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 .
Remarks
The claims being considered in this application are those submitted on 06/26/2023. Claims 1-20 are pending.
Priority
The applicant’s claim to priority of PRO63/355,242 on 06/24/2022 is acknowledged.
Information Disclosure Statement
The information disclosure statements filed on 01/16/2025 (x4) have been annotated and considered.
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.
Regarding Claim 12, the claim recites: “wherein the TCN further comprises a convolution layers following each of the plurality of residual blocks”. This language is unclear because of the “a” paired with the plural “convolution layers”. It is unclear if there is missing language, if this is intended to be a single convolution layer, or a plurality of convolution layers. Therefore, Claim 12 is indefinite.
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 1-2, and 4-6 are rejected under 35 U.S.C. 103 as being unpatentable over Mooney et. al. (US 20210369536 A1) in view of Jia et. al. (CN111582108A, Translation Attached).
Regarding Claim 1, Mooney discloses:
An exoskeleton control architecture of one or more memory structures and/or computer executable instructions stored on one or more non-transitory computer readable medium and executable by one or more processors, the exoskeleton control architecture comprising: (See at least Figure 3 which depicts the Exoskeleton Architecture via Processor 302, Memory 304, etc.)
a high-level control layer comprising a (See at least ¶0089 via "Exoskeleton controllers as described herein can convert real-time sensor data into motor commands. Exoskeleton controllers can be broken into three levels: high level, mid-level, and low level. High level controllers can include activity recognition (e.g., walking, running, sitting, etc.)." as well as ¶0004 via "The controller can receive sensor data associated with activity of the exoskeleton boot during a first time interval. The controller can determine, based on the sensor data input into a model trained via a machine learning technique based on historical motion capture data associated with one or more users performing one or more physical activities, one or more commands for a second time interval subsequent to the first time interval." and ¶0107 via " The sensor data 1420 can include data corresponding to steady state activities 1412 or transient activities 1412" as well as ¶0103 via "The machine learning device 1414 can identify patterns or similarities between different data points of the received input (e.g., sensor data 1420) and map the inputs to outputs that correspond to the identified patterns (e.g., ankle angle data, torque used to transition between walking and running in previous activities). The model 1424 can generate the commands 1426 based in part on the identified patterns in the received input data." **Wherein the Machine learning being used to identify patterns from sensor data/type of activity (steady state, transient) represents the user state estimate**)
a mid-level control layer configured to receive the user state estimate and generate a torque command for an actuator of the exoskeleton based on the user state estimate (See at least ¶0089 via "Mid-level controllers can include development of a torque profile based on recognized activity (e.g., converting activity into torque)").
However, although Mooney discloses using machine learning in at least ¶0090; Mooney does not explicitly disclose the convolutional neural network. Nevertheless, Jia--who is directed towards a gait recognition and intention perception method--discloses: a convolutional neural network (CNN) configured to receive exoskeleton sensor data from one or more sensors on an exoskeleton and generate a user state estimate (See at least ¶0028 via "The convolutional neural network model trained in the fourth step is saved for real-time gait phase recognition. Data transmitted in real time from the data acquisition system is input into the convolutional neural network model after feature mapping based on the mapping relationship established in the second step, which identifies the current gait phase of the human foot. Based on the correspondence between plantar pressure, ankle joint acceleration, angular velocity information and corresponding human posture established in the first step, the current human posture of the test subject is determined, thereby determining their movement intention." as well as ¶0066 via "The phase of human movement gait is used as the output of the convolutional neural network, including 6 gait types: heel strike, arch strike, flat foot, heel off the ground, arch off the ground, and toe off the ground").
Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to modify Mooney's exoskeleton control architecture in view of Jia's convolutional neural network in order to implement a method for estimating a user's locomotion state/activity in order to accurately determine appropriate torque commands in subsequent processing (Mooney's mid-level controller): "based on the correspondence between the plantar pressure, ankle joint acceleration, angular velocity information and the corresponding human posture established in step (1), the current human posture of the test subject is determined, and then the intention of the action is determined, which is used for a series of subsequent control decisions. The real-time recognition is highly accurate and fast" [Jia ¶0076].
Regarding Claim 2, Modified Mooney discloses the exoskeleton control architecture of Claim 1.
Furthermore, Mooney discloses: further comprising: a low-level control layer implemented on a motor driver of the actuator and configured to translate the torque command into an actuator action to supply a joint torque (See at least ¶0089 via "Low level controllers can include execution of the mid-level torque profile (e.g., motor commands, field oriented control of brushless DC motors, current causing torque or speed, etc.)." and ¶003 via "The electric motor can generate torque about an axis of rotation of an ankle joint of the user. " as well as ¶0047 via "The exoskeleton 100 can include an actuator belt 135 (e.g., belt drivetrain). The actuator belt 135 can include a shaft that is driven by the motor and winds the actuator belt 135 around itself.").
Regarding Claim 4, Modified Mooney discloses the exoskeleton control architecture of Claim 1.
Furthermore, Mooney discloses: wherein the exoskeleton is an autonomous robotic joint exoskeleton (See at least Figure 1 which depicts the autonomous robotic joint exoskeleton (ankle joint), and ¶0089 which discusses the controllers that illustrate the autonomy).
Regarding Claim 5, Modified Mooney discloses the exoskeleton control architecture of Claim 1.
Furthermore, Mooney discloses: wherein the one or more sensors include an encoder configured to measure a joint position and/or angular velocity and/or one or more inertial measurement units (IMUs) configured to measure joint position and/or kinematics (See at least ¶0040 via "The exoskeleton 100 can include a rotary encoder 155 (e.g., shaft encoder, first rotary encoder, or motor encoder)" and ¶0041 via " The exoskeleton 100 can include a second rotary encoder 160 (e.g., ankle encoder). The second rotary encoder 160 can measure an angle of the ankle joint"; as well as ¶0107 via "The sensor data 1420 can include, but is not limited to, motion data, force data, torque data, temperature data, speed, gait transitions, angle measurements (e.g., of different joints of the user 1402)…The sensor data 1420 can include ankle joint data, inertial measurement unit data, and/or battery data.").
Regarding Claim 6, Modified Mooney discloses the exoskeleton control architecture of Claim 5.
Furthermore, Mooney discloses: wherein the exoskeleton sensor data comprises measured sensor data and/or derived sensor data including one or more of position, velocity, and/or acceleration (See at least ¶0069 via "The encoder chip can measure the angular position of the rotary encoder 155", ¶0041 via "The second rotary encoder 160 can detect the angle of the ankle joint while the rotary encoder 155 can detect the angle of the electric motor.", ¶0107 via " The sensor data 1420 can include, but is not limited to, motion data, force data, torque data, temperature data, speed, gait transitions, angle measurements (e.g., of different joints of the user 1402)…The sensor data 1420 can include ankle joint data, inertial measurement unit data, and/or battery data.").
Claims 3 is rejected under 35 U.S.C. 103 as being unpatentable over Mooney et. al. (US 20210369536 A1) and Jia et. al. (CN111582108A, Translation Attached) in view of Lerner et. al. (US 20210291355 A1).
Regarding Claim 3, Modified Mooney discloses the exoskeleton control architecture of Claim 2.
Furthermore, Mooney discloses the lower-level control layer (See at least Mooney ¶0089).
However, Modified Mooney does not explicitly disclose the actuator action being a motor current.
Nevertheless, Lerner--who is directed towards control for an exoskeleton controller--discloses: wherein the actuator action is a motor current, wherein the low-level control layer uses closed-loop current-feedback control to translate the torque command into the motor current (See at least ¶0084 via "Motor current to produce a desired torque will generally be directly proportional to desired torque and inversely proportional to various system losses." and ¶0082 via "Exemplary data showing T.sub.set and T.sub.meas for a device controlled according to this closed-loop control methodology during walking is shown in FIG. 7C. As can be seen, closed-loop control does a good job of minimizing the error between T.sub.set and T.sub.meas throughout the walking motion, which is why it is regarded as the “gold standard” of conventional control methodologies of devices like that described above." and also ¶0011 via " In one case, a linear relationship is assumed between user-applied torque and desired assistive torque, and then a linear relationship is assumed between this parameter, T.sub.set, and the actuator motor current (or some other parameter relevant to the force supplied by the actuator).").
Therefore, it would have been obvious to one of the ordinary skill in the art prior to the effective filing date of the given invention to modify Modified Mooney in view of the closed-loop current feedback control to translate the torque command into the motor current such as in Lerner in order to minimize error: "As can be seen, closed-loop control does a good job of minimizing the error between T.sub.set and T.sub.meas throughout the walking motion, which is why it is regarded as the “gold standard” of conventional control methodologies of devices like that described above" [Lerner ¶0082].
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Mooney et. al. (US 20210369536 A1) and Jia et. al. (CN111582108A, Translation Attached) in view of Bulea et. al. (US 20210298984 A1) and Walsh et. al. (US 20210039248 A1).
Regarding Claim 7, Modified Mooney discloses the exoskeleton control architecture of Claim 1.
Furthermore, Mooney discloses: the mid-level control layer (See at least ¶0089 via "Mid-level controllers can include development of a torque profile based on recognized activity (e.g., converting activity into torque)").
However, Modified Mooney does not explicitly disclose the scaling, delaying, and filtering.
Nevertheless, Bulea--who is directed towards powered gait assistance systems--discloses: wherein the user state estimate is an estimated joint moment, and wherein the mid-level control layer is configured to scale, (See at least ¶0043 via "the sensors and controller can estimate the internal joint moment applied at the knee (i.e., the volitional muscle effort exerted by the user across the knee joint) and provide assistance based on that effort" as well as ¶0042 via "Internal knee extensor moment output scales with the degree of crouch and body mass, but rarely exceeds and is often much less than 50 Nm for children with mild to moderate crouch…in some embodiments, the powered device can be configured to provide about one third of the internal demand on knee extensors for patients with mild-moderate crouch (e.g., up to about 17 Nm of torque assistance)", ¶0054 via "PID gains can be adjusted prior to a patient donning the device, so that torque output reaches a stable response with minimal chatter and latency", as well as ¶0060 via " During the practice sessions, the system torque output was adjusted upwards from 1 Nm at 0.5 Nm increments. Final values of 3.5 Nm for stance phase and 2.625 Nm (75% of the stance value) for swing were established based on participant preference and visual feedback of participant comfort and walking stability" and ¶0061 which discloses filtering sensor signals that are used for state estimation: "EMG data were band-pass filtered at 15-380 Hz, full-wave rectified, and low-pass filtered at 7 Hz to create a linear envelope. Lower-extremity joint angles were computed from marker trajectories. Experimental data were time normalized to each gait cycle and averaged across the gait cycles for each walking condition").
Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to modify Modified Mooney in view of Bulea's scaling and filtering of the joint moment to generate an assistive torque command that has reduced noise and is proportional/customized to the user across gait phases for increased stability: "During the practice sessions, the system torque output was adjusted upwards from 1 Nm at 0.5 Nm increments. Final values of 3.5 Nm for stance phase and 2.625 Nm (75% of the stance value) for swing were established based on participant preference and visual feedback of participant comfort and walking stability." [Bulea ¶0060] and "The varying postures and physical deformities present in children with CP, combined with the heterogeneous causes of crouch gait, create a need for versatile, adjustable, and adjustable assistance devices, having human-machine interfaces that can be customized to each individual" [Bulea ¶0040].
However, Modified Mooney does not explicitly disclose the control layer being configured to delay.
Nevertheless, Walsh--who is directed towards assisting with human motion--discloses: delay (See at least ¶0197 via "Upon heel strike, the control scheme can use a look-up table to generate the required motor pull. The flat trajectory from 0-40% of the Gait Cycle (GC) acts as a delay, keeping the soft exosuit slack as the foot is planted on the ground and the user's hip pivots into position above the foot. Starting at 40%, the motor pulls the cable in and tensions the soft exosuit to the maximum level at 62.5% GC when toe off occurs. After a period of holding, the motor then unwinds the cable back down to zero at 83% GC and resets for a new cycle.").
Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to further modify Modified Mooney in view of Walsh's controlled delay in order to avoid a reduced delivering power due to inexact timing: "…methods do not account for the physiological step-to-step variability, and, due to inexact timing, can result in less positive power delivered to the wearer, and in a reduction of negative power absorption by biological structures." [Walsh ¶0235].
Claim 8 is rejected under 35 U.S.C. 103 as being unpatentable over Mooney et. al. (US 20210369536 A1) and Jia et. al. (CN111582108A, Translation Attached) in view of Walsh et. al. (US 20210039248 A1).
Regarding Claim 8, Modified Mooney discloses the exoskeleton control architecture of Claim 1.
Furthermore, Mooney discloses: wherein the mid-level control layer is configured to generate the torque command (See at least ¶0089 via "Mid-level controllers can include development of a torque profile based on recognized activity (e.g., converting activity into torque)").
However, although Mooney discloses the torque being based on user activity, Mooney does not explicitly disclose the user state estimate being an estimated gait phase or the command being a function of estimated gait phase based on an assistance profile.
Nevertheless, Jia discloses: wherein the user state estimate is an estimated gait phase (See at least ¶0066 via "The phase of human movement gait is used as the output of the convolutional neural network, including 6 gait types: heel strike, arch strike, flat foot, heel off the ground, arch off the ground, and toe off the ground").
Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to modify Modified Mooney in view of Jia's estimated gait phase in order to improve the estimation user's locomotion state/activity in order to accurately determine appropriate torque commands in subsequent processing (Mooney's mid-level controller): "based on the correspondence between the plantar pressure, ankle joint acceleration, angular velocity information and the corresponding human posture established in step (1), the current human posture of the test subject is determined, and then the intention of the action is determined, which is used for a series of subsequent control decisions. The real-time recognition is highly accurate and fast" [Jia ¶0076].
However, although Modified Mooney discloses generating a torque command (See at least Mooney ¶0089), Modified Mooney does not explicitly disclose the command being generated as a function of gait phase based on an assistance profile.
Nevertheless, Walsh discloses: generate the torque command as a function of the estimated gait phase based on an assistance profile (See at least ¶0069 via "Thus, were tension to be applied to the connection member 107 between 30-70% of the gait cycle, the connection member 107 provides a small moment extending the knee at 30-40% of the gait cycle, provides almost no moment at the knee at 50% of the gait cycle, and provides a larger moment at 60-70% of the gait cycle." as well as ¶0197 via "Upon heel strike, the control scheme can use a look-up table to generate the required motor pull. The flat trajectory from 0-40% of the Gait Cycle (GC) acts as a delay, keeping the soft exosuit slack as the foot is planted on the ground and the user's hip pivots into position above the foot. Starting at 40%, the motor pulls the cable in and tensions the soft exosuit to the maximum level at 62.5% GC when toe off occurs. After a period of holding, the motor then unwinds the cable back down to zero at 83% GC and resets for a new cycle.").
Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to further modify Modified Mooney in view of Walsh's assistance profile in order to determine when and how much of the assistance should be applied throughout a gait phase in order to avoid a reduced delivering power due to inexact timing: "…methods do not account for the physiological step-to-step variability, and, due to inexact timing, can result in less positive power delivered to the wearer, and in a reduction of negative power absorption by biological structures." [Walsh ¶0235].
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over Mooney et. al. (US 20210369536 A1), Jia et. al. (CN111582108A, Translation Attached), and Walsh et. al. (US 20210039248 A1) in view of Zhang et al. ("Human-in-the-Loop Optimization of Exoskeleton Assistance During Walking", IDS).
Regarding Claim 9, Modified Mooney discloses the exoskeleton control architecture of Claim 8.
Furthermore, Mooney discloses: the mid-level control layer (See at least ¶0089 via "Mid-level controllers can include development of a torque profile based on recognized activity (e.g., converting activity into torque)").
However, Mooney does not explicitly disclose the assistance profile.
Nevertheless, Walsh discloses: wherein timing and magnitude of nodes of the assistance profile represent control parameters, (See at least ¶0197 via "Upon heel strike, the control scheme can use a look-up table to generate the required motor pull. The flat trajectory from 0-40% of the Gait Cycle (GC) acts as a delay, keeping the soft exosuit slack as the foot is planted on the ground and the user's hip pivots into position above the foot. Starting at 40%, the motor pulls the cable in and tensions the soft exosuit to the maximum level at 62.5% GC when toe off occurs. After a period of holding, the motor then unwinds the cable back down to zero at 83% GC and resets for a new cycle." **Wherein the timing parameters are represented as the percents of the gait cycle, and the magnitude is represented by the levels slack, maximum level, and zero)
Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to modify Modified Mooney in view of Walsh's assistance profile in order to determine when and how much of the assistance should be applied throughout a gait phase in order to avoid a reduced delivering power due to inexact timing: "…methods do not account for the physiological step-to-step variability, and, due to inexact timing, can result in less positive power delivered to the wearer, and in a reduction of negative power absorption by biological structures." [Walsh ¶0235].
However, Modified Mooney does not explicitly disclose the human-in-the-loop optimization process that updates a cost landscape.
Nevertheless, Zhang--who is directed towards human-in-the-loop optimization of exoskeleton assistance during walking--discloses: a human-in-the-loop optimization process that updates a cost landscape based on walking speed and samples the control parameters that increases walking speed improvement based on the updated cost landscape (See at least Page 2 via "We have developed approaches in which device control is systematically varied during use so as to maximize human performance, which we call “human-in-the-loop optimization” (Fig. 1A).", as well as Page 3 via "The exoskeleton periodically changes the pattern of assistance, defined by a control law, while metabolic rate is measured. Steady-state metabolic energy cost is estimated for each control law by fitting a first-order dynamical model to 2 min of transient metabolic data (fig. S1). After a prescribed number of control laws have been evaluated, forming one generation, a covariance matrix adaptation evolution strategy (CMA-ES) (28) is used to calculate the next generation of control laws to be tested." and Figure 1's description via "(A) Measurements of human performance are used to update device control so as to improve performance in the human portion of the system."; as well as Page 8 via "Successfully reducing both metabolic rate and muscle activity suggests that alternate objective functions with similar properties could be optimized—for example, related to speed (40)").
PNG
media_image1.png
416
718
media_image1.png
Greyscale
Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to modify Modified Mooney in view of Zhang's human-in-the-loop optimization process in order to make the assistance more individualized for the user: "Finding a good generic assistance pattern, customizing it to individual needs, and helping users learn to take advantage of the device all contributed to improved economy" [Zhang, Abstract].
Claims 10-13 are rejected under 35 U.S.C. 103 as being unpatentable over Mooney et. al. (US 20210369536 A1) and Jia et. al. (CN111582108A, Translation Attached) in view of Kong et. al. (CN 112487902 A, Translation Attached).
Regarding Claim 10, Modified Mooney discloses the exoskeleton control architecture of Claim 1.
However, although Mooney discloses machine learning, and Jia discloses a CNN, Modified Mooney does not explicitly disclose the TCN /specific architecture as claimed.
Nevertheless, Kong--who is directed towards an exoskeleton oriented gait phase classification method based on TCN-HMM--discloses: wherein the CNN is a temporal convolutional network (TCN) and the TCN comprises: a series of a plurality of residual blocks and skip connections, (See at least ¶n0001 via " a human walking gait phase classification method based on hybrid TCN (Temporal Convolutional Networks) and HMM (Hidden Markov Model) models" as well as ¶n0023 via "TCN networks include a one-dimensional fully convolutional network structure, causal convolution, dilated convolution, and residual connections.")
wherein an output of a previous residual block is summed elementwise with an output of a following residual block via the skip connections, (¶n0067 via "The TCN network constructed in this invention includes a one-dimensional fully convolutional network (1D FCN) structure, causal convolutions, dilated convolutions, and residual connections." **Wherein the residual connection includes the summation of the skip outputs)
wherein each of the plurality of residual blocks comprises one or more convolutional layers, (See at least ¶n0029 via "Residual connection: Residual modules are used instead of convolutional layers. The residual modules mainly consist of two layers of dilated causal convolution and rectified linear units (ReLU)." **Wherein the residual modules consist of layers and are utilized rather than using only convolutional layers)
wherein a dilation factor of the one or more convolutional layers increases with one or more subsequent residual blocks in the series of the plurality of residual blocks (See at least ¶n0072 via "Where d is the dilation factor; (x*<sub>d</sub>f) represents the dilated convolution operation with dilation factor d; k is the size of the convolution kernel; f(i) represents the i-th weight of the convolution kernel; x<sub>s-d·i</sub> represents the element numbered s-d·i in the input one-dimensional sequence; and s is the position where the dilated convolution operation is to be performed. The expansion is equivalent to introducing a fixed step size between every two filters.").
Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to modify Modified Mooney in view of Kong's temporal convolutional network in order to provide specific architecture for gait phase classification that has improved accuracy and control of stability: "This model organically combines the temporal and spatial features of motion data and discriminates gait phase information. Not only did it obtain highly accurate gait phase classification results, but it also suppressed incorrect classifications based on discriminative learning, which is of great significance for the stable control of lower limb exoskeleton devices." [Kong ¶n0042].
Regarding Claim 11, Modified Mooney discloses the exoskeleton control architecture of Claim 10.
Furthermore, Kong discloses: wherein each of the plurality of residual blocks comprises two convolutional layers that are each followed by a weight normalization layer and an activation layer (See at least ¶n0029 via "Residual connection: Residual modules are used instead of convolutional layers. The residual modules mainly consist of two layers of dilated causal convolution and rectified linear units (ReLU). To perform normalization, the weights are normalized and applied to the convolution filter." **Wherein the rectified linear units represents an activation layer).
Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to modify Modified Mooney in view of Kong's temporal convolutional network in order to provide specific architecture for gait phase classification that has improved accuracy and control of stability: "This model organically combines the temporal and spatial features of motion data and discriminates gait phase information. Not only did it obtain highly accurate gait phase classification results, but it also suppressed incorrect classifications based on discriminative learning, which is of great significance for the stable control of lower limb exoskeleton devices." [Kong ¶n0042].
Regarding Claim 12, Modified Mooney discloses the exoskeleton control architecture of Claim 11.
Furthermore, Kong discloses: wherein the TCN further comprises a convolution layers following each of the plurality of residual blocks (See at least ¶n0029 via "In addition, to ensure that the input and the output of the residual module have the same width, an additional 1×1 convolution is used." as well as ¶n0024 via "1) One-dimensional fully convolutional network structure: In a one-dimensional fully convolutional network structure, each hidden layer has the same length as the input layer, and subsequent layers are zero-padding to keep their length the same as the previous layer.")
Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to modify Modified Mooney in view of Kong in order to ensure the signal data parameters are compatible: "the input and the output of the residual module have the same width" [Kong ¶n0029].
Regarding Claim 13, Modified Mooney discloses the exoskeleton control architecture of Claim 11.
Furthermore, Kong discloses: wherein the TCN further comprises a fully connected output layer (See at least ¶n0031 via "Each unit of the TCN network's output layer is associated with a specific state of the model, and the network is trained to estimate the posterior probability of each state. This is achieved by using a softmax activation function in the output layer to obtain the distribution of states y∈{1,…,Q}.").
Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to modify Modified Mooney in view of Kong's temporal convolutional network's fully connected output layer in order to provide specific architecture for gait phase classification that has improved accuracy and control of stability: "This model organically combines the temporal and spatial features of motion data and discriminates gait phase information. Not only did it obtain highly accurate gait phase classification results, but it also suppressed incorrect classifications based on discriminative learning, which is of great significance for the stable control of lower limb exoskeleton devices." [Kong ¶n0042].
Claims 14 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Mooney et. al. (US 20210369536 A1) and Jia et. al. (CN111582108A, Translation Attached) in view of Ren (CN114038005A, Translation Attached) and Lyons (US 20200327418 A1).
Regarding Claim 14, Modified Mooney discloses the exoskeleton control architecture of Claim 1.
Furthermore, Modified Mooney discloses the high-level control layer (See at least Mooney ¶0089).
However, Modified Mooney does not explicitly disclose, but Ren--who teaches labeling and peak-based relabeling--discloses: (See at least ¶n0019 via "Before the design and application of the gait phase recognition algorithm, the gait phase in the gait data needs to be labeled offline first. The dataset used to train the classification and recognition algorithm needs to be ground truth processed so that the supervised machine learning method can map an input to an output… Threshold-based methods are usually used to determine the "on/off state" of pressure sensor signals. By setting a corresponding signal threshold for each sensing unit individually, it is determined whether the sensing unit is in an "idle state" or a "working state"…" **Wherein the thresholds correspond to the signal peaks: ¶n0074 via "Where Dthreshold is the threshold of the FSR sensing unit in the gait data, S<sub>max</sub> is the maximum value in the gait data, S<sub>min</sub> is the minimum value in the gait data, and δ is an adjustable threshold coefficient used to represent the percentage of the threshold between the peak values.").
Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to modify Modified Money in view of Ren's labeling in order to increase the accuracy and optimize the neural network recognition in order for the most optimal control: "The accuracy of the gait phase recognition model increases with the optimization of the number of nodes in the hidden layer of the neural network. It has a certain degree of effectiveness in recognizing gait phases in various movement modes based on plantar pressure signals." [Ren ¶n0061] and "By comparing the impact of different network structures on recognition accuracy, the neural network recognition model is optimized and adjusted." [Ren ¶n0003].
However, Modified Mooney does not explicitly disclose, but Lyons discloses: backward labeler (See at least ¶0104 via "A feedback loop is introduced where; the CNN infers key-point labels, the key-point labels are refined using a GA, the refined labels are used to update CNN weights using backpropagation.")
wherein the high-level control layer further comprises a real-time adaptation trainer configured to train the CNN in a single epoch of backpropagation with the ground truth gait phase (See at least ¶0104 via "A feedback loop is introduced where; the CNN infers key-point labels, the key-point labels are refined using a GA, the refined labels are used to update CNN weights using backpropagation." and 0019 via "The labeled real depth frames are then used to continue training the CNN, improving accuracy beyond that achievable by using only unlabeled real depth frames for domain adaptation.").
Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to modify Modified Mooney in view of Lyons in order to automate the training as well as continuously train/improve the model: "The labeled real depth frames are then used to continue training the CNN, improving accuracy beyond that achievable by using only unlabeled real depth frames for domain adaptation. The merits of this approach are that no manual effort is required to label depth frames and the 3D model fitting algorithm does not have any real-time constraints." [Lyons ¶0019].
Regarding Claim 16, Modified Mooney discloses the exoskeleton control architecture of Claim 14.
Furthermore, Lyons discloses: wherein the backward labeler and the real time adaptation trainer operate in parallel (See at least ¶0104 via "A feedback loop is introduced where; the CNN infers key-point labels, the key-point labels are refined using a GA, the refined labels are used to update CNN weights using backpropagation." and 0019 via "The labeled real depth frames are then used to continue training the CNN, improving accuracy beyond that achievable by using only unlabeled real depth frames for domain adaptation. " **Wherein the system is both predicting labels and adapting/being trained within a feedback loop, representing the parallel operation).
Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to modify Modified Mooney in view of Lyons in order to automate the training as well as continuously train/improve the model: "The labeled real depth frames are then used to continue training the CNN, improving accuracy beyond that achievable by using only unlabeled real depth frames for domain adaptation." [Lyons ¶0019].
Claim 17 is rejected under 35 U.S.C. 103 as being unpatentable over Mooney et. al. (US 20210369536 A1) and Jia et. al. (CN111582108A, Translation Attached) in view of Nagraj Rao et. al. (US 20220383040 A1).
Regarding Claim 17, Modified Mooney discloses the exoskeleton control architecture of Claim 1.
Furthermore, Mooney discloses the high-level control layer (See at least Mooney ¶0089) and also an exoskeleton (See at least Mooney Figure 1), and Jia discloses a CNN (See at least Jia ¶0028 via "The convolutional neural network model trained in the fourth step is saved for real-time gait phase recognition….").
However, modified Mooney does not explicitly disclose the second exoskeleton or the transformation matrix.
Nevertheless, Nagraj Rao--who is directed towards sensor domain adaptation--discloses: wherein the CNN is trained based on sensor data from a second (See at least ¶0017 via "inputting the first sensor data and the second sensor data to a machine learning program to train the machine learning program to determine a domain translation of data from the first data space to the second data space" and ¶0047 via "With reference to FIG. 1, a rigid body transformation matrix R specifies a transformation from a first sensor 130A coordinate system 160A to a second sensor 130B coordinate system 160B" as well as ¶0053 via "The machine learning program Z is then trained using the collected data D.sub.1, D.sub.2 and ground truth data, e.g., as shown in Table 2, for transforming any other sensor data, e.g., pre-recorded training data, generated based on the first data space S.sub.1 of the first sensor 130A to data that is based on the second data space S.sub.2. In other words, the machine learning program Z provides a non-linear domain translation T to transform data from the first data space S1 to the second data space S.sub.2.").
Therefore, it would have been obvious to one of ordinary skill in the at prior to the effective filing date of the given invention to modify Modified Mooney in view of Nagraj Rao's transformation of sensor data to a second space to train the machine learning in order to apply the concept of the sensor domain adaptation to exoskeleton devices in order to prevent the need for large amounts of data to be recaptured: "a first sensor having a first data space can be replaced with a second sensor having a second data space without a need to recapture large amounts of training data to retrain a machine learning program trained to use data from the first sensor. The pre-recorded training data can be transformed to new training data based on the second data space. Thereby, foregoing a need to recapture the entire set of pre-recorded training data." [Nagraj Rao ¶0070], thus saving time and computational cost whilst improving the machine learning.
Claims 18-20 are rejected under 35 U.S.C. 103 as being unpatentable over Mooney et. al. (US 20210369536 A1) and Jia et. al. (CN111582108A, Translation Attached) in view of Kim (CN 114444672 A, Translation Attached).
Regarding Claim 18, Modified Mooney discloses the exoskeleton control architecture of Claim 1.
However, although Mooney discloses a processor (See at least Mooney Figure 3) and execute the mid-level control layer (See at least Mooney ¶0089), and Jia discloses a CNN (See at least Jia ¶0028 via "The convolutional neural network model trained in the fourth step is saved for real-time gait phase recognition…."), Modified Mooney does not explicitly disclose the two separate processors.
Nevertheless, Kim--who is directed towards neural networks--discloses: further comprising: a first processor configured to execute an inference process with the CNN as a dedicated process; and (See at least ¶n0474 via "For example, the convolution operation of the first layer is processed by the first processor" **Wherein the convolution operations are apart of the inference)
a second processor configured to execute (See at least ¶n0474 via " the convolution operation of the second layer is processed by the second processor").
Therefore it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to modify Modified Mooney in view of Kim's separate processors in order to separate the execution of the torque control of Mooney from the CNN inference processing in order to reduce computational load: "In particular, since artificial neural network computation involves a huge amount of data, efficient scheduling can significantly improve the computational processing speed of artificial neural networks" [Kim ¶n0931].
Regarding Clam 19, Modified Mooney discloses the exoskeleton control architecture of Claim 18.
Furthermore, Mooney discloses the exoskeleton sensor data (See at least Mooney ¶0004 via "he controller can receive sensor data associated with activity of the exoskeleton boot during a first time interval.")
However, Mooney does not explicitly disclose th I/O process as claimed.
Nevertheless, Kim discloses: wherein the first processor is further configured to execute an I/O process configured to receive and supply the (See at least ¶n0944 via "Buffer memory can be implemented in a first-in-first-out (FIFO) format. When the buffer memory is full, the AMC switches to standby mode. When the buffer memory transfers data to the NPU, the AMC reads the data from the main memory based on the ANN data locality information and stores the data in the buffer memory. AMC can swap the first data stored at the first memory address with the second data stored at the second memory address").
Therefore, it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to modify Modified Mooney in view of Kim's FIFP format (I/O process) in order to avoid latency in the main memory: " by providing AMC, the operation sequence of the artificial neural network model processed by the NPU can be essentially eliminated by the RAS latency and CAS latency that may be caused by the main memory." [Kim ¶n0940].
Regarding Claim 20, Modified Mooney discloses the exoskeleton control architecture of Claim 19.
Furthermore, Mooney discloses: supply the torque command to the actuator of the exoskeleton (See at least ¶0089 via "Low level controllers can include execution of the mid-level torque profile (e.g., motor commands, field oriented control of brushless DC motors, current causing torque or speed, etc.)." and ¶003 via "The electric motor can generate torque about an axis of rotation of an ankle joint of the user. " as well as ¶0047 via "The exoskeleton 100 can include an actuator belt 135 (e.g., belt drivetrain). The actuator belt 135 can include a shaft that is driven by the motor and winds the actuator belt 135 around itself.").
However, Mooney does not disclose this being executed on a second processor.
Nevertheless, Kim discloses: wherein the second processor is further configured to (See at least ¶n0474 via " the convolution operation of the second layer is processed by the second processor").
Therefore it would have been obvious to one of ordinary skill in the art prior to the effective filing date of the given invention to modify Modified Mooney in view of Kim's separate processors in order to separate the execution of the torque control of Mooney from the CNN inference processing in order to reduce computational load: "In particular, since artificial neural network computation involves a huge amount of data, efficient scheduling can significantly improve the computational processing speed of artificial neural networks" [Kim ¶n0931].
Allowable Subject Matter
Claim 15 is objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KAYLA RENEE DOROS whose telephone number is (703)756-1415. The examiner can normally be reached Generally: M-F (8-5) 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, Abby Lin can be reached on (571) 270-3976. 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.
/K.R.D./Examiner, Art Unit 3657
/ABBY LIN/Supervisory Patent Examiner, Art Unit 3657