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 .
This is the first office action on the merits and is responsive to the papers filed 01/12/2024. Claims 1-20 are currently pending and examined below.
Priority
Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d).
Information Disclosure Statement
The information disclosure statements submitted by Applicant are in compliance with the provision of 37 CFR 1.97, 1.98 and MPEP § 609. They have been placed in the application file and the information referred to therein has been considered as to the merits.
Claim Objections
Claim 16 is objected to because of the following informalities:
Claim 16, line 2 “increase” should be –increases—
Appropriate correction is required.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 5, 9-10, 14, 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Li et al. (CN 111896973 A, “Li”) in view of Templeton et al. (US 20160274589 A1, “Templeton”).
Regarding claim 1, Li teaches a ranging method for a LiDAR (Li (Fig. 1, [59]): the active detection subsystem is a “single photon lidar” including a laser, single-photon detector, time-flight instrument, and main control chip; [60-61]: laser light is transmitted toward a target, the reflected echo is received by the single-photon detector, and the detector output is supplied to the time-flight instrument as a stop signal; [64]: the time-flight instrument may use FPGA and TDC chips to obtain ranging accuracy), comprising:
predicting, (based on at least part of previous k frames of the detection data), a position where an obstacle is located in the three-dimensional environment during a ((k + 1) th) detection, (wherein k is an integer, and k ≥ 1) (Li further teaches predicting a position where an obstacle is located in the three-dimensional environment during a subsequent detection. In particular, Li ([105]) teaches that the main control chip predicts the three-dimensional motion trajectory of the target based on active distance data and passive motion trajectory data. Li ([112-115]) identify the target position, distance, and speed at time t and provide a motion model in which the target position is determined at t+T0. Li ([119-125]) further teaches that the state variables include position and speed, that Kalman-filter prediction iteratively determines the predicted target state based on observed values, and that the spherical coordinates of the target position after the predicted interval T0 are obtained.);
when performing the ((k + 1) th) detection, changing, based on predicted position information of the obstacle, a detection window of the LiDAR for at least one point on the obstacle (Li ([105]) states that the laser pointer is adjusted through the predicted target position and that the distance gate of the single-photon detector is reduced. More specifically, Li ([126]) teaches that the “range gate of the single photon detector can be reset according to the predicted position” and that the distance-gate width “in the next sampling interval” is set according to the predicted target speed. Thus, Li's range/distance gate corresponds to the claimed detection window, and Li's next sampling interval corresponds to the subsequent detection. Li ([82]) additionally teaches extracting point-cloud data near the target and determining the target's distance using the single-photon detector and time-flight instrument.);
and calculating ranging information of the at least one point only based on echo information within a range of the changed detection window (Li ([61]) teaches receiving the target-reflected echo at the single-photon detector and supplying the detector output to the time-flight instrument as a stop signal. Li ([71]) teaches extracting signal photons and calculating the distance and movement speed of the target, followed by reducing the distance-gate length to reduce noise interference. Li ([106]) explains that a wide range gate produces a large amount of useless noise point-cloud data, while Li ([126]) teaches narrowing/resetting the range gate according to the predicted position, thereby significantly reducing the noise point cloud. Accordingly, Li teaches determining the target range from echo information admitted within the changed range gate while excluding unwanted out-of-range information.).
Li does not explicitly teach: “acquiring multiple frames of detection data of a three-dimensional environment” and, more particularly, Li does not explicitly teach that the predicted target position is: “based on at least part of previous k frames of the detection data, wherein k is an integer and k ≥ 1.”
Li generates point-cloud data and predicts a future three-dimensional target position, but Li's disclosed prediction relies on the combination of active LiDAR distance information and passive CCD trajectory information. Li does not clearly disclose using the claimed multiple LiDAR detection frames themselves as the basis for the prediction.
However, Templeton teaches acquiring multiple frames of detection data of a three-dimensional environment. (Templeton [0137] and Fig. 7B teach analyzing LiDAR point-cloud data from “one or more previous scans” and explain that objects appear in: “more than one frame of a time series of generated point clouds” and are associated from one frame to the next based on point-cloud shape and/or relative position.
Templeton further teaches predicting, based on at least part of previous k frames of the detection data, a position where the obstacle is located during the (k+1) th detection. Templeton [0137] teaches that the same object is identified and position-mapped at a series of time points, thereby determining a motion profile, after which: “The locations of such moving objects during the next scan can be predicted.” Templeton [0139] further teaches predicting the future location of the moving car based on its motion profile, such as its speed and direction. Templeton claim 6 recites: “predicting a position of the reflective feature in motion, based on one or more previous point cloud data indicating previous positions of the reflective feature in motion” and determining the subsequent region based on the predicted position.
Thus, Templeton's “one or more previous point cloud data” corresponds to the claimed previous k frames, where k ≥ 1, and Templeton's prediction of the location “during the next scan” corresponds to the claimed (k+1) th detection.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify Li's predictive LiDAR ranging method to acquire and use the multiple LiDAR point-cloud detection frames taught by Templeton for predicting the target position because Templeton teaches associating the same moving object across successive point-cloud frames, determining a motion profile from the object's previous positions, and predicting the object's position during the next scan. Such a modification would have predictably provided Li with LiDAR-derived prior target-position information for determining where the target is expected during the next detection, thereby allowing Li's range gate to be reset according to the predicted position as taught by Li ([126]), while reducing unwanted noise data and maintaining tracking of the moving target.
Regarding claim 5, Li, in view of Templeton, teaches the ranging method of claim 1, wherein k > 1, and wherein predicting the position where the obstacle is located in the three-dimensional environment during the (k + 1)th detection comprises: predicting, based on a relative position change of the obstacle during previous k detections and a time interval between adjacent detections, the position where the obstacle is located during the (k + 1)th detection (Templeton [0103] teaches associating the position of a perceived object in each captured point cloud with a frame number or frame time and associating the same object across successive scans to track the object in time. Templeton [0104] further teaches combining two or more successively captured point clouds to determine translation information for the object and using the object's velocity and/or acceleration to predict its location during a subsequent scan. Templeton [0099] teaches that the point-cloud frames are captured at predetermined time intervals, such as 100 ms, 33 ms, 1 ms, or 1 s. Templeton [0137], Fig. 7B, further teaches associating the same object according to its relative position from one frame to the next, mapping the object's position at a series of time points to determine a motion profile, and predicting the object's location during the next scan.).
Regarding claim 9, Li, in view of Templeton, teaches the ranging method of claim 1, further comprising: in response to determining that no obstacle is detected within the range of the changed detection window during the (k + 1)th detection, changing the range of the detection window during a (k + 2)th detection to a range of an original detection window, wherein the range of the original detection window is associated with a predetermined maximum detection distance of the LiDAR.
Li teaches an original wide detection window and a narrow tracking window. In particular, Li [71] teaches that, when searching for the target, the single-photon system distance gate is set at the kilometer level, and that after the target is captured and its subsequent position is predicted, the distance-gate length is reduced for precise tracking. Li [74] further teaches predetermining, based on prior knowledge, the longest distance at which the target may appear. Li [106] teaches that when the specific target location is not known, the active detection subsystem uses a very wide range gate to search for the target location. Templeton [0092] teaches when an emitted LiDAR pulse does not receive a reflected signal, the corresponding distance may be set to the maximum distance sensitivity of the LiDAR, which is defined according to the maximum time delay for which the optical sensor waits for a return signal.
It would have been obvious to one of ordinary skill in the art before the effective filing date to configure Li's tracking method such that, when the target is not detected within Li's narrowed predicted-position range gate, the next detection returns to Li's wider search gate extending to the predetermined maximum target range, in view of Templeton's teaching that absence of a reflected return is treated using the maximum distance sensitivity of the LiDAR. Such a modification would have predictably allowed Li to reach a target that moved outside the narrowed predicted range instead of continuing to search only a range in which the target was no longer detected, while retaining Li's narrow range gate once the target is reacquired to reduce background-noise data.
Regarding claim 10, Li teaches a LiDAR, comprising:
a transmitter configured to transmit a detection laser beam for detecting a three-dimensional environment (Li [59], Fig. 1, teaches a single-photon LiDAR active detection subsystem including a laser, galvanometer, lens, mirror, single-photon detector, time-flight instrument, and main control chip. Li [60] teaches that the laser emits light under control of the main control chip, with the galvanometer controlling scanning of the laser beam and the resulting beam being directed onto the target surface.);
a receiver comprising one or more photodetectors configured to receive an echo from an obstacle and convert the echo into an electrical signal (Li [61] teaches that laser light reflected from the target is received by the optical telescope and coupled to the single photon detector, whose output signal is supplied to the time flight instrument as a stop signal.);
a signal processor coupled to the receiver and configured to receive the electrical signal and calculate ranging information of the obstacle based on the electrical signal (Li [61] teaches that laser light reflected from the target is received by the optical telescope and coupled to the single photon detector, whose output signal is supplied to the time flight instrument as a stop signal.); and
a controller coupled to the receiver and the signal processor (Li [59] identifies the main control chip as part of the LiDAR system. Li [60] teaches that the main control chip controls the laser and galvanometer, while [61] teaches the single photon detector coupled to the time-flight instrument. Li [64] states that the time-flight instrument and main control chip are directly connected, and [65] teaches that the main control chip may be implemented by a computer and constitutes the system processing/control portion.) and configured to perform operations comprising:
predicting, (based on at least part of previous k frames of the detection data), a position where an obstacle is located in the three-dimensional environment during a ((k + 1) th) detection, (wherein k is an integer, and k ≥ 1) (Li further teaches predicting a position where an obstacle is located in the three-dimensional environment during a subsequent detection. In particular, Li ([105]) teaches that the main control chip predicts the three-dimensional motion trajectory of the target based on active distance data and passive motion trajectory data. Li ([112-115]) identify the target position, distance, and speed at time t and provide a motion model in which the target position is determined at t+T0. Li ([119-125]) further teach that the state variables include position and speed, that Kalman-filter prediction iteratively determines the predicted target state based on observed values, and that the spherical coordinates of the target position after the predicted interval T0 are obtained.); and
when performing the ((k + 1) th) detection, changing, based on predicted position information of the obstacle, a detection window of the LiDAR for at least one point on the obstacle (Li ([105]) states that the laser pointer is adjusted through the predicted target position and that the distance gate of the single-photon detector is reduced. More specifically, Li ([126]) teaches that the “range gate of the single photon detector can be reset according to the predicted position” and that the distance-gate width “in the next sampling interval” is set according to the predicted target speed. Thus, Li's range/distance gate corresponds to the claimed detection window, and Li's next sampling interval corresponds to the subsequent detection. Li ([82]) additionally teaches extracting point-cloud data near the target and determining the target's distance using the single-photon detector and time-flight instrument.),
wherein the signal processor is configured to, when performing the (k + 1)th detection, calculate ranging information of the at least one point on the obstacle only based on echo information within a range of the changed detection window (Li [61] teaches that the single photon detector supplies its output to the time-flight instrument for TOF processing, while [126] teaches that the range gate of that single photon detector is reset/narrowed according to the predicted target position. Li [106] explains that a wide gate generates large amounts of unwanted noise point-cloud data, whereas [126] explains that narrowing the gate significantly reduces that noise.
Accordingly, because Li's photodetector is range gated before its output is supplied to the time-flight processor, Li teaches that the ranging processor determines range based on detector/echo information admitted within the changed range gate.).
Li does not explicitly teach: “acquiring multiple frames of detection data of a three-dimensional environment” and, more particularly, Li does not explicitly teach that the predicted target position is: “based on at least part of previous k frames of the detection data, wherein k is an integer and k ≥ 1.”
Li generates point-cloud data and predicts a future three-dimensional target position, but Li's disclosed prediction relies on the combination of active LiDAR distance information and passive CCD trajectory information. Li does not clearly disclose using the claimed multiple LiDAR detection frames themselves as the basis for the prediction.
However, Templeton teaches acquiring multiple frames of detection data of a three-dimensional environment. (Templeton [0137] and Fig. 7B teach analyzing LiDAR point-cloud data from “one or more previous scans” and explain that objects appear in: “more than one frame of a time series of generated point clouds” and are associated from one frame to the next based on point-cloud shape and/or relative position.
Templeton further teaches predicting, based on at least part of previous k frames of the detection data, a position where the obstacle is located during the (k+1) th detection. Templeton [0137] teaches that the same object is identified and position-mapped at a series of time points, thereby determining a motion profile, after which: “The locations of such moving objects during the next scan can be predicted.” Templeton [0139] further teaches predicting the future location of the moving car based on its motion profile, such as its speed and direction. Templeton claim 6 recites: “predicting a position of the reflective feature in motion, based on one or more previous point cloud data indicating previous positions of the reflective feature in motion” and determining the subsequent region based on the predicted position.
Thus, Templeton's “one or more previous point cloud data” corresponds to the claimed previous k frames, where k ≥ 1, and Templeton's prediction of the location “during the next scan” corresponds to the claimed (k+1) th detection.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify Li's predictive LiDAR ranging method to acquire and use the multiple LiDAR point-cloud detection frames taught by Templeton for predicting the target position because Templeton teaches associating the same moving object across successive point-cloud frames, determining a motion profile from the object's previous positions, and predicting the object's position during the next scan. Such a modification would have predictably provided Li with LiDAR-derived prior target-position information for determining where the target is expected during the next detection, thereby allowing Li's range gate to be reset according to the predicted position as taught by Li ([126]), while reducing unwanted noise data and maintaining tracking of the moving target.
Regarding claim 14, Li, in view of Templeton, teaches the LiDAR of claim 10, wherein k > 1, and wherein the controller is configured to predict a distance from the obstacle during the (k + 1)th detection by predicting, based on a relative position change of the obstacle during previous k detections and a time interval between adjacent detections, the position where the obstacle is located during the (k + 1)th detection (Templeton [0103] teaches associating the position of a perceived object in each captured point cloud with a frame number or frame time and associating the same object across successive scans to track the object in time. Templeton [0104] further teaches combining two or more successively captured point clouds to determine translation information for the object and using the object's velocity and/or acceleration to predict its location during a subsequent scan. Templeton [0099] teaches that the point-cloud frames are captured at predetermined time intervals, such as 100 ms, 33 ms, 1 ms, or 1 s. Templeton [0137], Fig. 7B, further teaches associating the same object according to its relative position from one frame to the next, mapping the object's position at a series of time points to determine a motion profile, and predicting the object's location during the next scan.).
Regarding claim 19, Li, in view of Templeton, teaches the LiDAR of claim 10, wherein the controller is configured to: in response to determining that no obstacle is detected within the range of the changed detection window during the (k + 1)th detection, change the range of the detection window during a (k + 2)th detection to a range of an original detection window, wherein the range of the original detection window is associated with a predetermined maximum detection distance of the LiDAR.
Li teaches an original wide detection window and a narrowed tracking window. In particular, Li [71] teaches that, when searching for the target, the single-photon system distance gate is set at the kilometer level, and that after the target is captured and its subsequent position is predicted, the distance-gate length is reduced for precise tracking. Li [74] further teaches predetermining, based on prior knowledge, the longest distance at which the target may appear. Li [106] teaches that when the specific target location is not known, the active detection subsystem uses a very wide range gate to search for the target location. Templeton [0092] teaches when an emitted LiDAR pulse does not receive a reflected signal, the corresponding distance may be set to the maximum distance sensitivity of the LiDAR, which is defined according to the maximum time delay for which the optical sensor waits for a return signal.
It would have been obvious to one of ordinary skill in the art before the effective filing date to configure Li's tracking method such that, when the target is not detected within Li's narrowed predicted-position range gate, the next detection returns to Li's wider search gate extending to the predetermined maximum target range, in view of Templeton's teaching that absence of a reflected return is treated using the maximum distance sensitivity of the LiDAR. Such a modification would have predictably allowed Li to reach a target that moved outside the narrowed predicted range instead of continuing to search only a range in which the target was no longer detected, while retaining Li's narrow range gate once the target is reacquired to reduce background-noise data.
Regarding claim 20, Li teaches computer-executable software (Li teaches implementing its LiDAR ranging and predictive tracking process using computer executable software. Li [65] teaches that the main control chip may be implemented by a computer and constitutes the processing and control portion of the system. Li [128] further states that: “the above process can be implemented automatically by computer software.”) to perform operations comprising:
predicting, (based on at least part of previous k frames of the detection data), a position where an obstacle is located in the three-dimensional environment during a ((k + 1)th ) detection, (wherein k is an integer, and k ≥ 1) (Li further teaches predicting a position where an obstacle is located in the three-dimensional environment during a subsequent detection. In particular, Li ([105]) teaches that the main control chip predicts the three-dimensional motion trajectory of the target based on active distance data and passive motion trajectory data. Li ([112-115]) identify the target position, distance, and speed at time t and provide a motion model in which the target position is determined at t+T0. Li ([119-125]) further teaches that the state variables include position and speed, that Kalman-filter prediction iteratively determines the predicted target state based on observed values, and that the spherical coordinates of the target position after the predicted interval T0 are obtained.);
when performing the ((k + 1)th) detection, changing, based on predicted position information of the obstacle, a detection window of a LiDAR for at least one point on the obstacle (Li ([105]) states that the laser pointer is adjusted through the predicted target position and that the distance gate of the single-photon detector is reduced. More specifically, Li (Fig. 11, [126]) teaches that the “range gate of the single photon detector can be reset according to the predicted position” and that the distance-gate width “in the next sampling interval” is set according to the predicted target speed. Thus, Li's range/distance gate corresponds to the claimed detection window, and Li's next sampling interval corresponds to the subsequent detection. Li ([82]) additionally teaches extracting point-cloud data near the target and determining the target's distance using the single-photon detector and time-flight instrument.);
and calculating ranging information of the at least one point only based on echo information within a range of the changed detection window ((Li ([61]) teaches receiving the target-reflected echo at the single-photon detector and supplying the detector output to the time-flight instrument as a stop signal. Li ([71]) teaches extracting signal photons and calculating the distance and movement speed of the target, followed by reducing the distance-gate length to reduce noise interference. Li ([106]) explains that a wide range gate produces a large amount of useless noise point-cloud data, while Li ([126]) teaches narrowing/resetting the range gate according to the predicted position, thereby significantly reducing the noise point cloud. Accordingly, Li teaches determining the target range from echo information admitted within the changed range gate while excluding unwanted out of range information.).).
Li does not explicitly teach: “the software stored on a non-transitory computer-readable storage medium”, “acquiring multiple frames of detection data of a three-dimensional environment” and, more particularly, Li does not explicitly teach that the predicted target position is: “based on at least part of previous k frames of the detection data, wherein k is an integer and k ≥ 1.”
Li generates point-cloud data and predicts a future three-dimensional target position, but Li's disclosed prediction relies on the combination of active LiDAR distance information and passive CCD trajectory information. Li does not clearly disclose using the claimed multiple LiDAR detection frames themselves as the basis for the prediction.
However, Templeton teaches Templeton [0176]- [0177], Fig. 12, teaches a system includes one or more processors, memory, and machine-readable instructions which, when executed, cause performance of the disclosed functions, and the disclosed techniques may be implemented by “computer program instructions encoded on a non-transitory computer-readable storage media.”. See also, [0073]
Templeton also teaches acquiring multiple frames of detection data of a three-dimensional environment. (Templeton [0137] and Fig. 7B teach analyzing LiDAR point-cloud data from “one or more previous scans” and explain that objects appear in: “more than one frame of a time series of generated point clouds” and are associated from one frame to the next based on point-cloud shape and/or relative position.
Templeton further teaches predicting, based on at least part of previous k frames of the detection data, a position where the obstacle is located during the (k+1) th detection. Templeton [0137] teaches that the same object is identified and position-mapped at a series of time points, thereby determining a motion profile, after which: “The locations of such moving objects during the next scan can be predicted.” Templeton [0139] further teaches predicting the future location of the moving car based on its motion profile, such as its speed and direction. Templeton claim 6 recites: “predicting a position of the reflective feature in motion, based on one or more previous point cloud data indicating previous positions of the reflective feature in motion” and determining the subsequent region based on the predicted position.
Thus, Templeton's “one or more previous point cloud data” corresponds to the claimed previous k frames, where k ≥ 1, and Templeton's prediction of the location “during the next scan” corresponds to the claimed (k+1) th detection.
It would have been obvious to one of ordinary skill in the art before the effective filing date to implement Li's computer-software-controlled predictive LiDAR ranging process using Templeton's non-transitory computer-readable storage medium and processor architecture because Templeton teaches storing machine-readable instructions on a non-transitory medium for execution by a processor to perform LiDAR processing functions. It further would have been obvious to configure those instructions to use Templeton's successive LiDAR point-cloud frames for Li's target-position prediction because Templeton teaches determining the motion of the same object from previous point-cloud positions and predicting its position during the next scan. Such modifications would have predictably provided a conventional stored-software implementation of Li's disclosed computer-controlled process and LiDAR-derived prior target-position information for positioning Li's range gate during the subsequent detection.
Claims 2, 6-8, 11, 15-18 are rejected under 35 U.S.C. 103 as being unpatentable over Li in view of Templeton and Michael Hall (US 20190293796 A1, “Hall”).
Regarding claim 2, Li, in view of Templeton, teaches the ranging method of claim 1, wherein the detection data comprises at least one of a relative orientation or a distance from the LiDAR (Li [69]: the azimuth and elevation angles of the target relative to the system are determined, and the photon-counting LiDAR obtains the distance between the target and the system. Li further teaches an initial broad detection range used before predicted-position tracking. Li [71] teaches that, when searching for the target, the single-photon system distance gate is set at the kilometer level and is subsequently reduced after the target is captured and its subsequent position is predicted. Li [74] further teaches determining, based on prior knowledge, the longest distance at which the target may appear. Li [106] further teaches using a very wide range gate when the specific target location is not yet known.),
Li, in view of Templeton, fails to explicitly teach wherein acquiring the multiple frames of the detection data of the three-dimensional environment comprises: acquiring, based on a range of an original detection window, k frames of the detection data of the three-dimensional environment, wherein the range of the original detection window is associated with a predetermined maximum detection distance of the LiDAR.
However, Hall teaches a LiDAR detector operated according to a “time gate, or window signal” (Hall [0050]) and teaches an initial proximity detection mode in which the time gate is opened over a predetermined depth range, e.g., 0.5 m to 1.0 m, to determine whether an object is located within that range (Hall [0058], Fig. 7).
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify Li's initial broad search range to employ Hall's time-gate/window implementation because Hall teaches that such a gate defines the depth range from which reflected photons are accepted. Such a modification would have provided a known temporal-gating implementation for Li's predetermined initial search range before Li narrows the gate after target localization.
Regarding claim 6, Li, in view of Templeton, teaches the ranging method of claim 1, wherein changing the detection window of the LiDAR for the at least one point on the obstacle comprises: obtaining, based on the predicted position information of the obstacle, corresponding predicted time of flight (TOF) for a point on the obstacle (teaches determining the predicted spherical coordinates of the target after a predicted time interval, resetting the single-photon detector range gate according to the predicted position, and determining the gate width for the next sampling interval based on the predicted relative speed of the target. Li further employs a time-flight instrument and TDC for determining target range from the received echo ([61], [64]).).
Li, in view of Templeton, fails to explicitly teach changing a central position of a corresponding detection window for the point on the obstacle to the corresponding predicted TOF, and changing a range of the corresponding detection window to a range from a difference between the corresponding predicted TOF and a time window to a sum of the corresponding predicted TOF and the time window, wherein the time window is a predetermined value or is associated with at least one of a size or a speed of the obstacle.
Hall teaches that the elapsed time from an emitted pulse to a detector time gate is related to the distance/depth being probed and that the time gate constitutes a window during which reflected photons are detected (Hall [0051]- [0052], Fig. 6). Hall further teaches changing the delay/time position of the gate relative to the emitted pulse to probe a selected depth or distance ([0053]- [0054], Fig. 6). Accordingly, implementing Li's predicted position-based range gate using Hall's temporal gate would provide a predicted TOF corresponding to Li's predicted range and position the temporal detection window about that predicted TOF. A finite time window centered at the predicted TOF has the range, with Li further teaching that the corresponding gate width is associated with target speed.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify Li’s predicted position based range gating technique to implement the range gate using Hall’s TOF based temporal gating, because Hall teaches that the temporal delay of a detector time gate relative to an emitted light pulse corresponds to the particular depth or distance being probed and that the time gate defines the window during which reflected photons are detected. Such a modification would have allowed Li’s predicted target range to be converted into a corresponding expected echo arrival time and the detection gate to be positioned at that expected TOF, while retaining Li’s speed dependent gate width. Li teaches resetting the range gate according to the predicted target position and selecting the gate width for the next sampling interval according to predicted target speed in order to reduce background light interference and noise point-cloud data. Therefore, the modification would have been a predictable use of Hall’s known TOF gating technique in Li’s LiDAR system to improve selective ranging at the predicted target distance and further Li’s stated objective of reducing unwanted noise.
Regarding claim 7, Li, in view of Templeton and Hall, teaches the ranging method of claim 6, wherein the time window increases as at least one of the size or the speed of the obstacle increases (Li further teaches that the detection window width increases as the speed of the obstacle increases. Specifically, Li [126], Fig. 11, teaches determining the range gate width for the next sampling interval based on the predicted relative speed of the target. Thus, for a given sampling interval, a higher predicted target speed results in a wider detection gate.).
Regarding claim 8, Li, in view of Templeton and Hall, teaches the ranging method of claim 7, wherein the LiDAR comprises a receiver that comprises one or more photodetectors, a time-to-digital converter, and a memory, and wherein the one or more photodetectors is configured to receive an echo and convert the echo into an electrical signal, the time to digital converter is configured to receive the electrical signal and output TOF of the echo, and the memory is configured to store the TOF of the echo, and wherein the ranging method comprises one of: during the (k + 1)th detection, turning on a photodetector of the LiDAR within the range of the changed detection window, and turning off a photodetector outside the range of the changed detection window; during the (k + 1)th detection, always keeping the one or more photodetectors and the time-to-digital converter on, and storing, by the memory, only the TOF of the echo outputted by the time-to-digital converter within the range of the changed detection window; or during the (k + 1)th detection, always keeping the one or more photodetectors on, and turning on the time-to-digital converter only within the range of the changed detection window.
Li teaches a receiver including a single-photon detector and a time-flight instrument, wherein the single-photon detector receives the reflected echo and provides an output signal to the time-flight instrument as a stop signal, and the time-flight instrument includes a TDC for determining ranging information (Li [61, 64]).
Hall teaches SPAD photodetectors, a TDC, and memory (Hall [0052]- [0053], [0069]-[0070], [0075], Figs. 6, 12, 14). Hall teaches that a time gate constitutes a window during which a SPAD responds to received reflected photons by providing electrical current, thereby temporally enabling the photodetector within the gate; the TDC is triggered by the SPAD output and determines a digital photon-arrival/elapsed-time value; and memory 1440 is capable of storing data generated by the depth-measurement system. Thus, Hall teaches or suggests the first recited alternative of temporally enabling the photodetector within the changed detection window and disabling/gating detection outside that window.
It would have been obvious to one of ordinary skill in the art before the effective filing date to further modify Li's predicted position based single photon range gate to employ Hall's gated SPAD/TDC receiver architecture, because Hall teaches using a time gate to control the interval during which reflected photons are detected and processed by the SPAD/TDC ranging circuitry. Such a modification would have provided a known hardware implementation for Li's range gate, allowing Li's single photon detector to accept reflected light signals only during the predicted target window while rejecting out of window returns, thereby furthering Li’s objective of reducing background light noise and unnecessary point-cloud data.
Regarding claim 11, Li, in view of Templeton, teaches the LiDAR of claim 10, wherein the detection data comprises at least one of a relative orientation or a distance from the LiDAR (Li [69]: the azimuth and elevation angles of the target relative to the system are determined, and the photon-counting LiDAR obtains the distance between the target and the radar).
Li, in view of Templeton, fails to explicitly teach wherein acquiring the multiple frames of the detection data of the three-dimensional environment comprises: acquiring, based on a range of an original detection window, k frames of the detection data of the three-dimensional environment, wherein the range of the original detection window is associated with a predetermined maximum detection distance of the LiDAR.
However, Hall teaches a LiDAR detector operated according to a “time gate, or window signal” (Hall [0050]) and teaches an initial proximity-detection mode in which the time gate is opened over a predetermined depth range, e.g., 0.5 m to 1.0 m, to determine whether an object is located within that range (Hall [0058], Fig. 7).
It would have been obvious to modify the Li LiDAR method to employ Hall's predetermined initial depth/time gate range when acquiring the previous detection data because Hall teaches that such an extended gate permits detection of whether an object is present within a predetermined depth range before subsequently performing more precise depth probing, thereby providing efficient initial target detection followed by more targeted ranging.
Regarding claim 15, Li, in view of Templeton, teaches the LiDAR of claim 10, wherein the controller is configured to change the range and a position of the detection window during the (k + 1)th detection by obtaining, based on the predicted position information of the obstacle, corresponding predicted time of flight (TOF) for a point on the obstacle (Li [125-126], Fig. 11, teaches determining the predicted spherical coordinates of the target after a predicted time interval, resetting the single-photon detector range gate according to the predicted position, and determining the gate width for the next sampling interval based on the predicted relative speed of the target. Li further employs a time-flight instrument and TDC for determining target range from the received echo ([61], [64]);
Li, in view of Templeton, fails to explicitly teach changing a central position of a corresponding detection window for the point on the obstacle to the corresponding predicted TOF; and changing a range of a corresponding detection window for the point on the obstacle to a range from a difference between the corresponding predicted TOF and a time window to a sum of the corresponding predicted TOF and the time window, wherein the time window is a predetermined value or is associated with at least one of a size or a speed of the obstacle.
Hall teaches that the elapsed time from an emitted pulse to a detector time gate is related to the distance/depth being probed and that the time gate constitutes a window during which reflected photons are detected (Hall [0051]- [0052], Fig. 6). Hall further teaches changing the delay/time position of the gate relative to the emitted pulse to probe a selected depth or distance ([0053]- [0054], Fig. 6). Accordingly, implementing Li's predicted position based range gate using Hall's temporal gate would provide a predicted TOF corresponding to Li's predicted range and position the temporal detection window about that predicted TOF. A finite time window centered at the predicted TOF has the range, with Li further teaching that the corresponding gate width is associated with target speed.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify Li’s predicted position based range gating technique to implement the range gate using Hall’s TOF based temporal gating, because Hall teaches that the temporal delay of a detector time gate relative to an emitted light pulse corresponds to the particular depth or distance being probed and that the time gate defines the window during which reflected photons are detected. Such a modification would have allowed Li’s predicted target range to be converted into a corresponding expected echo arrival time and the detection gate to be positioned at that expected TOF, while retaining Li’s speed dependent gate width. Li teaches resetting the range gate according to the predicted target position and selecting the gate width for the next sampling interval according to predicted target speed in order to reduce background light interference and noise point-cloud data. Therefore, the modification would have been a predictable use of Hall’s known TOF gating technique in Li’s LiDAR system to improve selective ranging at the predicted target distance and further Li’s stated objective of reducing unwanted noise.
Regarding claim 16, Li, in view of Templeton and Hall, teaches the LiDAR of claim 15, wherein the time window increases as at least one of the size or the speed of the obstacle increase (Li further teaches that the detection window width increases as the speed of the obstacle increases. Specifically, Li [126], Fig. 11, teaches determining the range gate width for the next sampling interval based on the predicted relative speed of the target. Thus, for a given sampling interval, a higher predicted target speed results in a wider detection gate.).
Regarding claim 17, Li, in view of Templeton and Hall, teaches the LiDAR of claim 16, wherein the receiver further comprises a time-to-digital converter and a memory, and wherein the time-to-digital converter is configured to receive the electrical signal and output TOF of the echo, and the memory is configured to store the TOF of the echo.
Li teaches a receiver including a single-photon detector and a time flight instrument, wherein the singlephoton detector receives the reflected echo and provides an output signal to the time-flight instrument as a stop signal, and the time-flight instrument includes a TDC for determining ranging information (Li [61, 64]).
Hall [0069]- [0070], Fig. 12, teaches one or more TDCs coupled to SPAD photodetectors, wherein the TDC provides a digital value representative of an elapsed/arrival time associated with the received reflected light, and the resulting photon arrival times are used to generate an overall arrival time and corresponding measured depth. Hall [0075], Fig. 14, further teaches memory 1440 capable of storing data for the depth measurement system, and Hall [0083] teaches storing data resulting from processing received light information into depth measurement information.
It would have been obvious to store Hall's TDC-generated digital arrival-time value in Hall's memory because that timing value is generated for determining the measured depth and Hall provides memory for storing data generated and processed by the depth-measurement system. Such storage would have predictably retained the TOF information for ranging calculations and subsequent processing.
Regarding claim 18, Li, in view of Templeton and Hall, teaches the LiDAR of claim 17, wherein the LiDAR is configured such that, during the (k + 1)th detection, a photodetector of the LiDAR within the range of the changed detection window is turned on, and a photodetector outside the range of the changed detection window is turned off; or during the (k + 1)th detection, the photodetectors and the time-to-digital converter are always kept on, and the memory stores only the TOF of the echo outputted by the time-to-digital converter within the range of the changed detection window; or during the (k + 1)th detection, the photodetectors are always kept on, and the time-to-digital converter is turned on only within the range of the changed detection window.
Li further teaches a single-photon detector whose range gate is reset according to the predicted target position (Li [126], Fig. 11), thereby providing the changed detection window used for the subsequent detection.
Hall [0050] teaches that each SPAD photodetector generates an electrical output when a photon is received during a “time gate,” or window signal during which the SPAD is enabled. Hall [0051]- [0052], Fig. 6, further teaches that the capture timing controls the SPAD array to establish the time gate and that the time gate represents a window during which the corresponding SPAD outputs electrical current in response to receiving reflected photons. Hall further teaches that start and stop timing signals may collectively define the time gate.
Accordingly, when Hall's gated-SPAD implementation is applied to Li's changed predicted-position range gate, the photodetector is enabled within the temporal range of the changed detection window and gated/disabled outside that window, thereby teaching or suggesting the first alternative of claim 18.
It would have been obvious to one of ordinary skill in the art before the effective filing date to implement Li's predicted position based single photon range gate using Hall's gated SPAD photodetector, because Hall teaches enabling the SPAD during a time gate/window corresponding to the depth range being probed. Such a modification would have provided a known detector gating implementation for Li's changed range window, allowing reflected photons to be detected during the predicted target interval while rejecting out of window returns, thereby furthering Li's objective of reducing background light interference and unnecessary noise point-cloud data.
Claims 3-4 are rejected under 35 U.S.C. 103 as being unpatentable over Li in view of Templeton and Lages et al. (DE 10148071 A1, “Lages”).
Regarding claim 3, Li, in view of Templeton, fails to explicitly teach but Lages teaches the ranging method of claim 1, wherein predicting the position where the obstacle is located in the three-dimensional environment during the (k + 1)th detection comprises: identifying a type of the obstacle (Lages teaches assigning an object detected in successive depth-resolved laser-scanner images to an object class, such as a person, passenger car, or truck, and assigning a model type and model parameters according to that object class (Lages [16], [26-32], [75], [80]; claims 1-3);
calculating a speed of the obstacle based on the type of the obstacle and the previous k frames of the detection data (Lages further teaches that successive laser scanner images are acquired at time intervals , that object tracking is performed on the basis of previous images, and that a Kalman-filter state vector includes object position and speed components , with the applicable model and dynamic parameters depending on the identified object class ([32], [57], [59-60], [81-85], [90-100]; Figs. 1-3).);
and predicting, based on the speed of the obstacle, the position where the obstacle is located during the (k + 1)th detection (Lages further teaches predicting the subsequent object position using the state transition model, wherein the state used for the prediction includes the object's speed ([15], [32], [59-60]; claim 4).).
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the Li moving object prediction method to employ Lages's object class dependent motion models because Lages teaches that different classes of objects, such as persons, passenger cars, and trucks, exhibit different dynamic behavior and that assigning a model type and model parameters according to the identified object class provides a more accurate prediction of the object's motion and subsequent position. Such a modification would have predictably improved the accuracy of the speed and subsequent position determination used by Li to predict the target position and position its range gate.
Regarding claim 4, Li, in view of Templeton and Lages, teaches the ranging method of claim 3, wherein predicting the position where the obstacle is located in the three-dimensional environment during the (k + 1)th detection further comprises: determining at least one of a size or a motion parameter of the obstacle based on a mutual correlation between multiple points in the detection data in conjunction with an object identification technique.
Templeton further teaches determining a size of the obstacle based on a mutual correlation between multiple points in the detection data in conjunction with an object identification technique (Templeton [0134]- [0136], Fig. 7A). In particular, Templeton teaches estimating the dimensional extent of an identified object from point-cloud groupings; determining object edges from discontinuities in distance measurements of adjacent or nearby points; grouping the points into a cluster defining a surface of a physical object; and using an object detection and/or recognition module to associate the point cluster with the environmental object. Templeton further teaches determining object dimensions using standard object dimension information, such as car dimensions or pedestrian heights. See also Templeton claim 4, which identifies a reflective feature based on discontinuities in the distances of angularly adjacent point-cloud points.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify Li’s LiDAR target-tracking method to determine the size of the tracked target using Templeton’s point cloud grouping and object recognition technique because Templeton teaches grouping adjacent or nearby LiDAR points that define a common object surface, associating the point cluster with an environmental object using object detection/recognition, and estimating the dimensional extent of the identified object. Such a modification would have predictably improved Li’s ability to distinguish the intended target from surrounding objects and noise and to maintain reliable tracking of the identified target, which is consistent with Li’s concern that uncertainty in the target location and other objects in the point cloud may cause the system to follow the wrong target.
Claims 12-13 are rejected under 35 U.S.C. 103 as being unpatentable over Li in view of Templeton, Hall and Lages.
Regarding claim 12, Li, in view of Templeton and Hall, fails to explicitly teach but Lages teaches the LiDAR of claim 11, wherein the controller is configured to predict the position where the obstacle is located during the (k + 1)th detection by identifying a type of the obstacle (Lages teaches assigning an object detected in successive depth-resolved laser-scanner images to an object class, such as a person, passenger car, or truck, and assigning a model type and model parameters according to that object class (Lages [16], [26-32], [75], [80]; claims 1-3);
calculating a speed of the obstacle based on the type of the obstacle and the previous k frames of the detection data (Lages further teaches that successive laser scanner images are acquired at time intervals , that object tracking is performed on the basis of previous images, and that a Kalman-filter state vector includes object position and speed components , with the applicable model and dynamic parameters depending on the identified object class ([32], [57], [59-60], [81-85], [90-100]; Figs. 1-3).); and
predicting, based on the speed of the obstacle, the position where the obstacle is located during the (k + 1)th detection (Lages further teaches predicting the subsequent object position using the state transition model, wherein the state used for the prediction includes the object's speed ([15], [32], [59-60]; claim 4).).
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the Li moving object prediction method to employ Lages's object class dependent motion models because Lages teaches that different classes of objects, such as persons, passenger cars, and trucks, exhibit different dynamic behavior and that assigning a model type and model parameters according to the identified object class provides a more accurate prediction of the object's motion and subsequent position. Such a modification would have predictably improved the accuracy of the speed and subsequent position determination used by Li to predict the target position and position its range gate.
Regarding claim 13, Li, in view of Templeton, Hall and Lages, teaches the LiDAR of claim 12, wherein the controller is configured to determine at least one of a size or a motion parameter of the obstacle based on a mutual correlation between multiple points in the detection data in conjunction with an object identification technique.
Templeton further teaches determining a size of the obstacle based on a mutual correlation between multiple points in the detection data in conjunction with an object identification technique (Templeton [0134]- [0136], Fig. 7A). In particular, Templeton teaches estimating the dimensional extent of an identified object from point-cloud groupings; determining object edges from discontinuities in distance measurements of adjacent or nearby points; grouping the points into a cluster defining a surface of a physical object; and using an object detection and/or recognition module to associate the point cluster with the environmental object. Templeton further teaches determining object dimensions using standard object dimension information, such as car dimensions or pedestrian heights. See also Templeton claim 4, which identifies a reflective feature based on discontinuities in the distances of angularly adjacent point-cloud points.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify Li’s LiDAR target-tracking method to determine the size of the tracked target using Templeton’s point cloud grouping and object recognition technique because Templeton teaches grouping adjacent or nearby LiDAR points that define a common object surface, associating the point cluster with an environmental object using object detection/recognition, and estimating the dimensional extent of the identified object. Such a modification would have predictably improved Li’s ability to distinguish the intended target from surrounding objects and noise and to maintain reliable tracking of the identified target, which is consistent with Li’s concern that uncertainty in the target location and other objects in the point cloud may cause the system to follow the wrong target.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Lippert et al. (US 6657581 B1), teaches Automotive Lane Changing Aid Indicator
Young Shin Kim (US 20150146189 A1), teaches Lidar sensor system
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JEMPSON NOEL whose telephone number is (571) 272-3376. The examiner can normally be reached on Monday-Friday 9:30-5:30.
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, Yuqing Xiao can be reached on (571) 270-3603. 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.
/JEMPSON NOEL/Examiner, Art Unit 3645
Prosecution Insights