DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application is being examined under the pre-AIA first to invent provisions.
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.
Status of the Claims
Claims 1-10 filed on 13 FEB 2025 are currently pending and have been examined.
Priority
The pending application 18/435,902, filed on 13 FEB 2025, claims priority from foreign application CN202311006764.7, filed on 8 AUG 2023.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 25 FEB 2026 has been considered by the examiner.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 3-10 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding claim 3, the equation includes an “H” superscript that is not defined. It is unclear to the examiner if the “H” superscript is intended to represent a variable or an operation to be performed. For the purpose of prosecution, the “H” has been interpreted as a transpose operation.
Claims 4-10 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being depending on rejected claim 3 and for failing to cure the deficiencies listed above.
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 1-10 is/are rejected under 35 U.S.C. 103 as being unpatentable over Ma et al. (NPL “Deep Learning Assisted mmWave Beam Prediction for Heterogeneous Networks: A Dual-Band Fusion Approach,” cited by applicant in IDS filed 25 FEB 2026) in view of Khan et al. (NPL “Machine Learning for Millimeter Wave and Terahertz Beam Management: A Survey and Open Challenges”).
Regarding claim 1, Ma et al. discloses:
[Note: what is not explicitly taught by Ma et al. has been struck-through]
A millimeter-wave beam tracking method in a microwave and millimeter wave heterogeneous network scenario (Ma et al. “we propose to fuse sub-6 GHz channel information and mmWave low-overhead measurement to predict the optimal mmWave beam in heterogeneous networks (HetNets) and reduce the overhead of both mmWave BS selection and beam training.” – abstract), comprising:
step S1: establishing a millimeter-wave beam tracking problem in the microwave and millimeter wave heterogeneous network scenario (Ma et al. III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, A. Motivation and Problem Formulation, p. 118-119);
step S2: for the millimeter-wave beam tracking problem, constructing a deep neural network (DNN) fusion model (Ma et al. “we propose to adopt deep learning to extract the complex relationship between sub-6 GHz CSI and mmWave low-overhead measurement…” – 1. Introduction, p. 116-117)
step S3: pre-training the DNN fusion model by using training data (Ma et al. “In the training stage, the training dataset is constructed for optimizing the deep learning model, where each sample comprises the sub-6 GHz CSI together with the received signal vector of mmWave wide beams as the prediction input, and the index of the optimal mmWave narrow beam as the prediction label.” – II. Dual-Band Fusion for mmWave Beam Prediction in HetNets, A. Motivation and Problem Formulation, p. 119); and
step S4: configuring a pre-trained beam tracking deep learning model in a base station controller of an actual access network (Ma et al. the deep learning model can be deployed at the sub-6 GHz BS, IV. Simulation Study, B. Simulation Results, 1) Complexity of the Proposed Model, p. 123), performing millimeter-wave beam tracking according to actually input microwave channel information (Ma et al. “Then the sub-6 GHz CSI and the received signal vector of the measured mmWave wide beams are adopted to jointly predict the optimal mmWave narrow beam.” – III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, A. Motivation and Problem Formulation, p. 119) and (Ma et al. “In order to adapt to dynamic environmental fluctuations, online training can be employed during the predicting stage, where the deep learning model is continuously optimized by the training data collected from a small part of users that adopt conventional mmWave beam search.” – III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, A. Motivation and Problem Formulation, p. 119).
Although Ma et al. does not explicitly disclose millimeter-wave beam tracking by using time sequence microwave channel information and user location information, Ma et al. does disclose that “On the other hand, the stability of UE movements in the dynamic scenarios can be further utilized to track the UE trajectory and enhance the accuracy of beam prediction, where recurrent neural networks (RNN), e.g., gated recurrent networks [48] and long short-term memory networks [49], [50], have shown impressive performance in tracking the beam variations. Therefore, integrating the proposed dual-band fusion approach with RNN is a very interesting direction for future research in the highly-dynamic scenarios.” (Ma et al. III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, H. Extensions to More Scenarios, 2) Highly-Dynamic Scenarios, p. 122)
Khan et al. discloses:
step S2: for the millimeter-wave beam tracking problem, constructing a deep neural network (DNN) fusion model that performs millimeter-wave beam tracking by using time sequence microwave channel information (Khan et al. “a DNN was proposed to perform EBS during the training phase that estimates the correlation between sub-7GHz power delay profile and the best beam in the mmWave link.” - III. State of the Art, B. Side-Information Assisted Supervised Learning, 2) Low-Frequency Channel State Information, p. 11889; “LSTM uses the received signal and previous AoA estimates as an input and exploits the fact that mobility over specific paths generates sequential UE parameters that evolve over time.” – III. State of the Art, A. Non-Side-Information-Assisted Supervised Learning, 2) Long Short-Term Memory Networks, p. 11886) and user location information (Khan et al. “The beam measurements overhead can be greatly reduced if the UE can provide any additional side information (location, orientation).” - III. State of the Art, B. Side-Information Assisted Supervised Learning, p. 11887)
performing millimeter-wave beam tracking according to actually input microwave channel information and user location information (Khan et al. “Recently, authors in [55] proposed a location-assisted ML-based alignment framework that predicts the optimal serving BS and narrows down the best candidate beams based on the receiver’s location.” - III. State of the Art, B. Side-Information Assisted Supervised Learning, 1) Location Information, p. 11888)
It would have been obvious to someone with ordinary skill in the art prior to the effective filing date of the claimed invention to incorporate the features as disclosed by Khan et al. into the invention of Ma et al. to yield the invention of claim 1 above. Both Ma et al. and Khan et al. are considered analogous arts to the claimed invention as they both disclose using machine learning methods to determine an optimal beam in heterogenous networks. Ma et al. discloses the limitations outlined above. Although Ma et al. does not explicitly disclose millimeter-wave beam tracking by using time sequence microwave channel information and user location information, Ma et al. considers tracking the UE trajectory for future research. Ma et al. also fails to explicitly disclose the user location as input to the DNN fusion model. This feature is disclosed by Khan et al. where “The beam measurements overhead can be greatly reduced if the UE can provide any additional side information (location, orientation).” (Khan et al. III. State of the Art, B. Side-Information Assisted Supervised Learning, p. 11887). The combination of Ma et al. and Khan et al. would be obvious with a reasonable expectation of success to improve beam prediction accuracy by utilizing contextual information (Khan et al., III. State of the Art, B. Side-Information-Assisted Supervised Learning, 1) Location Information, p. 11887) and predict the mmWave beam with higher probability (Khan et al., III. State of the Art, B. Side-Information-Assisted Supervised Learning, 2) Low-Frequency Channel State Information, p. 11889).
Regarding claim 2, Ma et al. as modified above discloses:
The millimeter-wave beam tracking method in the microwave and millimeter wave heterogeneous network scenario according to claim 1, wherein
the millimeter-wave beam tracking problem is: tracking beams from a plurality of millimeter-wave base stations (Ma et al. six mmWave BSs, Fig. 1(b)) by using one microwave base station (Ma et al. Sub-6 GHz BS, Fig. 1(b)) in the microwave and millimeter wave heterogeneous network scenario in which microwave base stations and millimeter-wave base stations are independently and separately deployed and the quantity and density of the microwave base stations are less than the quantity and density of the millimeter-wave base stations (Ma et al. “mmWave BSs are usually deployed more densely than sub-6 GHz BSs, which forms the heterogeneous network (HetNet) [23], [24], as shown in Fig. 1(b).” – I. Introduction, p. 116).
Regarding claim 3, Ma et al. as modified above discloses:
[Note: what is not explicitly taught by Ma et al. has been struck-through]
The millimeter-wave beam tracking method in the microwave and millimeter wave heterogeneous network scenario according to claim 2, wherein a mathematical expression of the millimeter-wave beam tracking problem is:
k
*
t
,
m
*
t
=
argmax
kϵ
1,2
,
…
,
K
mϵ
1,2
,
…
,
M
H
m
m
W
(
k
)
t
H
f
m
(
k
)
2
;
wherein
k
*
(
t
)
represents the index of the millimeter-wave optimum base station at moment t, and
m
*
(
t
)
represents the index of the millimeter-wave optimum beam at moment
t
;
H
m
m
W
k
(
t
)
is the channel state of the kth millimeter-wave base station at moment
t
; and
f
m
(
k
)
is the mth candidate beam of the base station, K is the quantity of the millimeter-wave base stations, and M is the quantity of beams.
(Ma et al. “Let
f
n
(
j
)
be the n-th transmit beam of the j-th mmWave BS. In HetNets beam training targets to find the optimal beam
f
n
*
(
j
*
)
with the maximum beamforming gain from the candidate beams of all the mmWave BSs, which can be expressed as
j
*
,
n
*
=
argmax
jϵ
1,2
,
…
,
J
nϵ
1,2
,
…
,
N
h
(
j
)
T
f
n
(
j
)
2
.” - II. System Model, B. mmWave Training Model, p. 118; “On the other hand, the stability of UE movements in the dynamic scenarios can be further utilized to track the UE trajectory and enhance the accuracy of beam prediction, where recurrent neural networks (RNN), e.g., gated recurrent networks [48] and long short-term memory networks [49], [50], have shown impressive performance in tracking the beam variations. Therefore, integrating the proposed dual-band fusion approach with RNN is a very interesting direction for future research in the highly-dynamic scenarios.” - III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, H. Extensions to More Scenarios, 2) Highly-Dynamic Scenarios, p. 122; Examiner notes that it would be obvious to one of ordinary skill in the art to modify the equation taught by Ma et al. to be a function of time when modeling time sequence scenarios.)
Regarding claim 4, Ma et al. as modified above discloses:
[Note: what is not explicitly taught by Ma et al. has been struck-through]
The millimeter-wave beam tracking method in the microwave and millimeter wave heterogeneous network scenario according to claim 3, wherein the DNN fusion model comprises a microwave feature extraction module (Ma et al. Sub-6 GHz feature extraction model, Fig. 2, a(Ma et al. millimeter wave feature extraction model, Fig. 2), and a feature fusion module (Ma et al. Feature fusion model, Fig. 2);
the microwave feature extraction module is configured to perform feature extraction (Ma et al. “the optimal narrow beam index is first predicted based on the sub-6 GHz CSI, where the output is expressed as the probability that each candidate narrow beam is the optimal one.” – III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, B. sub-6 GHz Feature Extraction, p. 120);
the feature fusion module is configured to perform feature fusion by using the first group of optimum beam probabilities and the second group of optimum beam probabilities as inputs (Ma et al. “The two predicted probability vectors based on the sub-6 GHz CSI
h
-
and the measured mmWave wide beams
y
w
p
can be regarded as the extracted features of the two frequency bands, which are further fused for enhance the prediction accuracy. Specifically, we concatenate the two predicted probability vectors as the fused feature vector
p
^
f
=
p
^
s
u
b
-
6
T
p
^
M
M
T
T
, which is fed into the feature fusion model to predict the narrow beam.” – III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, E. Feature Fusion, p. 121) to obtain and output millimeter-wave optimum beam probabilities (Ma et al. “the output is expressed as the narrow beam probability vector
p
^
=
p
^
1
p
^
2
…
p
^
J
N
T
.” - III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, E. Feature Fusion, p. 121)
Although Ma et al. does not explicitly disclose the beam probability are obtained by tracking, Ma et al. does consider tracking the UE trajectory for future research (Ma et al. “On the other hand, the stability of UE movements in the dynamic scenarios can be further utilized to track the UE trajectory and enhance the accuracy of beam prediction, where recurrent neural networks (RNN), e.g., gated recurrent networks [48] and long short-term memory networks [49], [50], have shown impressive performance in tracking the beam variations. Therefore, integrating the proposed dual-band fusion approach with RNN is a very interesting direction for future research in the highly-dynamic scenarios.” - III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, H. Extensions to More Scenarios, 2) Highly-Dynamic Scenarios, p. 122).
Khan et al. discloses:
the location feature extraction module is configured to perform feature extraction by using historical time sequence user location information as an input to obtain a second group of optimum beam probabilities (Khan et al. “The beam measurements overhead can be greatly reduced if the UE can provide any additional side information (location, orientation).” - III. State of the Art, B. Side-Information Assisted Supervised Learning, p. 11887; “LSTM uses the received signal and previous AoA estimates as an input and exploits the fact that mobility over specific paths generates sequential UE parameters that evolve over time.” – III. State of the Art, A. Non-Side-Information-Assisted Supervised Learning, 2) Long Short-Term Memory Networks, p. 11886; “For the BM framework this can be utilized by estimating UE mobility patterns from historical data, which can help in reducing the BM overhead. For example, angle of departure (AoD) and angle of arrival (AoA) from previous beam measurements can be utilized to direct fewer beams towards the anticipated direction of movement” – II. Key Characteristics of an Ideal Beam Management Framework, G. History Utilization, p. 11884)
It would have been obvious to someone with ordinary skill in the art prior to the effective filing date of the claimed invention to incorporate the features as disclosed by Khan et al. into the invention of Ma et al. to yield the invention of claim 4 above. Both Ma et al. and Khan et al. are considered analogous arts to the claimed invention as they both disclose using machine learning methods to determine an optimal beam in heterogenous networks. Ma et al. as modified above discloses the invention of claim 3. Although Ma et al. does not explicitly disclose the location feature extraction module, Ma et al. considers tracking the UE trajectory for future research. Ma et al. also fails to explicitly disclose the location feature extraction module is configured to perform feature extraction by using historical time sequence user location information as an input to obtain a second group of optimum beam probabilities. This feature is disclosed by Khan et al. where “The beam measurements overhead can be greatly reduced if the UE can provide any additional side information (location, orientation).” (Khan et al. III. State of the Art, B. Side-Information Assisted Supervised Learning, p. 11887). The combination of Ma et al. and Khan et al. would be obvious with a reasonable expectation of success to include the location feature extraction model in order to improve beam prediction accuracy by utilizing contextual information (Khan et al., III. State of the Art, B. Side-Information-Assisted Supervised Learning, 1) Location Information, p. 11887) and predict the mmWave beam with higher probability (Khan et al., III. State of the Art, B. Side-Information-Assisted Supervised Learning, 2) Low-Frequency Channel State Information, p. 11889).
Regarding claim 5, Ma et al. as modified above discloses:
[Note: what is not explicitly taught by Ma et al. has been struck-through]
The millimeter-wave beam tracking method in the microwave and millimeter wave heterogeneous network scenario according to claim 4, wherein
the microwave feature extraction module comprises (Ma et al. Sub-6 GHz feature extraction model comprises a convolution module, Fig. 2), a first decoder (Ma et al. FC layer of output module, Fig. 2), and a first softmax module (Ma et al. softmax activation layer of output module, Fig. 2) that are cascaded, wherein
the first neural network (Ma et al. “CNN is adopted to extract the features from the sub-6 GHz CSI…” - III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, B. sub-6 GHz Feature Extraction, p. 120)
the first decoder is a fully connected neural network (FCNN) and is configured to convert the (Ma et al. “The fully-connected (FC) layer is exploited after the convolution module for transforming the extracted features to the candidate narrow beams.” - III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, B. sub-6 GHz Feature Extraction, 3) Output Module, p. 120); and
the first softmax module is configured to classify the beam features output by the first decoder and output the first group of optimum beam probabilities (Ma et al. “Then, a softmax activation layer is applied to normalize the outputs of the FC layer into probabilities.” - III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, B. sub-6 GHz Feature Extraction, 3) Output Module, p. 120);
the second (Ma et al. mmWave feature extraction model comprises convolution module, Fig. 2), a second decoder (Ma et al. FC layer of output module, Fig. 2), and a second softmax module (Ma et al. softmax activation layer of output module, Fig. 2) that are cascaded, wherein
the second neural network (Ma et al. “Similar to the sub-6 GHz feature extraction model, CNN is adopted to extract the mmWave features from the received signal vector of the measured mmWave wide beams
y
w
p
.” - III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, D. MmWave Feature Extraction, p. 121)
the second decoder is an FCNN and is configured to convert the (Ma et al. FC layer of the mmWAve feature extraction model, Fig. 2); and
the second softmax module is configured to classify the beam features output by the second decoder and output the second group of optimum beam probabilities (Ma et al. softmax activation layer of the mmWave feature extraction model, Fig. 2); and
the feature fusion module comprises a batch normalization module (Ma et al. BatchNorm, Fig. 3), an attention module (Ma et al. attention module, Figs. 2-3), and a classifier module (Ma et al. convolution module of the feature fusion model, Fig. 2), wherein
the batch normalization module is configured to normalize the input first and second groups of optimum beam probabilities (Ma et al. “Batch normalization is adopted both before the FC layer and after the weighting operation, in order to adjust the distribution of the fused feature vector for facilitating model learning.” - III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, E. Feature Fusion, p. 121);
the attention module is configured to adjust the weights for the normalized first and second groups of optimum beam probabilities (Ma et al. “the attention mechanism is introduced to adaptively weight the elements in the fuse feature vector
p
^
f
, as depicte din Fig. 3” - III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, E. Feature Fusion, p. 121); and
the classifier module is configured to fuse and analyze, based on the FCNN, the first and second groups of optimum beam probabilities that are normalized and whose weights are adjusted to output a final optimum beam probability (Ma et al. “To predict the optimal narrow beam according to the weighted fused feature vector
p
^
f
,
w
, CNN is adopted after the attention module to implement the classification, where the output is expressed as the narrow beam probability vector
p
^
=
p
^
1
p
^
2
…
p
^
J
N
T
.” Finally, the narrow beam having the highest predicted probability is chosen…” - III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, E. Feature Fusion, p. 121).
Although Ma et al. does not explicitly disclose the use of a long short-term memory as part of the proposed prediction model shown in Fig. 2, Ma et al. does disclose consider the use of long short-term memory for future research in highly dynamic scenarios (Ma et al. “On the other hand, the stability of UE movements in the dynamic scenarios can be further utilized to track the UE trajectory and enhance the accuracy of beam prediction, where recurrent neural networks (RNN), e.g., gated recurrent networks [48] and long short-term memory networks [49], [50], have shown impressive performance in tracking the beam variations. Therefore, integrating the proposed dual-band fusion approach with RNN is a very interesting direction for future research in the highly-dynamic scenarios.” - III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, H. Extensions to More Scenarios, 2) Highly-Dynamic Scenarios, p. 122).
Khan et al. discloses:
the first LSTM module is configured to extract the temporal dynamic features of a microwave channel (Khan et al. “Channel prediction is enabled in the second stage during which the LSTM model predicts the channel for the next beam coherence time (an effective measure of beam alignment frequency [112]), i.e., hi(t+1) as shown in Fig. 5.” - III. State of the Art, A. Non-Side-Information-Assisted Supervised Learning, 2) Long Short-Term Memory Networks, p. 11887)
the second LSTM module is configured to extract the temporal dynamic features of a user location from changes in input historical time sequence user locations (Khan et al. “LSTM uses the received signal and previous AoA estimates as an input and exploits the fact that mobility over specific paths generates sequential UE parameters that evolve over time.” – III. State of the Art, A. Non-Side-Information-Assisted Supervised Learning, 2) Long Short-Term Memory Networks, p. 11886);
It would have been obvious to someone with ordinary skill in the art prior to the effective filing date of the claimed invention to incorporate the features as disclosed by Khan et al. into the invention of Ma et al. to yield the invention of claim 5 above. Both Ma et al. and Khan et al. are considered analogous arts to the claimed invention as they both disclose using machine learning methods to determine an optimal beam in heterogenous networks. Ma et al. as modified above discloses the invention of claim 4. Although Ma et al. does not explicitly disclose the location feature extraction module, Ma et al. considers tracking the UE trajectory for future research. Ma et al. also fails to explicitly disclose the location feature extraction module is configured to perform feature extraction by using historical time sequence user location information as an input to obtain a second group of optimum beam probabilities. This feature is disclosed by Khan et al. where “The beam measurements overhead can be greatly reduced if the UE can provide any additional side information (location, orientation).” (Khan et al. III. State of the Art, B. Side-Information Assisted Supervised Learning, p. 11887). The combination of Ma et al. and Khan et al. would be obvious with a reasonable expectation of success to include the LSTM modules in order to enhance beam prediction accuracy (Ma et al. III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, H. Extensions to More Scenarios, 2) Highly-Dynamic Scenarios, p. 122), improve beam prediction accuracy by utilizing contextual information (Khan et al., III. State of the Art, B. Side-Information-Assisted Supervised Learning, 1) Location Information, p. 11887) and predict the mmWave beam with higher probability (Khan et al., III. State of the Art, B. Side-Information-Assisted Supervised Learning, 2) Low-Frequency Channel State Information, p. 11889).
Regarding claim 6, Ma et al. as modified above discloses:
[Note: what is not explicitly taught by Ma et al. has been struck-through]
The millimeter-wave beam tracking method in the microwave and millimeter wave heterogeneous network scenario according to claim 5, wherein in step S3, the DNN fusion model is pre-trained in a self-supervised or supervised training manner;
in the self-supervised training manner, the training data is historical time sequence microwave channel information and user location information; and
in the supervised training manner, the training data is historical (Ma et al. “In the training stage, the training dataset is constructed for optimizing the deep learning model, where each sample comprises the sub-6 GHz CSI together with the received signal vector of mmWave wide beams as the prediction input, and the index of the optimal mmWave narrow beam as the prediction label.” – III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, A. Motivation and Problem Formulation, p. 119).
Khan et al. discloses:
in the supervised training manner, the training data is historical time sequence user location information, and the corresponding optimum beam label (Khan et al. “LSTM uses the received signal and previous AoA estimates as an input and exploits the fact that mobility over specific paths generates sequential UE parameters that evolve over time.” – III. State of the Art, A. Non-Side-Information-Assisted Supervised Learning, 2) Long Short-Term Memory Networks, p. 11886; “During the training phase, all BSs perform EBS to identify the best BS and beam pair at each location.” - III. State of the Art, B. Side-Information Assisted Supervised Learning, 1) Location Information, p. 11888; “Thus, historical data of UE movement can be utilized to significantly reduce the BM overhead.” – V. Existing Challenges and Recommendations, E. High Mobility, p. 11986)
It would have been obvious to someone with ordinary skill in the art prior to the effective filing date of the claimed invention to incorporate the features as disclosed by Khan et al. into the invention of Ma et al. to yield the invention of claim 6 above. Both Ma et al. and Khan et al. are considered analogous arts to the claimed invention as they both disclose using machine learning methods to determine an optimal beam in heterogenous networks. Ma et al. as modified above discloses the invention of claim 5. Although Ma et al. does not explicitly disclose the location feature extraction module, Ma et al. considers tracking the UE trajectory for future research. Ma et al. also fails to explicitly disclose in the supervised training manner, the training data is historical time sequence user location information, and the corresponding optimum beam label. This feature is disclosed by Khan et al. where LSTM networks are trained on received signals and labels (Khan et al. III. State of the Art, A. Non-Side-Information-Assisted Supervised Learning, 2) Long Short-Term Memory Networks, p. 11886). The combination of Ma et al. and Khan et al. would be obvious with a reasonable expectation of success to improve beam prediction accuracy by utilizing contextual information (Khan et al., III. State of the Art, B. Side-Information-Assisted Supervised Learning, 1) Location Information, p. 11887) and predict the mmWave beam with higher probability (Khan et al., III. State of the Art, B. Side-Information-Assisted Supervised Learning, 2) Low-Frequency Channel State Information, p. 11889).
Regarding claim 7, Ma et al. as modified above discloses:
[Note: what is not explicitly taught by Ma et al. has been struck-through]
The millimeter-wave beam tracking method in the microwave and millimeter wave heterogeneous network scenario according to claim 6
L
s
=
L
s
s
u
b
-
6
+
L
s
l
o
c
+
L
s
f
u
=
-
H
m
m
W
k
s
u
b
-
6
(
t
)
f
m
s
u
b
-
6
k
s
u
b
-
6
2
-
H
m
m
W
k
l
o
c
(
t
)
f
m
l
o
c
k
l
o
c
2
-
H
m
m
W
k
f
u
(
t
)
f
m
f
u
k
f
u
2
L
s
s
u
b
-
6
L
s
l
o
c
L
s
f
u
k
s
u
b
-
6
m
s
u
b
-
6
k
l
o
c
m
l
o
c
k
f
u
m
f
u
Regarding claim 7, the limitation(s) recited is/are not required to be part of the claimed invention. Parent claim 6 teaches alternative limitations, i.e., “self-supervised or supervised training.” If a parent claim includes alternative limitations, and the reference teaches one of them, further limitations to the other alternative(s) in dependent claims are not required limitations. See Ex parte Werner, Appeal 2019-001448, Application No. 15/109,888, March 23, 2020, 15 pages. Here, Ma et al. as modified above teaches supervised training, as detailed in the rejection of claim 6. Claim 7 is based on another alternative/other alternatives, i.e., self-supervised training.
Regarding claim 8, Ma et al. as modified above discloses:
[Note: what is not explicitly taught by Ma et al. has been struck-through]
The millimeter-wave beam tracking method in the microwave and millimeter wave heterogeneous network scenario according to claim 6,
L
=
L
C
E
s
u
b
-
6
+
L
C
E
l
o
c
+
L
C
E
f
u
=
-
∑
m
=
1
K
M
δ
m
t
log
2
p
m
s
u
b
-
6
+
log
2
p
m
l
o
c
+
log
2
p
m
L
C
E
s
u
b
-
6
L
C
E
l
o
c
L
C
E
f
u
δ
1
t
,
…
,
δ
m
t
,
…
,
δ
K
M
t
t
m
*
t
=
m
δ
m
t
=
1
δ
m
t
=
0
K
M
=
K
×
M
p
m
s
u
b
-
6
p
m
l
o
c
p
m
(Ma et al. “The cross entropy loss is used to optimize the proposed model, which can be expressed as
l
o
s
s
=
-
∑
n
=
1
J
N
p
n
log
p
^
n
, where
p
n
=
1
if the n-th candidate mmWave narrow beam is the actual optimal beam, and otherwise,
p
n
=
0
… In order to fully guide the parameter optimization of the whole model, the losses of the narrow beam predictions according to the sub-6 GHz CSI, the measured mmWave wide beams and the fused feature vector are added up to train the model, which can be written as
l
o
s
s
=
-
∑
n
=
1
J
N
p
n
(
λ
sub
-
6
log
p
^
s
u
b
-
6
,
n
+
λ
MM
log
p
^
M
M
,
n
+
λ
f
log
p
^
,
n
)
, where
λ
s
u
b
-
6
,
λ
M
M
and
λ
f
denote the corresponding weight coefficients.” - III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, F. Model Training, p. 122; One of ordinary skill in the art would recognize the
p
n
as the one-hot vector and the use of log base 2 for binary coding. Examiner notes that it would be obvious to one of ordinary skill in the art to modify the equation taught by Ma et al. to be a function of time when modeling time sequence scenarios.).
Khan et al. discloses:
the location feature extraction module (Khan et al. “The beam measurements overhead can be greatly reduced if the UE can provide any additional side information (location, orientation).” - III. State of the Art, B. Side-Information Assisted Supervised Learning, p. 11887)
It would have been obvious to someone with ordinary skill in the art prior to the effective filing date of the claimed invention to incorporate the features as disclosed by Khan et al. into the invention of Ma et al. to yield the invention of claim 8 above. Both Ma et al. and Khan et al. are considered analogous arts to the claimed invention as they both disclose using machine learning methods to determine an optimal beam in heterogenous networks. Ma et al. as modified above discloses the invention of claim 6. Although Ma et al. does not explicitly disclose the location feature extraction module, Ma et al. considers tracking the UE trajectory for future research and discloses an equation for the cross entropy loss function, which sums the losses of the sub-6 GHz feature extraction model, the mmWave feature extraction model and the fused feature extraction model. Khan et al. discloses the location feature extraction module where “The beam measurements overhead can be greatly reduced if the UE can provide any additional side information (location, orientation).” (Khan et al. III. State of the Art, B. Side-Information Assisted Supervised Learning, p. 11887). It would be obvious to one of ordinary skill in the art that replacing a feature extraction model or including another feature extraction model would require modification of the entropy loss function. In the case of substituting the millimeter wave feature extraction model with a location feature extraction model, the loss function would be similarly modified. The combination of Ma et al. and Khan et al. would be obvious with a reasonable expectation of success to improve beam prediction accuracy by utilizing contextual information (Khan et al., III. State of the Art, B. Side-Information-Assisted Supervised Learning, 1) Location Information, p. 11887) and predict the mmWave beam with higher probability (Khan et al., III. State of the Art, B. Side-Information-Assisted Supervised Learning, 2) Low-Frequency Channel State Information, p. 11889).
Regarding claim 9, Ma et al. as modified above discloses:
[Note: what is not explicitly taught by Ma et al. has been struck-through]
The millimeter-wave beam tracking method in the microwave and millimeter wave heterogeneous network scenario according to claim 6, wherein based on the established beam tracking problem, a mapping from historical (Ma et al. “There exists a mapping function
Υ
from the sub-6 GHz CSI vector to the indices of the optimal mmWave BS and beam
Υ
:
h
-
→
j
*
,
n
*
.” - III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, A. Motivation and Problem Formulation, p. 118), and an optimum beam label that is corresponding to historical (Ma et al. “In the training stage, the training dataset is constructed for optimizing the deep learning model, where each sample comprises the sub-6 GHz CSI together with the received signal vector of mmWave wide beams as the prediction input, and the index of the optimal mmWave narrow beam as the prediction label.” - III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, A. Motivation and Problem Formulation, p. 119); and
the established mapping from the historical
Φ
:
H
s
u
b
-
6
t
-
τ
.
q
(
t
-
τ
)
τ
=
1
T
→
k
*
t
,
m
*
(
t
)
T
H
s
u
b
-
6
t
-
τ
t
-
τ
q
t
-
τ
k
*
t
t
m
*
(
t
)
t
Khan et al. discloses:
wherein based on the established beam tracking problem, a mapping from historical time sequence microwave channel information and user location information to the millimeter-wave optimum beam is established (Khan et al. “In general, these approaches either utilize UE location or sub-7 GHz CSI to maintain a database and train the ML algorithm to map this information for beam prediction as shown in Fig. 7.” - III. State of the Art, B. Side-Information-Assisted Supervised Learning, p. 11887; “Recently, authors in [55] proposed a location-assisted ML-based beam alignment framework that predicts the optimal serving BS and narrows down the best candidate beams based on the receiver’s location…During the training phase, all BSs perform EBS to identify the best BS and beam pair at each location.” - III. State of the Art, B. Side-Information-Assisted Supervised Learning, 1) Location Information, p. 11889; “LSTM uses the received signal and previous AoA estimates as an input and exploits the fact that mobility over specific paths generates sequential UE parameters that evolve over time.” – III. State of the Art, A. Non-Side-Information-Assisted Supervised Learning, 2) Long Short-Term Memory Networks, p. 11886)
It would have been obvious to someone with ordinary skill in the art prior to the effective filing date of the claimed invention to incorporate the features as disclosed by Khan et al. into the invention of Ma et al. to yield the invention of claim 9 above. Both Ma et al. and Khan et al. are considered analogous arts to the claimed invention as they both disclose using machine learning methods to determine an optimal beam in heterogenous networks. Ma et al. as modified above discloses the invention of claim 6. Although Ma et al. does not explicitly disclose the location feature extraction module, Ma et al. discloses mapping the sub-6 GHz CSI vector to the indices of the optimal beam (Ma et al. III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, A. Motivation and Problem Formulation, p. 118) and considers integrating the dual-band approach with a RNN for tracking UE trajectory for future research (Ma et al. III. Dual-Band Fusion for mmWave Beam Prediction in HetNets, H. Extensions to More Scenarios, 2) Highly-Dynamic Scenarios, p. 122). Khan et al. discloses mapping sub-7 GHz CSI or UE location to the optimum beams, and using estimating UE patterns from historical data (Khan et al. III. State of the Art, B. Side-Information Assisted Supervised Learning, p. 11887; II. Key Characteristics of an Ideal Beam Management Framework, G. History Utilization, p. 11884). The above claimed mapping would be an obvious result of creating the models based on the microwave extracted features and the user location extracted features in a dynamic scenario. The combination of Ma et al. and Khan et al. would be obvious with a reasonable expectation of success to improve beam prediction accuracy by utilizing contextual information (Khan et al., III. State of the Art, B. Side-Information-Assisted Supervised Learning, 1) Location Information, p. 11887) and predict the mmWave beam with higher probability (Khan et al., III. State of the Art, B. Side-Information-Assisted Supervised Learning, 2) Low-Frequency Channel State Information, p. 11889).
Regarding claim 10, Ma et al. as modified above discloses:
[Note: what is not explicitly taught by Ma et al. has been struck-through]
The millimeter-wave beam tracking method in the microwave and millimeter wave heterogeneous network scenario according to claim 9, wherein the mapping from the historical (Ma et al. “we propose to adopt deep learning to extract the complex relationship between sub-6 GHz CSI and mmWave low-overhead measurement…” – 1. Introduction, p. 116-117); and
the corresponding optimum beam is obtained according to the historical
Khan et al. discloses:
wherein the mapping from the historical time sequence information and the user location information to the millimeter-wave optimum beam is established by using a DNN (Khan et al. LSTM, III. State of the Art, A. Non-Side-Information-Assisted Supervised Learning, 2) Long Short-Term Memory Networks, p. 11886); and
the corresponding optimum beam is obtained according to the historical time sequence microwave channel information (Khan et al. “LSTM uses the received signal and previous AoA estimates as an input and exploits the fact that mobility over specific paths generates sequential UE parameters that evolve over time.” – III. State of the Art, A. Non-Side-Information-Assisted Supervised Learning, 2) Long Short-Term Memory Networks, p. 11886) and the user location information (Khan et al. “Recently, authors in [55] proposed a location-assisted ML-based beam alignment framework that predicts the optimal serving BS and narrows down the best candidate beams based on the receiver’s location…” - III. State of the Art, B. Side-Information-Assisted Supervised Learning, 1) Location Information, p. 11889).
It would have been obvious to someone with ordinary skill in the art prior to the effective filing date of the claimed invention to incorporate the features as disclosed by Khan et al. into the invention of Ma et al. to yield the invention of claim 10 above. Both Ma et al. and Khan et al. are considered analogous arts to the claimed invention as they both disclose using machine learning methods to determine an optimal beam in heterogenous networks. Ma et al. as modified above discloses the invention of claim 9. Ma et al. does not explicitly disclose that the microwave channel information is historical time sequence information, and the corresponding optimum beam is obtained according to the historical time sequence microwave channel information and the user location information. Khan et al. discloses mapping sub-7 GHz CSI or UE location to the optimum beams, and using estimating UE patterns from historical data (Khan et al. III. State of the Art, B. Side-Information Assisted Supervised Learning, p. 11887; II. Key Characteristics of an Ideal Beam Management Framework, G. History Utilization, p. 11884). The combination of Ma et al. and Khan et al. would be obvious with a reasonable expectation of success to improve beam prediction accuracy by utilizing contextual information (Khan et al., III. State of the Art, B. Side-Information-Assisted Supervised Learning, 1) Location Information, p. 11887) and predict the mmWave beam with higher probability (Khan et al., III. State of the Art, B. Side-Information-Assisted Supervised Learning, 2) Low-Frequency Channel State Information, p. 11889).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NAOMI M WOLFORD whose telephone number is (571)272-3929. The examiner can normally be reached Monday - Friday, 8:30 am - 4:30 pm 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, Resha Desai can be reached at (571)270-7792. 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.
NAOMI M. WOLFORD
Examiner
Art Unit 3648
/N.M.W./Examiner, Art Unit 3648
4 SEP 2026
/RESHA DESAI/Supervisory Patent Examiner, Art Unit 3648
Prosecution Insights