statistical distribution of the traffic, the number of intervals in which the experiment is divided and the duration, and the type of service that it is emulating. Currently, there are defined three scenarios: traffic classification, crypto mining malware detection, and DoH attacks.” Bonati teaches generating UE trajectories in an emulator by varying distances and mobility of the UEs. Mozo’s experiment launcher shows such patterns are part of a configurable offline simulation module for emulation.)
simulating radio frequency (RF) propagation within the offline simulation module, (Pg. 4 of Bonati states “Colosseum includes 128 compute nodes, called Standard Radio Nodes (SRNs), equipped with USRP X310 SDRs that can be used to run generic protocol stacks. These are connected in a mesh topology through 128 additional USRPs X310 of the Massive Channel Emulator (MCHEM) for emulating realistic RF scenarios. The wireless channel between each pair of devices is modeled through complex-valued finite impulse response filter taps. In this way, scenarios are able to capture effects such as path loss, multi-path and fading as if the SDRs were operating in a real RF environment.”)
and loading the UE mobility pattern… to generate training or testing data to the AI model or the ML model under training or testing in run time. (Pg. 5 of Bonati states “We have emulated a 5G network with 4 BSs and 40 UEs (Fig. 3, left) in the dense urban scenario of Rome, Italy.” Pg. 6 of Bonati states “Agents have been trained under network configurations obtained by varying the distance between BSs and UEs and the mobility of the UEs.” Pg. 6 of Mozo states “One of the corner stones of B5GEMINI is the Mouseworld Lab [22], a controlled environment set up in the Telefónica I+D premises for running experiments that allow deploying complex network scenarios in a controlled way and generate realistic labeled data sets for training supervised ML components and validate supervised and unsupervised solutions.” Pg. 7 of Mozo states “The Experiment Launcher makes day-2 configurations and triggers the emulations functions for dataset generation, and it is described with OSM’s ProxyCharms or with dedicated scripts. The current version uses a configuration file to define the statistical distribution of the traffic, the number of intervals in which the experiment is divided and the duration, and the type of service that it is emulating. Currently, there are defined three scenarios: traffic classification, crypto mining malware detection, and DoH attacks.”)
However, the combination of Mozo and Bonati does not explicitly teach that
generating an RF map at least partially representing power and interference at each location within a geographical area;
and loading the RF map generated
Yilmaz teaches that
generating an RF map at least partially representing power and interference at each location within a geographical area; (Pg. 162 of Yilmaz states “In that regard, Radio Environment Map (REM) is a promising tool that provides a practical means for the realization of cognitive radio networks (CRNs). It constructs a comprehensive map of the CRN by utilizing multi-domain information from geolocation databases, characteristics of spectrum use, geographical terrain models, propagation environment, and regulations.” Pg. 163 of Yilmaz states “Essential functionality of a REM is the construction of dynamic interference map for each frequency at each location of interest.”)
and loading the RF map generated (Pg. 162 of Yilmaz states “In that regard, Radio Environment Map (REM) is a promising tool that provides a practical means for the realization of cognitive radio networks (CRNs). It constructs a comprehensive map of the CRN by utilizing multi-domain information from geolocation databases, characteristics of spectrum use, geographical terrain models, propagation environment, and regulations.” Pg. 163 of Yilmaz states “Essential functionality of a REM is the construction of dynamic interference map for each frequency at each location of interest.”)
It would have been obvious to one with ordinary skill in the art before the effective filing date of the invention to combine the teachings of Yilmaz into the combination of Mozo and Bonati. Bonati teaches an O-RAN architecture where ML based agents are trained using data generated on a large RF emulation and then deployed in the non-RT and near-RT RICs to optimize RAN performance. Mozo teaches using a network digital twin to deploy complex network scenarios in a controlled environment and generate realistic labeled datasets for training and validating ML components, and emphasize two way coupling between the digital twin and the real network. Xia teaches a networking domain curriculum generation technique in which monitored RL model performance is used to select and promote network environment configurations for subsequent training. Yilmaz teaches the well known concept of a Radio Environment Map, which constructs a map of the radio network to construct a dynamic interference map for each frequency at each location of interest. One with ordinary skill in the art would be motivated to incorporate the teachings of Yilmaz into the teachings of Mozo, Bonati, and Xia in order to capture the simulated RF conditions in a precise, reusable form indexed by geolocation to manage RF environments across scenarios, and provide location aware features to the ML training pipeline. It would have been predictable to combine to improve the RF propagation already simulated in Bonati’s emulator with RF map representation of Yilmaz.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BYUNGKWON HAN whose telephone number is (571)272-5294. The examiner can normally be reached M-F: 9:00AM-6PM PST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Li B Zhen can be reached at (571)272-3768. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/BYUNGKWON HAN/ Examiner, Art Unit 2121
/Li B. Zhen/ Supervisory Patent Examiner, Art Unit 2121
Prosecution Insights