DETAILED ACTION
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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 06/14/2024, 10/26/2024 were filed. The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner
Claim Interpretation
During examination, claim terms are given their broadest reasonable interpretation consistent with the specification as that specification would be read by one of ordinary skill in the art (MPEP 2111). The constructions below govern the rejections that follow.
Artificial intelligence (AI) model. Claims 8 and 20 recite output layer nodes of the AI model, so an AI model is construed as an inference function whose results are produced at output nodes, that is, a neural network or a comparable trained model that maps input information to output information. No particular architecture, training regime or degree of autonomy is required, and a neural network encoder is such a model.
Association relationship. Claims 2, 14 and 26 define the term by three alternatives, an index of the resource set and an index of the model being carried in the same report configuration information, the two indexes satisfying a given relationship, or the two being indicated by the same association indication information. Any one of the three satisfies the term, so a correspondence conveyed to the terminal in the configuration that tells it which model to run is an association relationship.
The conditional and mathematical limitation of claims 3, 15 and 27. The limitation is cast as mutually exclusive conditional branches, keyed to whether the number N of first resource sets is greater than, equal to, or less than the number M of AI models. Only the branch whose condition is met governs a given configuration, and in claims 3 and 15, which are method claims, a step recited only within a branch whose condition is not met need not be carried out (MPEP 2111.04, second paragraph). The claim also supplies its own operators, reciting that ⌈ ⌉ indicates rounding up, and indicates rounding down. Read with those definitions, the recited series ⌊N/M⌋ · i through ⌊N/M⌋ · (i + 1) − 1, and the parallel series formed with ⌈N/M⌉, describe a contiguous run of resource-set numbers whose length is N divided by M and rounded; the expressions therefore recite nothing beyond dividing the N resource sets among the M models in contiguous blocks of as nearly equal size as the division allows. Where N equals M that division degenerates to one resource set for each model, which is the case the claim states in words rather than symbols, an i-th AI model is associated with an i-th first resource set. A configuration in which each model is supplied and selected for the measurements of one function accordingly falls within the claim.
Memory, transceiver and processor. Claims 25 and 37 recite a memory, a transceiver and a processor. These are structural terms rather than generic placeholders and the claims use no means language, so 35 U.S.C. 112(f) is not invoked and the limitations are given their plain structural meaning as a programmed hardware processor with its memory and transceiver.
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.
Claims 1-6, 8, 9, 13-18, 20, 21, 25-27 and 37 are rejected under 35 U.S.C. 103 as being unpatentable over Zeng (US 2021/0273706 A1) in view of Ryu (US 2021/0328630 A1).
Regarding claim 1, Zeng discloses: A method for measurement report, comprising: determining at least one first resource set and at least one artificial intelligence (AI) model, because Zeng teaches a user equipment that takes its feedback from the channel state information reference signal resources it observes and runs a neural network encoder over what those resources gave it, (i.e., “determining at least one first resource set” as claimed) so that the terminal settles both the set of resources it will measure and the model it will apply to them: (Zeng, para [0070] “for each CSI feedback instance, a network node (e.g., UE 115) may provide feedback of payload based on CSI-RS observations to a corresponding network node of a communication link”; Zeng, para [0059] “CSI encoders and decoders used by network nodes may implement channel compression/reconstruction based upon NN training of collected channels”).
Furthermore, Zeng discloses: wherein the first resource set is a measurement resource set, because Zeng teaches that what the model is run on is the channel the terminal estimated from the reference signals it observed, so the resources standing behind the report are measurement resources: (Zeng, para [0079] “using RS estimated channel as an input to generate low dimensional information”).
Moreover, Zeng discloses: determining report information based on the at least one first resource set and the at least one AI model, and transmitting the report information to a network device, wherein the report information comprises part or all of output information of the at least one AI model., because Zeng teaches that the estimated channel is passed through the terminal's neural network encoder and that the encoder's own output, the size reduced encoded payload, (i.e., “output information of the at least one AI model” as claimed) is what travels back to the base station as the feedback: (Zeng, para [0070] “may be encoded by an instance of CSI encoder 311 implemented by UE 115 and utilizing the above described encoder parameters (e.g., implementing neural-network based channel compression) to provide size reduced CSI encoded payload” . . . [0088] “the encoded CSI may, for example, comprise information regarding a reference signal observed by the first network node that is compressed by a CSI encoder using neural-network based channel compression, wherein the CSI information is information regarding the estimated channel other than quantized parameters determined from information regarding the estimated channel as observed by the first network node”)
Although Zeng teaches a terminal that measures a set of reference signal resources, runs a neural network encoder over the channel it estimated from them, and feeds the encoder's own output back to the base station as its report: (Zeng, para [0059], para [0070]), Zeng does not explicitly disclose that the model run on those resources is one of several artificial intelligence models and that it stands in an association relationship with the measurement resource set it is applied to.
However, Zeng in view of Ryu discloses having an association relationship with the at least one first resource set because Ryu teaches a base station that develops several artificial intelligence models for each of several functions, hands those models to the terminal, and then, on the strength of the very channel measurements the terminal makes and reports, tells the terminal which of the models to use for each functionality, (i.e., “having an association relationship with” as claimed) so that a given model stands in a settled correspondence with the measurement resources it is to be run on (Ryu, para [0053], “multiple neural network (NN), artificial intelligence (AI), or machine learning (ML) models may be generated for each of multiple different functions”; Ryu, para [0054], “the serving base station may provide the UE with a prioritized list of predictive models to be used by the UE for each functionality”).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to hold Zeng's neural network encoder as one of the several models Ryu's base station supplies, and to keep the correspondence Ryu already keeps between a model and the measurements it is chosen for. Zeng runs a single encoder over whatever the terminal observed and is silent on how that encoder is picked; Ryu addresses exactly that question, supplying several models per function and choosing among them from the terminal's own channel measurements, because a model fitted to the conditions the terminal is actually in predicts better than one that is not. Applying Ryu's model selection to Zeng's encoder is the use of a known technique to improve a comparable system in the same way, and it would have yielded no more than the predictable result of an encoder chosen to suit the resources it encodes. One of ordinary skill would have had a reasonable expectation of success, since the combination changes only which encoder the terminal loads and leaves Zeng's measurement, encoding and feedback path untouched.
Regarding claim 2, which depends on claim 1, Zeng in view of Ryu discloses The method of claim 1, wherein the association relationship is determined based on at least one of the following: an index of the at least one first resource set and an index of the at least one AI model are comprised in same report configuration information; an index of the at least one first resource set and an index of the at least one AI model satisfy a given relationship; or, an index of the at least one first resource set and an index of the at least one AI model are indicated by same association indication information, as Ryu further discloses the base station carrying, in the very configuration it sends the terminal, which of the models is to be used for which functionality, so that the tie between a resource set and a model is conveyed by that same configuration rather than worked out separately by the terminal (Ryu, para [0054], “the serving base station may provide the UE with a prioritized list of predictive models to be used by the UE for each functionality”).
Claim 2 recites its limitations in the alternative, in the form at least one of the following, so the claim is satisfied when any one of the listed items is satisfied. The alternative relied on is the recitation of an index of the at least one first resource set and an index of the at least one AI model are comprised in same report configuration information, which is the limitation mapped above. No reference is relied on as teaching the remaining alternatives, and none is needed to sustain the rejection. Accordingly, Zeng and Ryu are combined for the reasons set forth in the rejection of claim 1 above.
Regarding claim 3, which depends on claim 2, Zeng in view of Ryu discloses The method of claim 2, wherein a quantity of the at least one first resource set is N, a quantity of the at least one AI model is M, N and M are both positive integers, and the association relationship comprises, as Ryu in paragraph [0053] further discloses a base station that develops a number of artificial intelligence models for each of a number of functions and hands them to the terminal, so the terminal holds some number of models alongside the resource sets (“at least one first resource set is N” as claimed) (i.e., “a wireless device, a remote device, a handheld device, or a subscriber device” in para [0064]) it measures, each being a positive whole number (Ryu, para [0053], “multiple neural network (NN), artificial intelligence (AI), or machine learning (ML) models may be generated for each of multiple different functions . . . [0064] a wireless device, a remote device, a handheld device, or a subscriber device, or some other suitable terminology, where the “device” may also be referred to as a unit, a station, a terminal, or a client,
Zeng in view of Ryu discloses if N is equal to M, an i-th AI model is associated with an i-th first resource set and if N is equal to M, an AI model with index j is associated with a first resource set with index j, as Ryu further discloses a base station that builds its models function by function and then, once the terminal has measured the channel and reported what those measurements were, names for the terminal the model it is to use for each functionality, so that exactly one model stands designated for each function the terminal performs and the model so designated is the one run on the measurements taken for that function; the terminal carrying one measurement resource set for each such function, the quantity of first resource sets and the quantity of designated models are the same quantity (i.e., “an i-th AI model is associated with an i-th first resource set” as claimed) and each model is paired with the single resource set whose measurements it is applied to, which is the equal-quantity correspondence these branches recite (Ryu, para [0052] “base station may transmit a number of beams in a beam sweeping procedure, and a UE may measure received signals to identify a preferred beam, and the base station and UE may proceed to establish a beam pair link” . . . para [0053], “multiple neural network (NN), artificial intelligence (AI), or machine learning (ML) models may be generated for each of multiple different functions”; Ryu, para [0054], “the serving base station may provide the UE with a prioritized list of predictive models to be used by the UE for each functionality”). The claim states the equal-quantity case twice, once on an index i running from 1 to M and once on an index j running from 0 to M-1, which differs only in where the count begins; the one-to-one pairing just described satisfies each of them.
The correspondence just set out is the association relationship established in the rejection of claim 1 above and defined in the rejection of claim 2 above, now fixed as a quantity. Arriving at the equal-quantity configuration would have required no more than ordinary skill: Ryu designates a model for each functionality and Zeng reports on the resources it measures, so a terminal performing some number of functions holds one model and one resource set for each of them, and the two quantities come out equal in the ordinary course rather than by inventive selection. The claim admits of only three possibilities, N greater than, equal to, or less than M, so settling on the equal case is a selection from a finite number of identified, predictable solutions made with a reasonable expectation of success, which is a recognized rationale for a conclusion of obviousness (MPEP 2143(I)(E)), and the claim identifies no result that turns on which of the three is chosen. The association relationship is recited in claims 1, 2, 3, 13, 14, 15, 25, 26 and 27, each rejected on this same configuration, the network-side claims and the claims of a different statutory class carrying the construction forward without restating it.
The remaining branches of claim 3, if N is greater than M, an i-th AI model is associated with first resource sets numbered ⌊N/M⌋ · i, ⌊N/M⌋ · i + 1, …, ⌊N/M⌋ · (i + 1) − 1, or associated with first resource sets numbered ⌈N/M⌉ · i, ⌈N/M⌉ · i + 1, …, ⌈N/M⌉ · (i + 1) − 1, if N is less than M, an i-th first resource set is associated with AI models numbered ⌊M/N⌋ · i, ⌊M/N⌋ · i + 1, …, ⌊M/N⌋ · (i + 1) − 1, or associated with AI models numbered ⌈M/N⌉ · i, ⌈M/N⌉ · i + 1, …, ⌈M/N⌉ · (i + 1) − 1, or, if N is greater than M, an AI model with index j is associated with first resource sets with indexes of ⌊N/M⌋ · j, ⌊N/M⌋ · j + 1, …, ⌊N/M⌋ · (j + 1) − 1, or associated with first resource sets with indexes of ⌈N/M⌉ · j, ⌈N/M⌉ · j + 1, …, ⌈N/M⌉ · (j + 1) − 1 and if N is less than M, a first resource set with index j is associated with AI models with indexes of ⌊M/N⌋ · j, ⌊M/N⌋ · j + 1, …, ⌊M/N⌋ · (j + 1) − 1, or associated with AI models with indexes of ⌈M/N⌉ · j, ⌈M/N⌉ · j + 1, …, ⌈M/N⌉ · (j + 1) − 1, are conditional limitations and are not reached. Given their broadest reasonable interpretation in light of the specification, as set out in the Claim Interpretation section above, each of these branches is keyed to a condition, N greater than M or N less than M, that the configuration relied on does not satisfy, and a step recited only within a branch whose condition is not met need not be carried out. In the relied-on configuration the base station names a single model for each functionality the terminal performs (Ryu, para [0054], “the serving base station may provide the UE with a prioritized list of predictive models to be used by the UE for each functionality”), so N equals M and none of these branches is entered. It is noted, without reliance, that were either branch reached it would not distinguish the claim, Ryu contemplating grouping in both directions: a same function may have several different models provided to the terminal, which is the many-models-to-one-resource-set case, and a single model selection may rest on the measurements the terminal takes on a serving cell together with other cells it can receive, which is the many-resource-sets-to-one-model case (Ryu, para [0053], “for a same function (e.g., a beam prediction function to identify a transmit/receive beam for communications) may have multiple different models, which may be provided to the UE by a base station”; Ryu, para [0054], “a UE may measure a channel between the UE and a serving base station or cell, and optionally one or more other base stations or cells from which the UE can receive a signal, and use the channel measurements for model selection”). What these branches add to that grouping is the arithmetic by which the groups are numbered, which the claim defines for itself and which, as construed above, allocates the resource sets among the models in contiguous blocks of as nearly equal size as the division allows, an allocation calling for no more than ordinary skill once the grouping is contemplated.
Regarding the recitation wherein ⌈ ⌉ indicates rounding up, ⌊ ⌋ indicates rounding down, i is an integer from 1 to M, and j is an integer from 0 to M-1, this limitation is definitional. It fixes the meaning of the rounding operators and the ranges of the indices i and j, and imposes no requirement on the association relationship beyond the one construed above.
Thus, Zeng and Ryu are combined for the reasons set forth in the rejection of claim 1 above.
Regarding claim 4, which depends on claim 2, Zeng in view of Ryu discloses The method of claim 2, wherein the association indication information is used to indicate indexes of one or more AI models associated with each first resource set, and/or the association indication information is used to indicate one or more first resource sets associated with each AI model, as Ryu further discloses the prioritized list the base station sends naming, for each functionality, which model the terminal is to use, so that the indication itself is what points from a resource set to the model or models that go with it (Ryu, para [0054], “the serving base station may provide the UE with a prioritized list of predictive models to be used by the UE for each functionality”).
Claim 4 recites its limitations in the alternative, joined by and/or, so the claim is satisfied when either of them is satisfied. The alternative relied on is the recitation of the association indication information is used to indicate indexes of one or more AI models associated with each first resource set, which is the limitation mapped above. No reference is relied on as teaching the remaining alternatives, and none is needed to sustain the rejection.
Consequently, Zeng and Ryu are combined for the reasons set forth in the rejection of claim 1 above.
Regarding claim 5, which depends on claim 1, Zeng in view of Ryu discloses The method of claim 1, wherein determining the report information based on the at least one first resource set and the at least one AI model comprises: determining the report information based on the at least one first resource set and one or more AI models in the at least one AI model, as Ryu further discloses the terminal being given a number of models and determining which of them to run, so the report is produced from the resource set together with one or more of the models it holds (Ryu, para [0055], “the UE may receive a number of models for a number of functions, and the UE may determine which to select for communications”).
For these reasons Zeng and Ryu are combined for the reasons set forth in the rejection of claim 1 above.
Regarding claim 6, which depends on claim 5, Zeng discloses The method of claim 5, wherein determining the report information comprises: determining the report information based on the at least one first resource set and one AI model in the at least one AI model, wherein an input of the one AI model comprises part or all of first resource sets in the at least one first resource set, the report information comprises part or all of output information of the one AI model, as Zeng further discloses a terminal that runs a single neural network encoder over the channel it estimated from the resource sets it observed and returns that encoder's own output as its report, which is the report information determined from the resource set and the one model (Zeng, para [0070] “may be encoded by an instance of CSI encoder 311 implemented by UE 115 and utilizing the above described encoder parameters (e.g., implementing neural-network based channel compression) to provide size reduced CSI encoded payload”).
Zeng in view of Ryu discloses the one AI model is selected by a terminal device or determined based on model indication information, or the one AI model is the only AI model comprised in the at least one AI model; or, as Ryu further discloses the terminal being handed a number of models and either determining for itself which to run or following the list the base station indicates, which is the model being selected by the terminal or determined from model indication information (Ryu, para [0055], “the UE may receive a number of models for a number of functions, and the UE may determine which to select for communications”; Ryu, para [0054], “the serving base station may provide the UE with a prioritized list of predictive models to be used by the UE for each functionality”).
Regarding the recitation determining the report information based on the at least one first resource set and multiple AI models in the at least one AI model, wherein an input of each AI model among the multiple AI models comprises part or all of first resource sets in the at least one first resource set, and the report information comprises part or all of output information of the each AI model; or, this alternative need not be reached. Claim 6 sets out its alternatives in the disjunctive, each closing with the word or, so the claim is satisfied when any one of them is satisfied, and the first alternative is satisfied as set out above. It is noted, without reliance, that the recited plurality of models would not distinguish the claim, Ryu disclosing that a same function may have several different models which the base station provides to the terminal (Ryu, para [0053], “for a same function (e.g., a beam prediction function to identify a transmit/receive beam for communications) may have multiple different models, which may be provided to the UE by a base station”). No mapping is asserted for this alternative.
Likewise, the recitation an input of a first AI model among the multiple AI models comprises part or all of first resource sets in the at least one first resource set, an input of an (m + n)-th AI model comprises part or all of output information of an m-th to an (m + n - 1)-th AI models, and the report information comprises part or all of output information of a last AI model among the multiple AI models, m = 1, 2,…M-1, n is a predefined or configured positive integer, m + n is an integer from 2 to M, and M is a quantity of the multiple AI models is a further alternative of the same disjunctive recitation and need not be reached. No reference is relied on as teaching the cascade of models it recites.
Zeng in view of Ryu discloses wherein the model indication information and/or the input information and the output information of each AI model is determined based on at least one of the following: pre-definition in a protocol; a radio resource control (RRC) message configuration; a media access control - control element (MAC CE) indication; or a downlink control information (DCI) indication, as Ryu further discloses the base station signalling to the terminal, in the configuration it sends, which model is to be used for each functionality, so the model indication information reaches the terminal by network configuration (Ryu, para [0054], “the serving base station may provide the UE with a prioritized list of predictive models to be used by the UE for each functionality”).
Therefore Zeng and Ryu are combined for the reasons set forth in the rejection of claim 1 above.
Regarding claim 8, which depends on claim 1, Zeng discloses The method of claim 1, wherein the output information of the AI model in the report information comprises at least one of the following: first channel state information of the first resource set; second channel state information of a second resource set, wherein the second resource set corresponds to the output information of the AI model in the report information; identifiers of part of output layer nodes of the AI model and/or output values of the part of output layer nodes, wherein the identifiers and/or output values of the part of output layer nodes satisfy a given condition; or output values of all of output layer nodes of the AI model, as Zeng further discloses the encoded payload the terminal returns being the channel state information of the very resources it observed, carried in the encoder's output rather than as a conventionally computed measurement quantity (Zeng, para [0070] “may be encoded by an instance of CSI encoder 311 implemented by UE 115 and utilizing the above described encoder parameters (e.g., implementing neural-network based channel compression) to provide size reduced CSI encoded payload”).
Claim 8 recites its limitations in the alternative, in the form at least one of the following, so the claim is satisfied when any one of the listed items is satisfied. The alternative relied on is the recitation of first channel state information of the first resource set, which is the limitation mapped above. No reference is relied on as teaching the remaining alternatives, and none is needed to sustain the rejection.
Thus Zeng and Ryu are combined for the reasons set forth in the rejection of claim 1 above.
Regarding claim 9, which depends on claim 8, Zeng discloses The method of claim 8, wherein the second resource set is different from the first resource set in at least one of the following: a type of resource set is different; a type of reference signal is different; a reference signal is different; a time domain resource for transmission is different; a frequency domain resource for transmission is different; a transmission port is different; a transmission beam is different; a reception beam is different; the first resource set and the second resource set are independently configured; or, the first resource set is a subset of the second resource set; and/or, the second channel state information of the second resource set comprises at least one of the following: identifiers of K resources satisfying the given condition in the second resource set; confidence or probability corresponding to K resources; inference reference signal received power (RSRP) corresponding to K resources; inference reference signal received quality (RSRQ) corresponding to K resources; or, inference signal to interference and noise ratio (SINR) corresponding to K resources; wherein K is a predefined or configured positive integer, and a value of K does not exceed a quantity of the second resource sets; and/or, the first channel state information of the first resource set comprises at least one of the following: a channel estimation result obtained based on the at least one first resource set; channel estimation result compression information obtained based on the at least one first resource set; a channel estimation result and channel estimation result compression information obtained based on the at least one first resource set; or, information for channel estimation or channel recovery corresponding to a resource in the first resource set, as Zeng further discloses the terminal's report carrying the compressed form of the channel estimation result it obtained from the observed reference signal resources, which is channel estimation result compression information obtained based on that resource set (Zeng, para [0070] “may be encoded by an instance of CSI encoder 311 implemented by UE 115 and utilizing the above described encoder parameters (e.g., implementing neural-network based channel compression) to provide size reduced CSI encoded payload”; Zeng, para [0079] “using RS estimated channel as an input to generate low dimensional information”).
Claim 9 recites its limitations in the alternative throughout, its groups joined by and/or and each group introduced by at least one of the following, so the claim is satisfied when any one of the listed items is satisfied. The alternative relied on is the recitation of channel estimation result compression information obtained based on the at least one first resource set, which is the limitation mapped above. No reference is relied on as teaching the remaining alternatives, and none is needed to sustain the rejection.
Consequently Zeng and Ryu are combined for the reasons set forth in the rejection of claim 1 above.
Regarding claim 13, Zeng discloses: A method for measurement report, comprising: receiving report information transmitted from a terminal device, wherein the report information is determined based on at least one first resource set and at least one artificial intelligence (AI) model, because Zeng teaches the same exchange seen from the network end, where the base station is the corresponding node that receives the feedback payload the terminal produced from its reference signal observations by way of the neural network encoder: (Zeng, para [0070] “for each CSI feedback instance, a network node (e.g., UE 115) may provide feedback of payload based on CSI-RS observations to a corresponding network node of a communication link”).
Furthermore, Zeng discloses: the report information comprises part or all of output information of the at least one AI model, and the first resource set is a measurement resource set., because Zeng teaches that the payload arriving at the base station is the encoder's own output, produced from the channel the terminal estimated on the reference signals it measured: (Zeng, para [0070] “may be encoded by an instance of CSI encoder 311 implemented by UE 115 and utilizing the above described encoder parameters (e.g., implementing neural-network based channel compression) to provide size reduced CSI encoded payload”; Zeng, para [0079] “using RS estimated channel as an input to generate low dimensional information”).
Although Zeng teaches a terminal that measures a set of reference signal resources, runs a neural network encoder over the channel it estimated from them, and feeds the encoder's own output back to the base station as its report: (Zeng, para [0059], para [0070]), Zeng does not explicitly disclose that the model run on those resources is one of several artificial intelligence models and that it stands in an association relationship with the measurement resource set it is applied to.
Nonetheless, Zeng in view of Ryu discloses the at least one first resource set and the at least one AI model have an association relationship because Ryu teaches a base station that develops several artificial intelligence models for each of several functions, hands those models to the terminal, and then, on the strength of the very channel measurements the terminal makes and reports, tells the terminal which of the models to use for each functionality, so that a given model stands in a settled correspondence with the measurement resources it is to be run on (Ryu, para [0053], “multiple neural network (NN), artificial intelligence (AI), or machine learning (ML) models may be generated for each of multiple different functions”; Ryu, para [0054], “the serving base station may provide the UE with a prioritized list of predictive models to be used by the UE for each functionality”).
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to hold Zeng's neural network encoder as one of the several models Ryu's base station supplies, and to keep the correspondence Ryu already keeps between a model and the measurements it is chosen for. Zeng runs a single encoder over whatever the terminal observed and is silent on how that encoder is picked; Ryu addresses exactly that question, supplying several models per function and choosing among them from the terminal's own channel measurements, because a model fitted to the conditions the terminal is actually in predicts better than one that is not. Applying Ryu's model selection to Zeng's encoder is the use of a known technique to improve a comparable system in the same way, and it would have yielded no more than the predictable result of an encoder chosen to suit the resources it encodes. One of ordinary skill would have had a reasonable expectation of success, since the combination changes only which encoder the terminal loads and leaves Zeng's measurement, encoding and feedback path untouched.
Regarding claim 14, which depends on claim 13, Zeng in view of Ryu discloses The method of claim 13, wherein the association relationship is determined based on at least one of the following: an index of the at least one first resource set and an index of the at least one AI model are comprised in same report configuration information; an index of the at least one first resource set and an index of the at least one AI model satisfy a given relationship; or, an index of the at least one first resource set and an index of the at least one AI model are indicated by same association indication information, as Ryu further discloses the base station carrying, in the very configuration it sends the terminal, which of the models is to be used for which functionality, so that the tie between a resource set and a model is conveyed by that same configuration rather than worked out separately by the terminal (Ryu, para [0054], “the serving base station may provide the UE with a prioritized list of predictive models to be used by the UE for each functionality”).
Claim 14 recites its limitations in the alternative, in the form at least one of the following, so the claim is satisfied when any one of the listed items is satisfied. The alternative relied on is the recitation of an index of the at least one first resource set and an index of the at least one AI model are comprised in same report configuration information, which is the limitation mapped above. No reference is relied on as teaching the remaining alternatives, and none is needed to sustain the rejection.
Consequently Zeng and Ryu are combined for the reasons set forth in the rejection of claim 13 above.
Regarding claim 15, which depends on claim 14, Zeng in view of Ryu discloses The method of claim 14, wherein a quantity of the at least one first resource set is N, a quantity of the at least one AI model is M, N and M are both positive integers, and the association relationship comprises, as Ryu further discloses a base station that develops a number of artificial intelligence models for each of a number of functions and hands them to the terminal, so the terminal holds some number of models alongside the resource sets it measures, each being a positive whole number (Ryu, para [0053], “multiple neural network (NN), artificial intelligence (AI), or machine learning (ML) models may be generated for each of multiple different functions”).
Zeng in view of Ryu discloses if N is equal to M, an i-th AI model is associated with an i-th first resource set and if N is equal to M, an AI model with index j is associated with a first resource set with index j, as Ryu further discloses a base station that builds its models function by function and then, once the terminal has measured the channel and reported what those measurements were, names for the terminal the model it is to use for each functionality, so that exactly one model stands designated for each function the terminal performs and the model so designated is the one run on the measurements taken for that function; the terminal carrying one measurement resource set for each such function, the quantity of first resource sets and the quantity of designated models are the same quantity and each model is paired with the single resource set whose measurements it is applied to, which is the equal-quantity correspondence these branches recite (Ryu, para [0053], “multiple neural network (NN), artificial intelligence (AI), or machine learning (ML) models may be generated for each of multiple different functions”; Ryu, para [0054], “the serving base station may provide the UE with a prioritized list of predictive models to be used by the UE for each functionality”). The claim states the equal-quantity case twice, once on an index i running from 1 to M and once on an index j running from 0 to M-1, which differ only in where the count begins; the one-to-one pairing just described satisfies each of them.
The correspondence just set out is the association relationship established in the rejection of claim 13 above and defined in the rejection of claim 14 above, now fixed as a quantity. Arriving at the equal-quantity configuration would have required no more than ordinary skill: Ryu designates a model for each functionality and Zeng reports on the resources it measures, so a terminal performing some number of functions holds one model and one resource set for each of them, and the two quantities come out equal in the ordinary course rather than by inventive selection. The claim admits of only three possibilities, N greater than, equal to, or less than M, so settling on the equal case is a selection from a finite number of identified, predictable solutions made with a reasonable expectation of success, which is a recognized rationale for a conclusion of obviousness (MPEP 2143(I)(E)), and the claim identifies no result that turns on which of the three is chosen. The association relationship is recited in claims 1, 2, 3, 13, 14, 15, 25, 26 and 27, each rejected on this same configuration, the network-side claims and the claims of a different statutory class carrying the construction forward without restating it.
The remaining branches of claim 15, if N is greater than M, an i-th AI model is associated with first resource sets numbered ⌊N/M⌋ · i, ⌊N/M⌋ · i + 1, …, ⌊N/M⌋ · (i + 1) − 1, or associated with first resource sets numbered ⌈N/M⌉ · i, ⌈N/M⌉ · i + 1, …, ⌈N/M⌉ · (i + 1) − 1, if N is less than M, an i-th first resource set is associated with AI models numbered ⌊M/N⌋ · i, ⌊M/N⌋ · i + 1, …, ⌊M/N⌋ · (i + 1) − 1, or associated with AI models numbered ⌈M/N⌉ · i, ⌈M/N⌉ · i + 1, …, ⌈M/N⌉ · (i + 1) − 1, or, if N is greater than M, an AI model with index j is associated with first resource sets with indexes of ⌊N/M⌋ · j, ⌊N/M⌋ · j + 1, …, ⌊N/M⌋ · (j + 1) − 1, or associated with first resource sets with indexes of ⌈N/M⌉ · j, ⌈N/M⌉ · j + 1, …, ⌈N/M⌉ · (j + 1) − 1 and if N is less than M, a first resource set with index j is associated with AI models with indexes of ⌊M/N⌋ · j, ⌊M/N⌋ · j + 1, …, ⌊M/N⌋ · (j + 1) − 1, or associated with AI models with indexes of ⌈M/N⌉ · j, ⌈M/N⌉ · j + 1, …, ⌈M/N⌉ · (j + 1) − 1, are conditional limitations and are not reached. Given their broadest reasonable interpretation in light of the specification, as set out in the Claim Interpretation section above, each of these branches is keyed to a condition, N greater than M or N less than M, that the configuration relied on does not satisfy, and a step recited only within a branch whose condition is not met need not be carried out. In the relied-on configuration the base station names a single model for each functionality the terminal performs (Ryu, para [0054], “the serving base station may provide the UE with a prioritized list of predictive models to be used by the UE for each functionality”), so N equals M and none of these branches is entered. No reference is relied on as teaching them. It is noted, without reliance, that were either branch reached it would not distinguish the claim, Ryu contemplating grouping in both directions: a same function may have several different models provided to the terminal, which is the many-models-to-one-resource-set case, and a single model selection may rest on the measurements the terminal takes on a serving cell together with other cells it can receive, which is the many-resource-sets-to-one-model case (Ryu, para [0053], “for a same function (e.g., a beam prediction function to identify a transmit/receive beam for communications) may have multiple different models, which may be provided to the UE by a base station”; Ryu, para [0054], “a UE may measure a channel between the UE and a serving base station or cell, and optionally one or more other base stations or cells from which the UE can receive a signal, and use the channel measurements for model selection”). What these branches add to that grouping is the arithmetic by which the groups are numbered, which the claim defines for itself and which, as construed above, allocates the resource sets among the models in contiguous blocks of as nearly equal size as the division allows, an allocation calling for no more than ordinary skill once the grouping is contemplated.
Regarding the recitation wherein ⌈ ⌉ indicates rounding up, ⌊ ⌋ indicates rounding down, i is an integer from 1 to M, and j is an integer from 0 to M-1, this limitation is definitional. It fixes the meaning of the rounding operators and the ranges of the indices i and j, and imposes no requirement on the association relationship beyond the one construed above.
For these reasons Zeng and Ryu are combined for the reasons set forth in the rejection of claim 13 above.
Regarding claim 16, which depends on claim 14, Zeng in view of Ryu discloses The method of claim 14, wherein the association indication information is used to indicate indexes of one or more AI models associated with each first resource set, and/or the association indication information is used to indicate one or more first resource sets associated with each AI model, as Ryu further discloses the prioritized list the base station sends naming, for each functionality, which model the terminal is to use, so that the indication itself is what points from a resource set to the model or models that go with it (Ryu, para [0054], “the serving base station may provide the UE with a prioritized list of predictive models to be used by the UE for each functionality”).
Claim 16 recites its limitations in the alternative, joined by and/or, so the claim is satisfied when either of them is satisfied. The alternative relied on is the recitation of the association indication information is used to indicate indexes of one or more AI models associated with each first resource set, which is the limitation mapped above. No reference is relied on as teaching the remaining alternatives, and none is needed to sustain the rejection.
Therefore Zeng and Ryu are combined for the reasons set forth in the rejection of claim 13 above.
Regarding claim 17, which depends on claim 13, Zeng in view of Ryu discloses The method of claim 13, wherein determining the report information based on the at least one first resource set and the at least one AI model comprises: determining the report information based on the at least one first resource set and one or more AI models in the at least one AI model, as Ryu further discloses the terminal being given a number of models and determining which of them to run, so the report is produced from the resource set together with one or more of the models it holds (Ryu, para [0055], “the UE may receive a number of models for a number of functions, and the UE may determine which to select for communications”).
Accordingly Zeng and Ryu are combined for the reasons set forth in the rejection of claim 13 above.
Regarding claim 18, which depends on claim 13, Zeng discloses The method of claim 13, wherein determining the report information comprises: determining the report information based on the at least one first resource set and one AI model in the at least one AI model, wherein an input of the one AI model comprises part or all of first resource sets in the at least one first resource set, the report information comprises part or all of output information of the one AI model, as Zeng further discloses a terminal that runs a single neural network encoder over the channel it estimated from the resource sets it observed and returns that encoder's own output as its report, which is the report information determined from the resource set and the one model (Zeng, para [0070] “may be encoded by an instance of CSI encoder 311 implemented by UE 115 and utilizing the above described encoder parameters (e.g., implementing neural-network based channel compression) to provide size reduced CSI encoded payload”).
Zeng in view of Ryu discloses the one AI model is selected by the terminal device or determined based on model indication information, or the one AI model is the only AI model comprised in the at least one AI model; or, as Ryu further discloses the terminal being handed a number of models and either determining for itself which to run or following the list the base station indicates, which is the model being selected by the terminal or determined from model indication information (Ryu, para [0055], “the UE may receive a number of models for a number of functions, and the UE may determine which to select for communications”; Ryu, para [0054], “the serving base station may provide the UE with a prioritized list of predictive models to be used by the UE for each functionality”).
Regarding the recitation determining the report information based on the at least one first resource set and multiple AI models in the at least one AI model, wherein an input of each AI model among the multiple AI models comprises part or all of first resource sets in the at least one first resource set, and the report information comprises part or all of output information of the each AI model; or, this alternative need not be reached. Claim 18 sets out its alternatives in the disjunctive, each closing with the word or, so the claim is satisfied when any one of them is satisfied, and the first alternative is satisfied as set out above. It is noted, without reliance, that the recited plurality of models would not distinguish the claim, Ryu disclosing that a same function may have several different models which the base station provides to the terminal (Ryu, para [0053], “for a same function (e.g., a beam prediction function to identify a transmit/receive beam for communications) may have multiple different models, which may be provided to the UE by a base station”). No mapping is asserted for this alternative.
Likewise, the recitation an input of a first AI model among the multiple AI models comprises part or all of first resource sets in the at least one first resource set, an input of an (m + n)-th AI model comprises part or all of output information of an m-th to an (m + n - 1)-th AI models, and the report information comprises part or all of output information of a last AI model among the multiple AI models, m = 1, 2,…M-1, n is a predefined or configured positive integer, m + n is an integer from 2 to M, and M is a quantity of the multiple AI models is a further alternative of the same disjunctive recitation and need not be reached. No reference is relied on as teaching the cascade of models it recites.
Zeng in view of Ryu discloses wherein the model indication information and/or the input information and the output information of each AI model is determined based on at least one of the following: pre-definition in a protocol; a radio resource control (RRC) message configuration; a media access control - control element (MAC CE) indication; or a downlink control information (DCI) indication, as Ryu further discloses the base station signalling to the terminal, in the configuration it sends, which model is to be used for each functionality, so the model indication information reaches the terminal by network configuration (Ryu, para [0054], “the serving base station may provide the UE with a prioritized list of predictive models to be used by the UE for each functionality”). Thus, Zeng and Ryu are combined for the reasons set forth in the rejection of claim 13 above.
Regarding claim 20, which depends on claim 13, Zeng discloses The method of claim 13, wherein the output information of the AI model in the report information comprises at least one of the following: first channel state information of the first resource set; second channel state information of a second resource set, wherein the second resource set corresponds to the output information of the AI model in the report information; identifiers of part of output layer nodes of the AI model and/or output values of the part of output layer nodes, wherein the identifiers and/or output values of the part of output layer nodes satisfy a given condition; or output values of all of output layer nodes of the AI model, as Zeng further discloses the encoded payload the terminal returns being the channel state information of the very resources it observed, carried in the encoder's output rather than as a conventionally computed measurement quantity (Zeng, para [0070] “may be encoded by an instance of CSI encoder 311 implemented by UE 115 and utilizing the above described encoder parameters (e.g., implementing neural-network based channel compression) to provide size reduced CSI encoded payload”).
Claim 20 recites its limitations in the alternative, in the form at least one of the following, so the claim is satisfied when any one of the listed items is satisfied. The alternative relied on is the recitation of first channel state information of the first resource set, which is the limitation mapped above. No reference is relied on as teaching the remaining alternatives, and none is needed to sustain the rejection.
For these reasons Zeng and Ryu are combined for the reasons set forth in the rejection of claim 13 above.
Regarding claim 21, which depends on claim 20, Zeng discloses The method of claim 20, wherein the second resource set is different from the first resource set in at least one of the following: a type of resource set is different; a type of reference signal is different; a reference signal is different; a time domain resource for transmission is different; a frequency domain resource for transmission is different; a transmission port is different; a transmission beam is different; a reception beam is different; the first resource set and the second resource set are independently configured; or, the first resource set is a subset of the second resource set; and/or, the second channel state information of the second resource set comprises at least one of the following: identifiers of K resources satisfying the given condition in the second resource set; confidence or probability corresponding to K resources; inference reference signal received power (RSRP) corresponding to K resources; inference reference signal received quality (RSRQ) corresponding to K resources; or, inference signal to interference and noise ratio (SINR) corresponding to K resources; wherein K is a predefined or configured positive integer, and a value of K does not exceed a quantity of the second resource sets; and/or, the first channel state information of the first resource set comprises at least one of the following: a channel estimation result obtained based on the at least one first resource set; channel estimation result compression information obtained based on the at least one first resource set; a channel estimation result and channel estimation result compression information obtained based on the at least one first resource set; or, information for channel estimation or channel recovery corresponding to a resource in the first resource set, as Zeng further discloses the terminal's report carrying the compressed form of the channel estimation result it obtained from the observed reference signal resources, which is channel estimation result compression information obtained based on that resource set (Zeng, para [0070] “may be encoded by an instance of CSI encoder 311 implemented by UE 115 and utilizing the above described encoder parameters (e.g., implementing neural-network based channel compression) to provide size reduced CSI encoded payload”; Zeng, para [0079] “using RS estimated channel as an input to generate low dimensional information”).
Claim 21 recites its limitations in the alternative throughout, its groups joined by and/or and each group introduced by at least one of the following, so the claim is satisfied when any one of the listed items is satisfied. The alternative relied on is the recitation of channel estimation result compression information obtained based on the at least one first resource set, which is the limitation mapped above. No reference is relied on as teaching the remaining alternatives, and none is needed to sustain the rejection.
Therefore, Zeng and Ryu are combined for the reasons set forth in the rejection of claim 13 above.
Regarding claim 25, the claim recites: A terminal device, comprising a memory, a transceiver and a processor, wherein the memory is used for storing a computer program, the transceiver is used for transmitting and receiving data under control of the processor, and the processor is used for reading the computer program in the memory and performing the following operations: determining at least one first resource set and at least one artificial intelligence (AI) model having an association relationship with the at least one first resource set, wherein the first resource set is a measurement resource set; and determining report information based on the at least one first resource set and the at least one AI model, and transmitting the report information to a network device, wherein the report information comprises part or all of output information of the at least one AI model. Claim 25 is analogous to claim 1 and is rejected for the same reasons.
Regarding claim 26, which depends on claim 25, the claim recites: The terminal device of claim 25, wherein the association relationship is determined based on at least one of the following: an index of the at least one first resource set and an index of the at least one AI model are comprised in same report configuration information; an index of the at least one first resource set and an index of the at least one AI model satisfy a given relationship; or, an index of the at least one first resource set and an index of the at least one AI model are indicated by same association indication information. Claim 26 is analogous to claim 2 and is rejected for the same reasons.
Regarding claim 27, which depends on claim 26, the claim recites: The terminal device of claim 26, wherein a quantity of the at least one first resource set is N, a quantity of the at least one AI model is M, N and M are both positive integers, and the association relationship comprises: if N is greater than M, an i-th AI model is associated with first resource sets numbered ⌊N/M⌋ · i, ⌊N/M⌋ · i + 1, …, ⌊N/M⌋ · (i + 1) − 1, or associated with first resource sets numbered ⌈N/M⌉ · i, ⌈N/M⌉ · i + 1, …, ⌈N/M⌉ · (i + 1) − 1; if N is equal to M, an i-th AI model is associated with an i-th first resource set; if N is less than M, an i-th first resource set is associated with AI models numbered ⌊M/N⌋ · i, ⌊M/N⌋ · i + 1, …, ⌊M/N⌋ · (i + 1) − 1, or associated with AI models numbered ⌈M/N⌉ · i, ⌈M/N⌉ · i + 1, …, ⌈M/N⌉ · (i + 1) − 1; or, if N is greater than M, an AI model with index j is associated with first resource sets with indexes of ⌊N/M⌋ · j, ⌊N/M⌋ · j + 1, …, ⌊N/M⌋ · (j + 1) − 1, or associated with first resource sets with indexes of ⌈N/M⌉ · j, ⌈N/M⌉ · j + 1, …, ⌈N/M⌉ · (j + 1) − 1; if N is equal to M, an AI model with index j is associated with a first resource set with index j; if N is less than M, a first resource set with index j is associated with AI models with indexes of ⌊M/N⌋ · j, ⌊M/N⌋ · j + 1, …, ⌊M/N⌋ · (j + 1) − 1, or associated with AI models with indexes of ⌈M/N⌉ · j, ⌈M/N⌉ · j + 1, …, ⌈M/N⌉ · (j + 1) − 1; wherein ⌈ ⌉ indicates rounding up, ⌊ ⌋ indicates rounding down, i is an integer from 1 to M, and j is an integer from 0 to M-1. Claim 27 is analogous to claim 3 and is rejected for the same reasons.
Regarding claim 37, the claim recites: A network device, comprising a memory, a transceiver and a processor, wherein the memory is used for storing a computer program, the transceiver is used for transmitting and receiving data under control of the processor, and the processor is used for reading the computer program in the memory and performing the method of claim 13. Claim 37 is analogous to claim 13 and is rejected for the same reasons.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to CHONGSUH (John) PARK whose telephone number is 408-918-7574. The examiner can normally be reached Monday - Friday 8:00-5:30 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, Avellino, Joseph 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.
/CHONGSUH PARK/Examiner, Art Unit 2478
/JOSEPH E AVELLINO/Supervisory Patent Examiner, Art Unit 2478
Prosecution Insights