Neural network training using ground truth data augmented with map information for autonomous machine applications

Abstract

In various examples, training sensor data generated by one or more sensors of an autonomous machine may be positioned as high-definition (HD) map data, in embodiments, for example, to automatically enhance and / or generate ground truth data. The ground truth data may be associated with the training sensor data for training one or more deep neural networks (DNNs) to compute outputs corresponding to autonomous machine operations—e.g., object or feature detection, road feature detection and classification, waiting condition identification and classification, etc. Thus, the HD map data may be utilized during training so that the DNN, when deployed, may help the autonomous machine safely navigate an environment without having to rely on the HD map data to do so.

Description

Background Art

[0001] In order for an autonomous vehicle to safely navigate an environment, the vehicle may rely on a high-definition (HD) map corresponding to the area in which the vehicle is intended to operate. Due to the detailed, three-dimensional, high-precision nature of HD maps, navigation based on HD map data has proven to be effective for safe navigation in environments where HD map information is available. However, in some cases, HD map information is not available, inaccurate or not up-to-date, and / or HD maps may not be relied upon (at least completely) to help navigate in the environment. In this case, conventional methods utilize the vehicle's on-board sensors, such as visual sensors (e.g., cameras, LIDAR, RADAR, etc.) to detect objects, road features (e.g., lane markings, road edges, etc.), free space boundaries, waiting condition information, intersection structure and pose, etc. For example, a deep neural network (DNN) can be utilized to process sensor data from the on-board sensors and calculate outputs corresponding to any of the above operations.

[0002] However, in order to be successful and perform at the level of accuracy required for autonomous operation, DNNs require a large amount of diverse training data and ground truth data (e.g., labels, annotations, etc.) that corresponds to it during training. The ground truth data may include, for example, polylines that identify lane markings or road edges, bounding boxes around objects or environmental features, and / or other types of ground truth data. The process of generating ground truth data requires significant manual work and is often a significant component of the cost and development time of DNN-based products - for example, a single set of ground truth data for training data instances (e.g., images) may require more than 20 minutes of labeling or annotation work. In addition, manual labeling may not produce ground truth data that enables the DNN to perform as accurately as possible - for example, due to human error in the labeling or annotation process. Therefore, in situations where high-definition maps are not available, DNNs may not perform at the accuracy required for safe autonomous operation and / or may require significant cost and manual work to implement. Summary of the invention

[0003] Embodiments of the present disclosure relate to utilizing map information to generate ground truth data for training neural networks for autonomous machine applications. Systems and methods are disclosed for using positioning techniques to determine information corresponding to a high-definition (HD) map corresponding to training data (e.g., images, confidence maps, LIDAR data, RADAR data, etc.) captured using one or more sensors of an autonomous machine. Information from the HD map can then be utilized to generate training labels, annotations, or other ground truth data corresponding to the training data to train one or more neural networks to perform calculations corresponding to autonomous machine operations (e.g., object detection, waiting condition analysis, road structure determination, positioning, feature detection, etc.).

[0004] Compared to conventional systems such as those described above, embodiments of the present disclosure combine HD map data with the training of deep neural networks (DNNs) to address deficiencies or shortcomings in HD maps and conventional training processes for DNNs. For example, limited coverage of accurate and up-to-date HD maps can be addressed by using DNNs trained using ground truth data, which is generated using HD maps and corresponds to sensor data captured within areas where the HD maps are accurate and up-to-date. In addition, the expensive and potentially inaccurate ground truth generation process of conventional DNN training techniques can be remedied using accurate ground truth data automatically generated from HD map information. As a result, the cost of training DNNs in terms of time and labor may be greatly reduced, and the resulting DNNs that can be used to assist in various operations of autonomous machines may be more accurate and reliable - especially in locations where HD map information is not available or is not up-to-date. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] The present system and method for utilizing map information to generate ground truth data for training neural networks for autonomous machine applications is described in detail below with reference to the accompanying drawings, wherein:

[0006] Figure 1 An example data flow diagram of a process for localizing training data to high definition (HD) map data to enhance or generate labels for training a deep neural network (DNN) according to some embodiments of the present disclosure is included;

[0007] Figure 2A depicts an example visualization of automatically generated ground truth labels corresponding to road features according to some embodiments of the present disclosure;

[0008] Figure 2B depicts example visualizations of automatically generated ground truth labels corresponding to intersection detection and classification according to some embodiments of the present disclosure;

[0009] Figure 2C depicts example visualizations of automatically generated ground truth labels corresponding to environmental features including poles, road markings, lane lines, road boundaries, and crosswalks in accordance with some embodiments of the present disclosure;

[0010] Figure 3 depicts an example histogram corresponding to deviations from the center of a lane of travel or track according to some embodiments of the present disclosure;

[0011] Figure 4 is a flow chart illustrating a method for locating training data to HD map data to enhance or generate labels for training a DNN according to some embodiments of the present disclosure;

[0012] Figure 5A is an illustration of an example autonomous vehicle according to some embodiments of the present disclosure;

[0013] Figure 5B According to some embodiments of the present disclosure Figure 5A Examples of autonomous vehicle camera positions and fields of view;

[0014] Figure 5C According to some embodiments of the present disclosure Figure 5A a block diagram of an example system architecture for an example autonomous vehicle;

[0015] Figure 5D is a cloud-based server and Figure 5A A system diagram of an example of communication between autonomous vehicles; and

[0016] Figure 6 is a block diagram of an example computing device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0017] Systems and methods are disclosed that are related to utilizing map information to generate ground truth data for training neural networks for autonomous machine applications. Although the present disclosure may be related to an example autonomous vehicle 500 (referred to herein alternatively as “vehicle 500,” “ego vehicle 500,” “data collection vehicle 500,” or “dynamic actor 500”), the examples are related to Figures 5A-5D The present disclosure may be described with respect to autonomous driving or ADAS systems, but this is not intended to be limiting. For example, the systems and methods described herein may be used by, but are not limited to, non-autonomous vehicles, semi-autonomous vehicles (e.g., in one or more adaptive driver assistance systems (ADAS)), robots, warehouse vehicles, off-road vehicles, spacecraft, ships, shuttles, emergency response vehicles, motorcycles, electric or motorized bicycles, aircraft, engineering vehicles, underwater vehicles, drones, and / or other vehicle types. In addition, although the present disclosure may be described with respect to autonomous driving or ADAS systems, this is not intended to be limiting. For example, the systems and methods described herein may be used to generate ground truth data for training deep neural networks (DNNs) for implementation in simulated environments, robotics (e.g., using map information of indoor environments, outdoor environments, warehouses, etc.), aviation systems, marine systems, and / or other technical fields.

[0018] refer to Figure 1 , Figure 1 An example data flow diagram of a process 100 for locating training data to high definition (HD) map data to enhance or generate labels for training a deep neural network (DNN) according to some embodiments of the present disclosure is included. Although the autonomous vehicle 500 and / or the example computing device 600 ( Figure 6 ) to describe process 100, but this is not intended to be limiting. It should be understood that this and other arrangements described herein are presented only as examples. In addition to those shown or in place of those shown, other arrangements and elements (e.g., machines, interfaces, functions, sequences, functional groupings, etc.) can be used, and some elements can be omitted completely. In addition, many of the elements described herein are functional entities that can be implemented as discrete or distributed components or combined with other components, and implemented in any suitable combination and position. The various functions described herein as being performed by an entity can be performed by hardware, firmware and / or software. For example, the various functions can be performed by a processor that executes instructions stored in a memory.

[0019] Process 100 may include generating and / or receiving sensor data 102 from one or more sensors of a data collection vehicle 500 (which may be similar to vehicle 500, or may include a non-autonomous or semi-autonomous vehicle). Sensor data 102 may be used within process 100 for localization, correlation, and ground truth generation, as well as input data for a deep neural network (DNN) 116. Sensor data 102 may include, but is not limited to, sensor data 102 from any type of sensor, such as, but not limited to, those described herein with respect to vehicle 500 and / or other vehicles or objects (e.g., robotic devices, VR systems, AR systems, etc., in some examples). For non-limiting examples, and with reference to Figures 5A-5C , the sensor data 102 may include, but is not limited to, data generated by a global navigation satellite system (GNSS) sensor 558 (e.g., a global positioning system (GPS) sensor, a differential GPS (DGPS) sensor, etc.), a RADAR sensor 560, an ultrasonic sensor 562, a LIDAR sensor 564, an inertial measurement unit (IMU) sensor 566 (e.g., an accelerometer, a gyroscope, a magnetic compass, a magnetometer, etc.), a microphone 596, a stereo camera 568, a wide-angle camera 570 (e.g., a fisheye camera), an infrared camera 572, a surround camera 574 (e.g., a 360-degree camera), a long-range and / or mid-range camera 598, a speed sensor 544 (e.g., for measuring the speed and / or travel distance of the data collection vehicle), and / or other sensor types.

[0020] In some examples, sensor data 102 may include sensor data generated by one or more forward-facing sensors, side-view sensors, and / or rear-view sensors. The sensor data 102 may be used to identify, detect, classify, and / or track objects around the data collection vehicle within an environment—e.g., for localization, association with HD map 104, ground truth generation, etc. In embodiments, any number of sensors may be used to combine multiple fields of view (e.g., Figure 5BThe fields of view of the remote camera 598, the forward stereo camera 568, and / or the forward wide-angle camera 570) and / or the sensing field (e.g., the LIDAR sensor 564, the RADAR sensor 560, etc.).

[0021] In some embodiments, the sensor data 102 may include image data representing an image, image data representing a video (e.g., a snapshot of a video), and / or sensor data representing a representation of a sensing field of a sensor (e.g., a depth map for a LIDAR sensor, a value map for an ultrasonic sensor, etc.). Where the sensor data 102 includes image data, any type of image data format may be used, such as, but not limited to, compressed images such as Joint Photographic Experts Group (JPEG) or Luma / Chrominance (YUV) formats, compressed images as frames derived from a compressed video format (e.g., H.264 / Advanced Video Coding (AVC) or H.265 / High Efficiency Video Coding (HEVC)), raw images (e.g., derived from Red-Transparent-Blue (RCCB), Red-Transparent-Blue (RCCC), or other types of imaging sensors), and / or other formats. Furthermore, in some examples, the sensor data 102 may be used within the process 100 without any preprocessing (e.g., in a raw or captured format), while in other examples, the sensor data 102 may undergo preprocessing (e.g., noise balancing, demosaicing, scaling, cropping, enhancement, white balancing, tone curve adjustment, etc., such as using a sensor data preprocessor (not shown)). As used herein, the sensor data 102 may refer to unprocessed sensor data, preprocessed sensor data, or a combination thereof.

[0022] In addition, process 100 may include obtaining a map, such as HD map 104 (which may be similar to Figure 5CHD map 522) - generates and / or receives map data that can be accessed by and / or stored by the data collection vehicle 500. In some embodiments, the HD map 104 and / or the locator 106 can be components of the HD map manager. In some embodiments, the HD map 104 may include centimeter-level or finer accuracy, so that the data collection vehicle 500 can rely on the HD map 104 for precise instructions, planning, and positioning. The HD map 104 can represent lanes, road boundaries, road shapes, elevations, slopes and / or contours, heading information, waiting conditions, static object locations, and / or other information. Therefore, the process 100 can use information from the HD map 104 - such as the location and shape of the lanes - to generate ground truth data for training the DNN 116. Although described herein as an HD map 104, this is not intended to be limiting, and the map data can be generated from any type of map with higher or lower accuracy than the HD map 104. For example, and without departing from the scope of the present disclosure, map data generated from a navigation or GPS application may be used in addition to or instead of using map data generated from the high definition map 104 .

[0023] The sensor data 102 and map data (e.g., from the HD map 104) may be used by the locator 106 to locate (e.g., for position and / or orientation) the data collection vehicle 500 relative to the HD map 104. For example, the sensor data 102 generated by a location-based sensor (e.g., a GNSS sensor 558) may be used to determine an approximate location of the data collection vehicle 500 on the HD map 104. This information may be used to determine areas of the HD map 104 that should be analyzed to accurately locate the data collection vehicle 500 within the HD map 104. For example, the sensor data 102 generated by a vision-based sensor, such as a camera, a RADAR sensor 560, an ultrasonic sensor 562, a LIDAR sensor 564, and / or other sensors, may be used to identify features and / or objects within the environment that have a known and accurate location within the area of ​​the HD map 104 determined using the location-based sensor. Once the vehicle 500 is first located within the HD map 104, the position and / or orientation of the vehicle 500 over time may be tracked using the same localization techniques—e.g., using vision-based sensors to identify objects and / or features with known locations in the HD map 104—and / or may be tracked using sensor data 102 generated by the vehicle's 500 ego-motion sensors (e.g., IMU sensors 566, speed sensors 544, steering sensors 540 tracking steering wheel angles, etc.). For example, once the vehicle 500 is initially located, the ego-motion sensors may be used to accurately track changes in the position and / or orientation of the vehicle 500 over time. However, even in the case of using ego-motion sensors, localization techniques using vision-based sensors may be performed periodically (e.g., every 3 seconds, every 10 seconds, etc.) to recalibrate the system to locate the vehicle 500 within the HD map 104.

[0024] In some non-limiting embodiments, the sensor data 102 and / or information from the HD map 104 may be applied to a coordinate transformer to transform the sensor data 102 and / or the HD map 104 to the coordinate system of the vehicle 500 and / or to transform map data from the HD map 104 to the coordinate space of the sensor data 102 (e.g., to transform 3D world space map data to 2D image space or sensor space). In order for the coordinate transformer 108 to perform the transformations or shifts described herein, intrinsic (e.g., optical center, focal length, etc.) and / or extrinsic (e.g., position of the sensor on the vehicle 500, rotation, translation, etc.) sensor parameters may be used to determine a correlation between a 2D pixel location on the HD map 104 (or other image or sensor space location) and a 2D or 3D world space location.

[0025] For example, the coordinate transformer 108 may orient the HD map 104 relative to the vehicle 500 and / or relative to the field of view of the sensor that captured the instance of the sensor data 102. In some embodiments, the coordinate transformer 108 may shift the perspective of the map data relative to the location and / or orientation of the data collection vehicle 500 and / or its sensors. Thus, portions of the HD map 104 that may be used by the ground truth generator 112 to generate ground truth data may be shifted relative to the vehicle 500 (e.g., where the data collection vehicle 500 is at the center, at (x,y) coordinates (0,0), where y is the longitudinal dimension extending from the front to the rear of the vehicle and x is the lateral dimension perpendicular to y and extending from the left to the right of the vehicle) and / or its sensors (e.g., the field of view or sensing field of the sensor that generated the instance of the sensor data 102 corresponding to the HD map 104). In some embodiments, in addition to or in lieu of shifting the perspective or coordinate system relative to the vehicle 500 and / or its sensors, the coordinate transformer 108 may also shift the perspective to the same field of view for each type of data. For example, where the HD map 104 may generate data from a top-down view of the environment, the sensor generating the sensor data 102 may do so from a different view—e.g., head-on, sideways, tilted downward, tilted upward, etc. Thus, in order to generate ground truth data from the same view, the coordinate transformer 108 may adjust the sensor data 102 and / or the map data to the same view. In some non-limiting embodiments, each of the sensor data 102 and the HD map 104 may be transferred to a top-down view view or coordinate system of the HD map 104, a coordinate system or view of the sensor data 102, and / or other view or coordinate systems.

[0026] In addition to or in lieu of the coordinate transformer 108 transferring or transforming the coordinate system of the high-definition map 104 to the coordinate system of the vehicle 500 and / or its sensors, in some embodiments, the coordinate transformer 108 may transfer or transform the map data to the coordinate system or dimensions of the sensor data 102. For example, in the case of training the DNN 116 to compute the output 118 in a 2D image space, the map data may be transformed or transferred from 2D or 3D world space coordinates to 2D image or sensor space coordinates. As another example, in the case of training the DNN 116 to compute the output 118 in a 3D world space, the map data may not be transformed or transferred to the 2D image space even if the sensor data 102 input to the DNN 116 represents sensor data representation in a two-dimensional space. Therefore, the DNN 116 may be trained to compute the output 118 in a 2D or 3D world space coordinate, and therefore, the ground truth generated for training the DNN 116 may correspond to a 2D or 3D world space coordinate system. In at least some embodiments, some or all of the sensor data 102 corresponding to the ground truth (e.g., features, objects, etc. represented thereby) can be converted from 2D image space or sensor space to 2D or 3D world space by a coordinate transformer 108 to generate ground truth corresponding to the 2D or 3D world space.

[0027] Once the localizer 106 has performed localization, and / or the sensor data 102 and / or the map data have been transformed or transferred by the coordinate transformer 108, the correlator 110 (which may include or alternatively be referred to herein as a feature determiner 110) may associate the map data with the sensor data 102. For example, depending on the type of ground truth data to be generated, the correlator may determine a correlation between features and / or objects represented by the map data and features and / or objects represented by the sensor data 102. For example, in the case of training the DNN 116 to predict the location of lane lines, dividers, and / or other features of the driving surface, the map data representing the lane lines, dividers, and / or other features may be correlated with the sensor data 102 at each sensor data instance (e.g., each image or frame) (e.g., similar to Figure 2A As another example, in the case of training DNN 116 to generate trajectories, as described herein, correlator 110 may determine a track (e.g., center) of a lane corresponding to a ground truth track such that ground truth generator 112 may generate a final ground truth track that more closely corresponds to or is centered on the track of the lane (e.g., similar to Figure 2AAs a further example, where DNN 116 is trained to generate outputs corresponding to an intersection (e.g., boundary shape vertices corresponding to a boundary shape surrounding the intersection), correlator 110 may determine each feature of the intersection corresponding to the intersection (e.g., a traffic light, a traffic sign, a label or marking on a driving surface, etc.), so that ground truth generator 112 may generate a boundary shape containing each feature (e.g., similar to Figure 2B Visualization 240).

[0028] Depending on the output 118 that the DNN 116 is trained to compute and the format of the output 118, the ground truth generator 112 can generate the ground truth data using the sensor data 102 and / or the map data from the HD map 104. For example, in the case where the DNN 116 is trained to generate outputs corresponding to points of a polyline defining a lane, the ground truth generator 112 can generate the ground truth data from the map data by determining points along the lane markings (in some increment) from the map data and associating these points with the points of the polyline. Thus, when the DNN 116 generates outputs 118 corresponding to polyline points, the ground truth data generated by the ground truth generator 112 can be compared to the points from the output 118 by the training engine 114 (e.g., using one or more loss functions) to determine updates to the parameters (e.g., weights and biases) of the DNN 116.

[0029] Similarly, where the DNN 116 is trained to predict an output 118 corresponding to a boundary shape of an intersection, the ground truth generator 112 may generate the ground truth boundary shape according to the format predicted by the DNN 116. For example, the DNN 116 may be trained to output boundary shape coordinates as pixel locations of two or more vertices of the boundary shape in 2D image space, and thus, the ground truth generator 112 may generate ground truth data according to such format—e.g., by generating ground truth data corresponding to the locations of the two or more vertices of the boundary shape. As another example, the DNN 116 may be trained to output boundary shape coordinates as pixel locations of the centroid of the boundary shape and dimensions (e.g., height and width) in 2D image space, and thus, the ground truth generator 112 may generate ground truth data according to such format—e.g., by generating ground truth data corresponding to the location of the centroid and the dimensions of the boundary shape. As a further example, DNN 116 may be trained to output bounding shape coordinates as the locations of two or more vertices of the bounding shape in 3D world space, and thus, ground truth generator 112 may generate ground truth data according to this format—e.g., by generating ground truth data corresponding to the locations of two or more vertices of the bounding shape in 3D world space.

[0030] In some embodiments, depending on any pre-processing of the sensor data 102, the ground truth generator 112 may compensate for the pre-processing in the generation of the ground truth data. For example, where instances of the sensor data 102 (e.g., images, depth maps, etc.) are pre-processed—e.g., by adjusting spatial resolution, flipping, rotating, cropping, scaling, augmenting, etc.—the ground truth generator 112 may account for these changes in generating the ground truth data. For example, in some embodiments, each instance of the sensor data 102 may be adjusted in some manner. In such an example, the ground truth generator 112 may adjust the ground truth data accordingly for each instance of the ground truth data (e.g., where each image represented by the sensor data 102 is cropped, each instance of the ground truth data may be adjusted to account for the cropping). In other examples, some instances of the sensor data 102 may be adjusted while other instances may not—e.g., to train the DNN 116 to not overfit and accurately compute outputs 118 that vary across any sensor data instance. In such an example, the ground truth generator 112 may receive data representing adjustments to instances of the sensor data 102 and may compensate for these adjustments when generating the ground truth data. For example, if a first instance of the sensor data 102 has not changed, the ground truth data may be generated normally, but if a second instance of the sensor data 102 is rotated prior to input into the DNN 116, the ground truth generator 112 may account for this by similarly rotating the ground truth data to correspond to the rotated instance of the sensor data 102.

[0031] Once the ground truth data is generated—automatically in embodiments—the ground truth data may be used by the training engine 114 to train the DNN 116. For example, the sensor data 102 may be applied to the DNN 116, the DNN 116 may generate an output 118, the training engine 114 may analyze the output 118 in light of the ground truth data from the ground truth generator 112 using one or more loss functions, and the DNN 116 may be updated using the computations of the training engine 114 until the DNN 116 converges to a desired or acceptable accuracy.

[0032] Depending on the embodiment, the DNN 116 may include any type of DNN or machine learning model. For example, but not limited to, the DNN 116 may 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, loops, perceptrons, long / short-term memory / LSTM, Hopfield, Boltzmann, deep belief, deconvolution, generative adversarial, liquid state machines, etc.), lane detection algorithms, computer vision algorithms, and / or other types of machine learning models.

[0033] As an example, such as where the DNN 116 includes a CNN, the DNN 116 may include any number of layers. One or more of the multiple layers may include an input layer. The input layer may store values ​​associated with the sensor data 102 (e.g., before or after post-processing). For example, when the sensor data 102 is an image, the input layer may store values ​​representing raw pixel values ​​of the image as a volume (e.g., width, height, and color channels (e.g., RGB), such as 32x32x3).

[0034] One or more layers may include a convolutional layer. A convolutional layer may compute the output of neurons connected to local regions in the input layer, with each neuron computing the dot product between their weights and the small region they are connected to in the input volume. The result of a convolutional layer may be another volume where one dimension is based on the number of filters applied (e.g., width, height, and number of filters, such as 32x32x12 if 12 is the number of filters).

[0035] One or more layers may include a rectified linear unit (ReLU) layer. For example, a ReLU layer may apply an element-wise activation function, such as max(0, x), such as a threshold set to zero. The resulting volume of the ReLU layer may be the same as the volume of the input to the ReLU layer.

[0036] One or more layers may include a pooling layer. The pooling layer may perform a downsampling operation along the spatial dimensions (e.g., height and width), which may result in a smaller volume than the input to the pooling layer (e.g., 16x16x12 from a 32x32x12 input volume).

[0037] One or more layers may include one or more fully connected layers. Each neuron in a fully connected layer may be connected to each neuron in a precursor. A fully connected layer may calculate class scores, and the resulting body may be 1x1x number of classes. In some examples, a CNN may include one or more fully connected layers, such that the output of one or more layers of a CNN may be provided as input to one or more fully connected layers of a CNN. In some examples, one or more convolutional streams may be implemented by a DNN 116, and some or all convolutional streams may include corresponding fully connected layers.

[0038] In some non-limiting embodiments, DNN 116 may include a series of convolutional layers and max pooling layers to facilitate image feature extraction, followed by multi-scale dilated convolutional layers and upsampling layers to facilitate global context feature extraction.

[0039] Although input layers, convolutional layers, pooling layers, ReLU layers, and fully connected layers are discussed herein with respect to DNN 116, this is not intended to be limiting. For example, additional or alternative layers, such as normalization layers, SoftMax layers, and / or other layer types, may be used in DNN 116.

[0040] In embodiments where DNN 116 includes a CNN, different orders and numbers of layers of the CNN may be used, depending on the embodiment. In other words, the order and number of layers of DNN 116 are not limited to any one architecture.

[0041] In addition, some layers may include parameters (e.g., weights and / or biases), such as convolutional layers and fully connected layers, while other layers may not include, such as ReLU layers and pooling layers. In some examples, the parameters can be learned by DNN 116 during training. In addition, some layers may include additional hyperparameters (e.g., learning rate, stride, epoch, etc.), such as convolutional layers, fully connected layers, and pooling layers, while other layers may not include, such as ReLU layers. Parameters and hyperparameters are not limited and may vary depending on the embodiment.

[0042] Reference now Figures 2A-2C , Figures 2A-2C Example visualizations are included that represent various instances of automatically generating ground truth using process 100. The visualizations described herein are for example purposes only and are not intended to limit the scope of the present disclosure. Figure 2A , Figure 2AAn example visualization 200 of automatically generated ground truth labels corresponding to road features in accordance with some embodiments of the present disclosure is depicted. Visualization 200 may represent an instance of sensor data 102 (e.g., an image) and corresponding ground truth data that may be generated using map data from HD map 104 and / or trajectory information from ego-motion sensor data. For example, where DNN 116 is trained to predict road features (e.g., lane tracks or centers, lane dividers, road boundaries, etc.), the ground truth data generated from the map data may include each of the ground truth labels in visualization 200. In one or more embodiments, the ground truth labels may include all labels except ego-trajectory 202 (described in more detail herein). For example, after the vehicle 500 is localized to the HD map 104, the coordinate transformer 108 orients the map data relative to the coordinate system or dimensional space of the vehicle 500, its sensors, and / or the sensor data 102, and the correlator 110 can determine each road boundary 204 (e.g., 204A and 204B), each lane separator 206 (e.g., 206A-206D), each lane track 210 (e.g., 210A-210C, 208), and / or other features of the road and their respective locations relative to the instance of the sensor data. The ground truth generator 112 can then generate ground truth data from each label corresponding to these features in the format that the DNN 116 is trained to predict output, as described herein.

[0043] As another example, refer again to Figure 2A , map data from the HD map can be used to augment other ground truth data. For example, where the ego trajectory 202 is automatically generated using an ego-motion sensor, the ego trajectory 202 can be adjusted to more closely correspond to a lane track 208 corresponding to the lane of travel of the vehicle 500. Thus, the ground truth generator 112 can generate a final ground truth trajectory corresponding to the lane track 208 rather than the ego trajectory 202 - or data representative thereof, such as points along a broken line defining the final ground truth trajectory. In some non-limiting embodiments, the ego trajectory can be generated using methods and systems as described in U.S. Non-Provisional Application No. 16 / 409,056, filed on May 10, 2019, which is incorporated herein by reference in its entirety. For example, and with reference to Figure 3 Because the ego trajectory 202 may be generated based on a driving path of a human driver, the ego trajectory may deviate from the center or track of the driving lane. Figure 3Graph 300 of illustrates the offsets corresponding to generated ego trajectories, depicting trajectories that deviate from the center or track of the lane of travel by more than half a meter or more. Where these ego trajectories are used directly to train the DNN 116 to compute future trajectories for the vehicle, the computed trajectories in deployment may similarly include offsets from the track or center of the lane of travel. Thus, by augmenting the ego trajectories with map data to generate ground truth data that more closely follows the center or track of the lane of travel, the computed trajectories of the deployed vehicle may also more closely follow the center or track of the lane of travel. Furthermore, since the trajectory may include lane changes, turns, and / or the like, adjustments to the ego trajectories used to generate the ground truth may also enable turns, lane changes, and / or other operations of the computed trajectory of the DNN 116 to be transferred from the center or track of the lane of travel to the center or track of other lanes of travel. As a result, the computed trajectory may place the deployed vehicle at a safer distance from surrounding objects as the vehicle traverses the driving surface more closely along the center or track of the lane of travel.

[0044] Reference now Figure 2B , Figure 2B An example visualization 240 of automatically generated ground truth labels corresponding to intersection detection and classification according to some embodiments of the present disclosure is depicted. For example, the DNN 116 can be trained to generate outputs 118 representing a boundary shape corresponding to an intersection (e.g., boundary shape 242 of visualization 240), a classification corresponding to the intersection (e.g., intersection, traffic light controlled), a distance to the intersection, and / or other information corresponding to the intersection. Because map data from an HD map can represent features of an intersection and their locations, the map data can be used to generate ground truth data. For example, where the boundary shape 242 will contain each feature of the intersection - e.g., traffic light 244, crosswalks 246A and 246B, intersection entry line 248, etc. - the location and existence of these features can be determined from the map determination data, and the correlator 110 can determine the dimensions of the boundary shape 242 that encompasses all of them, the classification of the intersection, and the distance to the intersection (e.g., to the entry line in intersection entry lines 248). Thus, the ground truth generator 212 can generate ground truth data corresponding to the classification information, boundary shape dimensions and / or vertices, and / or distance to the intersection. In some non-limiting embodiments, the ground truth generated for the intersection can be similar to the ground truth data generated in U.S. Non-Provisional Application No. 16 / 814,351, filed on March 10, 2020, which is incorporated herein by reference in its entirety.

[0045] Reference now Figure 2C , Figure 2CAn example visualization 280 of automatically generated ground truth labels corresponding to features of an environment (including poles, road markings, lane lines, road boundaries, and crosswalks) according to some embodiments of the present disclosure is depicted. For example, the DNN 116 can be trained to compute outputs 118 corresponding to features of the environment (e.g., lines, markings, poles, etc.). Thus, map data from the HD map 104 can be used to generate ground truth data that can represent poles 282, crosswalks 284, lane markings 286, road markings 288, and / or other features of the environment. Thus, after positioning, coordinate transformation, and / or association, the ground truth generator 112 can generate ground truth data corresponding to each feature of the environment that the DNN 116 is trained to compute. In a non-limiting example, the generated ground truth can be similar to that described in U.S. Non-Provisional Application No. 16 / 514,230, filed on July 17, 2019, which is incorporated herein by reference in its entirety.

[0046] Reference now Figure 4 , each block of the method 400 described herein includes a computing process that can be performed using any combination of hardware, firmware, and / or software. For example, various functions can be performed by a processor executing instructions stored in a memory. The method 400 can also be embodied as computer usable instructions stored on a computer storage medium. The method 400 can be provided by a stand-alone application, a service or a hosted service (standalone or in combination with another hosted service), or a plug-in to another product, to name a few. In addition, by way of example, with respect to Figure 1 The process of 100 and Figures 5A-5D Method 400 is described with reference to autonomous vehicle 500. However, these methods may additionally or alternatively be performed by any system or within any process or any combination of systems and processes, including but not limited to those described herein.

[0047] Figure 4 4 is a flow chart illustrating a method 400 for locating training data to HD map data to enhance or generate labels for training a DNN according to some embodiments of the present disclosure. At block B402, the method 400 includes receiving HD map data corresponding to an area including a location of a dynamic participant at a certain time. For example, map data corresponding to an area of ​​the HD map 104 including the data collection vehicle 500 may be generated and / or received—e.g., using a HD map manager.

[0048] At block B404 , the method 400 includes locating the dynamic participant relative to the HD map data. For example, the locator 106 may locate the data collection vehicle 500 within the HD map 104 .

[0049] At block B406 , the method 400 includes receiving sensor data generated by a sensor of a dynamic participant at a certain time. For example, an instance of the sensor data 102 generated by a sensor of the data collection vehicle 500 may be generated and / or received.

[0050] At block B408, method 400 includes generating ground truth data corresponding to the sensor data based at least in part on the HD map data and the positioning. For example, ground truth generator 112 may use the map data to generate ground truth data corresponding to an instance of sensor data 102. In some embodiments, this may include transforming or transferring the coordinate system of high-definition map 104 to the coordinate system of vehicle 500, associating map data with sensor data 102, and / or other processes described herein with respect to coordinate transformer 108, correlator 110, and / or ground truth generator 112.

[0051] At block B410, method 400 includes training a neural network using sensor data and ground truth data. For example, DNN 116 may be trained using sensor data 102 and ground truth data generated using ground truth generator 112—eg, using training engine 114.

[0052] Example Autonomous Vehicle

[0053] Figure 5A is an illustration of an example autonomous vehicle 500 according to some embodiments of the present disclosure. Autonomous vehicle 500 (alternatively referred to herein as "vehicle 500") may include, but is not limited to, a passenger vehicle, such as a car, a truck, a bus, an emergency vehicle, a shuttle, an electric or motorized bicycle, a motorcycle, a fire truck, a police car, an ambulance, a boat, an engineering vehicle, an underwater vessel, a drone, and / or other types of vehicles (e.g., unmanned and / or capable of accommodating 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). The vehicle 500 may be capable of implementing functions consistent with one or more of Levels 3-5 of autonomous driving levels. For example, depending on the embodiment, the vehicle 500 may be capable of conditional automation (Level 3), high automation (Level 4), and / or full automation (Level 5).

[0054] The vehicle 500 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 500 may include a propulsion system 550, such as an internal combustion engine, a hybrid power plant, an all-electric engine, and / or another type of propulsion system. The propulsion system 550 may be connected to a drive train of the vehicle 500, which may include a transmission, to achieve propulsion of the vehicle 500. The propulsion system 550 may be controlled in response to receiving a signal from a throttle / accelerator 552.

[0055] A steering system 554, which may include a steering wheel, may be used to steer vehicle 500 (e.g., along a desired path or route) when propulsion system 550 is operating (e.g., when the vehicle is in motion). Steering system 554 may receive signals from steering actuator 556. For fully automated (Level 5) functionality, a steering wheel may be optional.

[0056] Brake sensor system 546 may be used to operate vehicle brakes in response to receiving signals from brake actuator 548 and / or brake sensors.

[0057] May include one or more system on chip (SoC) 504 ( Figure 5C ) and / or one or more controllers 536 of one or more GPUs can provide signals (e.g., representing commands) to one or more components and / or systems of the vehicle 500. For example, the one or more controllers can send signals to operate vehicle brakes via one or more brake actuators 548, to operate the steering system 554 via one or more steering actuators 556, and to operate the propulsion system 550 via one or more throttles / accelerators 552. The one or more controllers 536 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 achieve autonomous driving and / or assist a human driver in driving the vehicle 500. The one or more controllers 536 may include a first controller 536 for autonomous driving functions, a second controller 536 for functional safety functions, a third controller 536 for artificial intelligence functions (e.g., computer vision), a fourth controller 536 for infotainment functions, a fifth controller 536 for redundancy in emergency situations, and / or other controllers. In some examples, a single controller 536 may handle two or more of the above functions, two or more controllers 536 may handle a single function, and / or any combination thereof.

[0058] The one or more controllers 536 may provide signals for controlling one or more components and / or systems of the vehicle 500 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 558 (e.g., a global positioning system sensor), a RADAR sensor 560, an ultrasonic sensor 562, a LIDAR sensor 564, an inertial measurement unit (IMU) sensor 566 (e.g., an accelerometer, a gyroscope, a magnetic compass, a magnetometer, etc.), a microphone 596, a stereo camera 568, a wide angle camera 570 (e.g., a fisheye camera), an infrared camera 572, a surround camera 574 (e.g., a 360 degree camera), a long-range and / or mid-range camera 598, a speed sensor 544 (e.g., for measuring the velocity of the vehicle 500), a vibration sensor 542, a steering sensor 540, a brake sensor (e.g., as part of a brake sensor system 546), and / or other sensor types.

[0059] One or more of the controllers 536 may receive inputs (e.g., represented by input data) from the instrument cluster 532 of the vehicle 500 and provide outputs (e.g., represented by output data, display data, etc.) via a human machine interface (HMI) display 534, an audible annunciator, a speaker, and / or via other components of the vehicle 500. These outputs may include information such as vehicle speed, velocity, time, map data (e.g., Figure 5C The HMI display 534 may include information such as the HD map 522 of the vehicle 500, location data (e.g., the location of the vehicle 500 on the map), directions, locations of other vehicles (e.g., occupancy grid), information about objects and states of objects as sensed by the controller 536, etc. For example, the HMI display 534 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.).

[0060] The vehicle 500 further includes a network interface 524 that can communicate via one or more networks using one or more wireless antennas 526 and / or a modem. For example, the network interface 524 may be capable of communicating via LTE, WCDMA, UMTS, GSM, CDMA2000, etc. The one or more wireless antennas 526 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.

[0061] Figure 5BAccording to some embodiments of the present disclosure, Figure 5A An example of camera positions and fields of view for an example autonomous vehicle 500 is shown. 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 500.

[0062] 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 500. 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), 520fps, 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 Foveon X3 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.

[0063] 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 of the cameras) can simultaneously record and provide image data (e.g., video).

[0064] 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.

[0065] 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 500 can be used for surround view 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 536 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.

[0066] 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 570, which may be used to sense objects (e.g., pedestrians, intersection traffic, or bicycles) entering the field of view from the periphery. Although Figure 5B Only one wide-angle camera is shown in the figure, but there may be any number of wide-angle cameras 570 on the vehicle 500. In addition, long-range cameras 598 (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. Long-range cameras 598 may also be used for object detection and classification and basic object tracking.

[0067] One or more stereo cameras 568 may also be included in the front configuration. The stereo camera 568 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 568 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 568 may be used in addition to or alternatively to those described herein.

[0068] A camera (e.g., a side view camera) having a field of view of a portion of the environment including the sides of the vehicle 500 can be used for surround viewing, providing information used to create and update the occupancy grid and generate side impact collision warnings. Figure 5BThe four surround cameras 574 shown in FIG. 5 can be placed on the vehicle 500. The surround cameras 574 can include a wide-angle camera 570, a fisheye camera, a 360-degree camera, and / or the like. For example, the four fisheye cameras can be placed in front, behind, and on the sides of the vehicle. In an alternative arrangement, the vehicle can use three surround cameras 574 (e.g., left, right, and rear), and can utilize one or more other cameras (e.g., a forward-facing camera) as a fourth surround camera.

[0069] A camera having a field of view that includes a portion of the environment behind the vehicle 500 (e.g., a rear view camera) can be used to assist with parking, surround view, rear collision warning, and creating and updating an occupancy grid. A variety of cameras can be used, including but not limited to cameras that are also suitable as front-facing cameras as described herein (e.g., long-range and / or mid-range cameras 598, stereo cameras 568, infrared cameras 572, etc.).

[0070] Figure 5C According to some embodiments of the present disclosure, Figure 5A Block diagram of an example system architecture of an example autonomous vehicle 500. 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.

[0071] Figure 5C Each of the components, features, and systems of the vehicle 500 is illustrated as being connected via a bus 502. The bus 502 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 500 that assists in controlling various features and functions of the vehicle 500, 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., a 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.

[0072] Although bus 502 is described as a CAN bus here, this is not intended to be limiting. For example, in addition to or alternatively to the CAN bus, FlexRay and / or Ethernet can be used. In addition, although bus 502 is represented by a single line, this is not intended to be limiting. For example, there can be any number of buses 502, 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 502 can be used to perform different functions, and / or can be used for redundancy. For example, a first bus 502 can be used for a collision avoidance function, and a second bus 502 can be used for drive control. In any example, each bus 502 can communicate with any component of the vehicle 500, and two or more buses 502 can communicate with the same component. In some examples, each SoC 504, each controller 536, and / or each computer in the vehicle can have access to the same input data (e.g., input from a sensor of the vehicle 500), and can be connected to a common bus such as a CAN bus.

[0073] The vehicle 500 may include one or more controllers 536, such as those described herein. Figure 5A Those controllers described. Controller 536 can be used for a variety of functions. Controller 536 can be coupled to any other different components and systems of vehicle 500, and can be used for control of vehicle 500, artificial intelligence of vehicle 500, infotainment for vehicle 500, and / or the like.

[0074] The vehicle 500 may include one or more system on chip (SoC) 504. The SoC 504 may include a CPU 506, a GPU 508, a processor 510, a cache 512, an accelerator 514, a data store 516, and / or other components and features not shown. The SoC 504 may be used to control the vehicle 500 in a variety of platforms and systems. For example, one or more SoCs 504 may be combined with an HD map 522 in a system (e.g., a system of the vehicle 500), and the HD map may be downloaded from one or more servers (e.g., a server) via a network interface 524. Figure 5D one or more servers 578) to obtain map refreshes and / or updates.

[0075] CPU 506 may include a CPU cluster or CPU complex (alternatively, referred to herein as "CCPLEX"). CPU 506 may include multiple cores and / or L2 caches. For example, in some embodiments, CPU 506 may include eight cores in a coherent multiprocessor configuration. In some embodiments, CPU 506 may include four dual-core clusters, each with a dedicated L2 cache (e.g., 2MB L2 cache). CPU 506 (e.g., CCPLEX) may be configured to support simultaneous cluster operations, such that any combination of CPU 506 clusters can be active at any given time.

[0076] CPU 506 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. CPU 506 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.

[0077] The GPU 508 may include an integrated GPU (alternatively, referred to herein as an "iGPU"). The GPU 508 may be programmable and efficient for parallel workloads. In some examples, the GPU 508 may use an enhanced tensor instruction set. The GPU 508 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, the GPU 508 may include at least eight streaming microprocessors. The GPU 508 may use a computing application programming interface (API). In addition, the GPU 508 may use one or more parallel computing platforms and / or programming models (e.g., NVIDIA's CUDA).

[0078] In the case of automotive and embedded use, GPU 508 can be power optimized to achieve optimal performance. For example, GPU 508 can be manufactured on fin field effect transistors (FinFETs). However, this is not intended to be limiting, and GPU 508 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 can include independent parallel integer and floating point data paths to provide efficient execution of workloads using a mix of calculations and addressing calculations. 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.

[0079] GPU 508 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.

[0080] The GPU 508 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 can be used to allow the GPU 508 to directly access the CPU 506 page table. In such an example, when the GPU 508 memory management unit (MMU) experiences a miss, an address translation request can be transmitted to the CPU 506. In response, the CPU 506 can look up the virtual-physical mapping for the address in its page table and transmit the translation back to the GPU 508. In this way, unified memory technology can allow a single unified virtual address space to be used for the memory of both the CPU 506 and the GPU 508, thereby simplifying GPU 508 programming and porting applications to the GPU 508.

[0081] In addition, GPU 508 can include access counters that can track how often GPU 508 accesses the memory of other processors. Access counters can help ensure that memory pages are moved to the physical memory of the processor that accesses them most frequently.

[0082] SoC 504 may include any number of caches 512, including those described herein. For example, cache 512 may include an L3 cache available to both CPU 506 and GPU 508 (e.g., connected to both CPU 506 and GPU 508). Cache 512 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.

[0083] The SoC 504 may include one or more arithmetic logic units (ALUs) that may be used to perform processing for any of a variety of tasks or operations with respect to the vehicle 500, such as processing a DNN. Additionally, the SoC 504 may include a floating point unit (FPU) or other math coprocessor or digital coprocessor type for performing mathematical operations within the system. For example, the SoC 504 may include one or more FPUs integrated as execution units within the CPU 506 and / or GPU 508.

[0084] SoC 504 may include one or more accelerators 514 (e.g., hardware accelerators, software accelerators, or combinations thereof). For example, SoC 504 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 GPU 508 and offload some tasks of GPU 508 (e.g., free up more cycles of GPU 508 for performing other tasks). As an example, accelerator 514 may be used for targeted workloads (e.g., perception, convolutional neural networks (CNNs), etc.) that are sufficiently stable to be easily controlled for 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).

[0085] The accelerator 514 (e.g., a hardware acceleration cluster) may include a deep learning accelerator (DLA). The 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. The TPU may be an accelerator configured to perform image processing functions (e.g., for CNN, RCNN, etc.) and optimized for performing image processing functions. The DLA may be further optimized for a specific set of neural network types and floating point operations and reasoning. The design of the DLA may provide higher performance per millimeter than a general-purpose GPU, and far exceeds the performance of the CPU. The 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.

[0086] 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.

[0087] The DLA can perform any function of the GPU 508, and by using an inference accelerator, for example, the designer can target any function to either the DLA or the GPU 508. For example, the designer can focus the processing of CNNs and floating point operations on the DLA, and leave other functions to the GPU 508 and / or other accelerators 514.

[0088] The accelerator 514 (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.

[0089] 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.

[0090] The DMA may enable components of the PVA to access system memory independently of the CPU 506. 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.

[0091] 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.

[0092] 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.

[0093] The accelerator 514 (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 514. 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).

[0094] The on-chip computer vision network may include an interface that determines that both the 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 ISO 26262 or IEC 61508 standards, but other standards and protocols may also be used.

[0095] In some examples, SoC 504 may include a real-time ray tracing hardware accelerator such as described in U.S. Patent Application No. 16 / 101,232 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 purposes. In some embodiments, one or more tree traversal units (TTUs) may be used to perform one or more ray tracing related operations.

[0096] The accelerator 514 (e.g., a hardware accelerator cluster) 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.

[0097] 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.

[0098] 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.

[0099] 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 a confidence value can be interpreted as a probability, or as providing a relative "weight" of each detection compared to other detections. This 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 will 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 can take at least some subset of parameters as its input, such as bounding box dimensions, ground plane estimates obtained (e.g., from another subsystem), inertial measurement unit (IMU) sensor 566 output related to vehicle 500 orientation and distance, 3D position estimates of objects obtained from neural networks and / or other sensors (e.g., LIDAR sensor 564 or RADAR sensor 560), etc.

[0100] SoC 504 may include one or more data stores 516 (e.g., memory). Data store 516 may be on-chip memory of SoC 504 that may store neural networks to be executed on the GPU and / or DLA. In some examples, data store 516 may be large enough to store multiple instances of the neural network for redundancy and safety. Data store 512 may include an L2 or L3 cache 512. References to data store 516 may include references to memory associated with PVA, DLA, and / or other accelerators 514 as described herein.

[0101] SoC 504 may include one or more processors 510 (e.g., embedded processors). Processor 510 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 SoC 504 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, SoC 504 thermal and temperature sensor management, and / or SoC 504 power state management. Each temperature sensor may be implemented as a ring oscillator whose output frequency is proportional to the temperature, and SoC 504 may use the ring oscillator to detect the temperature of CPU 506, GPU 508, and / or accelerator 514. If it is determined that the temperature exceeds a threshold, the boot and power management processor may enter a temperature fault routine and place SoC 504 in a lower power state and / or place vehicle 500 in a driver safety parking mode (e.g., parking vehicle 500 safely).

[0102] Processor 510 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.

[0103] The processor 510 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.

[0104] The processor 510 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.

[0105] Processor 510 may further include a real-time camera engine, which may include a dedicated processor subsystem for handling real-time camera management.

[0106] Processor 510 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 the camera processing pipeline.

[0107] The processor 510 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 for 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 570, the surround camera 574, and / or for an 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 calls, dictate emails, change vehicle destinations, 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.

[0108] 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.

[0109] 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 508 does not need to continuously render new surfaces. Even when the GPU 508 is powered on and active for 3D rendering, the video image compositor may be used to offload the GPU 508 to improve performance and responsiveness.

[0110] SoC 504 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 504 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.

[0111] SoC 504 may further include a wide range of peripheral interfaces to enable communication with peripherals, audio codecs, power management and / or other devices. SoC 504 may be used to process data from cameras (connected via Gigabit multimedia serial link and Ethernet), sensors (e.g., LIDAR sensor 564, RADAR sensor 560, etc., which may be connected via Ethernet), data from bus 502 (e.g., speed of vehicle 500, steering wheel position, etc.), data from GNSS sensor 558 (connected via Ethernet or CAN bus). SoC 504 may further include dedicated high-performance mass storage controllers, which may include their own DMA engines, and which may be used to free up CPU 506 from routine data management tasks.

[0112] SoC 504 can be an end-to-end platform with a flexible architecture that spans automation levels 3-5, providing a comprehensive functional safety architecture that leverages and efficiently uses computer vision and ADAS technologies to achieve diversity and redundancy, together with deep learning tools to provide a platform for a flexible and reliable driving software stack. SoC 504 can be faster, more reliable, and even more energy efficient and space efficient than conventional systems. For example, when combined with CPU 506, GPU 508, and data storage 516, accelerator 514 can provide a fast and efficient platform for level 3-5 autonomous vehicles.

[0113] 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.

[0114] 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., GPU 520) 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 that semantic understanding to a path planning module running on the CPU complex.

[0115] 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 identified 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 the DLA and / or on the GPU 508.

[0116] In some examples, a CNN for facial recognition and owner recognition can use data from the camera sensor to identify the presence of an authorized driver and / or owner of the vehicle 500. The 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 504 provides security against theft and / or carjacking.

[0117] In another example, a CNN for emergency vehicle detection and identification can use data from microphone 596 to detect and identify emergency vehicle sirens. In contrast to conventional systems that use general classifiers to detect sirens and manually extract features, SoC 504 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 558. 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 562, 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.

[0118] The vehicle may include a CPU 518 (e.g., a discrete CPU or dCPU) that may be coupled to the SoC 504 via a high-speed interconnect (e.g., PCIe). The CPU 518 may include, for example, an X86 processor. The CPU 518 may be used to perform any of a variety of functions, including, for example, arbitrating potentially inconsistent results between ADAS sensors and the SoC 504, and / or monitoring the status and health of the controller 536 and / or the infotainment SoC 530.

[0119] The vehicle 500 may include a GPU 520 (e.g., a discrete GPU or dGPU) that may be coupled to the SoC 504 via a high-speed interconnect (e.g., NVIDIA's NVLINK). The GPU 520 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 on inputs (e.g., sensor data) from sensors of the vehicle 500.

[0120] The vehicle 500 may further include a network interface 524, which may include one or more wireless antennas 526 (e.g., one or more wireless antennas for different communication protocols, such as cellular antennas, Bluetooth antennas, etc.). The network interface 524 can be used to enable wireless connections with the cloud (e.g., with a server 578 and / or other network devices), with other vehicles, and / or with 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 500 with information about vehicles approaching the vehicle 500 (e.g., vehicles in front of, to the side of, and / or behind the vehicle 500). This functionality can be part of the collaborative adaptive cruise control functionality of the vehicle 500.

[0121] The network interface 524 may include a SoC that provides modulation and demodulation functions and enables the controller 536 to communicate over a wireless network. The network interface 524 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.

[0122] The vehicle 500 may further include data storage 528, which may include off-chip storage (e.g., outside the SoC 504). The data storage 528 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.

[0123] The vehicle 500 may further include a GNSS sensor 558. The GNSS sensor 558 (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 558 may be used, including, for example and without limitation, GPS using a USB connector with an Ethernet to serial (RS-232) bridge.

[0124] The vehicle 500 may further include a RADAR sensor 560. The RADAR sensor 560 may be used by the vehicle 500 for remote vehicle detection even in darkness and / or inclement weather conditions. The RADAR functional safety level may be ASIL B. The RADAR sensor 560 may use CAN and / or bus 502 (e.g., to transmit data generated by the RADAR sensor 560) 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 560 may be suitable for front, rear, and side RADAR use. In some examples, a pulse Doppler RADAR sensor is used.

[0125] The RADAR sensor 560 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 browsing. The RADAR sensor 560 may help distinguish between static objects 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 500 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 500.

[0126] As an example, a medium-range RADAR system may include a range of up to 560m (front) or 80m (rear) and a field of view of up to 42 degrees (front) or 550 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.

[0127] Short-range RADAR systems can be used in ADAS systems for blind spot detection and / or lane change assistance.

[0128] The vehicle 500 may further include ultrasonic sensors 562. The ultrasonic sensors 562, which may be placed on the front, rear, and / or sides of the vehicle 500, may be used for parking assistance and / or creating and updating occupancy grids. A variety of ultrasonic sensors 562 may be used, and different ultrasonic sensors 562 may be used for different detection ranges (e.g., 2.5 m, 4 m). The ultrasonic sensors 562 may operate at a functional safety level of ASIL B.

[0129] The vehicle 500 may include a LIDAR sensor 564. The LIDAR sensor 564 may be used for object and pedestrian detection, emergency braking, collision avoidance, and / or other functions. The LIDAR sensor 564 may be ASIL B for functional safety level. In some examples, the vehicle 500 may include multiple LIDAR sensors 564 (e.g., two, four, six, etc.) that may use Ethernet (e.g., to provide data to a Gigabit Ethernet switch).

[0130] In some examples, the LIDAR sensor 564 may be able to provide a list of objects and their distances for a 360-degree field of view. Commercially available LIDAR sensors 564 may have, for example, an advertised range of approximately 500m, an accuracy of 2cm-3cm, and support for 500Mbps Ethernet connections. In some examples, one or more non-protruding LIDAR sensors 564 may be used. In such examples, the LIDAR sensor 564 may be implemented as a small device that can be embedded in the front, back, side, and / or corner of the vehicle 500. In such examples, the LIDAR sensor 564 may provide a field of view of up to 520 degrees horizontally and 35 degrees vertically, with a range of 200m, even for low reflectivity objects. The front-mounted LIDAR sensor 564 may be configured for a horizontal field of view between 45 and 135 degrees.

[0131] 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 500. Available 3D flash LIDAR systems include solid-state 3D staring array LIDAR cameras (e.g., non-browsing 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 564 may be less susceptible to motion blur, vibration, and / or shock.

[0132] The vehicle may further include an IMU sensor 566. In some examples, the IMU sensor 566 may be located at the center of the rear axle of the vehicle 500. The IMU sensor 566 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 566 may include an accelerometer and a gyroscope, and in a nine-axis application, the IMU sensor 566 may include an accelerometer, a gyroscope, and a magnetometer.

[0133] In some embodiments, the IMU sensor 566 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 566 can enable the vehicle 500 to estimate heading by directly observing and correlating velocity changes from the GPS to the IMU sensor 566 without the need for input from a magnetic sensor. In some examples, the IMU sensor 566 and the GNSS sensor 558 can be combined into a single integrated unit.

[0134] The vehicle may include microphones 596 positioned in and / or around the vehicle 500. The microphones 596 may be used for, among other things, emergency vehicle detection and identification.

[0135] The vehicle may further include any number of camera types, including a stereo camera 568, a wide angle camera 570, an infrared camera 572, a surround camera 574, a long-range and / or mid-range camera 598, and / or other camera types. These cameras may be used to capture image data around the entire periphery of the vehicle 500. The type of camera used depends on the embodiment and the requirements of the vehicle 500, and any combination of camera types may be used to provide the necessary coverage around the vehicle 500. In addition, the number of cameras may 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 may support Gigabit Multimedia Serial Link (GMSL) and / or Gigabit Ethernet. Each of the cameras described herein may be described in detail with respect to Figure 5A and Figure 5B Described in more detail.

[0136] The vehicle 500 may further include a vibration sensor 542. The vibration sensor 542 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 542 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).

[0137] The vehicle 500 may include an ADAS system 538. In some examples, the ADAS system 538 may include a SoC. The ADAS system 538 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.

[0138] The ACC system may use a RADAR sensor 560, a LIDAR sensor 564, 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 500, 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 500 change lanes when necessary. The lateral ACC is related to other ADAS applications such as LCA and CWS.

[0139] 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 524 and / or a wireless antenna 526. 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 500 and in the same lane as it), while the I2V communication concept provides information about traffic farther ahead. The CACC system may include either or both of the I2V and V2V information sources. Given information about the vehicle ahead of the vehicle 500, CACC can be more reliable, and it is possible to improve the smoothness of traffic flow and reduce road congestion.

[0140] 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 560 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.

[0141] The AEB system detects an impending front 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 560 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.

[0142] The LDW system provides visual, audible, and / or tactile warnings, such as steering wheel or seat vibrations, to alert the driver when the vehicle 500 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.

[0143] The LKA system is a variation of the LDW system. If the vehicle 500 begins to leave the lane, the LKA system provides steering input or braking to correct the vehicle 500.

[0144] The BSW system detects and warns the driver of vehicles in the car's blind spot. The BSW system can provide visual, audible, and / or tactile alerts to indicate that it is unsafe to merge or change lanes. 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 560 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.

[0145] The RCTW system can provide visual, audible, and / or tactile notifications when an object is detected outside the range of the rear camera while the vehicle 500 is in reverse. Some RCTW systems include AEB to ensure that the vehicle brakes are applied to avoid a crash. The RCTW system can use one or more rear RADAR sensors 560 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.

[0146] 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 actually exists and take action accordingly. However, in the autonomous vehicle 500, in the case of conflicting results, the vehicle 500 itself must decide whether to pay attention to the results from the main computer or the auxiliary computer (e.g., the first controller 536 or the second controller 536). For example, in some embodiments, the ADAS system 538 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 538 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.

[0147] 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.

[0148] The supervisory MCU may 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 on outputs from the primary computer and the auxiliary computer. Thus, the neural network in the supervisory MCU may learn when the output of the auxiliary computer may be trusted and when it may not. For example, when the auxiliary computer is a RADAR-based FCW system, the neural network in the supervisory MCU may learn when the FCW system is identifying a metal object that is not actually dangerous, such as a drain grate 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 may learn to ignore the LDW when a cyclist or pedestrian is present and lane departure is actually the safest strategy. In embodiments that include 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 with associated memory. In a preferred embodiment, the supervisory MCU may include and / or be included as a component of the SoC 504.

[0149] In other examples, the ADAS system 538 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.

[0150] In some examples, the output of the ADAS system 538 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 538 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.

[0151] The vehicle 500 may further include an infotainment SoC 530 (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 530 may include a combination of hardware and software that may 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, Wi-Fi, 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 500. For example, the infotainment SoC 530 may include a radio, a disc player, a navigation system, a video player, USB and Bluetooth connectivity, an onboard computer, onboard entertainment, Wi-Fi, steering wheel audio controls, hands-free voice controls, a head-up display (HUD), an HMI display 534, 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 530 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 538, 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.

[0152] The infotainment SoC 530 may include GPU functionality. The infotainment SoC 530 may communicate with other devices, systems, and / or components of the vehicle 500 via a bus 502 (e.g., a CAN bus, Ethernet, etc.). In some examples, the infotainment SoC 530 may be coupled to a supervisory MCU so that in the event of a failure of a main controller 536 (e.g., a main and / or backup computer of the vehicle 500), the GPU of the infotainment system may perform some self-driving functions. In such an example, the infotainment SoC 530 may place the vehicle 500 in a driver safety parking mode as described herein.

[0153] The vehicle 500 may further include an instrument cluster 532 (e.g., a digital instrument panel, an electronic instrument cluster, a digital instrument panel, etc.). The instrument cluster 532 may include a controller and / or a supercomputer (e.g., a separate controller or a supercomputer). The instrument cluster 532 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, lighting controls, safety system controls, navigation information, and the like. In some examples, information may be displayed and / or shared between the infotainment SoC 530 and the instrument cluster 532. In other words, the instrument cluster 532 may be included as part of the infotainment SoC 530, or vice versa.

[0154] Figure 5D For a cloud-based server and Figure 5A 500. System 576 may include a server 578, a network 590, and a vehicle including the vehicle 500. The server 578 may include a plurality of GPUs 584(A)-584(H) (collectively referred to herein as GPUs 584), PCIe switches 582(A)-582(H) (collectively referred to herein as PCIe switches 582), and / or CPUs 580(A)-580(B) (collectively referred to herein as CPUs 580). The GPUs 584, CPUs 580, and PCIe switches may be interconnected with a high-speed interconnect and / or PCIe connection 586 such as, for example and without limitation, an NVLink interface 588 developed by NVIDIA. In some examples, the GPUs 584 are connected via NVLink and / or NVSwitch SoCs, and the GPUs 584 and PCIe switches 582 are connected via PCIe interconnects. Although eight GPUs 584, two CPUs 580, and two PCIe switches are illustrated, this is not intended to be limiting. Depending on the embodiment, each of the servers 578 may include any number of GPUs 584, CPUs 580, and / or PCIe switches. For example, each of the servers 578 may include eight, sixteen, thirty-two, and / or more GPUs 584.

[0155] Server 578 may receive image data over network 590 and from a vehicle, the image data representing images showing unexpected or changed road conditions, such as recently begun road work. Server 578 may transmit neural network 592, updated neural network 592, and / or map information 594, including information about traffic and road conditions, over network 590 and to the vehicle. Updates to map information 594 may include updates to HD map 522, such as information about construction sites, potholes, curves, flooding, or other obstacles. In some examples, neural network 592, updated neural network 592, and / or map information 594 may have been generated from new training and / or data received from any number of vehicles in the environment and / or based on experience from training performed at a data center (e.g., using server 578 and / or other servers).

[0156] Server 578 can be used to train a machine learning model (e.g., a neural network) based on training data. The training data can be generated by the vehicle, and / or can be generated in a simulation (e.g., using a game engine). In some examples, the training data is labeled (e.g., when 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., when the neural network does not require supervised learning). Training can be performed according to any one or more classes of machine learning techniques, including but not limited to the following classes: supervised training, semi-supervised training, unsupervised training, self-learning, reinforcement learning, joint learning, transfer learning, feature learning (including principal component and cluster analysis), multilinear subspace learning, manifold learning, representation learning (including alternate dictionary learning), rule-based machine learning, anomaly detection, and any variant or combination thereof. 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 590), and / or the machine learning model can be used by server 578 to remotely monitor the vehicle.

[0157] In some examples, server 578 can receive data from the vehicle and apply the data to the latest real-time neural network for real-time intelligent reasoning. Server 578 may include a deep learning supercomputer and / or a dedicated AI computer powered by GPU 584, such as DGX and DGX Station machines developed by NVIDIA. However, in some examples, server 578 may include a deep learning infrastructure of a data center powered only by CPUs.

[0158] The deep learning infrastructure of server 578 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 500. For example, the deep learning infrastructure may receive periodic updates from vehicle 500, such as a sequence of images and / or objects located in the sequence of images that vehicle 500 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 500, and if the results do not match and the infrastructure concludes that the AI ​​in vehicle 500 has failed, then server 578 may transmit a signal to vehicle 500 instructing the fail-safe computer of vehicle 500 to take control, notify passengers, and complete a safe parking maneuver.

[0159] For reasoning, the server 578 may include a GPU 584 and one or more programmable reasoning accelerators (e.g., NVIDIA's TensorRT 3). 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.

[0160] Example computing device

[0161] Figure 6 6 is a block diagram of an example computing device 600 suitable for implementing some embodiments of the present disclosure. The computing device 600 may include a bus 602 that directly or indirectly couples the following devices: a memory 604, one or more central processing units (CPUs) 606, one or more graphics processing units (GPUs) 608, a communication interface 610, an input / output (I / O) port 612, an input / output component 614, a power supply 616, and one or more presentation components 618 (e.g., a display).

[0162] although Figure 6 The various blocks of are shown as being connected via a bus 602 having wires, but this is not intended to be limiting and is merely for clarity. For example, in some embodiments, a presentation component 618 such as a display device may be considered an I / O component 614 (e.g., if the display is a touch screen). As another example, CPU 606 and / or GPU 608 may include memory (e.g., memory 604 may represent a storage device in addition to the memory of GPU 608, CPU 606, and / or other components). In other words, Figure 6The 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 Figure 6 within the range of computing devices.

[0163] The bus 602 may represent one or more buses, such as an address bus, a data bus, a control bus, or a combination thereof. The bus 602 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.

[0164] Memory 604 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 600. 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.

[0165] 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 604 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 disk (DVD) or other optical disk storage, cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by computing device 600. As used herein, computer storage media does not include the signals themselves.

[0166] 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.

[0167] The CPU 606 may be configured to execute computer-readable instructions to control one or more components of the computing device 600 to perform one or more of the methods and / or processes described herein. Each of the CPUs 606 may include one or more cores (e.g., one, two, four, eight, twenty-eight, seventy-two, etc.) capable of simultaneously processing a large number of software threads. The CPU 606 may include any type of processor, and may include different types of processors, depending on the type of computing device 600 implemented (e.g., a processor with fewer cores for mobile devices and a processor with more cores for servers). For example, depending on the type of computing device 600, 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). In addition to one or more microprocessors or supplementary coprocessors such as math coprocessors, the computing device 600 may also include one or more CPUs 606.

[0168] The computing device 600 can use GPU 608 to render graphics (e.g., 3D graphics). GPU 608 may include hundreds or thousands of cores capable of processing hundreds or thousands of software threads simultaneously. GPU 608 may generate pixel data for outputting an image in response to a rendering command (e.g., a rendering command from CPU 606 received via a host interface). GPU 608 may include a graphics memory such as a display memory for storing pixel data. Display memory may be included as part of memory 604. GPU 708 may include two or more GPUs operating in parallel (e.g., via a link). When combined together, each GPU 608 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.

[0169] In examples where computing device 600 does not include GPU 608 , CPU 606 may be used to render graphics.

[0170] The communication interface 610 may include one or more receivers, transmitters, and / or transceivers that enable the computing device 700 to communicate with other computing devices via an electronic communication network, including wired and / or wireless communications. The communication interface 610 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.

[0171] The I / O port 612 can enable the computing device 600 to be logically coupled to other devices including I / O components 614, presentation components 618, and / or other components, some of which can be built into (e.g., integrated into) the computing device 600. Illustrative I / O components 614 include microphones, mice, keyboards, joysticks, game pads, game controllers, satellite dishes, browsers, printers, wireless devices, and the like. The I / O components 614 can provide a natural user interface (NUI) that processes air gestures, voice, or other physiological inputs generated by the 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 the computing device 600 (as described in more detail below). The computing device 600 may include a depth camera such as a stereo camera system, an infrared camera system, an RGB camera system, touch screen technology, and a combination of these for gesture detection and recognition. Additionally, computing device 600 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 600 to render immersive augmented reality or virtual reality.

[0172] The power supply 616 may include a hardwired power supply, a battery power supply, or a combination thereof. The power supply 616 may provide power to the computing device 600 to enable the components of the computing device 600 to operate.

[0173] The presentation component 618 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 618 may receive data from other components (e.g., GPU 608, CPU 606, etc.) and output the data (e.g., as an image, video, sound, etc.).

[0174] 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.

[0175] 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.

[0176] 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 method for training a neural network using ground truth data augmented with map information for autonomous machine applications, comprising: receiving map data corresponding to an area including a location of a dynamic participant at a certain time; positioning the dynamic participant relative to the map data; receiving sensor data generated by a sensor of the dynamic participant at the time; associating the map data with the sensor data based at least in part on the positioning; generating ground truth data corresponding to a predicted output format of a neural network trained with the sensor data using the sensor data and the map data based at least on the associating, generating ground truth data by at least associating one or more first locations of one or more first features from the map with one or more second locations associated with one or more second features within the sensor data; generating, for the sensor data, one or more boundary shapes or one or more line segments indicating the one or more second locations within the sensor data based at least on the associating; generating one or more second labels for the sensor data using the one or more first labels from the map, the second labels classifying the one or more second feature types associated with the one or more second features; and generating the ground truth data to represent the one or more boundary shapes or the one or more line segments and the one or more second labels; as well as One or more parameters of the one or more neural networks are updated based at least on a comparison of the ground truth data with one or more predictions of the one or more neural networks corresponding to the sensor data.

2. The method of claim 1, wherein generating the ground truth data comprises: transforming the map data to a coordinate system of the dynamic participant; determining one or more features represented by the map data; and generating labels or annotations corresponding to the one or more features within the coordinate system, wherein the ground truth data represents the labels or the annotations.

3. The method of claim 1, further comprising: The area including the location is determined based at least in part on at least one of Global Navigation Satellite System (GNSS) data, Global Positioning System (GPS) data, or Differential GPS (DGPS) data generated using one or more location-based sensors of the dynamic participant.

4. The method of claim 1 , wherein updating one or more parameters of the one or more neural networks comprises updating one or more parameters of the one or more neural networks such that the one or more neural networks perform inference for at least one of: object detection, feature detection, wait condition detection or classification, or future trajectory generation.

5. The method of claim 1, further comprising: Data representing a trajectory of the dynamic participant through at least a portion of a field of view represented by the sensor data is received, wherein generating the ground truth data includes adjusting the trajectory based at least in part on the map data.

6. The method of claim 5, wherein adjusting the trajectory comprises: A track that moves the trajectory toward the dynamic participant's lane of travel.

7. The method of claim 1, wherein the sensor data comprises at least one of image data, LIDAR data, RADAR data, SONAR data, or ultrasonic data.

8. The method of claim 1, wherein generating the ground truth data comprises: The map data is transformed from a first coordinate system of the map to a second coordinate system of the sensor data, wherein the ground truth data corresponds to the second coordinate system. 9 . The method of claim 8 , wherein the first coordinate system is a three-dimensional (3D) world space coordinate system and the second coordinate system is a two-dimensional (2D) image space coordinate system.

10. A method for neural network training using ground truth data augmented with map information for autonomous machine applications, comprising: receiving map data representing a high definition (HD) map; receiving first sensor data generated by at least one first sensor of a dynamic participant; positioning the dynamic participant relative to the high definition map based at least in part on the first sensor data; receiving second sensor data generated by a second sensor of the dynamic participant; associating the map data with the second sensor data based at least in part on the location; and generating, based at least in part on the association, ground truth data corresponding to a predicted output format of a neural network trained with the second sensor data from the second sensor data and the map data, generating ground truth data by at least associating one or more first locations of one or more first features from the map with one or more second locations associated with one or more second features within the second sensor data; generating, for the second sensor data, one or more boundary shapes or one or more line segments indicating the one or more second locations within the second sensor data based at least on the associating; generating one or more second labels for the second sensor data using the one or more first labels from the map, the second labels classifying the one or more second feature types associated with the one or more second features; and generating the ground truth data to represent the one or more boundary shapes or the one or more line segments and the one or more second labels; as well as Based at least on a comparison of the ground truth data with one or more predictions of the one or more neural networks corresponding to the second sensor data, one or more parameters of the one or more neural networks are updated.

11. The method of claim 10, wherein the associating comprises: The map data is transformed into a coordinate space oriented relative to the dynamic participant.

12. The method of claim 11, wherein the associating further comprises: After the transformation, the map data are converted from world space coordinates to sensor space coordinates corresponding to the second sensor data.

13. The method of claim 10, wherein the first sensor data represents one or more features within a field of view of the one or more first sensors, and locating the dynamic participant comprises: The one or more features are associated with one or more corresponding features represented by the high definition map.

14. The method of claim 10, further comprising: Receiving third sensor data generated by one or more third sensors of the dynamic participant and representing a trajectory of the dynamic participant through at least a portion of the field of view represented by the second sensor data, wherein generating the ground truth data includes: adjusting the trajectory based at least in part on the map data.

15. The method of claim 10, wherein the ground truth data and the output of the neural network are generated in three-dimensional (3D) world space coordinates.

16. A system for neural network training using ground truth data augmented with map information for autonomous machine applications, comprising: a sensor for generating sensor data representing a sensor data representation; A computing device comprising one or more processor devices and one or more memory devices communicatively coupled to the one or more processing devices and having stored thereon programming instructions, the programming instructions, when executed by the processors, resulting in an instance of: a map manager for locating a dynamic participant relative to a high definition (HD) map; a feature determiner for determining one or more features represented by the HD map; a ground truth generator for associating a first position of the one or more features with a second position of the one or more features within the sensor data representation; and generating ground truth data corresponding to a predicted output format of the second position of the one or more features trained by the neural network based at least in part on the association; and a training engine for training a neural network using the sensor data and the ground truth data; generating ground truth data by associating, based at least on the positioning, one or more first locations of one or more features represented by the map with one or more second locations of the one or more features in the sensor data, the one or more features comprising static objects; generating, based at least on the associating, one or more ground truth labels corresponding to the one or more second locations of the one or more features in the sensor data; and updating one or more parameters of the one or more neural networks based on a comparison of the ground truth data with one or more predictions of the one or more neural networks corresponding to the sensor data.

17. The system of claim 16, wherein the one or more features include an object, a road feature, or a waiting condition.

18. The system of claim 16, further comprising: A trajectory generator is configured to generate a trajectory corresponding to a field of view of the sensor represented by the sensor data, wherein generating the ground truth data comprises adjusting the trajectory based at least in part on the one or more features.

19. The system of claim 18, wherein the one or more features include a track of a lane of travel of the dynamic participant.

20. The system of claim 16, wherein the map manager is further used to orient the high definition map relative to the position and orientation of the dynamic participants.

21. The system of claim 16, wherein the one or more features include at least one of a road marking, a traffic signal, or a traffic sign.

Images