Object detection using tilted polygons for parking space detection

Abstract

The neural network can be used to determine the corner points of the inclined polygon (e.g., as the displacement values ​​of the corner points of the anchor box) that accurately depict the area in the image that defines the parking space. In addition, the neural network can output a confidence value for the likelihood that the corner points of the predicted anchor box correspond to the entrance to the parking space. The confidence value can be used to select a subset of the corner points of the anchor box and / or the inclined polygon in order to define the entrance to the parking space. The minimum convergence distance between the corner points of the inclined polygon predicted using the CNN and the ground truth corner points of the parking space can be used to simplify the determination of whether the anchor box should be used as a positive sample for training.

Description

Background Art

[0001] Accurate and efficient image processing (e.g., for recognition and classification) by machines (e.g., computers programmed with trained neural networks) is important in a variety of situations. For example, an autonomous vehicle (e.g., a vehicle equipped with an advanced driver assistance system (ADAS)) or a drone may analyze image data (e.g., images representing a road and / or parking lot captured by a camera) in real time to formulate driving maneuvers (e.g., turning the steering device to the left, activating the braking system, etc.). In one such instance, the vehicle may analyze the image data to detect a parking space while performing a parking maneuver, and identify the parking space's attributes, such as location, size, and orientation. To facilitate this process, the vehicle may include an object detector that uses a convolutional neural network (CNN) to detect the presence of a parking space in an image.

[0002] A traditional CNN for detecting parking spaces may use an axis-aligned rectangular anchor box (all four corners are right angles) as a form of detection output. However, due to the perspective projection of the sensor, the parking spaces present in the sensor data are usually not rectangular or axis-aligned. Therefore, once the sensor data is detected, additional processing is required to accurately identify the boundaries of each parking space. For example, a camera on a vehicle may capture an image of a parking space, and based on the perspective of the camera's field of view, the parking space may not be depicted as an axis-aligned rectangle in the image. A traditional CNN may provide an axis-aligned rectangular anchor box as a form of detection output, in which case additional processing is required to accurately depict the parking space in the image. When training a traditional CNN, the intersection over union (IoU) between the anchor box output and the ground truth output from the CNN can be used to identify positive samples. The IoU calculation can be simple because both the anchor box output and the ground truth are axis-aligned rectangles. Summary of the invention

[0003] The present disclosure relates to object detection using tilted polygons (e.g., quadrilaterals) suitable for parking space detection. For example, in some cases, at least one convolutional neural network (CNN) can be used to detect and / or delineate one or more parking spaces represented in image data. The CNN output can be post-processed and provided to downstream systems (e.g., vehicle control modules) to inform subsequent operations.

[0004] Aspects of the present disclosure may use a CNN to determine the corner points of an inclined polygon (e.g., as displacement or offset values ​​of the corner points of an anchor shape) that accurately depicts the area in the image that defines the parking space. In addition, the present disclosure provides a CNN that outputs a confidence value that predicts the likelihood that the corner points of the anchor shape define or otherwise correspond to the entrance to the parking space. The confidence value may be used to select a subset of the corner points of the anchor shape and / or the inclined polygon in order to define the entrance to the parking space. According to an embodiment of the present disclosure, a CNN may be used to predict the likelihood that a particular corner point of an anchor shape corresponds to the entrance to a parking space and to predict the displacement values ​​of the corner points that depict the boundary of the parking space.

[0005] The present disclosure also provides calculating the distance (e.g., minimum convergence distance) between the corner points of the inclined polygon predicted using the CNN and the ground truth corner points of the parking space to determine whether the anchor shape should be used as a positive training sample. For example, a positive sample can be identified based at least in part on the distance being below a threshold. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] The present system and method for object detection using tilted polygons suitable for parking space detection are described in detail below with reference to the accompanying drawings, which are incorporated herein by reference, wherein:

[0007] Figure 1 is a diagram including an example object detection system according to some embodiments of the present disclosure;

[0008] Figure 2 is a flow chart illustrating an example process for identifying one or more parking spaces according to some embodiments of the present disclosure;

[0009] Figure 3 is an illustration of an image representable by image data processed by an object detector, a grid of spatial elements of the object detector, and a set of anchor shapes that may be associated with one or more of the spatial elements in accordance with some embodiments of the present disclosure;

[0010] Figure 4 is an illustration of an image overlaid with visual elements for different spatial element resolutions according to some embodiments of the present disclosure;

[0011] Figure 5A is an illustration of a neural network for detecting parking spaces according to some embodiments of the present disclosure;

[0012] Figure 5B is an illustration of an image with an entry line depiction and a parking space depiction according to some embodiments of the present disclosure;

[0013] Figure 6 is an illustration of an image with ground truth data and corner points of a tilted quadrilateral for training an object detector according to some embodiments of the present disclosure;

[0014] Figure 7 is a block diagram illustrating a method of training a machine learning model to provide corner points of a parking space according to some embodiments of the present disclosure;

[0015] Figure 8 is a block diagram illustrating a method for determining corner points of a parking space using a neural network according to some embodiments of the present disclosure;

[0016] Fig. 9 is a block diagram illustrating a method of determining a parking space entrance using a neural network according to some embodiments of the present disclosure;

[0017] Fig.10 is an illustration of an example operating environment suitable for implementing some embodiments of the present disclosure;

[0018] Fig.11A is an illustration of an example autonomous vehicle according to some embodiments of the present disclosure;

[0019] Fig. 11B For use according to some embodiments of the present disclosure Fig.11A Examples of autonomous vehicle camera positions and fields of view;

[0020] Fig. 11C For use according to some embodiments of the present disclosure Fig.11A a block diagram of an example system architecture for an example autonomous vehicle;

[0021] Fig.11D According to some embodiments of the present disclosure Fig.11A A system diagram of communications between a cloud-based server and an example autonomous vehicle; and

[0022] Fig.12 is a block diagram of an example computing device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0023] The present disclosure relates to object detection using tilted polygons (e.g., quadrilaterals) suitable for parking space detection. The disclosed methods may be applicable to driving operations (e.g., autonomous driving, advanced driver assistance systems (ADAS)), etc.) for detecting parking spaces, as well as other applications (e.g., robotics, video analysis, weather forecasting, medical imaging, etc.) to detect objects (e.g., buildings, windows, doors, driveways, intersections, curbs, real estate areas, surface areas or regions, etc.) corresponding to tilted polygons in image and / or sensor data.

[0024] The present disclosure may be described with respect to an example autonomous vehicle 1100 (alternatively referred to herein as “vehicle 1100 ” or “autonomous vehicle 1100 ”), examples of which are described herein with reference to Figures 11A-11DDescribed in more detail. Although the present disclosure primarily provides examples using autonomous vehicles, the various methods described herein may be implemented using other types of devices, such as robots, unmanned aerial vehicles, camera systems, weather forecasting equipment, medical imaging equipment, etc. In addition, these methods may be used to control autonomous vehicles, or for other purposes, such as but not limited to video surveillance, video or image editing, parking space occupancy monitoring, identification and / or detection, video or image search or retrieval, object tracking, weather forecasting (e.g., using RADAR data) and / or medical imaging (e.g., using ultrasound or magnetic resonance imaging (MRI) data).

[0025] Although parking spaces are primarily described as detected objects, the disclosed methods may generally be applied to objects that may appear as tilted polygons (such as quadrilaterals or other shapes) in the sensor's field of view and / or image data (e.g., these objects may be rectangular in the real world but appear as tilted quadrilaterals due to perspective). Although the disclosed methods are described using tilted quadrilaterals and four corner points, the disclosed concepts may be applied to any number of shapes and points (e.g., corner points) that define those shapes. Furthermore, although this document primarily defines an entrance as being defined by two points (e.g., corner points), in other examples, any number of points (e.g., corner points) may be used to define an entrance. Furthermore, although the present disclosure focuses on object detectors implemented using neural networks, other types of machine learning models may be employed in some embodiments.

[0026] Compared to conventional methods that can use CNNs to predict axis-aligned rectangular anchor boxes that generally indicate the size and location of parking spaces, aspects of the present disclosure can use CNNs to determine the corner points of a tilted quadrilateral (e.g., as displacement or offset values ​​of the anchor box corner points) that accurately depicts the area defining the parking space in the image. Therefore, in some embodiments, the tilted quadrilateral can be directly consumed by downstream systems without requiring additional or significant processing to identify the boundaries of the parking space. By reducing subsequent processing, the disclosed method may be more efficient and faster than conventional methods.

[0027] In addition, compared to conventional methods, the present disclosure provides one or more CNNs that output confidence values ​​that predict the likelihood that the corner points of the anchor box define or otherwise correspond to the entrance to the parking space. The confidence values ​​can be used to select a subset of the corner points of the anchor box and / or the oblique quadrilateral to define the entrance to the parking space. According to an embodiment of the present disclosure, processing can be further reduced by using a CNN to predict the likelihood that a particular corner point of the anchor box corresponds to the entrance to the parking space and predicting the displacement values ​​of the corner points that delineate the boundary of the parking space.

[0028] On the other hand, while conventional CNNs use intersection over union (IoU) to determine whether an axis-aligned rectangular anchor box output is a positive sample, the present disclosure provides for calculating the minimum convergence distance between the corner points of the tilted quadrilateral predicted using the CNN and the ground truth corner points of the parking space to determine whether the anchor box should be used as a positive sample for training. For example, a positive sample may be identified based at least in part on the minimum convergence distance (e.g., after normalization) being below a threshold. Calculating the minimum convergence distance may be more direct than calculating the IoU of the tilted quadrilateral, thereby reducing processing time.

[0029] Exemplary Parking Space Detector

[0030] Now refer to Figure 1 , Figure 1 A diagram including an example object detection system 100 according to some embodiments of the present disclosure is shown. It should be understood that this and other arrangements described herein are merely set forth as examples. In addition to or in lieu of what is shown, other arrangements and elements (such as machines, interfaces, functions, sequences, and functional groupings, etc.) may be used, and some elements may be completely omitted for clarity. Further, many of the elements described herein are functional entities that may be implemented as discrete or distributed components or in conjunction with other components, and implemented in any appropriate combination and position. The various functions performed by one or more entities described herein may be implemented by hardware, firmware, and / or software. For example, some functions may be implemented by a processor that executes instructions stored in a memory.

[0031] In one or more embodiments, the object detection system 100 includes, for example, a communication manager 104, an object detector 106, a feature determiner 108, a confidence score generator 110, a displacement value generator 112, a skewed quadrilateral generator 114, and an entry determiner 126. Some examples described in this disclosure use quadrilaterals (e.g., regular, skewed, irregular, box, etc.), and the described systems and methods may similarly use other polygons.

[0032] The communication manager 104 may be configured to manage communications received by the object detection system 100 (e.g., including sensor data and / or image data) and / or communications provided by the object detection system 100 (e.g., including confidence scores, displacement scores, corner points of the skewed quadrilateral, and / or information derived therefrom). Additionally or alternatively, the communication manager 104 may manage communications within the object detection system 100, such as communications between any of the object detector 106, the confidence score generator 110, the displacement value generator 112, the skewed quadrilateral generator 114, the entry determiner 126, and / or other components that may be included in the object detection system 100 or that may communicate with the object detection system 100 (e.g., downstream system components that consume output from the object detection system 100).

[0033] refer to Figure 2 , Figure 2 2 is a flowchart illustrating an example process 200 for identifying one or more parking spaces according to some embodiments of the present disclosure. The object detector 106 may be configured to analyze input data received from the communication manager 104, such as sensor data and / or image data representing any number of parking spaces (or no parking spaces), and generate object detection data for any number of detected objects captured in the input data. To this end, the object detector 106 may use a feature determiner 108, a displacement value generator 112, and a confidence score generator 110. The feature determiner 108 may be configured to generate or determine features of the input data as inputs to the confidence score generator 110 and the displacement value generator 112. The confidence score generator 110 may be configured to generate or determine confidence scores 118 for one or more anchor boxes based on data from the feature determiner 108. The confidence score 118 for each anchor box may predict the likelihood that the corresponding anchor box corresponds to a parking space detected in the input data.

[0034] The displacement value generator 112 may be configured to generate or determine displacement values ​​122 to the corner points of each anchor box based on data from the feature determiner 108. The slanted quadrilateral generator 114 may receive as input any of the various outputs from the object detector 106, such as the confidence value 118 and the displacement value 122 for each anchor box. The slanted quadrilateral generator 114 may generate and / or determine the slanted quadrilaterals based on the input using any suitable technique, such as non-maximum suppression (NMS). This may include the slanted quadrilateral generator 144 determining the corner points of the slanted quadrilateral from any number of anchor boxes based on the displacement values ​​122 (e.g., provided by the displacement value generator 112) and the corner points of the anchor boxes. As a non-limiting example, the slanted quadrilateral generator 114 may determine which anchor boxes have confidence values ​​118 that exceed a threshold (if any). Based on those anchor boxes, the slanted quadrilateral generator 114 may filter and / or cluster the candidate detections into one or more output object detections, and determine the corner points of the slanted quadrilaterals corresponding to those output object detections (eg, using corresponding displacement values ​​122 ).

[0035] In addition to or in lieu of the confidence score generator 110 generating or determining a confidence score 118 to predict the likelihood that the corresponding anchor box corresponds to a parking space detected in the input data, the confidence score generator 110 may generate or determine a confidence score 116 to predict the likelihood that the corresponding corner point corresponds to a detected parking space entrance represented in the input data. The entrance determiner 126 may use at least the confidence score 116 to determine one or more entrances to one or more parking spaces. As a non-limiting example, the entrance determiner 126 may define an entrance for each object detection output by the slanted quadrilateral generator 114 by selecting a set of corner points for each slanted quadrilateral (e.g., two corner points) having the highest confidence value (e.g., optionally requiring those confidence values ​​116 to exceed a threshold). The selected corner points may then be used to define an entrance to the corresponding parking space (e.g., an entrance line connecting the selected corner points). As shown in FIG. Figure 2 As shown in dashed lines, in other examples, the oblique quadrilateral generator 114 may not be implemented in an object detection system 100 having an entrance determiner 126 and / or used by the entrance determiner 126 to identify and / or define an entrance to a parking space or other detection object area.

[0036] According to some embodiments of the present disclosure, the object detection system 100 may be configured to: Fig.10 is implemented in the example operating environment 1000. For example, Figure 1The components of the object detection system 100 may generally be implemented using any combination of one or more client devices 1020, one or more server devices 1060, or one or more data stores 1050. Thus, the object detection system 100 may be provided via multiple devices arranged in a distributed environment that collectively provide the functionality described herein, or may be embodied on a single device (e.g., a vehicle 1100). Thus, while some examples used to describe the object detection system 100 may relate to specific devices and / or configurations, it is contemplated that those examples may be more generally applicable to any potential combination of devices and configurations described herein. For example, in some embodiments, at least some of the sensors 1080 used to generate one or more portions of sensor data input to the object detector 106 may be distributed among multiple devices and / or objects in the environment, and / or at least one of the sensors 1080 may be included in the vehicle 1100.

[0037] As mentioned herein, the communication manager 104 can be configured to manage communications received by the object detection system 100 (e.g., including sensor data and / or image data) and / or communications provided by the object detection system 100 (e.g., including confidence scores or values, displacement values, corner points to tilted quadrilaterals, detected object data, and / or information derived therefrom). Additionally or alternatively, the communication manager 104 can manage communications within the object detection system 100.

[0038] Where communications are received and / or provided as network communications, the communications manager 104 may include a network interface that may utilize one or more wireless antennas ( Fig.11A The network interface may include one or more wireless antennas 1126 and / or one or more modems to communicate over one or more networks. For example, the network interface may be capable of communicating over Long Term Evolution (LTE), Wideband Code Division Multiple Access (WCDMA), Universal Mobile Telecommunications Service (UMTS), Global System for Mobile Communications (GSM), CDMA2000, etc. The network interface may also use one or more local area networks such as Bluetooth, Bluetooth Low Energy (LE), Z-wave, ZigBee, etc. and / or one or more low power wide area networks (LPWAN) such as Long Range Wide Area Network (LoRaWAN), SigFox, etc. to enable communication between objects in the environment (e.g., vehicles, mobile devices, etc.). However, for example, in the case where the object detection system 100 is implemented entirely on an autonomous vehicle (e.g., vehicle 1100), the communication manager 104 does not have to include a network interface. In some examples, one or more of the communications described herein may be between components of the computing device 1200 via Fig.12 The bus 1202 is carried out.

[0039] Sensor data received by the communications manager 104 may be used Fig.10 For example, the sensor data may include image data representing one or more images, image data representing a video (e.g., a video snapshot), and / or sensor data representing a field of view of a sensor (e.g., LIDAR data from one or more LIDAR sensors 1164, RADAR data from one or more RADAR sensors 1160, Fig. 11B image data from one or more cameras, etc.).

[0040] The sensor data and / or image data provided by the communication manager 104 to the object detector 106 may be generated in a physical or virtual environment and may include image data representing a field of view of a camera. For example, in aspects of the present disclosure, the communication manager 104 provides the object detector 106 with image data generated by a camera of the vehicle 1100 in the physical environment.

[0041] Although some examples of machine learning models that may be used for object detector 106 and / or other components described herein may refer to a particular type of machine learning model (e.g., a neural network), it is contemplated that the machine learning models described herein may, for example but not limited to, include any type of machine learning model, such as machine learning models using linear regression, logistic regression, decision trees, support vector machines (SVM), naive Bayes, k-nearest neighbors (Knn), K-means clustering, random forests, dimensionality reduction algorithms, gradient boosting algorithms, neural networks (e.g., autoencoders, convolutions, recurrent, perceptrons, long / short term memory (LSTM), Hopfield, Boltzmann, deep belief, deconvolution, generative adversarial, liquid machines, etc.), and / or other types of machine learning models.

[0042] refer to Figure 3 , Figure 3 is an illustration of an image represented by image data that may be processed by an object detector, a grid of spatial elements of the object detector, and a set of anchor boxes that may be associated with one or more of the spatial elements according to some embodiments of the present disclosure. For example, Figure 3 A depiction of an image 310 that may be generated by a camera of a vehicle 1100 in a physical environment and provided to an object detector 106, which may analyze the image data to generate object detection data. The object detection data may represent detections of objects (also referred to as detected objects) in the image 310 by the object detector 106. The detected objects may or may not correspond to actual objects depicted in the image 310. For example, some detected objects may correspond to false detections made by the object detector 106. Furthermore, some detected objects may correspond to the same objects depicted in the image 310.

[0043] The object detector 106 may include one or more machine learning models that are trained to generate object detection data based on features extracted from sensor data (e.g., image data). In some examples, the object detector 106 is configured to determine a set of object detection data (e.g., confidence values ​​and displacement values ​​representing corner points) for each spatial element of the field of view and / or image and / or one or more corresponding anchor boxes thereof. In various examples, the spatial element may also refer to a grid cell, an output cell, a super pixel, and / or an output pixel of the object detector 106.

[0044] In various examples, the spatial elements may form a grid of spatial element regions. For example, Figure 3 A grid 312 visually indicates spatial elements of the object detector 106 that can be logically applied to the sensor data (e.g., representing the image 310). Figure 3 , the grid 312 is depicted separately from the image 310 so as not to obscure the image 310, and Figure 4 An overlay depiction 402 is provided in FIG. 4. A spatial element, such as a grid cell 311, may be defined by a position in a grid. For example, each grid cell may contain a spatial element region of a spatial element. In other examples, grid-based spatial elements may not be used. Furthermore, a spatial element may not necessarily define a continuous spatial element region, may not necessarily define a rectangular-shaped spatial element region, and / or may not cover all areas of a field of view and / or an image.

[0045] In some examples, for a single image or frame (e.g., image 310) or a group of images or frames, each spatial element of the object detector 106 may provide object detection data for one or more corresponding anchor boxes. In other examples, one or more spatial elements may not provide object detection data. The object detection data may represent, for example, a confidence value 118, a displacement value 122, and / or a confidence value 116 for each anchor box of the spatial element, which may or may not correspond to the field of view and / or the parking space in the image 310.

[0046] Figure 3 A set of anchor boxes 314 is shown, where each spatial element applied to the image 310 can be associated with a corresponding set of anchor boxes 314. Eight anchor boxes are shown, but any number of anchor boxes can be used for spatial elements and the anchor boxes used for different spatial elements can differ from each other in shape, size, number, etc. The anchor boxes can be of various sizes and shapes, such as regular rectangles (e.g., equiangular rectangles); compared to some traditional systems, the anchor boxes can also include one or more tilted quadrilaterals, such as irregular rectangles (e.g., no congruent angles); rhombuses; kites; trapezoids; parallelograms; isosceles trapezoids; tilted quadrilaterals; or any combination thereof. In Figure 3In FIG. 3 , the anchor box 314 is depicted as being separate from the image 310 so as not to obscure the image 310 and in FIG. Figure 4 An overlay depiction 402 is provided.

[0047] As stated in this article, Figure 4 An overlay depiction 402 is provided in which an image 310 is overlaid with a grid 312, and an anchor box 314 of a single spatial element is located at a grid cell 311, which indicates a corresponding position relative to the image 310. As described herein, a confidence score and a displacement value may be generated for each grid cell and / or each anchor box of a spatial element. For example, the anchor box 314 is depicted for only one grid cell 311, and in other aspects, the anchor box 314 (or a variation thereof) may be used for multiple grid cells of the grid 312, or for each grid cell of the grid 312. The anchor boxes 314 of different grid cells 311 may be at positions corresponding to the grid cells 311 (or more generally, spatial elements). The grid 312 is an example of one size or resolution of a spatial element. As a non-limiting example, the grid 312 is 10x6 with 60 grid cells, so if each grid cell is associated with eight anchor boxes, confidence scores and displacement values ​​may be generated for 480 different anchor boxes.

[0048] In other aspects, the grid or other arrangement of spatial element regions can have different sizes or resolutions, with more spatial regions or fewer spatial regions, in which case the proportion of anchor boxes can be increased (e.g., a coarser grid with fewer, larger spatial regions) or decreased (e.g., a finer grid with more, smaller spatial regions). For example, the overlay depiction 404 includes an image 310 overlaid with a courser resolution grid 412 (e.g., 2×2) and with a set of different anchor boxes 414, which can be consistent with the anchor box 314 (e.g., the same shape and size), similar to the anchor box 314 (e.g., the same shape and / or different size), or different from the anchor box 314 (e.g., different shape and / or different size). In some aspects of the present disclosure, the object detector can apply multiple resolutions of spatial element regions (e.g., a grid) to the same input data, each spatial element region corresponding to a corresponding set of anchor boxes. Among other potential advantages, using multiple resolutions can increase the likelihood that the object detector 106 is accurate for both larger parking spaces and smaller parking spaces, whether in the same image (e.g., based on perspective, parking spaces closer to the camera may appear larger and parking spaces farther from the camera may appear smaller) or different images. In some cases, the actual set of spatial element regions (e.g., grids) used to analyze the input data may be much finer in resolution than the grids 312 and 412. Furthermore, any number of sets of spatial element regions may be employed.

[0049] As described herein, based on object detection data provided by object detector 106, slanted quadrilateral generator 114 may generate and / or identify one or more slanted quadrilaterals corresponding to one or more parking spaces and entrance determiner 126 may determine and / or identify one or more entrances to the one or more parking spaces.

[0050] refer to Figure 5A , Figure 5A At least a portion of an example object detector 106 implemented using a neural network (e.g., CNN) is depicted. For example, the object detector 106 includes a feature backbone network 506, such as ResNet 50 or another feature backbone network. In addition, the neural network includes a feature pyramid network 508. In addition, the neural network includes a classification subnetwork 510.

[0051] In an embodiment, the feature backbone network 506 and the feature pyramid network 508 may correspond to Figure 1 The feature determiner 108, the classification subnetwork 510 may correspond to Figure 1 The confidence score generator 110 of , the regression subnetwork 512 may correspond to Figure 1 The displacement value generator 112. However, Figure 5A The description of the neural network in is not intended to limit the object detector 106 to the neural network shown. In addition, the classification subnetwork 510 is shown as outputting a representation confidence score 514 (which may correspond to Figure 2 Although not shown for simplicity, in an embodiment of detecting a parking space entrance, the classification subnetwork 510 may additionally or alternatively output a representation of Figure 1 The confidence score 116 may be used to represent the data, or another classification subnetwork may be used. The regression subnetwork 512 is shown as outputting a displacement value 516 (which may correspond to Figure 2 The displacement value in 122) is shown in Figure 1. Figure 5A The output of the object detector 106 described in can be provided to each predefined anchor box.

[0052] In another aspect of the present disclosure, the tilted quadrilateral generator 518 may correspond to Figure 1The slanted quadrilateral generator 114 in the apparatus may generate and / or identify one or more slanted quadrilaterals based on the output from the object detector 106. For example, based on the displacement values ​​516 (e.g., Δx1, Δy1..., Δx4, Δy4) and the confidence values ​​514, the slanted quadrilateral generator 518 may select an anchor box and adjust the corner positions or points (e.g., x1, y1..., x4, y4) of the anchor box 522 to generate corner points (e.g., the adjusted corner points 520 include [x'1, y'1..., x'4, y'4] of the slanted quadrilateral). Data representing the slanted quadrilateral (e.g., the adjusted corner points 520) may be provided to various downstream components or systems. As shown, in various embodiments, confidence map classification may be performed, such as classifying the anchor box as a positive or negative parking space detection (e.g., using binary classification), and the slanted quadrilateral generator 518 may utilize this information. For example, the object detection system 100 can compare the confidence value 514 of each anchor box with a threshold value. When the confidence value 514 is greater than the threshold value, the anchor box can generate a positive detection, and when the confidence value is less than the threshold value, a negative detection can be generated.

[0053] As a non-limiting example, the tilted quadrilateral generator 114 can generate and / or determine any number of tilted quadrilaterals by applying a clustering algorithm to the output of the object detector 106 of the detected object (e.g., after filtering out negative detections using the confidence value 514) to form any number of detected object clusters. In order to cluster the detected objects, the tilted quadrilateral generator 114 can cluster the positions of the detected objects (e.g., candidate tilted quadrilaterals) together. For example, this can be based at least in part on the confidence value 514 associated with the detected object and / or other detected object data described herein. In some examples, the tilted quadrilateral generator 114 uses a density-based application space clustering (DBSCAN) algorithm with noise. Other examples include NMS or a modified group rectangle algorithm. A tilted quadrilateral can be selected, determined, and / or generated from each cluster as an output object detection (e.g., using one or more algorithms and / or neural networks).

[0054] Data representing the adjusted corner points 520 and / or each tilted quadrilateral determined by the tilted quadrilateral generator 114 can be provided to various downstream components or systems. For example, in one case, the corner points of the tilted quadrilateral can be provided to a vehicle control module, which can directly consume the corner points by converting the two-dimensional corner point coordinates to three-dimensional coordinates or otherwise processing the data, such as to coordinate a parking maneuver for the vehicle. In another aspect, the corner points of the tilted quadrilateral can be provided to an instrument cluster control module having a video or image monitor for displaying a representation of one or more parking spaces. For example, the corner points can be used to annotate the image 502 and / or a corresponding image or image 502 having the corner points depicted—e.g., Figure 5A The annotated image 525 in FIG. 5 has an outline (eg, indicated by a dashed line) 526 of a tilted quadrilateral having adjusted corner points 520 .

[0055] In another aspect, the corner points of the tilted quadrilateral, the displacement values ​​122, and / or the confidence values ​​(e.g., confidence values ​​116) corresponding to the corner points of the anchor box can be provided as input to the entrance determiner 126 to detect and / or define one or more entrances of one or more parking spaces (e.g., parking spaces identified by the tilted quadrilateral generator 114). For example, the entrance line can be detected and / or defined by selecting two corner points (e.g., four corner points) with the highest confidence values ​​in the confidence values ​​116 of the anchor box. In some examples, the selection can also be based on the confidence value being greater than a threshold (e.g., indicating that each of the corner points may correspond to an entrance). The entrance of the parking space can be defined as an entrance line or otherwise determined and / or defined using the positions of the selected corner points.

[0056] As such, the entrance information determined by the entrance determiner 126 can be provided to various downstream components or systems. For example, in some cases, the corner points identified as corresponding entrances can be provided to a vehicle control module, which can directly consume the corner points by converting the two-dimensional corner point coordinates to three-dimensional coordinates or otherwise processing the corner points. In another aspect, the corner points can be provided to an instrument cluster control module having a video or image monitor for displaying representations of one or more entrances to one or more parking spaces. For example, the corner points can be used to annotate the image 502 and / or a corresponding image or image 502 having the corner points and / or entrances depicted—e.g., Figure 5B The annotated image 530 in FIG. 5 may include a depiction of an entrance line to a parking space (e.g., indicated by a dashed line) 532. Optionally, a parking space outline 534 (e.g., a dashed line) may also be provided. In one example, the outline of the entrance and / or parking space may include a colored line or other suitable image annotation.

[0057] Example of training a machine learning model for object detection

[0058] The object detector 106 can be trained using a variety of possible methods. In some examples, the object detector 106 can be trained in a fully supervised manner. The training images and their labels can be grouped in mini-batches, where the size of the mini-batch can be an adjustable hyperparameter. Each mini-batch can be passed to an online data augmentation layer, which can apply a transformation to the images in the mini-batch. Data augmentation can be used to mitigate possible overfitting of the training data by the object detector 106. The data augmentation transformation can include (but is not limited to) spatial transformations such as left-right flipping, zooming in / out, random translation, etc., color transformations such as hue, saturation, and contrast adjustment, or additive noise. The labels can be transformed to reflect the corresponding transformations made to the training images.

[0059] The augmented image may be passed to the object detector 106 to perform a forward pass computation. The object detector 106 may perform feature extraction and prediction (e.g., prediction associated with an anchor box) on a per-spatial element basis. In the task of predicting the various outputs (e.g., the confidence value and displacement value of each anchor box), the loss function may simultaneously measure the error.

[0060] The component losses for the various outputs may be combined together in a single loss function applied to the entire mini-batch. A backward pass computation may then be performed to recursively compute the gradient of the cost function with respect to the trainable parameters (typically at least the weights and biases of the object detector 106, but not limited thereto, as there may be other trainable parameters when batch normalization is used, for example). The forward and backward pass computations may typically be handled by a deep learning framework and the software stack below.

[0061] Then, a parameter update for the object detector 106 can occur. An optimizer can be used to adjust the trainable parameters. Examples include stochastic gradient descent, or stochastic gradient descent with a momentum term. The main hyperparameter connected to the optimizer can be the learning rate. Depending on the optimizer, there may also be other hyperparameters.

[0062] For each epoch during training, the images in the dataset may be presented in a random order, which may result in faster convergence. An epoch may refer to the number of forward / backward pass iterations used to show each image of the dataset once to the object detector 106 being trained. The entire process "forward pass - backward pass - parameter update" may be iterated until the trained parameters converge. Convergence may be assessed by observing that the value of the loss function decreases to a sufficiently low value on the training set and validation set and determining that further iterations do not further reduce the loss. Other metrics may be used to assess convergence, such as the average precision calculated on the validation set.

[0063] During training, validation may be performed periodically, which may involve checking the average of the loss function for images in a validation set (separate from the training set). As described herein, each output of the object detector 106 (e.g., confidence scores for each anchor box, displacement values ​​for each anchor box, etc.) may be associated with a separate loss function for training. Any suitable loss function may be used.

[0064] According to one aspect of the present disclosure, the ground truth data of the parking space may include the corner position of the parking space, and the corner position may form or define a tilted quadrilateral. In addition, when the tilted quadrilateral angle of the anchor frame is sufficiently similar to the ground truth angle position, for example, based on the matching cost being less than a threshold, a positive training sample can be identified from the output of the object detector 106. In one aspect of the present disclosure, various types of anchor frames can be used to train the neural network and identify positive samples. For example, in one aspect, the predefined anchor frame may include a rectangle (e.g., a rectangle). In addition, the predefined anchor frame may include a rotated rectangle. In addition or alternatively, one or more anchor frames may include tilted and rotated rectangles. Examples of tilted rectangles include irregular rectangles (e.g., no congruent angles); rhombuses; kites; trapezoids; parallelograms; isosceles trapezoids; tilted quadrilaterals; and any combination thereof. The predefined anchor frame may be manually designed or obtained from ground truth labels and may be used to calculate ground truth displacement values ​​for training the object detector 106. An anchor box obtained from the ground truth labeling can be referred to as a "data-driven anchor box", which is generated by clustering or otherwise analyzing the ground truth samples. For example, ground truth samples (e.g., including tilted quadrilaterals) can be generated for one or more images. The ground truth samples can then be clustered into one or more clusters, and at least one data-driven anchor box can be generated, selected and / or determined from the samples of each of the one or more clusters. In some examples, the data-driven anchor box can have a shape calculated from one or more samples of the cluster (e.g., corresponding to an average or otherwise statistically derived shape of the cluster). In various examples, spectral clustering can be performed, such as by calculating an affinity matrix for the ground truth samples using a shape similarity function, and performing spectral clustering using the affinity matrix with k clusters, where k is the number of clusters to be generated.

[0065] In one aspect, the matching cost for identifying positive samples from the output of the object detector 106 is based at least in part on the minimum converged distance between the predefined anchor box corners and the ground truth corner positions adjusted by the corresponding displacement values ​​output by the object detector 106. This is in contrast to determining positive samples based on intersection over union (IOU), and may be more straightforward than IOU because the compared corner points may not define a regular rectangle (but rather a skewed quadrilateral).

[0066] The minimum aggregation distance can be calculated in various ways. For example, refer to Figure 6 , depicting an image 610 showing ground truth corner points (B1, B2, B3, and B4) of a depicted parking space 602. The image 610 may be used as a training input for the object detector 106. As a result, the object detector 106 may provide displacement values ​​to the corner points of the anchor boxes for calculating the adjusted corner points (A1, A2, A3, and A4) of the anchor boxes, as shown. Figure 6 Only the corner points of a single anchor box are shown to simplify the illustration, and otherwise similar information is available for each anchor box described herein.

[0067] In one aspect of the present disclosure, calculating the minimum aggregation distance includes calculating the minimum average distance. For example, the first aggregation distance can be calculated by determining the distances between (A1, B1), (A2, B2), (A3, B3), and (A4, B4), and then deriving the first aggregation distance from these distance statistics, such as using the mean. The second, third, and fourth aggregation distances can also be calculated by changing the association between the corner points of each data set (e.g., for each possible combination)-for example, using (A1, B2), (A2, B3), (A3, B4), and (A4, B1) to calculate the second aggregation distance; using (A1, B3), (A2, B4), (A3, B1), and (A4, B2) to calculate the third aggregation distance; and using (A1, B4), (A2, B1), (A3, B2), and (A4, B3) to calculate the fourth aggregation distance. The minimum aggregation distance can then be selected from the various aggregation distances and used to determine whether the anchor box is a positive training sample (e.g., similar to IOU). For example, a positive sample may be selected based at least in part on an average aggregate distance being less than a threshold. In other aspects, an average mean distance or other statistical quantification may be selected and used to determine whether the matching cost is less than a threshold.

[0068] In some aspects of the present disclosure, a minimum aggregation distance may be determined for any number of anchor boxes associated with the object detector 106 to determine whether the anchor box corresponds to a positive sample for training. The confidence value 118 may be used to filter the anchor boxes from being considered as positive samples. For example, a minimum aggregation distance may be determined for an anchor box based at least in part on the confidence value 118 associated with the anchor box. In some examples, a minimum aggregation distance may be determined for each anchor box having a confidence value 118 that exceeds a threshold (e.g., indicating a positive detection).

[0069] In another aspect of the present disclosure, the minimum aggregation distance of each anchor box can be normalized based at least in part on the size and / or area defined by the ground truth corner points (e.g., the ground truth slanted quadrilateral). The normalized minimum aggregation distance can be used to account for size differences between anchor boxes, for example, when different anchor box sizes and / or spatial element area (e.g., grid) resolutions are used. According to the present disclosure, when the matching cost (e.g., based at least in part on the normalized minimum aggregation distance) is less than a certain (e.g., predetermined) threshold, the anchor box can be identified as a positive sample. The positive sample can then be used to update the parameters of the object detector 106 (e.g., CNN) being trained.

[0070] Reference now Figure 7 , Figure 7 is a flowchart illustrating a method 700 for training a machine learning model to provide corner points of parking spaces according to some embodiments of the present disclosure. Each box of method 700, as well as other methods described herein, includes a computational process that can be performed using any combination of hardware, firmware, and / or software. For example, various functions can be performed by a processor that executes instructions stored in a memory. Method 700 can also be embodied as computer-usable instructions stored on a computer storage medium. Method 700 can be provided by a standalone application, a service or a hosted service (standalone or in combination with another hosted service), or a plug-in for another product, to name a few. The methods described herein may be performed additionally or alternatively by any one system or any combination of systems, including but not limited to those described herein and not limited to specific examples.

[0071] At block B702, method 700 includes applying image data representing a parking space to a neural network. For example, image 502 may be applied to object detector 106, image 502 depicting at least one parking space.

[0072] At block B704 , the method 700 includes receiving data generated from the image data and representing displacement values ​​to corner points of the anchor shape using a neural network. For example, the regression subnetwork 512 may output displacement values ​​516 associated with the predefined anchor box 522 and generated from the image data representing the image 502 .

[0073] At block B706, method 700 includes determining the corner points of the tilted polygon from the displacement values ​​to the corner points of the anchor shape. For example, the tilted quadrilateral generator 518 (or at least other components for training) can determine the adjusted corner points 520 of the tilted quadrilateral based on the displacement values ​​516 associated with the predefined anchor box 522.

[0074] At block B708, the method 700 includes calculating a first distance between the corner points of the inclined polygon and the ground truth corner points of the parking space. For example, the minimum convergence distance between (A1, A2, A3, and A4) and (B1, B2, B3, and B4) may be calculated, such as Figure 6 as described.

[0075] At block B710, method 700 includes determining a sample ranking based on the first distance. For example, the sample ranking may be the minimum aggregate distance or some derivative thereof (eg, based on ground truth size normalization).

[0076] At block B712, method 700 includes updating parameters of the neural network using the anchor shape as a positive training sample based on the sample rating exceeding (e.g., falling below) a threshold. For example, an anchor can be defined as a positive sample when the matching cost (e.g., based on the sample rating) is less than a threshold.

[0077] Reference now Figure 8 , Figure 8 8 is a flow chart illustrating a method 800 for determining corner points of a parking space using a neural network according to some embodiments of the present disclosure. At block B802, the method 800 includes applying sensor data representing a field of view of at least one sensor in the environment to the neural network. For example, sensor data representing an image 502 representing a field of view of a camera of the vehicle 1100 may be applied to the object detector 106.

[0078] At block B804, the method 800 includes receiving, from the neural network, first data representing displacement values ​​to corner points of the anchor shape and second data representing a confidence value of a likelihood of predicting that the anchor shape corresponds to a parking space in a field of view of at least one sensor. For example, the regression subnetwork 512 may output data representing displacement values ​​516 associated with the predefined anchor box 522 and generated from the sensor data representing the image 502. Additionally, the classification subnetwork 510 may output data representing a confidence score 514 of a likelihood of predicting that the anchor box 522 corresponds to a parking space in the image 502.

[0079] At block B806, method 800 includes determining corner points of the tilted polygon based on displacement values ​​to the corner points of the anchor shape based on the confidence value exceeding the threshold. For example, the tilted quadrilateral generator 518 may determine data representing adjusted corner points 520 of the tilted quadrilateral based at least in part on the confidence value 514 exceeding the threshold based on the displacement values ​​516 associated with the predefined anchor shape 522, such as Figure 5A shown.

[0080] Reference now Fig. 9 , Fig. 9is a flow chart illustrating a method 900 for determining a parking space entrance using a neural network according to some embodiments of the present disclosure. At block B 902, the method 900 includes applying sensor data representing a field of view of at least one sensor in the environment to the neural network. For example, sensor data representing image 502 may be applied to object detector 106, the sensor data representing a field of view of a camera of vehicle 1100.

[0081] At block B904, method 900 includes receiving first data and second data generated from the image data from a neural network. The first data represents displacement values ​​to corner points of an anchor shape, and the second data represents a confidence value predicting the likelihood that the corner points of the anchor shape define a parking space entrance in the field of view of at least one sensor. For example, regression subnetwork 512 may output data representing displacement values ​​516 associated with predefined anchor boxes 522 and generated from sensor data. Additionally, classification subnetwork 510 (or another similar network) may output Figure 2 The confidence score 116 predicts the likelihood that the corner point of the anchor box represents at least a portion of the parking space entrance.

[0082] At block B906 , the method 900 includes selecting a subset of corner points of the anchor shape based on the confidence value. For example, the entry determiner 126 may filter the corner points to determine and / or select the corner points with the highest confidence scores 116 .

[0083] At block B908, method 900 includes identifying an entrance to the parking space from the subset of corner points. For example, once two corner points with the highest confidence scores are selected, entrance determiner 126 may be designated to define an entrance and / or entrance line to the parking space.

[0084] Example operating environment

[0085] According to some embodiments of the present disclosure, the object detection system 100 and / or Figure 5A The neural network can be Fig.10 The example operating environment 1000 is implemented in FIG.

[0086] In addition to other components not shown, operating environment 1000 includes one or more client devices 1020, one or more networks 1040, one or more server devices 1060, one or more sensors 1080, and one or more data stores 1050. It should be understood that Fig.10 Operating environment 1000 shown in FIG. 1 is an example of a suitable operating environment. Fig.10 Each component shown in can be implemented via any type of computing device, such as, for example, a combination of Fig.12One or more computing devices 1200 described. These components can communicate with each other via a network 1040, which can be wired, wireless, or both. The network 1040 may include multiple networks or a network of networks, but is shown in a simplified form so as not to obscure aspects of the present disclosure. For example, the network 1040 may include one or more wide area networks (WANs), one or more local area networks (LANs), one or more public networks such as the Internet, and / or one or more private networks. In the case where the network 1040 includes a wireless telecommunications network, components such as base stations, communication towers, or even access points (and other components) can provide wireless connections. In any example, at least one network 1040 can be connected to a wireless network as further described below. Fig.11D Corresponding to one or more networks 1190.

[0087] It should be understood that within the scope of the present disclosure, any number of client devices 1020, server devices 1060, sensors 1080, and data stores 1050 may be employed within operating environment 1000. Each may be configured as a single device or as multiple devices cooperating in a distributed environment.

[0088] One or more client devices 1020 may include the Fig.12 At least some of the components, features, and functions of the example computing device 1200 described herein. By way of example and not limitation, the client device 1020 may be embodied as a personal computer (PC), a laptop computer, a mobile device, a smart phone, a tablet computer, a smart watch, a wearable computer, a personal digital assistant (PDA), an MP3 player, a global positioning system (GPS) or device, a video player, a handheld communication device, a gaming device or system, an entertainment system, an in-vehicle computer system, an embedded system controller, a remote control, an appliance, a consumer electronic device, a workstation, any combination of these depicted devices, or any other suitable device. In any example, at least one client device 1020 may be such as described in further detail herein. Figures 11A-11D A portion of a vehicle such as vehicle 1100.

[0089] One or more client devices 1020 may include one or more processors and one or more computer-readable media. The computer-readable media may include computer-readable instructions that can be executed by one or more processors. These instructions, when executed by one or more processors, may cause the one or more processors to perform any combination and / or portion of the methods described herein, and / or implement Figure 1 Any part of the functionality of object detection system 100.

[0090] The one or more server devices 1060 may also include one or more processors and one or more computer-readable media. The computer-readable media includes computer-readable instructions that can be executed by one or more processors. These instructions, when executed by one or more processors, may cause the one or more processors to perform any combination and / or portion of the methods described herein, and / or implement Figure 1 Any portion of the functionality of the object detection system 100. In any example, at least one server device 1060 may be associated with a server device described in further detail herein. Fig.11D Corresponding to one or more servers 1178.

[0091] The one or more data stores 1050 may include one or more computer-readable media. The computer-readable media may include computer-readable instructions that can be executed by one or more processors. These instructions, when executed by one or more processors, may cause the one or more processors to perform any combination and / or portion of the methods described herein, and / or implement Figure 1 Any portion of the functionality of object detection system 100 of A. One or more data stores 1050 (or computer data storage devices) are depicted as a single component, but may be embodied as one or more data stores (e.g., databases) and may be at least partially in the cloud. One or more of the one or more data stores 1050 may be connected to Fig. 11C corresponds to one or more of the data stores 1050 .

[0092] Although depicted as being external to the one or more server devices 1060 and the one or more client devices 1020, the one or more data stores 1050 may be implemented at least in part on any combination of the one or more server devices 1060 and / or the one or more client devices 1020 (e.g., as memory 1204 ( Fig.12 )). For example, some information may be stored on one or more client devices 1020, and other and / or duplicate information may be stored externally (e.g., on one or more server devices 1060). Thus, it should be appreciated that the information in the one or more data stores 1050 may be distributed across one or more data stores for storage (which may be hosted externally) in any suitable manner. For example, the one or more data stores 1050 may include at least some of the one or more computer-readable media of the one or more server devices 1060 and / or at least some of the one or more computer-readable media of the one or more client devices 1020.

[0093] The one or more sensors 1080 include at least one sensor capable of generating sensor data representative of at least some aspects of the environment. For example, the one or more sensors 1080 may generate Figure 1 A's sensor data 102. The one or more sensors 1080 may include one or more global navigation satellite system (GNSS) sensors (e.g., global positioning system (GPS) sensors), RADAR sensors, ultrasonic sensors, LIDAR sensors, inertial measurement unit (IMU) sensors (e.g., accelerometers, gyroscopes, magnetic compasses, magnetometers, etc.), microphones, stereo cameras, wide-angle cameras (e.g., fisheye cameras), infrared cameras, surround cameras (e.g., 360-degree cameras), long-range and / or mid-range cameras, speed sensors (e.g., for measuring the speed of the vehicle 1100), vibration sensors, steering sensors, brake sensors (e.g., as part of a brake sensor system), and / or any combination of other sensor types.

[0094] Reference Figures 11A-11C , the sensor data 102 may be generated by, for example, but not limited to, one or more global navigation satellite system (GNSS) sensors 1168 (e.g., global positioning system sensors), RADAR sensors 1160, ultrasonic sensors 1162, LIDAR sensors 1164, inertial measurement unit (IMU) sensors 1166 (e.g., accelerometers, gyroscopes, magnetic compasses, magnetometers, etc.), microphones 1196, stereo cameras 1168, wide-angle cameras 1170 (e.g., fisheye cameras), infrared cameras 1172, surround cameras 1174 (e.g., 360-degree cameras), long-range and / or mid-range cameras 1198, speed sensors 1144 (e.g., for measuring the speed of the vehicle 1100), vibration sensors 1142, steering sensors 1140, brake sensors (e.g., as part of a brake sensor system 1146), and / or other sensor types.

[0095] In some examples, sensor data 102 may be generated by forward-facing and / or side-facing cameras such as wide-angle camera 1170, surround camera 1174, stereo camera 1168, and / or long-range or mid-range camera 1198. In some examples, more than one camera or other sensor may be used to combine multiple fields of view (e.g., Fig. 11B field of view of remote camera 1198, forward stereo camera 1168 and / or forward wide-angle camera 1170).

[0096] Example Autonomous Vehicle

[0097] Fig.11A1 is a diagram of an example autonomous vehicle 1100 according to some embodiments of the present disclosure. Autonomous vehicle 1100 (alternatively, referred to herein as "vehicle 1100") may include a passenger vehicle, such as a car, a truck, a bus, and / or another type of vehicle that accommodates one or more passengers. Autonomous vehicles are generally described according to the automation levels defined by the National Highway Traffic Safety Administration (NHTSA), a division of the U.S. Department of Transportation, and the Society of Automotive Engineers (SAE) "Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles" (Standard No. J3016-201806 issued on June 15, 2018, Standard No. J3016-201609 issued on September 30, 2016, and previous and future versions of the standard). Vehicle 1100 may be able to implement one or more functions that meet the autonomous driving level 3-5. For example, depending on the embodiment, the vehicle 1100 may be capable of conditional automation (Level 3), high automation (Level 4), and / or full automation (Level 5).

[0098] The vehicle 1100 may include components such as a chassis, a body, wheels (e.g., 2, 4, 6, 8, 18, etc.), tires, axles, and other components of the vehicle. The vehicle 1100 may include a propulsion system 1150, such as an internal combustion engine, a hybrid power plant, an all-electric engine, and / or another propulsion system type. The propulsion system 1150 may be connected to a drive train of the vehicle 1100, which may include a transmission, to achieve propulsion of the vehicle 1100. The propulsion system 1150 may be controlled in response to receiving a signal from a throttle / accelerator 1152.

[0099] A steering system 1154, which may include a steering wheel, may be used to steer the vehicle 1100 (e.g., along a desired path or route) when the propulsion system 1150 is operating (e.g., when the vehicle is in motion). The steering system 1154 may receive signals from a steering actuator 1156. For fully automated (Level 5) functionality, a steering wheel may be optional.

[0100] Brake sensor system 1146 may be used to operate vehicle brakes in response to receiving signals from brake actuator 1148 and / or brake sensors.

[0101] May include one or more system on chip (SoC) 1104 ( Fig. 11C) and / or one or more controllers 1136 of one or more GPUs may provide signals (e.g., representing commands) to one or more components and / or systems of the vehicle 1100. For example, the one or more controllers may send signals to operate the vehicle brakes via one or more brake actuators 1148, to operate the steering system 1154 via one or more steering actuators 1156, and to operate the propulsion system 1150 via one or more throttles / accelerators 1152. The one or more controllers 1136 may include one or more onboard (e.g., integrated) computing devices (e.g., supercomputers) that process sensor signals and output operating commands (e.g., signals representing commands) to enable autonomous driving and / or assist a human driver in driving the vehicle 1100. The one or more controllers 1136 may include a first controller 1136 for autonomous driving functions, a second controller 1136 for functional safety functions, a third controller 1136 for artificial intelligence functions (e.g., computer vision), a fourth controller 1136 for infotainment functions, a fifth controller 1136 for redundancy in emergency situations, and / or other controllers. In some examples, a single controller 1136 may handle two or more of the above functions, two or more controllers 1136 may handle a single function, and / or any combination thereof.

[0102] The one or more controllers 1136 may provide signals for controlling one or more components and / or systems of the vehicle 1100 in response to sensor data (e.g., sensor input) received from one or more sensors. The sensor data may be received from, for example and without limitation, a global navigation satellite system sensor 1158 (e.g., a global positioning system sensor), a RADAR sensor 1160, an ultrasonic sensor 1162, a LIDAR sensor 1164, an inertial measurement unit (IMU) sensor 1166 (e.g., an accelerometer, a gyroscope, a magnetic compass, a magnetometer, etc.), a microphone 1196, a stereo camera 1168, a wide angle camera 1170 (e.g., a fisheye camera), an infrared camera 1172, a surround camera 1174 (e.g., a 360 degree camera), a long-range and / or mid-range camera 1198, a speed sensor 1144 (e.g., for measuring the velocity of the vehicle 1100), a vibration sensor 1142, a steering sensor 1140, a brake sensor (e.g., as part of a brake sensor system 1146), and / or other sensor types.

[0103] One or more of the controllers 1136 may receive inputs (e.g., represented by input data) from the instrument cluster 1132 of the vehicle 1100 and provide outputs (e.g., represented by output data, display data, etc.) via a human machine interface (HMI) display 1134, an audible annunciator, a speaker, and / or via other components of the vehicle 1100. These outputs may include information such as vehicle speed, velocity, time, map data (e.g., Fig. 11C The HMI display 1134 may include information such as the HD map 1122 of the vehicle 1100, location data (e.g., the location of the vehicle 1100 on the map), directions, locations of other vehicles (e.g., occupancy grid), information about objects and states of objects as sensed by the controller 1136, etc. For example, the HMI display 1134 may display information about the presence of one or more objects (e.g., street signs, warning signs, traffic light changes, etc.) and / or information about driving maneuvers that the vehicle has made, is making, or will make (e.g., changing lanes now, leaving 34B in two miles, etc.).

[0104] The vehicle 1100 further includes a network interface 1124 that can communicate via one or more networks using one or more wireless antennas 1126 and / or a modem. For example, the network interface 1124 may be capable of communicating via LTE, WCDMA, UMTS, GSM, CDMA2000, etc. The one or more wireless antennas 1126 may also enable communication between objects in the environment (e.g., vehicles, mobile devices, etc.) using one or more local area networks such as Bluetooth, Bluetooth LE, Z-wave, ZigBee, etc. and / or one or more low power wide area networks (LPWANs) such as LoRaWAN, SigFox, etc.

[0105] Fig. 11B For use according to some embodiments of the present disclosure Fig.11A 100. The cameras and respective fields of view are an example embodiment and are not intended to be limiting. For example, additional and / or replaceable cameras may be included and / or the cameras may be located at different locations on the vehicle 1100.

[0106] The camera type for the camera may include, but is not limited to, a digital camera that may be suitable for use with components and / or systems of the vehicle 1100. The camera may operate at an automotive safety integrity level (ASIL) B and / or at another ASIL. The camera type may have any image capture rate, such as 60 frames per second (fps), 1520fps, 240fps, etc., depending on the embodiment. The camera may be capable of using a rolling shutter, a global shutter, another type of shutter, or a combination thereof. In some examples, the color filter array may include a red-white-white-white (RCCC) color filter array, a red-white-white-blue (RCCB) color filter array, a red-blue-green-white (RBGC) color filter array, a FoveonX3 color filter array, a Bayer sensor (RGGB) color filter array, a monochrome sensor color filter array, and / or another type of color filter array. In some embodiments, a clear pixel camera such as a camera with an RCCC, RCCB, and / or RBGC color filter array may be used in an effort to improve light sensitivity.

[0107] In some examples, one or more of the cameras can be used to perform advanced driver assistance system (ADAS) functions (e.g., as part of a redundant or fail-safe design). For example, a multi-function monocular camera can be installed to provide functions including lane departure warning, traffic sign assistance, and intelligent headlight control. One or more of the cameras (e.g., all cameras) can simultaneously record and provide image data (e.g., video).

[0108] One or more of the cameras may be mounted in a mounting assembly such as a custom designed (3-D printed) assembly to cut off stray light and reflections from within the car (e.g., reflections from the dashboard reflected in the windshield mirror) that may interfere with the camera's image data capture capabilities. With respect to the wing mirror mounting assembly, the wing mirror assembly may be custom 3-D printed so that the camera mounting plate matches the shape of the wing mirror. In some examples, one or more cameras may be integrated into the wing mirror. For side view cameras, one or more cameras may also be integrated into the four pillars at each corner of the cab.

[0109] A camera (e.g., a front-facing camera) having a field of view that includes a portion of the environment in front of the vehicle 1100 can be used for surround vision to help identify the forward path and obstacles, as well as assist in providing information critical to generating an occupancy grid and / or determining a preferred vehicle path with the help of one or more controllers 1136 and / or control SoCs. The front-facing camera can be used to perform many of the same ADAS functions as LIDAR, including emergency braking, pedestrian detection, and collision avoidance. The front-facing camera can also be used for ADAS functions and systems, including lane departure warning ("LDW"), autonomous cruise control ("ACC"), and / or other functions such as traffic sign recognition.

[0110] A variety of cameras may be used in the front-facing configuration, including, for example, a monocular camera platform including a CMOS (complementary metal oxide semiconductor) color imager. Another example may be a wide-angle camera 1170, which may be used to sense objects (e.g., pedestrians, intersection traffic, or bicycles) entering the field of view from the periphery. Fig. 11B Only one wide-angle camera is shown, but there may be any number of wide-angle cameras 1170 on the vehicle 1100. In addition, remote cameras 1198 (e.g., a long-view stereo camera pair) may be used for depth-based object detection, especially for objects for which a neural network has not been trained. Remote cameras 1198 may also be used for object detection and classification and basic object tracking.

[0111] One or more stereo cameras 1168 may also be included in the front configuration. The stereo camera 1168 may include an integrated control unit including an expandable processing unit that may provide a multi-core microprocessor and programmable logic (FPGA) with an integrated CAN or Ethernet interface on a single chip. Such a unit may be used to generate a 3-D map of the vehicle environment, including distance estimates for all points in the image. An alternative stereo camera 1168 may include a compact stereo vision sensor that may include two camera lenses (one on each side) and an image processing chip that may measure the distance from the vehicle to the target object and use the generated information (e.g., metadata) to activate autonomous emergency braking and lane departure warning functions. Other types of stereo cameras 1168 may be used in addition to or alternatively to those described herein.

[0112] A camera having a field of view that includes portions of the environment to the sides of the vehicle 1100 (e.g., a side view camera) can be used for surround viewing, providing information used to create and update an occupancy grid and generate side impact collision warnings. Fig. 11BThe four surround cameras 1174 shown in FIG. 1174 may be placed on the vehicle 1100. The surround cameras 1174 may include a wide-angle camera 1170, a fisheye camera, a 360-degree camera, and / or the like. For example, the four fisheye cameras may be placed in front, at the rear, and on the sides of the vehicle. In an alternative arrangement, the vehicle may use three surround cameras 1174 (e.g., left, right, and rear), and may utilize one or more other cameras (e.g., a forward-facing camera) as a fourth surround camera.

[0113] A camera having a field of view that includes a portion of the environment behind the vehicle 1100 (e.g., a rear view camera) can be used to assist with parking, surround view, rear collision warning, and creating and updating occupancy grids. A variety of cameras can be used, including but not limited to cameras that are also suitable as front cameras as described herein (e.g., long-range and / or mid-range cameras 1198, stereo cameras 1168, infrared cameras 1172, etc.).

[0114] Fig. 11C For use according to some embodiments of the present disclosure Fig.11A Block diagram of an example system architecture of an example autonomous vehicle 1100. It should be understood that this arrangement and other arrangements described herein are set forth merely as examples. Other arrangements and elements (e.g., machines, interfaces, functions, sequences, functional groupings, etc.) may be used in addition to or in place of those shown, and some elements may be omitted entirely. Further, many of the elements described herein are functional entities that may be implemented as discrete or distributed components or in combination with other components, and in any appropriate combination and location. The various functions described herein as being performed by an entity may be implemented by hardware, firmware, and / or software. For example, the various functions may be implemented by a processor executing instructions stored in a memory.

[0115] Fig. 11C Each of the components, features, and systems of the vehicle 1100 is illustrated as being connected via a bus 1102. The bus 1102 may include a controller area network (CAN) data interface (alternatively, referred to herein as a "CAN bus"). The CAN may be a network inside the vehicle 1100 that assists in controlling various features and functions of the vehicle 1100, such as the actuation of brakes, acceleration, braking, steering, windshield wipers, and the like. The CAN bus may be configured to have tens or even hundreds of nodes, each with its own unique identifier (e.g., CAN ID). The CAN bus may be read to find steering wheel angle, ground speed, engine speed per minute (RPM), button position, and / or other vehicle status indicators. The CAN bus may be ASIL B compatible.

[0116] Although bus 1102 is described as a CAN bus here, this is not intended to be restrictive. For example, in addition to or alternatively to the CAN bus, FlexRay and / or Ethernet can be used. In addition, although bus 1102 is represented by a single line, this is not intended to be restrictive. For example, there can be any number of buses 1102, which may include one or more CAN buses, one or more FlexRay buses, one or more Ethernet buses, and / or one or more other types of buses using different protocols. In some examples, two or more buses 1102 can be used to perform different functions, and / or can be used for redundancy. For example, a first bus 1102 can be used for a collision avoidance function, and a second bus 1102 can be used for drive control. In any example, each bus 1102 can communicate with any component of the vehicle 1100, and two or more buses 1102 can communicate with the same component. In some examples, each SoC 1104 , each controller 1136 , and / or each computer within the vehicle may have access to the same input data (eg, input from sensors of the vehicle 1100 ) and may be connected to a common bus such as a CAN bus.

[0117] The vehicle 1100 may include one or more controllers 1136, such as those described herein. Fig.11A Those controllers described. Controller 1136 can be used for a variety of functions. Controller 1136 can be coupled to any other different components and systems of vehicle 1100, and can be used for control of vehicle 1100, artificial intelligence of vehicle 1100, infotainment for vehicle 1100, and / or the like.

[0118] The vehicle 1100 may include one or more system on chip (SoC) 1104. The SoC 1104 may include a CPU 1106, a GPU 1108, a processor 1110, a cache 1112, an accelerator 1114, a data store 1116, and / or other components and features not shown. The SoC 1104 may be used to control the vehicle 1100 in a variety of platforms and systems. For example, one or more SoCs 1104 may be combined with an HD map 1122 in a system (e.g., a system of the vehicle 1100), and the HD map may be downloaded from one or more servers (e.g., a server) via a network interface 1124. Fig.11D one or more servers 1178) to obtain map refreshes and / or updates.

[0119] CPU1106 may include a CPU cluster or CPU complex (alternatively, referred to herein as "CCPLEX"). CPU1106 may include multiple cores and / or L2 caches. For example, in some embodiments, CPU1106 may include eight cores in a consistent multiprocessor configuration. In some embodiments, CPU1106 may include four dual-core clusters, each of which has a dedicated L2 cache (e.g., 2MB L2 cache). CPU1106 (e.g., CCPLEX) may be configured to support simultaneous cluster operations so that any combination of CPU1106 clusters can be active at any given time.

[0120] CPU1106 may implement power management capabilities including one or more of the following features: each hardware block may be automatically clock gated when idle to save dynamic power; each core clock may be gated when the core is not actively executing instructions due to the execution of WFI / WFE instructions; each core may be independently power gated; each core cluster may be independently clock gated when all cores are clock gated or power gated; and / or each core cluster may be independently power gated when all cores are power gated. CPU1106 may further implement an enhanced algorithm for managing power states, in which allowed power states and expected wake-up times are specified, and hardware / microcode determines the optimal power state to enter for the core, cluster, and CCPLEX. The processing core may support a simplified power state entry sequence in software, with the work being offloaded to the microcode.

[0121] GPU1108 may include an integrated GPU (alternatively, referred to herein as an "iGPU"). GPU1108 may be programmable and efficient for parallel workloads. In some examples, GPU1108 may use an enhanced tensor instruction set. GPU1108 may include one or more streaming microprocessors, each of which may include an L1 cache (e.g., an L1 cache with at least 96KB storage capacity), and two or more of these streaming microprocessors may share an L2 cache (e.g., an L2 cache with 512KB storage capacity). In some embodiments, GPU1108 may include at least eight streaming microprocessors. GPU1108 may use a computing application programming interface (API). In addition, GPU1108 may use one or more parallel computing platforms and / or programming models (e.g., NVIDIA's CUDA).

[0122] In the case of automotive and embedded use, GPU1108 can be power optimized to achieve optimal performance. For example, GPU1108 can be manufactured on fin field effect transistors (FinFETs). However, this is not intended to be limiting, and GPU1108 can be manufactured using other semiconductor manufacturing processes. Each streaming microprocessor can merge several mixed precision processing cores divided into multiple blocks. For example and without limitation, 64 PF32 cores and 32 PF64 cores can be divided into four processing blocks. In such an example, each processing block can be allocated 16 FP32 cores, 8 FP64 cores, 16 INT32 cores, two mixed precision NVIDIA tensor cores for deep learning matrix arithmetic, L0 instruction cache, thread beam (warp) scheduler, dispatch unit and / or 64KB register file. In addition, the streaming microprocessor may include independent parallel integer and floating point data paths to provide efficient execution of workloads using a mix of computation and addressing computations. The streaming microprocessor may include independent thread scheduling capabilities to allow for finer-grained synchronization and collaboration between parallel threads. A streaming microprocessor may include a combined L1 data cache and shared memory unit to increase performance while simplifying programming.

[0123] GPU 1108 can include high bandwidth memory (HBM) and / or a 16GB HBM2 memory subsystem that provides a peak memory bandwidth of approximately 900 GB / s in some examples. In some examples, synchronous graphics random access memory (SGRAM), such as fifth generation graphics double data rate synchronous random access memory (GDDR5), can be used in addition to or in lieu of HBM memory.

[0124] The GPU 1108 may include unified memory technology that includes access counters to allow memory pages to be more accurately migrated to the processor that accesses them most frequently, thereby improving the efficiency of memory ranges shared between processors. In some examples, address translation service (ATS) support may be used to allow the GPU 1108 to directly access the CPU 1106 page table. In such an example, when the GPU 1108 memory management unit (MMU) experiences a miss, an address translation request may be transmitted to the CPU 1106. In response, the CPU 1106 may look for a virtual-physical mapping for the address in its page table and transmit the translation back to the GPU 1108. In this way, unified memory technology may allow a single unified virtual address space to be used for the memory of both the CPU 1106 and the GPU 1108, thereby simplifying GPU 1108 programming and porting applications to the GPU 1108.

[0125] In addition, the GPU 1108 may include access counters that can track how often the GPU 1108 accesses the memory of other processors. The access counters can help ensure that memory pages are moved to the physical memory of the processor that accesses them most frequently.

[0126] SoC 1104 may include any number of caches 1112, including those described herein. For example, cache 1112 may include an L3 cache available to both CPU 1106 and GPU 1108 (e.g., connected to both CPU 1106 and GPU 1108). Cache 1112 may include a write-back cache that may track the state of a line, for example, by using a cache coherence protocol (e.g., MEI, MESI, MSI, etc.). Depending on the embodiment, the L3 cache may include 4MB or more, although smaller cache sizes may also be used.

[0127] SoC1104 may include one or more accelerators 1114 (e.g., hardware accelerators, software accelerators, or a combination thereof). For example, SoC1104 may include a hardware acceleration cluster, which may include optimized hardware accelerators and / or large on-chip memory. The large on-chip memory (e.g., 4MB SRAM) may enable the hardware acceleration cluster to accelerate neural networks and other calculations. The hardware acceleration cluster may be used to supplement GPU1108 and offload some tasks of GPU1108 (e.g., freeing up more cycles of GPU1108 for performing other tasks). As an example, accelerator 1114 may be used for targeted workloads (e.g., perception, convolutional neural networks (CNNs), etc.) that are stable enough to easily control acceleration. When used herein, the term "CNN" may include all types of CNNs, including region-based or regional convolutional neural networks (RCNNs) and fast RCNNs (e.g., for object detection).

[0128] Accelerator 1114 (e.g., a hardware acceleration cluster) may include a deep learning accelerator (DLA). DLA may include one or more tensor processing units (TPUs) that may be configured to provide an additional 10 trillion operations per second for deep learning applications and reasoning. TPU may be an accelerator configured to perform image processing functions (e.g., for CNN, RCNN, etc.) and optimized for performing image processing functions. DLA may be further optimized for a specific set of neural network types and floating point operations and reasoning. The design of DLA may provide higher performance per millimeter than a general-purpose GPU, and far exceeds the performance of a CPU. TPU may perform several functions, including a single instance convolution function, support, for example, INT8, INT16, and FP16 data types for both features and weights, and post-processor functions.

[0129] The DLA can quickly and efficiently execute neural networks, particularly CNNs, on processed or unprocessed data for any of a wide variety of functions, such as, but not limited to: a CNN for object recognition and detection using data from a camera sensor; a CNN for distance estimation using data from a camera sensor; a CNN for emergency vehicle detection and identification and detection using data from a microphone; a CNN for facial recognition and vehicle owner identification using data from a camera sensor; and / or a CNN for safety and / or security-related events.

[0130] The DLA can perform any function of the GPU 1108, and by using an inference accelerator, for example, the designer can target any function to either the DLA or the GPU 1108. For example, the designer can focus the processing of CNNs and floating point operations on the DLA, and leave other functions to the GPU 1108 and / or other accelerators 1114.

[0131] The accelerator 1114 (e.g., a hardware acceleration cluster) may include a programmable vision accelerator (PVA), which may be alternatively referred to herein as a computer vision accelerator. The PVA may be designed and configured to accelerate computer vision algorithms for advanced driver assistance systems (ADAS), autonomous driving, and / or augmented reality (AR) and / or virtual reality (VR) applications. The PVA may provide a balance between performance and flexibility. For example, each PVA may include, for example and without limitation, any number of reduced instruction set computer (RISC) cores, direct memory access (DMA), and / or any number of vector processors.

[0132] The RISC core can interact with an image sensor (e.g., an image sensor of any camera described herein), an image signal processor, and / or the like. Each of these RISC cores can include any amount of memory. Depending on the embodiment, the RISC core can use any of a number of protocols. In some examples, the RISC core can execute a real-time operating system (RTOS). The RISC core can be implemented using one or more integrated circuit devices, application specific integrated circuits (ASICs), and / or storage devices. For example, the RISC core can include an instruction cache and / or a tightly coupled RAM.

[0133] The DMA may enable components of the PVA to access system memory independently of the CPU 1106. The DMA may support any number of features used to provide optimizations for the PVA, including but not limited to support for multi-dimensional addressing and / or circular addressing. In some examples, the DMA may support addressing in up to six or more dimensions, which may include block width, block height, block depth, horizontal block stepping, vertical block stepping, and / or depth stepping.

[0134] The vector processor can be a programmable processor that can be designed to efficiently and flexibly perform programming for computer vision algorithms and provide signal processing capabilities. In some examples, the PVA can include a PVA core and two vector processing subsystem partitions. The PVA core can include a processor subsystem, one or more DMA engines (e.g., two DMA engines), and / or other peripherals. The vector processing subsystem can operate as the main processing engine of the PVA, and can include a vector processing unit (VPU), an instruction cache, and / or a vector memory (e.g., VMEM). The VPU core can include a digital signal processor, such as, for example, a single instruction multiple data (SIMD), a very long instruction word (VLIW) digital signal processor. The combination of SIMD and VLIW can enhance throughput and rate.

[0135] Each of the vector processors may include an instruction cache and may be coupled to a dedicated memory. As a result, in some examples, each of the vector processors may be configured to execute independently of the other vector processors. In other examples, the vector processors included in a particular PVA may be configured to employ data parallelism. For example, in some embodiments, multiple vector processors included in a single PVA may execute the same computer vision algorithm, but on different regions of an image. In other examples, the vector processors included in a particular PVA may execute different computer vision algorithms simultaneously on the same image, or even execute different algorithms on sequence images or portions of an image. Among other things, any number of PVAs may be included in a hardware acceleration cluster, and any number of vector processors may be included in each of these PVAs. In addition, the PVA may include additional error correction code (ECC) memory to enhance overall system security.

[0136] The accelerator 1114 (e.g., a hardware acceleration cluster) may include an on-chip computer vision network and SRAM to provide high bandwidth, low latency SRAM for the accelerator 1114. In some examples, the on-chip memory may include at least 4MB of SRAM consisting of, for example and not limited to, eight field-configurable memory blocks, which may be accessed by both the PVA and the DLA. Each pair of memory blocks may include an advanced peripheral bus (APB) interface, configuration circuitry, a controller, and a multiplexer. Any type of memory may be used. The PVA and DLA may access the memory via a backbone that provides high-speed memory access to the PVA and DLA. The backbone may include an on-chip computer vision network that interconnects the PVA and DLA to the memory (e.g., using APB).

[0137] The on-chip computer vision network may include an interface that determines that both PVA and DLA provide ready and valid signals before transmitting any control signals / addresses / data. Such an interface may provide separate phases and separate channels for transmitting control signals / addresses / data, as well as burst communication for continuous data transmission. This type of interface may comply with ISO26262 or IEC612508 standards, but other standards and protocols may also be used.

[0138] In some examples, SoC 1104 may include a real-time ray tracing hardware accelerator such as described in U.S. Patent Application No. 16 / 101,1232 filed on August 10, 2018. The real-time ray tracing hardware accelerator may be used to quickly and efficiently determine the position and range of objects (e.g., within a world model) in order to generate real-time visualization simulations for RADAR signal interpretation, for sound propagation synthesis and / or analysis, for SONAR system simulations, for general wave propagation simulations, for comparison with LIDAR data for the purpose of positioning and / or other functions, and / or for other uses.

[0139] The accelerator 1114 (e.g., a cluster of hardware accelerators) has a wide range of uses in autonomous driving. The PVA can be a programmable visual accelerator that can be used for key processing stages in ADAS and autonomous vehicles. The capabilities of the PVA are a good match for algorithmic domains that require predictable processing, low power, and low latency. In other words, the PVA performs well on semi-dense or dense rule computations, and even on small data sets that require predictable runtimes with low latency and low power. Therefore, in the context of a platform for autonomous vehicles, the PVA is designed to run classic computer vision algorithms because they are efficient at object detection and integer math.

[0140] For example, according to one embodiment of the technology, PVA is used to perform computer stereo vision. In some examples, algorithms based on semi-global matching can be used, but this is not intended to be limiting. Many applications for level 3-5 autonomous driving require instant motion estimation / stereo matching (e.g., structure from motion, pedestrian recognition, lane detection, etc.). PVA can perform computer stereo vision functions on input from two monocular cameras.

[0141] In some examples, PVA can be used to perform dense optical flow. According to the process raw RADAR data (e.g., using 4D fast Fourier transform) to provide processed RADAR. In other examples, PVA is used for time-of-flight depth processing, which is for example by processing raw time-of-flight data to provide processed time-of-flight data.

[0142] DLA can be used to run any type of network to enhance control and driving safety, including, for example, a neural network that outputs a confidence measure for each object detection. Such confidence values ​​can be interpreted as probabilities, or as providing a relative "weight" of each detection compared to other detections. The confidence value enables the system to make further decisions about which detections should be considered true positive detections rather than false positive detections. For example, the system can set a threshold for confidence and only consider detections that exceed the threshold as true positive detections. In an automatic emergency braking (AEB) system, a false positive detection can cause the vehicle to automatically perform emergency braking, which is obviously undesirable. Therefore, only the most confident detections should be considered as triggers for AEB. DLA can run a neural network for regressing confidence values. The neural network may take as its input at least some subset of parameters, such as bounding box dimensions, a ground plane estimate obtained (e.g., from another subsystem), an inertial measurement unit (IMU) sensor 1166 output related to the orientation and distance of the vehicle 1100, a 3D position estimate of an object obtained from the neural network and / or other sensors (e.g., a LIDAR sensor 1164 or a RADAR sensor 1160), etc.

[0143] SoC 1104 may include one or more data stores 1116 (e.g., memory). Data store 1116 may be on-chip memory of SoC 1104 that may store neural networks to be executed on the GPU and / or DLA. In some examples, data store 1116 may be large enough to store multiple instances of a neural network for redundancy and safety. Data store 1112 may include an L2 or L3 cache 1112. References to data store 1116 may include references to memory associated with a PVA, DLA, and / or other accelerator 1114 as described herein.

[0144] SoC1104 may include one or more processors 1110 (e.g., embedded processors). Processor 1110 may include a boot and power management processor, which may be a dedicated processor and subsystem for handling boot power and management functions and related safety implementations. The boot and power management processor may be part of the SoC1104 boot sequence and may provide runtime power management services. The boot power and management processor may provide clock and voltage programming, auxiliary system low power state transitions, SoC1104 thermal and temperature sensor management, and / or SoC1104 power state management. Each temperature sensor may be implemented as a ring oscillator whose output frequency is proportional to the temperature, and SoC1104 may use the ring oscillator to detect the temperature of CPU1106, GPU1108, and / or accelerator 1114. If it is determined that the temperature exceeds a threshold, the boot and power management processor may enter a temperature fault routine and place SoC1104 in a lower power state and / or place the vehicle 1100 in a driver safety parking mode (e.g., parking the vehicle 1100 safely).

[0145] Processor 1110 may further include a set of embedded processors that may be used as an audio processing engine. The audio processing engine may be an audio subsystem that allows for full hardware support for multi-channel audio through multiple interfaces and a wide range of flexible audio I / O interfaces. In some examples, the audio processing engine is a dedicated processor core having a digital signal processor with dedicated RAM.

[0146] The processor 1110 may further include an always-on processor engine that may provide the necessary hardware features to support low-power sensor management and wake-up use cases. The always-on processor engine may include a processor core, tightly coupled RAM, supporting peripherals (e.g., timers and interrupt controllers), various I / O controller peripherals, and routing logic.

[0147] The processor 1110 may further include a safety cluster engine, which includes a dedicated processor subsystem that handles safety management of automotive applications. The safety cluster engine may include two or more processor cores, tightly coupled RAM, supporting peripherals (e.g., timers, interrupt controllers, etc.), and / or routing logic. In safety mode, the two or more cores may operate in lockstep mode and act as a single core with comparison logic to detect any differences between their operations.

[0148] Processor 1110 may further include a real-time camera engine, which may include a dedicated processor subsystem for handling real-time camera management.

[0149] Processor 1110 may further include a high dynamic range signal processor, which may include an image signal processor, which is a hardware engine that is part of a camera processing pipeline.

[0150] The processor 1110 may include a video image compositer, which may be a processing block (e.g., implemented on a microprocessor) that implements video post-processing functions required by the video playback application to produce a final image for the player window. The video image compositer may perform lens distortion correction for the wide-angle camera 1170, the surround camera 1174, and / or for the in-cab surveillance camera sensor. The in-cab surveillance camera sensor is preferably monitored by a neural network running on another instance of the advanced SoC, configured to recognize in-cab events and respond accordingly. The in-cab system may perform lip reading to activate mobile phone service and place a call, dictate an email, change the vehicle's destination, activate or change the vehicle's infotainment system and settings, or provide voice-activated web surfing. Certain functions are only available to the driver when the vehicle is operating in autonomous mode and are disabled in other circumstances.

[0151] The video image compositer may include enhanced temporal noise reduction for both spatial and temporal noise reduction. For example, in the case of motion in the video, the noise reduction appropriately weights the spatial information and reduces the weight of information provided by adjacent frames. In the case where an image or portion of an image does not include motion, the temporal noise reduction performed by the video image compositer may use information from previous images to reduce noise in the current image.

[0152] The video image compositor may also be configured to perform stereoscopic rectification on the input stereoscopic footage frames. The video image compositor may further be used for user interface composition when the operating system desktop is in use and the GPU 1108 does not need to continuously render new surfaces. Even when the GPU 1108 is powered on and active for 3D rendering, the video image compositor may be used to offload the GPU 1108 to improve performance and responsiveness.

[0153] SoC 1104 may further include a mobile industry processor interface (MIPI) camera serial interface for receiving video and input from a camera, a high-speed interface, and / or a video input block that may be used for camera and related pixel input functions. SoC 1104 may further include an input / output controller that may be controlled by software and may be used to receive I / O signals that are not committed to a specific role.

[0154] SoC1104 may further include a wide range of peripheral interfaces to enable communication with peripherals, audio codecs, power management and / or other devices. SoC1104 may be used to process data from cameras (connected via Gigabit multimedia serial links and Ethernet), sensors (e.g., LIDAR sensor 1164, RADAR sensor 1160, etc., which may be connected via Ethernet), data from bus 1102 (e.g., speed of vehicle 1100, steering wheel position, etc.), data from GNSS sensor 1158 (connected via Ethernet or CAN bus). SoC1104 may further include dedicated high-performance mass storage controllers, which may include their own DMA engines, and which may be used to free up CPU1106 from routine data management tasks.

[0155] SoC1104 can be an end-to-end platform with a flexible architecture that spans automation levels 3-5, thereby providing a comprehensive functional safety architecture that utilizes and efficiently uses computer vision and ADAS technologies to achieve diversity and redundancy, and provides a platform for flexible and reliable driving software stacks together with deep learning tools. SoC1104 can be faster, more reliable, and even more energy efficient and space efficient than conventional systems. For example, when combined with CPU1106, GPU1108, and data storage 1116, accelerator 1114 can provide a fast and efficient platform for level 3-5 autonomous vehicles.

[0156] The technology thus provides capabilities and functionality that cannot be achieved by conventional systems. For example, computer vision algorithms can be executed on CPUs, which can be configured using high-level programming languages ​​such as the C programming language to execute a wide variety of processing algorithms across a wide variety of visual data. However, CPUs often fail to meet the performance requirements of many computer vision applications, such as those related to, for example, execution time and power consumption. In particular, many CPUs are unable to execute complex object detection algorithms in real time, which is a requirement for in-vehicle ADAS applications and a requirement for practical Level 3-5 autonomous vehicles.

[0157] In contrast to conventional systems, by providing a CPU complex, a GPU complex, and a hardware acceleration cluster, the techniques described herein allow multiple neural networks to be executed simultaneously and / or sequentially, and the results to be combined together to achieve Level 3-5 autonomous driving functions. For example, a CNN executed on a DLA or dGPU (e.g., GPU1120) may include text and word recognition, allowing the supercomputer to read and understand traffic signs, including signs for which a neural network has not been specifically trained. The DLA may further include a neural network capable of recognizing, interpreting, and providing semantic understanding of the signs, and passing the semantic understanding to a path planning module running on the CPU complex.

[0158] As another example, as required for Level 3, 4, or 5 driving, multiple neural networks may be running simultaneously. For example, a warning sign consisting of "Caution: Flashing Lights Indicate Icing Conditions" along with electric lights may be interpreted by several neural networks independently or collectively. The sign itself may be recognized as a traffic sign by a deployed first neural network (e.g., a trained neural network), and the text "Flashing Lights Indicate Icing Conditions" may be interpreted by a deployed second neural network that informs the vehicle's path planning software (preferably executing on a CPU complex) that icing conditions exist when the flashing lights are detected. The flashing lights may be recognized by operating a deployed third neural network over multiple frames that informs the vehicle's path planning software of the presence (or absence) of the flashing lights. All three neural networks may be running simultaneously, for example within a DLA and / or on GPU 1108.

[0159] In some examples, a CNN for facial recognition and owner recognition can use data from a camera sensor to identify the presence of an authorized driver and / or owner of the vehicle 1100. An always-on sensor processing engine can be used to unlock the vehicle and turn on the lights when the owner approaches the driver's door, and in security mode, disable the vehicle when the owner leaves the vehicle. In this way, the SoC 1104 provides security against theft and / or carjacking.

[0160] In another example, a CNN for emergency vehicle detection and identification can use data from microphone 1196 to detect and identify emergency vehicle sirens. In contrast to conventional systems that use general classifiers to detect sirens and manually extract features, SoC1104 uses CNN to classify environmental and urban sounds and to classify visual data. In a preferred embodiment, the CNN running on the DLA is trained to identify the relative closing rate of emergency vehicles (e.g., by using the Doppler effect). The CNN can also be trained to identify emergency vehicles specific to the local area in which the vehicle is operating, as identified by the GNSS sensor 1158. Thus, for example, when operating in Europe, the CNN will seek to detect European sirens, and when in the United States, the CNN will seek to identify only North American sirens. Once an emergency vehicle is detected, with the assistance of the ultrasonic sensor 1162, the control program can be used to execute emergency vehicle safety routines to slow the vehicle, drive to the side of the road, stop the vehicle, and / or idle the vehicle until the emergency vehicle passes.

[0161] The vehicle may include a CPU 1118 (e.g., a discrete CPU or dCPU) that may be coupled to the SoC 1104 via a high-speed interconnect (e.g., PCIe). The CPU 1118 may include, for example, an X86 processor. The CPU 1118 may be used to perform any of a variety of functions, including, for example, arbitrating potentially inconsistent results between ADAS sensors and the SoC 1104, and / or monitoring the status and health of the controller 1136 and / or the infotainment SoC 1130.

[0162] The vehicle 1100 may include a GPU 1120 (e.g., a discrete GPU or dGPU) that may be coupled to the SoC 1104 via a high-speed interconnect (e.g., NVIDIA's NVLINK). The GPU 1120 may provide additional artificial intelligence functionality, for example, by executing redundant and / or different neural networks, and may be used to train and / or update the neural network based at least in part on input from sensors of the vehicle 1100 (e.g., sensor data).

[0163] The vehicle 1100 may further include a network interface 1124, which may include one or more wireless antennas 1126 (e.g., one or more wireless antennas for different communication protocols, such as a cellular antenna, a Bluetooth antenna, etc.). The network interface 1124 can be used to enable wireless connections to the cloud (e.g., with a server 1178 and / or other network devices), to other vehicles, and / or to computing devices (e.g., a passenger's client device) via the Internet. In order to communicate with other vehicles, a direct link can be established between the two vehicles, and / or an indirect link can be established (e.g., across a network and through the Internet). The direct link can be provided using a vehicle-to-vehicle communication link. The vehicle-to-vehicle communication link can provide the vehicle 1100 with information about vehicles approaching the vehicle 1100 (e.g., vehicles in front of, to the side of, and / or behind the vehicle 1100). This functionality can be part of the cooperative adaptive cruise control functionality of the vehicle 1100.

[0164] The network interface 1124 may include a SoC that provides modulation and demodulation functions and enables the controller 1136 to communicate over a wireless network. The network interface 1124 may include a radio frequency front end for up-conversion from baseband to radio frequency and down-conversion from radio frequency to baseband. The frequency conversion may be performed by a known process and / or may be performed using a super-heterodyne process. In some examples, the radio frequency front end function may be provided by a separate chip. The network interface may include a wireless function for communicating via LTE, WCDMA, UMTS, GSM, CDMA2000, Bluetooth, Bluetooth LE, Wi-Fi, Z-wave, ZigBee, LoRaWAN, and / or other wireless protocols.

[0165] The vehicle 1100 may further include data storage 1128, which may include off-chip storage (e.g., outside the SoC 1104). The data storage 1128 may include one or more storage elements, including RAM, SRAM, DRAM, VRAM, flash memory, hard disk, and / or other components and / or devices that can store at least one bit of data.

[0166] The vehicle 1100 may further include a GNSS sensor 1158. The GNSS sensor 1158 (e.g., GPS and / or assisted GPS sensor) is used to assist in mapping, perception, occupancy grid generation, and / or path planning functions. Any number of GNSS sensors 1158 may be used, including, for example and without limitation, GPS using a USB connector with an Ethernet to serial (RS-232) bridge.

[0167] The vehicle 1100 may further include a RADAR sensor 1160. The RADAR sensor 1160 may be used by the vehicle 1100 for remote vehicle detection even in darkness and / or inclement weather conditions. The RADAR functional safety level may be ASILB. The RADAR sensor 1160 may use CAN and / or bus 1102 (e.g., to transmit data generated by the RADAR sensor 1160) for control and access to object tracking data, accessing Ethernet in some examples to access raw data. A variety of RADAR sensor types may be used. For example and without limitation, the RADAR sensor 1160 may be suitable for front, rear, and side RADAR use. In some examples, a pulse Doppler RADAR sensor is used.

[0168] The RADAR sensor 1160 may include different configurations, such as long-range with a narrow field of view, short-range with a wide field of view, short-range side coverage, and the like. In some examples, the long-range RADAR may be used for adaptive cruise control functions. The long-range RADAR system may provide a wide field of view (e.g., within a range of 250m) achieved by two or more independent scans. The RADAR sensor 1160 may help distinguish between static and moving objects and may be used by the ADAS system for emergency braking assistance and forward collision warnings. The long-range RADAR sensor may include a single-station multimode RADAR with multiple (e.g., six or more) fixed RADAR antennas and high-speed CAN and FlexRay interfaces. In an example with six antennas, the central four antennas may create a focused beam pattern designed to record the surroundings of the vehicle 1100 at a higher rate with minimal traffic interference from adjacent lanes. The other two antennas may expand the field of view, making it possible to quickly detect vehicles entering or leaving the lane of the vehicle 1100.

[0169] As an example, a medium-range RADAR system may include a range of up to 1460m (front) or 80m (rear) and a field of view of up to 42 degrees (front) or 1450 degrees (rear). A short-range RADAR system may include, but is not limited to, a RADAR sensor designed to be mounted on both ends of a rear bumper. When mounted on both ends of a rear bumper, such a RADAR sensor system may create two beams that continuously monitor the blind spots behind and beside the vehicle.

[0170] Short-range RADAR systems can be used in ADAS systems for blind spot detection and / or lane change assistance.

[0171] The vehicle 1100 may further include an ultrasonic sensor 1162. The ultrasonic sensor 1162, which may be placed on the front, rear, and / or side of the vehicle 1100, may be used for parking assistance and / or creating and updating an occupancy grid. A variety of ultrasonic sensors 1162 may be used, and different ultrasonic sensors 1162 may be used for different detection ranges (e.g., 2.5 m, 4 m). The ultrasonic sensor 1162 may operate at a functional safety level of ASILB.

[0172] The vehicle 1100 may include a LIDAR sensor 1164. The LIDAR sensor 1164 may be used for object and pedestrian detection, emergency braking, collision avoidance, and / or other functions. The LIDAR sensor 1164 may be functionally safe at level ASILB. In some examples, the vehicle 1100 may include multiple LIDAR sensors 1164 (e.g., two, four, six, etc.) that may use Ethernet (e.g., to provide data to a Gigabit Ethernet switch).

[0173] In some examples, the LIDAR sensor 1164 may be able to provide a list of objects and their distances for a 360-degree field of view. Commercially available LIDAR sensors 1164 may have, for example, an advertised range of approximately 1400m, an accuracy of 2cm-3cm, and support for 1400Mbps Ethernet connections. In some examples, one or more non-obtrusive LIDAR sensors 1164 may be used. In such examples, the LIDAR sensor 1164 may be implemented as a small device that can be embedded in the front, back, side, and / or corner of the vehicle 1100. In such an example, the LIDAR sensor 1164 may provide a field of view of up to 1420 degrees horizontally and 35 degrees vertically, with a range of 200m, even for low reflectivity objects. The front-mounted LIDAR sensor 1164 may be configured for a horizontal field of view between 45 and 135 degrees.

[0174] In some examples, LIDAR technologies such as 3D flash LIDAR may also be used. 3D flash LIDAR uses a flash of laser as an emission source to illuminate the vehicle's surroundings up to about 200m. The flash LIDAR unit includes a receiver that records the laser pulse transmission time and the reflected light on each pixel, which in turn corresponds to the range from the vehicle to the object. Flash LIDAR can allow highly accurate and distortion-free images of the surrounding environment to be generated with each laser flash. In some examples, four flash LIDAR sensors can be deployed, one on each side of the vehicle 1100. Available 3D flash LIDAR systems include solid-state 3D staring array LIDAR cameras (e.g., non-scanning LIDAR devices) with no moving parts other than fans. The flash LIDAR device can use 5 nanosecond Class I (eye-safe) laser pulses per frame, and can capture the reflected laser light in the form of a 3D range point cloud and co-registered intensity data. By using flash LIDAR, and because flash LIDAR is a solid-state device with no moving parts, the LIDAR sensor 1164 may be less susceptible to motion blur, vibration, and / or shock.

[0175] The vehicle may further include an IMU sensor 1166. In some examples, the IMU sensor 1166 may be located at the center of the rear axle of the vehicle 1100. The IMU sensor 1166 may include, for example and without limitation, an accelerometer, a magnetometer, a gyroscope, a magnetic compass, and / or other sensor types. In some examples, for example, in a six-axis application, the IMU sensor 1166 may include an accelerometer and a gyroscope, and in a nine-axis application, the IMU sensor 1166 may include an accelerometer, a gyroscope, and a magnetometer.

[0176] In some embodiments, the IMU sensor 1166 can be implemented as a miniature high-performance GPS-assisted inertial navigation system (GPS / INS) that combines a micro-electromechanical system (MEMS) inertial sensor, a high-sensitivity GPS receiver, and an advanced Kalman filter algorithm to provide estimates of position, velocity, and attitude. Thus, in some examples, the IMU sensor 1166 can enable the vehicle 1100 to estimate heading by directly observing and correlating velocity changes from the GPS to the IMU sensor 1166 without the need for input from a magnetic sensor. In some examples, the IMU sensor 1166 and the GNSS sensor 1158 can be combined into a single integrated unit.

[0177] The vehicle may include microphones 1196 positioned in and / or around the vehicle 1100. The microphones 1196 may be used for, among other things, emergency vehicle detection and identification.

[0178] The vehicle may further include any number of camera types, including stereo cameras 1168, wide-angle cameras 1170, infrared cameras 1172, surround cameras 1174, long-range and / or mid-range cameras 1198, and / or other camera types. These cameras can be used to capture image data around the entire periphery of the vehicle 1100. The type of camera used depends on the embodiment and the requirements of the vehicle 1100, and any combination of camera types can be used to provide the necessary coverage around the vehicle 1100. In addition, the number of cameras can vary depending on the embodiment. For example, the vehicle may include six cameras, seven cameras, ten cameras, twelve cameras, and / or another number of cameras. As an example and not limitation, the cameras can support Gigabit Multimedia Serial Link (GMSL) and / or Gigabit Ethernet. Each of the cameras described herein is described with respect to Fig.11A and Fig. 11B Described in more detail.

[0179] The vehicle 1100 may further include a vibration sensor 1142. The vibration sensor 1142 may measure vibrations of components of the vehicle, such as an axle. For example, changes in vibration may indicate changes in the road surface. In another example, when two or more vibration sensors 1142 are used, the difference between the vibrations may be used to determine friction or slip of the road surface (e.g., when there is a vibration difference between a powered drive shaft and a free-spinning shaft).

[0180] The vehicle 1100 may include an ADAS system 1138. In some examples, the ADAS system 1138 may include a SoC. The ADAS system 1138 may include autonomous / adaptive / automatic cruise control (ACC), cooperative adaptive cruise control (CACC), forward collision warning (FCW), automatic emergency braking (AEB), lane departure warning (LDW), lane keeping assist (LKA), blind spot warning (BSW), rear cross traffic warning (RCTW), collision warning system (CWS), lane centering (LC), and / or other features and functions.

[0181] The ACC system may use a RADAR sensor 1160, a LIDAR sensor 1164, and / or a camera. The ACC system may include a longitudinal ACC and / or a lateral ACC. The longitudinal ACC monitors and controls the distance to the vehicle immediately in front of the vehicle 1100, and automatically adjusts the vehicle speed to maintain a safe distance from the vehicle in front. The lateral ACC performs distance keeping and suggests that the vehicle 1100 change lanes when necessary. The lateral ACC is related to other ADAS applications such as LCA and CWS.

[0182] CACC uses information from other vehicles, which can be received from other vehicles indirectly via a wireless link or through a network connection (e.g., through the Internet) via a network interface 1124 and / or a wireless antenna 1126. A direct link can be provided by a vehicle-to-vehicle (V2V) communication link, while an indirect link can be an infrastructure-to-vehicle (I2V) communication link. Typically, the V2V communication concept provides information about the vehicle immediately ahead (e.g., the vehicle immediately ahead of the vehicle 1100 and in the same lane as it), while the I2V communication concept provides information about traffic farther ahead. The CACC system can include either or both of the I2V and V2V information sources. Given information about the vehicle ahead of the vehicle 1100, CACC can be more reliable, and it is possible to improve the smoothness of traffic flow and reduce road congestion.

[0183] The FCW system is designed to alert the driver to hazards so that the driver can take corrective action. The FCW system uses a front camera and / or RADAR sensor 1160 coupled to a dedicated processor, DSP, FPGA and / or ASIC, which is electrically coupled to driver feedback such as a display, speaker and / or vibrating components. The FCW system can provide warnings in the form of, for example, sound, visual warnings, vibrations and / or rapid brake pulses.

[0184] The AEB system detects an impending forward collision with another vehicle or other object and can automatically apply the brakes if the driver does not take corrective action within specified time or distance parameters. The AEB system can use a front camera and / or RADAR sensor 1160 coupled to a dedicated processor, DSP, FPGA and / or ASIC. When the AEB system detects a hazard, it typically first alerts the driver to take corrective action to avoid the collision, and if the driver does not take corrective action, the AEB system can automatically apply the brakes in an effort to prevent or at least mitigate the effects of the predicted collision. The AEB system may include technologies such as dynamic brake support and / or collision approach braking.

[0185] The LDW system provides visual, audible, and / or tactile warnings, such as steering wheel or seat vibrations, to alert the driver when the vehicle 1100 crosses a lane marking. When the driver indicates an intention to leave the lane, by activating a turn signal, the LDW system is not activated. The LDW system may use a front-facing camera coupled to a dedicated processor, DSP, FPGA, and / or ASIC that is electrically coupled to driver feedback such as a display, speaker, and / or vibrating component.

[0186] The LKA system is a variation of the LDW system. If the vehicle 1100 begins to leave the lane, the LKA system provides steering input or braking to correct the vehicle 1100.

[0187] The BSW system detects and warns the driver of vehicles in the car's blind spot. The BSW system can provide visual, auditory, and / or tactile alerts to indicate that merging or changing lanes is unsafe. The system can provide additional warnings when the driver uses a turn signal. The BSW system can use a rear-facing camera and / or RADAR sensor 1160 coupled to a dedicated processor, DSP, FPGA, and / or ASIC that is electrically coupled to driver feedback such as a display, speaker, and / or vibrating component.

[0188] The RCTW system may provide visual, audible, and / or tactile notifications when an object is detected outside the range of the rear camera while the vehicle 1100 is in reverse. Some RCTW systems include AEB to ensure that the vehicle brakes are applied to avoid a crash. The RCTW system may use one or more rear RADAR sensors 1160 coupled to a dedicated processor, DSP, FPGA, and / or ASIC that is electrically coupled to driver feedback such as a display, speaker, and / or vibrating component.

[0189] Conventional ADAS systems may be prone to false positive results, which may annoy and distract the driver, but are typically not catastrophic because the ADAS system alerts the driver and allows the driver to decide whether the safety condition really exists and take action accordingly. However, in the autonomous vehicle 1100, in the case of conflicting results, the vehicle 1100 itself must decide whether to pay attention to the results from the main computer or the auxiliary computer (e.g., the first controller 1136 or the second controller 1136). For example, in some embodiments, the ADAS system 1138 can be a backup and / or auxiliary computer for providing perception information to the backup computer rationality module. The backup computer rationality monitor can run redundant and diverse software on hardware components to detect faults in perception and dynamic driving tasks. The output from the ADAS system 1138 can be provided to the supervisory MCU. If the outputs from the main computer and the auxiliary computer conflict, the supervisory MCU must determine how to coordinate the conflict to ensure safe operation.

[0190] In some examples, the master computer can be configured to provide a confidence score to the supervisory MCU, indicating the master computer's confidence in the selected result. If the confidence score exceeds a threshold, the supervisory MCU can follow the direction of the master computer, regardless of whether the auxiliary computer provides conflicting or inconsistent results. In the case where the confidence score does not meet the threshold and the master computer and the auxiliary computer indicate different results (e.g., conflicts), the supervisory MCU can arbitrate between these computers to determine the appropriate result.

[0191] The supervisory MCU can be configured to run a neural network that is trained and configured to determine conditions under which the auxiliary computer provides a false alarm based at least in part on outputs from the main computer and the auxiliary computer. Thus, the neural network in the supervisory MCU can learn when the output of the auxiliary computer can be trusted and when it cannot. For example, when the auxiliary computer is a RADAR-based FCW system, the neural network in the supervisory MCU can learn when the FCW system is identifying a metal object that is not actually dangerous, such as a drainage grid or manhole cover that triggers an alarm. Similarly, when the auxiliary computer is a camera-based LDW system, the neural network in the supervisory MCU can learn to ignore the LDW when a cyclist or pedestrian is present and lane departure is actually the safest strategy. In an embodiment including a neural network running on the supervisory MCU, the supervisory MCU may include at least one of a DLA or GPU suitable for running the neural network using associated memory. In a preferred embodiment, the supervisory MCU may include a component of SoC1104 and / or be included as a component of SoC1104.

[0192] In other examples, the ADAS system 1138 may include an auxiliary computer that uses traditional computer vision rules to perform ADAS functions. In this way, the auxiliary computer can use classic computer vision rules (if-then), and the presence of a neural network in the supervisory MCU can improve reliability, safety, and performance. For example, diverse implementations and intentional non-identity make the entire system more fault-tolerant, especially for failures caused by software (or software-hardware interface) functions. For example, if there is a software vulnerability or error in the software running on the main computer and the non-identical software code running on the auxiliary computer provides the same overall result, the supervisory MCU can be more confident that the overall result is correct and that the vulnerability in the software or hardware on the main computer does not cause a substantial error.

[0193] In some examples, the output of the ADAS system 1138 can be fed to the perception block of the main computer and / or the dynamic driving task block of the main computer. For example, if the ADAS system 1138 indicates a forward collision warning due to an object immediately ahead, the perception block can use this information in identifying the object. In other examples, the auxiliary computer can have its own neural network that is trained and thus reduces the risk of false positives as described herein.

[0194] The vehicle 1100 may further include an infotainment SoC 1130 (e.g., an in-vehicle infotainment system (IVI)). Although illustrated and described as an SoC, the infotainment system may not be an SoC and may include two or more discrete components. The infotainment SoC 1130 may include a combination of hardware and software that can be used to provide audio (e.g., music, personal digital assistant, navigation instructions, news, radio, etc.), video (e.g., TV, movies, streaming, etc.), phone (e.g., hands-free calling), network connectivity (e.g., LTE, WiFi, etc.), and / or information services (e.g., navigation system, rear parking assistance, radio data system, vehicle-related information such as fuel level, total distance covered, brake fuel level, oil level, door open / closed, air filter information, etc.) to the vehicle 1100. For example, the infotainment SoC 1130 may include a radio, a disk player, a navigation system, a video player, USB and Bluetooth connections, an onboard computer, onboard entertainment, WiFi, steering wheel audio controls, hands-free voice controls, a head-up display (HUD), an HMI display 1134, a telematics device, a control panel (e.g., for controlling and / or interacting with various components, features and / or systems), and / or other components. The infotainment SoC 1130 may further be used to provide information (e.g., visual and / or auditory) to a user of the vehicle, such as information from an ADAS system 1138, autonomous driving information such as planned vehicle maneuvers, trajectories, surrounding environment information (e.g., intersection information, vehicle information, road information, etc.), and / or other information.

[0195] The infotainment SoC 1130 may include GPU functionality. The infotainment SoC 1130 may communicate with other devices, systems, and / or components of the vehicle 1100 via a bus 1102 (e.g., a CAN bus, Ethernet, etc.). In some examples, the infotainment SoC 1130 may be coupled to a supervisory MCU so that in the event of a failure of a main controller 1136 (e.g., a main and / or backup computer of the vehicle 1100), the GPU of the infotainment system may perform some self-driving functions. In such an example, the infotainment SoC 1130 may place the vehicle 1100 in a driver-safe parking mode as described herein.

[0196] The vehicle 1100 may further include an instrument cluster 1132 (e.g., a digital instrument panel, an electronic instrument cluster, a digital instrument panel, etc.). The instrument cluster 1132 may include a controller and / or a supercomputer (e.g., a discrete controller or a supercomputer). The instrument cluster 1132 may include a set of instruments, such as a speedometer, a fuel level, an oil pressure, a tachometer, an odometer, a turn indicator, a shift position indicator, a seat belt warning light, a parking brake warning light, an engine fault light, an airbag (SRS) system information, a lighting control, a safety system control, navigation information, and the like. In some examples, information may be displayed and / or shared between the infotainment SoC 1130 and the instrument cluster 1132. In other words, the instrument cluster 1132 may be included as part of the infotainment SoC 1130, or vice versa.

[0197] Fig.11D For a cloud-based server and Fig.11A 1176. System 1176 may include a server 1178, a network 1190, and vehicles including vehicle 1100. Server 1178 may include multiple GPUs 1184(A)-1284(H) (collectively referred to herein as GPUs 1184), PCIe switches 1182(A)-1182(H) (collectively referred to herein as PCIe switches 1182), and / or CPUs 1180(A)-1180(B) (collectively referred to herein as CPUs 1180). GPUs 1184, CPUs 1180, and PCIe switches may be interconnected with a high-speed interconnect and / or PCIe connection 1186 such as, for example and without limitation, an NVLink interface 1188 developed by NVIDIA. In some examples, GPUs 1184 are connected via NVLink and / or NVSwitch SoC, and GPUs 1184 and PCIe switches 1182 are connected via PCIe interconnects. Although eight GPUs 1184, two CPUs 1180, and two PCIe switches are illustrated, this is not intended to be limiting. Depending on the embodiment, each of the servers 1178 may include any number of GPUs 1184, CPUs 1180, and / or PCIe switches. For example, each of the servers 1178 may include eight, sixteen, thirty-two, and / or more GPUs 1184.

[0198] Server 1178 may receive image data over network 1190 and from a vehicle, the image data representing images showing unexpected or changed road conditions, such as recently begun road work. Server 1178 may transmit neural network 1192, updated neural network 1192, and / or map information 1194, including information about traffic and road conditions, over network 1190 and to the vehicle. Updates to map information 1194 may include updates to HD map 1122, such as information about construction sites, potholes, curves, flooding, or other obstacles. In some examples, neural network 1192, updated neural network 1192, and / or map information 1194 may have been generated from new training and / or data received from any number of vehicles in the environment and / or based at least in part on experience of training performed at a data center (e.g., using server 1178 and / or other servers).

[0199] Server 1178 can be used to train a machine learning model (e.g., a neural network) based at least in part on training data. The training data can be generated by the vehicle, and / or can be generated in simulation (e.g., using a game engine). In some examples, the training data is labeled (e.g., where the neural network benefits from supervised learning) and / or undergoes other preprocessing, while in other examples, the training data is not labeled and / or preprocessed (e.g., where the neural network does not require supervised learning). Once the machine learning model is trained, the machine learning model can be used by the vehicle (e.g., transmitted to the vehicle via network 1190), and / or the machine learning model can be used by server 1178 to remotely monitor the vehicle.

[0200] In some examples, server 1178 can receive data from the vehicle and apply the data to the latest real-time neural network for real-time intelligent reasoning. Server 1178 can include a deep learning supercomputer and / or a dedicated AI computer powered by GPU 1184, such as DGX and DGX Station machines developed by NVIDIA. However, in some examples, server 1178 can include a deep learning infrastructure of a data center powered only by CPUs.

[0201] The deep learning infrastructure of server 1178 may be capable of rapid real-time inference, and may use this capability to assess and verify the health of the processors, software, and / or associated hardware in vehicle 1100. For example, the deep learning infrastructure may receive periodic updates from vehicle 1100, such as a sequence of images and / or objects located in the sequence of images that vehicle 1100 has located (e.g., via computer vision and / or other machine learning object classification techniques). The deep learning infrastructure may run its own neural network to identify objects and compare them to the objects identified by vehicle 1100, and if the results do not match and the infrastructure concludes that the AI ​​in vehicle 1100 has failed, then server 1178 may transmit a signal to vehicle 1100 instructing the fail-safe computer of vehicle 1100 to take control, notify passengers, and complete a safe parking maneuver.

[0202] For reasoning, server 1178 may include GPU 1184 and one or more programmable reasoning accelerators (e.g., NVIDIA's TensorRT3). The combination of GPU-powered servers and reasoning acceleration can make real-time responses possible. In other examples, such as when performance is not so important, CPU, FPGA, and other processor-powered servers can be used for reasoning.

[0203] Example computing device

[0204] Fig.12 is a block diagram of an example computing device 1200 suitable for implementing some embodiments of the present disclosure, such as the object detector 106 and Figure 5A The computing device 1200 may include a bus 1202 that directly or indirectly couples the following devices: a memory 1204, one or more central processing units (CPUs) 1206, one or more graphics processing units (GPUs) 1208, a communication interface 1210, an input / output (I / O) port 1212, an input / output component 1214, a power supply 1216, and one or more presentation components 1218 (e.g., a display).

[0205] although Fig.12 The various blocks of are shown as being connected via a bus 1202 having lines, but this is not intended to be limiting and is merely for clarity. For example, in some embodiments, a presentation component 1218 such as a display device may be considered an I / O component 1214 (e.g., if the display is a touch screen). As another example, CPU 1206 and / or GPU 1208 may include memory (e.g., memory 1204 may represent a storage device in addition to the memory of GPU 1208, CPU 1206, and / or other components). In other words, Fig.12The computing devices described herein are illustrative only. No distinction is made between categories such as "workstations," "servers," "laptops," "desktops," "tablets," "client devices," "mobile devices," "handheld devices," "game consoles," "electronic control units (ECUs)," "virtual reality systems," and / or other device or system types, as all are considered Fig.12 within the range of computing devices.

[0206] The bus 1202 may represent one or more buses, such as an address bus, a data bus, a control bus, or a combination thereof. The bus 1202 may include one or more bus types, such as an Industry Standard Architecture (ISA) bus, an Extended Industry Standard Architecture (EISA) bus, a Video Electronics Standards Association (VESA) bus, a Peripheral Component Interconnect (PCI) bus, a Peripheral Component Interconnect Express (PCIe) bus, and / or another type of bus.

[0207] Memory 1204 may include any of a variety of computer-readable media. Computer-readable media may be any available media that can be accessed by computing device 1200. Computer-readable media may include volatile and nonvolatile media and removable and non-removable media. By way of example and not limitation, computer-readable media may include computer storage media and communication media.

[0208] Computer storage media may include volatile and nonvolatile media and / or removable and non-removable media implemented in any method or technology for storing information such as computer-readable instructions, data structures, program modules, and / or other data types. For example, memory 1204 may store computer-readable instructions (e.g., representing programs and / or program elements, such as an operating system). Computer storage media may include, but are not limited to, RAM, ROM, EEPROM, flash memory or other storage technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage devices, cassettes, magnetic tape, magnetic disk storage devices or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by computing device 1200. As used herein, computer storage media does not include the signals themselves.

[0209] Communication media may contain computer readable instructions, data structures, program modules, and / or other data types in a modulated data signal such as a carrier wave or other transmission mechanism, and include any information transport medium. The term "modulated data signal" may refer to a signal that has one or more of its characteristics set or changed in such a way that information is encoded into the signal. By way of example and not limitation, communication media may include wired media such as a wired network or a direct wired connection, and wireless media such as sound, RF, infrared, and other wireless media. Any combination of the above should also be included within the scope of computer readable media.

[0210] The CPU 1206 may be configured to execute computer-readable instructions to control one or more components of the computing device 1200 to perform one or more of the methods and / or processes described herein (e.g., Figure 2 , 5A and 7-9). Each of the CPUs 1206 may include one or more cores (e.g., one, two, four, eight, twenty-eight, seventy-two, etc.) capable of processing a large number of software threads simultaneously. The CPUs 1206 may include any type of processor, and may include different types of processors, depending on the type of computing device 1200 implemented (e.g., a processor with fewer cores for a mobile device and a processor with more cores for a server). For example, depending on the type of computing device 1200, the processor may be an ARM processor implemented using reduced instruction set computing (RISC) or an x86 processor implemented using complex instruction set computing (CISC). The computing device 1200 may include one or more CPUs 1206 in addition to one or more microprocessors or supplementary coprocessors such as math coprocessors.

[0211] GPU1208 can be used by computing device 1200 to render graphics (e.g., 3D graphics). GPU1208 may include hundreds or thousands of cores capable of processing hundreds or thousands of software threads simultaneously. GPU1208 may generate pixel data for outputting an image in response to a rendering command (e.g., a rendering command from CPU1206 received via a host interface). GPU1208 may include a graphics memory such as a display memory for storing pixel data. Display memory may be included as part of memory 1204. GPU1208 may include two or more GPUs operating in parallel (e.g., via a link). When combined together, each GPU1208 may generate pixel data for different parts of an output image or for different output images (e.g., a first GPU for a first image, a second GPU for a second image). Each GPU may include its own memory, or may share memory with other GPUs.

[0212] In examples where computing device 1200 does not include GPU 1208 , CPU 1206 may be used to render graphics.

[0213] The communication interface 1210 may include one or more receivers, transmitters, and / or transceivers that enable the computing device 1200 to communicate with other computing devices via an electronic communication network, including wired and / or wireless communications. The communication interface 1210 may include components and functionality that enable communication over any of a number of different networks, such as a wireless network (e.g., Wi-Fi, Z-wave, Bluetooth, Bluetooth LE, ZigBee, etc.), a wired network (e.g., communicating over Ethernet), a low power wide area network (e.g., LoRaWAN, SigFox, etc.), and / or the Internet.

[0214] I / O ports 1212 can enable computing device 1200 to be logically coupled to other devices including I / O components 1214, presentation components 1218, and / or other components, some of which can be built into (e.g., integrated into) computing device 1200. Illustrative I / O components 1214 include microphones, mice, keyboards, joysticks, game pads, game controllers, satellite dishes, scanners, printers, wireless devices, and the like. I / O components 1214 can provide a natural user interface (NUI) that processes air gestures, voice, or other physiological inputs generated by a user. In some instances, the input can be transmitted to an appropriate network element for further processing. The NUI can implement any combination of voice recognition, stylus recognition, facial recognition, biometric recognition, gesture recognition on and adjacent to the screen, air gestures, head and eye tracking, and touch recognition associated with the display of computing device 1200 (as described in more detail below). Computing device 1200 may include a depth camera such as a stereo camera system, an infrared camera system, an RGB camera system, touch screen technology, and combinations of these for gesture detection and recognition. Additionally, computing device 1200 may include an accelerometer or gyroscope to enable motion detection (e.g., as part of an inertial measurement unit (IMU)). In some examples, the output of the accelerometer or gyroscope may be used by computing device 1200 to render immersive augmented reality or virtual reality.

[0215] The power supply 1216 may include a hardwired power supply, a battery power supply, or a combination thereof. The power supply 1216 may provide power to the computing device 1200 to enable the components of the computing device 1200 to operate.

[0216] The presentation component 1218 may include a display (e.g., a monitor, a touch screen, a television screen, a head-up display (HUD), other display types, or a combination thereof), a speaker, and / or other presentation components. The presentation component 1218 may receive data from other components (e.g., GPU 1208, CPU 1206, etc.) and output the data (e.g., as an image, video, sound, etc.). In one aspect, the presentation component may display an image (e.g., 525) depicting a parking space, a parking space entrance, or any combination thereof.

[0217] The present disclosure may be described in the general context of machine-usable instructions or computer codes executed by a computer or other machine such as a personal digital assistant or other handheld device, including computer-executable instructions such as program modules. Typically, program modules including routines, programs, objects, components, data structures, etc. refer to codes that perform specific tasks or implement specific abstract data types. The present disclosure may be practiced in a variety of system configurations, including handheld devices, consumer electronics, general-purpose computers, more specialized computing devices, etc. The present disclosure may also be practiced in a distributed computing environment in which tasks are performed by remote processing devices linked through a communication network.

[0218] As used herein, the description of "and / or" about two or more elements should be interpreted as referring to only one element or a combination of elements. For example, "element A, element B and / or element C" may include only element A, only element B, only element C, element A and element B, element A and element C, element B and element C, or elements A, B and C. In addition, "at least one of element A or element B" may include at least one of element A, at least one of element B, or at least one of element A and at least one of element B. Further, "at least one of element A and element B" may include at least one of element A, at least one of element B, or at least one of element A and at least one of element B.

[0219] The subject matter of the present disclosure is described in detail herein to meet statutory requirements. However, the description itself is not intended to limit the scope of the present disclosure. On the contrary, the inventors have contemplated that the claimed subject matter may also be embodied in other ways to include steps different from the steps described herein in conjunction with other current or future technologies or combinations of similar steps. Moreover, although the terms "step" and / or "block" may be used herein to imply different elements of the method employed, these terms should not be interpreted as implying any particular order among or between the various steps disclosed herein, unless the order of the steps is explicitly described.

Claims

1. A computer-implemented method comprising: Applying image data representing parking spaces to a neural network; receiving, using the neural network, data generated from the image data and representing displacement values ​​to corner points of an anchor shape; determining corner points of the inclined polygon based on the displacement values ​​to the corner points of the anchor shape; calculating a first distance between the corner point of the inclined polygon and a ground truth corner point of the parking space; determining a sample rating based at least in part on the first distance; and updating parameters of the neural network using the anchor shape as a positive training sample based on the sample rating being below a threshold.

2. The method of claim 1, wherein the first distance comprises a minimum aggregate distance, and wherein the sample rating is a normalized version of the minimum aggregate distance. 3 . The method of claim 1 , wherein determining the sample rating comprises normalizing the first distance based at least in part on an area of ​​a polygon defined by the ground truth corner points of the parking space.

4. The method of claim 1, wherein the inclined polygon is a first inclined quadrilateral and the anchor shape is a second inclined quadrilateral. The method of claim 1 , wherein the anchor shape is a data-driven anchor box generated from one or more ground truth samples. The method of claim 1 , wherein the first distance is a minimum average distance between different combinations of the corner points of the inclined polygon and the ground truth corner points of the parking space.

7. The method of claim 1, wherein the corner points of the inclined polygon include a first angle (A1), a second angle (A2), a third angle (A3) and a fourth angle (A4); wherein, The corner points in the ground truth corner points of the parking space include a fifth corner (B1), a sixth corner (B2), a seventh corner (B3) and an eighth corner (B4); and wherein calculating the first distance includes calculating a first normalized aggregate distance based on distances (A1, B1), (A2, B2), (A3, B3) and (A4, B4); calculating a second normalized aggregate distance based on distances (A1, B2), (A2, B3), (A3, B4) and (A4, B1); calculating a third normalized aggregate distance based on distances (A1, B3), (A2, B4), (A3, B1) and (A4, B2); and calculating a fourth normalized aggregate distance based on distances (A1, B4), (A2, B1), (A3, B2) and (A4, B3), and the first distance is the minimum value of the first normalized aggregate distance, the second normalized aggregate distance, the third normalized aggregate distance and the fourth normalized aggregate distance.

8. A computer-implemented method comprising: applying sensor data representing a field of view of at least one sensor in the environment to the neural network; receiving, from the neural network, first data and second data generated by the sensor data, the first data representing displacement values ​​to corner points of an anchor shape, and the second data representing a confidence value predicting a likelihood that the anchor shape corresponds to a parking space within the field of view of the at least one sensor; and determining a corner point of the tilted polygon corresponding to the displacement value to a corner point of the anchor shape based at least in part on the confidence value exceeding a threshold.

9. The method of claim 8, wherein the anchor shape is an anchor shape among a plurality of anchor shapes associated with spatial elements of the neural network, and for each given anchor shape among the plurality of anchor shapes, the neural network outputs data representing displacement values ​​to corner points of the given anchor shape and a confidence value predicting a corresponding likelihood that the given anchor shape corresponds to a corresponding parking space in the field of view of the at least one sensor.

10. The method of claim 8, wherein the anchor shape is an anchor shape from a plurality of anchor shapes associated with a grid of spatial elements of the neural network, and for each given anchor shape from the plurality of anchor shapes, the neural network outputs data representing displacement values ​​to corner points of the given anchor shape and a confidence value predicting a corresponding likelihood that the given anchor shape corresponds to a corresponding parking space in the field of view of the at least one sensor.

11. The method of claim 8, wherein the sensor data comprises image data representing a field of view of a camera.

12. The method of claim 8, wherein the anchor shape is an anchor shape of a plurality of anchor shapes associated with one or more spatial elements of the neural network, and the plurality of anchor shapes include tilted polygons of different shapes.

13. The method of claim 8, wherein the inclined polygon is a first inclined quadrilateral and the anchor shape is a second inclined quadrilateral.

14. The method of claim 8, further comprising: receiving, from the neural network, third data representing a confidence value predicting a likelihood that a corner point of the tilted polygon defines an entrance to the parking space in the field of view of the at least one sensor; selecting a subset of the corner points of the anchor shape based at least in part on the confidence value; and identifying the entrance to the parking space based on the subset of the corner points.

15. The method of claim 8, further comprising: One or more operations of an autonomous vehicle are controlled based at least in part on the corner points of the tilted polygon.

16. A computer-implemented method comprising: applying sensor data representing a field of view of at least one sensor in the environment to a neural network; receiving, from the neural network, first data and second data generated by the sensor data, the first data representing displacement values ​​to corner points of an anchor shape, and the second data representing confidence values ​​predicting a likelihood that the corner points of the anchor shape define an entrance to a parking space in the field of view of the at least one sensor; selecting a subset of the corner points of the anchor shape based on the confidence value; and identifying the entrance to the parking space based on the subset of the corner points.

17. The computer-implemented method of claim 16, further comprising: determining corner points of a tilted polygon based on the displacement values ​​to the corner points of the anchor shape; and controlling one or more operations of an autonomous vehicle based at least in part on corner points of the tilted polygon and the entrance to the parking space. The method of claim 16 , wherein the anchor shape is a tilted polygon.

19. The method of claim 16, wherein the anchor shape is an anchor shape among a plurality of anchor shapes associated with spatial elements of the neural network, and for each given anchor shape among the plurality of anchor shapes, the neural network outputs data representing the displacement values ​​to corner points of the given anchor shape and a confidence value predicting a corresponding likelihood that the corner points of the given anchor shape define a given entrance to a corresponding parking space in the field of view of the at least one sensor.

20. The method of claim 16, wherein the anchor shape is an anchor shape from a plurality of anchor shapes associated with a grid of spatial elements of the neural network, and for each given anchor shape from the plurality of anchor shapes, the neural network outputs data representing displacement values ​​to corner points of the given anchor shape and confidence values ​​predicting a corresponding likelihood that the corner points of the given anchor shape define a given entrance to a corresponding parking space in the field of view of the at least one sensor.

Images