DETAILED ACTION
This office action is responsive to communications filed on February 10, 2026. Claims 1, 4, 6-11, 13,15-18, 20, 21, and 25-30 have been amended. Claims 1, 2, and 4-30 are pending in the application.
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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant’s submission filed on February 10, 2026 has been entered.
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, 2, 4, and 6-30 are rejected under 35 U.S.C. 103 as being unpatentable over Guan et al. (US 2025/0260449) in view of Shi (US 2025/0247733) and Saber et al. (US 2023/0131694).
Regarding Claim 1, Guan teaches an apparatus of a user equipment (UE) for wireless communication, comprising: one or more memories; and one or more processors coupled to the one or more memories, the one or more processors individually or collectively configured to (“The device 700 can be considered as a further example implementation of the terminal device 110” – See [0265]; “As shown, the device 700 includes a processor 710, a memory 720 coupled to the processor 710, a suitable transmitter (TX) and receiver (RX) 740 coupled to the processor 710, and a communication interface coupled to the TX/RX 740. The memory 710 stores at least a part of a program 730” – See [0266]; “The program 730 is assumed to include program instructions that, when executed by the associated processor 710, enable the device 700 to operate in accordance with the embodiments of the present disclosure” – See [0267]) cause the UE to:
receive multiple reference signals (RSs) (“the terminal device 110 may receive 311 the set of RSs from the network device 120” – See [0115]; The UE receives a set of RSs);
measure the multiple RSs to generate measurements of the multiple RSs (“Then the terminal device 110 may perform 312 a channel measurement based on the set of RSs to obtain the CSI” – See [0115]; The UE measures the set of RSs);
encode the measurements of the multiple RSs using a machine learning model to generate codewords representing quantized versions of the measurements of the multiple RSs (“the terminal device 110 compresses 330 the CSI based on the compression method” – See [0189]; “the terminal device 110 may determine the compression method by determining a set of compression parameters. In some embodiments, the set of compression parameters may comprise at least one of a CR, the number of quantization bits, or the number of compressed bits. Of course, any other suitable compression parameters are also feasible” – See [0118]; “the AI or ML based CSI feedback process may be based on an autoencoder. The autoencoder is a type of artificial neural network used to learn efficient codings of unlabeled data (unsupervised learning)” – See [0105]; The UE compresses the CSI feedback by encoding the measurements using an AI/ML (machine learning) model to generate quantized codewords); and
transmit the codewords (“the terminal device 110 transmits 340 the compressed CSI to the network device 120” – See [0201]; The UE transmits the encoded CSI feedback to the base station).
Although Guan discloses that the measurements are reference signal received power measurements (See [0058]), Guan does not explicitly teach that the measurements of the multiple RSs are Layer 1 reference signal received power measurements.
However, Shi teaches that the measurements of the multiple RSs are Layer 1 reference signal received power measurements (“User equipment (UE) measures a layer 1 reference signal received power (L1-RSRP)” – See [0093]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Guan such that the measurements of the multiple RSs are Layer 1 reference signal received power measurements since Layer 1 reference signal received power is well-known in the art for being used to represent channel information such as an optimal beam and its quality to the network (See Shi, [0093]).
Guan does not explicitly teach that the machine learning model is trained using the quantized versions of the measurements of the multiple RSs.
However, Saber teaches that the machine learning model is trained using the quantized versions of the measurements of the multiple RSs (“CSI compression performance may be improved using AI and/or ML” – See [0046]; “In some embodiments of a joint training framework, the training node may train one or both models using a corresponding quantizer and/or dequantizer function (e.g., an approximated and/or differentiable quantizer and/or dequantizer function). A node that receives a trained model may also receive and use the corresponding quantizer and/or dequantizer function for further training, validation, testing, inference, and/or the like” – See [0106]; “The encoded measurement may then be quantized by a quantizer” – See [0078]; The AI/ML (machine learning) model is trained using quantized versions of the measurements of the multiple RS).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Guan such that the machine learning model is trained using the quantized versions of the measurements of the multiple RSs. Motivation for doing so would be to provide joint training of models at the UE and the network entity, that results in models that are jointly matched to a target task so that performance can be improved/optimized (See Saber, [0107]).
Regarding Claim 2, Guan in view of Shi and Saber teaches the apparatus of Claim 1. Guan further teaches that the multiple RSs are channel state information RSs or synchronization signal blocks (“The set of RSs may be a set of CSI-RSs or any other suitable RSs” – See [0115]).
Regarding Claim 4, Guan in view of Shi and Saber teaches the apparatus of Claim 1. Shi further teaches that the one or more processors, to encode the measurements of the multiple RSs, are individually or collectively configured to cause the UE to encode differential measurements that are each a difference between a respective measurement of the measurements of the multiple RSs and a strongest measurement of the measurements of the multiple RSs, and wherein the one or more processors, to transmit the codewords, are individually or collectively configured to cause the UE to transmit the codewords with an indication of the strongest measurement (“When a feedback report of the terminal includes only one L1-RSRP, a 7-bit quantization method is used, a quantization step is 1 dB, and a quantization range is −140 dBm to −44 dBm. When an indicated feedback report of the terminal includes multiple L1-RSRPs, or a group-based beam report is enabled, 7-bit quantization is used for strongest RSRP quantization, a 4-bit differential quantization method is used for remaining RSRP quantization, and a quantization step is 2 dB” – See [0094]; The UE encodes differential measurements that are a difference between a respective value and a strongest value, and the strongest measurement value is encoded/indicated using a 7-bit quantization value).
Regarding Claim 6, Guan in view of Shi and Saber teaches the apparatus of Claim 1. Saber further teaches that the one or more processors are individually or collectively configured to cause the UE to train the machine learning model using the measurements of the multiple RSs as inputs and the quantized versions of the measurements of the multiple RSs as outputs, wherein the machine learning model comprises a compression machine learning model that is jointly trained with a decompression machine learning model (“In some embodiments of a joint training framework, the training node may train one or both models using a corresponding quantizer and/or dequantizer function (e.g., an approximated and/or differentiable quantizer and/or dequantizer function). A node that receives a trained model may also receive and use the corresponding quantizer and/or dequantizer function for further training, validation, testing, inference, and/or the like” – See [0106]; “The encoded measurement may then be quantized by a quantizer 519 and transmitted back to the gNB 502 as a UL signal 520 (e.g., a bitstream). In some embodiments, the description of a model at a node may also include a quantizer and/or dequantizer description, for example, a function that may map channel information (e.g., a real CSI codeword) at the output of an encoder model to quantized values or a bit stream, and vice versa at a decoder model at the other node. The gNB 502 may apply the received UL signal 520 to a dequantizer 521 to generate an equivalent feature vector which may be fed to a model 504 to extract information 522 (e.g., necessary or optional information) relating to the DL physical layer” – See [0078]; See also Fig. 5; The machine learning model is trained using the RS measurements as inputs, wherein quantized versions of the measurements are generated and output to the gNB, wherein the machine learning model is a joint training model comprising a compression machine learning model on the UE side that is jointly trained with a decompression machine learning model on the gNB side).
Regarding Claim 7, Guan in view of Shi and Saber teaches the apparatus of Claim 6. Guan further teaches that the one or more processors are individually or collectively configured to cause the UE to transmit one or more of an indication of an architecture of the machine learning model or parameters of the machine learning model (“Some of key parameters of the AI/ML model may include an input/output dimension, a compression method and a quantization method, etc.” – See [0031]; “the terminal device 110 may transmit 331 an indication of the compression method to the network device 120. In this way, the terminal device 110 may report the compression method applied or to be applied” – See [0190]; The UE transmits an indication of parameters of the ML model, such as a compression method).
Saber further teaches that the decompression machine learning model is trained using reconstructed RS measurements as outputs (“In some embodiments of a joint training framework, the training node may train one or both models using a corresponding quantizer and/or dequantizer function (e.g., an approximated and/or differentiable quantizer and/or dequantizer function). A node that receives a trained model may also receive and use the corresponding quantizer and/or dequantizer function for further training, validation, testing, inference, and/or the like” – See [0106]; “The encoded measurement may then be quantized by a quantizer 519 and transmitted back to the gNB 502 as a UL signal 520 (e.g., a bitstream). In some embodiments, the description of a model at a node may also include a quantizer and/or dequantizer description, for example, a function that may map channel information (e.g., a real CSI codeword) at the output of an encoder model to quantized values or a bit stream, and vice versa at a decoder model at the other node. The gNB 502 may apply the received UL signal 520 to a dequantizer 521 to generate an equivalent feature vector which may be fed to a model 504 to extract information 522 (e.g., necessary or optional information) relating to the DL physical layer” – See [0078]; See also Fig. 5; The decompression machine learning model on the gNB side is trained using the dequantized/reconstructed measurement values as outputs).
Regarding Claim 8, Guan in view of Shi and Saber teaches the apparatus of Claim 6. Guan further teaches that the one or more processors, to train the machine learning model, are individually or collectively configured to cause the UE to train machine learning models for different quantities of RSs (“Some of key parameters of the AI/ML model may include an input/output dimension, a compression method and a quantization method, etc., and may need to be aligned at both NW and UE side” – See [0031]; ML models are provided by varying the ML parameters such as input/output dimension, compression method, quantization method, etc. for different quantities of RSs).
Regarding Claim 9, Guan in view of Shi and Saber teaches the apparatus of Claim 1. Guan further teaches that the one or more processors are individually or collectively configured to cause the UE to receive, in a reporting configuration, one or more of an indication of an architecture of the machine learning model or parameters of the machine learning model, and wherein the one or more processors, to encode the measurements of the multiple RSs, are individually or collectively configured to cause the UE to encode the measurements of the multiple RSs using the architecture or the parameters (“the terminal device 110 may transmit 331 an indication of the compression method to the network device 120” – See [0190]; “the terminal device 110 may receive 333 a confirmation for the indication from the network device 120 and start 334 the compression of the CSI at a predetermined timing after the reception of the confirmation” – See [0192]; The UE receives a confirmation/indication of the compression method (architecture/parameters of the ML model) from the network device and performs the compression/encoding of the measurements using the indicated parameters).
Regarding Claim 10, Guan in view of Shi and Saber teaches the apparatus of Claim 1. Guan further teaches that the one or more processors are individually or collectively configured to cause the UE to receive a reporting configuration that indicates a reporting option from among multiple reporting options, and wherein the one or more processors, to encode the measurements of the multiple RSs, are individually or collectively configured to cause the UE to encode the measurements of the multiple RSs based at least in part on the reporting option (“A higher CR, less compressed bits and/or less quantization bits may be determined for a more frequent CSI measurement or report” – See [0144]; The compression ratio (reporting option) is determined based on a reporting frequency, such that the measurements are compressed/encoded with a higher CR, fewer compressed bits, and/or fewer quantized bits when reporting is performed frequently).
Regarding Claim 11, Guan in view of Shi and Saber teaches the apparatus of Claim 1. Guan further teaches that the one or more processors are individually or collectively configured to cause the UE to receive a reporting configuration that indicates multiple reporting options, and wherein the one or more processors, to encode the measurements of the multiple RSs, are individually or collectively configured to cause the UE to select a reporting option from among the multiple reporting options and encode the measurements of the multiple RSs based at least in part on the selected reporting option (“the network device 120 may configure a larger or smaller periodicity of CSI measurement or report for the terminal device 110” – See [0193]; “A higher CR, less compressed bits and/or less quantization bits may be determined for a more frequent CSI measurement or report” – See [0144]; “Some examples of the relationship between the compression method (e.g., compression parameter) and the time-domain information (e.g., time-domain parameter) are shown in Tables 1 to 4 below by taking a CR as an example of a compression parameter and a RS or report periodicity as an example of a time-domain parameter” – See [0131]; The UE receives a configuration/mapping that indicates multiple compression ratios (reporting options), wherein the UE selects one of the compression ratios and compresses/encodes the measurements based on the reporting option).
Regarding Claim 12, Guan in view of Shi and Saber teaches the apparatus of Claim 11. Guan further teaches that the one or more processors are individually or collectively configured to cause the UE to transmit an indication of the selected reporting option (“the terminal device 110 may explicitly report CR and implicitly report quantization information. In some embodiments, Table 18 below may be used” – See [0204]; The UE indicates the selected compression ratio (reporting option)).
Regarding Claim 13, Guan in view of Shi and Saber teaches the apparatus of Claim 11. Guan further teaches that the one or more processors, to select the reporting option, are individually or collectively configured to cause the UE to select the reporting option based at least in part on one or more of an accuracy of the quantized versions of the measurements or a payload size of the quantized versions of the measurements of the multiple RSs (“The terms “a CR” may refer to one of the following definitions: Definition 1: CR=output dimension of encoder/input dimension of encoder. For example, the input dimension of encoder may refer to the number of ports X the number of sub-bands, and the output dimension of encoder may refer to the number of compressed bits or the number of compressed bits/the number of quantization bits” – See [0038]-[0039]; The CR/reporting option is based on a number of quantized bits (payload size of the quantized measurements)).
Regarding Claim 14, Guan in view of Shi and Saber teaches the apparatus of Claim 11. Guan further teaches that the one or more processors, to select the reporting option, are individually or collectively configured to cause the UE to select the reporting option based at least in part on information in the reporting configuration or a rule in the reporting configuration (“Some examples of the relationship between the compression method (e.g., compression parameter) and the time-domain information (e.g., time-domain parameter) are shown in Tables 1 to 4 below by taking a CR as an example of a compression parameter and a RS or report periodicity as an example of a time-domain parameter” – See [0131]; The UE selects one of the compression ratios (reporting options) based on a mapping (information/rule) in the reporting option).
Regarding Claim 15, Guan in view of Shi and Saber teaches the apparatus of Claim 1. Guan further teaches that the one or more processors are individually or collectively configured to cause the UE to transmit an indication of a UE capability for using machine learning to generate codewords representing the quantized versions of the measurements of the multiple RSs (“he terminal device 110 may transfer 210 information of UE capability with the network device 120” – See [0107]; “The information of UE capability may comprise AI/ML capability on NW or UE side, for example, a supported AI/ML model, algorithm, size or dimension, computational power or complexity or time” – See [0108]; The UE transmits an indication of its capability for performing AI/ML encoding of the measurements).
Regarding Claim 16, Guan teaches an apparatus of a network entity for wireless communication, comprising: one or more memories; and one or more processors coupled to the one or more memories, the one or more processors configured to (“The device 700 can be considered as a further example implementation of … the network device 120” – See [0265]; “As shown, the device 700 includes a processor 710, a memory 720 coupled to the processor 710, a suitable transmitter (TX) and receiver (RX) 740 coupled to the processor 710, and a communication interface coupled to the TX/RX 740. The memory 710 stores at least a part of a program 730” – See [0266]; “The program 730 is assumed to include program instructions that, when executed by the associated processor 710, enable the device 700 to operate in accordance with the embodiments of the present disclosure” – See [0267]) cause the network entity to:
transmit a reporting configuration associated with using machine learning for quantizing measurements of multiple reference signals (RSs) (“the terminal device 110 may receive, from the network device 120, a configuration for CSI measurement and report via a DL control channel transmission” – See [0101]; “information configured by the RRC (re) configuration message may comprise a CSI report configuration and an AI/ML model configuration” – See [0109]; The network node (network entity) transmits a reporting configuration associated with ML (machine learning) for quantizing measurements of multiple RSs);
transmit the multiple RSs (“the terminal device 110 may receive 311 the set of RSs from the network device 120” – See [0115]; The network node transmits a set of RSs);
receive codewords that represent quantized versions of the measurements of the multiple RSs (“the terminal device 110 transmits 340 the compressed CSI to the network device 120” – See [0201]; The network node receives the encoded CSI feedback from the UE); and
decode the codewords to obtain the measurements of the multiple RSs (“A network device determines the compression method based on the configuration and recoveries CSI based on the compression method and the compressed CSI” – See [0035]; The network node decodes the codewords to recover the CSI/measurements of the multiple RSs).
Although Guan discloses that the measurements are reference signal received power measurements (See [0058]), Guan does not explicitly teach that the measurements of the multiple RSs are Layer 1 reference signal received power measurements.
However, Shi teaches that the measurements of the multiple RSs are Layer 1 reference signal received power measurements (“User equipment (UE) measures a layer 1 reference signal received power (L1-RSRP)” – See [0093]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Guan such that the measurements of the multiple RSs are Layer 1 reference signal received power measurements since Layer 1 reference signal received power is well-known in the art for being used to represent channel information such as an optimal beam and its quality to the network (See Shi, [0093]).
Guan does not explicitly teach that the machine learning model is trained using the quantized versions of the measurements of the multiple RSs.
However, Saber teaches that the machine learning model is trained using the quantized versions of the measurements of the multiple RSs (“CSI compression performance may be improved using AI and/or ML” – See [0046]; “In some embodiments of a joint training framework, the training node may train one or both models using a corresponding quantizer and/or dequantizer function (e.g., an approximated and/or differentiable quantizer and/or dequantizer function). A node that receives a trained model may also receive and use the corresponding quantizer and/or dequantizer function for further training, validation, testing, inference, and/or the like” – See [0106]; “The encoded measurement may then be quantized by a quantizer” – See [0078]; The AI/ML (machine learning) model is trained using quantized versions of the measurements of the multiple RS).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Guan such that the machine learning model is trained using the quantized versions of the measurements of the multiple RSs. Motivation for doing so would be to provide joint training of models at the UE and the network entity, that results in models that are jointly matched to a target task so that performance can be improved/optimized (See Saber, [0107]).
Claims 17 and 27 are rejected based on reasoning similar to Claim 4.
Claim 18 is rejected based on reasoning similar to Claim 6.
Claims 19 and 28 are rejected based on reasoning similar to Claim 9.
Claim 20 is rejected based on reasoning similar to Claim 8.
Claim 21 is rejected based on reasoning similar to Claim 7.
Claim 22 is rejected based on reasoning similar to Claim 10.
Claim 23 is rejected based on reasoning similar to Claim 10.
Claim 24 is rejected based on reasoning similar to Claim 14.
Claims 25 and 30 are rejected based on reasoning similar to Claim 15.
Claim 26 is rejected based on reasoning similar to Claim 1.
Claim 29 is rejected based on reasoning similar to Claim 16.
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Guan et al. (US 2025/0260449) in view of Shi (US 2025/0247733) and Saber et al. (US 2023/0131694) and further in view of Song (US 2023/0040739).
Regarding Claim 5, Guan in view of Shi and Saber teaches the apparatus of Claim 1. Although Guan discloses a quantity of RSs being plural/multiple RSs (See above with respect to claim 1, wherein Guan discloses a “set of RSs”), Guan does not explicitly teach that a quantity of the multiple RSs is greater than 4 RSs.
However, Song teaches that a quantity of the multiple RSs is greater than 4 RSs (“NT represents the number of CSI-RSs” – See [0083]; “in our example NT=8” – See [0103]; The quantity of CSI-RSs is 8 (i.e., greater than 4)).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Guan such that a quantity of the multiple RSs is greater than 4 RSs. Motivation for doing so would be to enable a UE in a massive MIMO system having a large number of antenna ports to provide CSI feedback with good performance (See Song, [0002]-[0003]).
Response to Arguments
On page 12 of the remarks, Applicant argues in substance that Guan and Shi do not teach “wherein the machine learning model is trained using the quantized version of the measurements of the multiple RSs,” as recited in independent claims 1, 16, 26, and 29. Applicant’s arguments have been considered but are moot based on the new grounds of rejection. In response to the amended limitations, the Examiner relies upon the newly-cited Saber reference.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Scott M Sciacca whose telephone number is (571)270-1919. The examiner can normally be reached Monday thru Friday, 7:30 A.M. - 5:00 P.M. EST.
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, Joseph Avellino can be reached at (571) 272-3905. 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.
/SCOTT M SCIACCA/ Primary Examiner, Art Unit 2478
Prosecution Insights