Belief Propagation for Range Image Mapping in Autonomous Machine Applications

Abstract

To provide systems and methods for belief propagation for range image mapping in autonomous machine applications.SOLUTION: In various examples, systems and methods are described that generate a scene flow in a 3D space through simplifying 3D LiDAR data to a "2.5D" optical flow space (e.g., x, y, and depth flow). For example, LiDAR range images may be used to generate 2.5D representations of depth flow information between frames of LiDAR data, and two or more range images may be compared to generate depth flow information, and messages may be passed, e.g., using a belief propagation algorithm, to update pixel values in the 2.5D representation. The resulting images may be used to generate 2.5D motion vectors, and the 2.5D motion vectors may be converted to 3D space to generate a 3D scene flow representation of an environment around an autonomous machine.SELECTED DRAWING: Figure 1A

Description

[Technical Field]

[0001] To operate safely and effectively, autonomous and semi-autonomous vehicles need to identify obstacles or objects moving at different distances so that appropriate action can be taken in a timely manner. [Background technology]

[0002] For example, appropriate actions may include swerving, changing lanes, braking, and / or accelerating to avoid or move away from an identified obstacle. Furthermore, fast and accurate determination of obstacles and their movements allows time to determine and implement appropriate actions. Therefore, object detection and tracking must occur in real time or near real time, even when the vehicle is traveling at highway speeds or when cross traffic is present.

[0003] A representation of stationary and / or moving objects in three-dimensional (3D) space is sometimes called a scene flow. Existing scene flow systems are generally designed to be stationary and / or have a limited field of view. Due to the complexity and computational requirements of performing scene flow in 3D space, existing scene flow systems may also be unable to operate in real time or may only be able to operate in real time with limited resolution and / or accuracy. Therefore, due to these limitations, existing autonomous vehicles rely on other methods and data sources for obstacle detection, such as relying on camera-based neural networks, light detection and ranging (LiDAR) point clouds, and / or RADAR data. However, camera-based neural networks are accurate at short distances but are less accurate at longer distances due to the limited amount of pixels representing any given object or obstacle. Furthermore, LiDAR and RADAR sensors generally contain missing data points and can be noisy. These existing methods generally rely on analyzing a single frame of data, which may lack important context that can only be gained by analyzing multiple frames over time. Summary of the Invention [Means for solving the problem]

[0004] Embodiments of the present disclosure relate to belief propagation for range image mapping in autonomous machine applications. Systems and methods are disclosed that provide LiDAR-based scene flow in real time, at least in part by focusing on motion over time instead of single-frame detection. In contrast to previous systems, such as those described above, the current systems and methods generate scene flow in 3D space by simplifying 3D LiDAR data into an extended two-dimensional ("2.5D") depth flow space (e.g., x, y, and depth flows indicating changes in depth values ​​between LiDAR range images), passing messages between pixels (e.g., between nodes in a pixel representation, such as a matrix, grid, or table corresponding to the pixel), e.g., via a belief propagation algorithm, and calculating 3D motion vectors for the pixels by transforming the 2.5D information back into 3D space (e.g., based on known associations between 2.5D image space locations and 3D world space locations). As a result, noisy or missing information from LiDAR data can be estimated using message passing, e.g., in 2D or 2.5D space, to generate a denser representation of the scene, which can then be transformed back to 3D space for use in detecting, identifying, and / or tracking dynamic objects in the environment. Current systems can be configured for use on mobile platforms, such as autonomous machines (e.g., vehicles, robots, etc.), and can include, for example, spinning LiDAR sensors or other depth sensors that provide a wider field of view (e.g., up to 360 degrees) than existing systems. Furthermore, because a substantial portion of the computations are performed on projected images (e.g., LiDAR range images) instead of in 3D space, current systems can also operate in real time to identify moving obstacles.

[0005] Below, the present system and method for belief propagation for range image mapping in autonomous machine applications is described in detail with reference to the accompanying drawings. [Brief explanation of the drawings]

[0006] [Figure 1A] 1 is a data flow diagram illustrating a scene flow estimation system according to some embodiments of the present disclosure. [Figure 1B] 1 is a time flow diagram illustrating a process for scene flow generation according to some embodiments of the present disclosure. [Figure 2] 1A-1C illustrate labels on an exemplary pixel volume showing image displacement and depth flow, according to some embodiments of the present disclosure. [Figure 3] FIG. 10 illustrates an example cost function for message passing for depth flow, according to some embodiments of the present disclosure. [Figure 4] FIG. 1 illustrates message passing between a pixel node and neighboring pixel nodes, according to some embodiments of the present disclosure. [Figure 5] FIG. 1 illustrates two successive LiDAR range images and a synthetic depth flow image, according to some embodiments of the present disclosure. [Figure 6A] FIG. 10 illustrates a synthetic scene flow generated from depth flow images, according to some embodiments of the present disclosure. [Figure 6B] FIG. 10 illustrates a synthetic scene flow generated from depth flow images, according to some embodiments of the present disclosure. [Figure 7] 1 is a flow diagram illustrating a method for scene flow generation using belief propagation, according to some embodiments of the present disclosure. [Figure 8A] 1 is an illustration of an exemplary autonomous vehicle, according to some embodiments of the present disclosure. [Figure 8B] 8B is an illustration of camera positions and fields of view for the example autonomous vehicle of FIG. 8A, according to some embodiments of the present disclosure. [Figure 8C] FIG. 8B is a block diagram of an example system architecture of the example autonomous vehicle of FIG. 8A, in accordance with some embodiments of the present disclosure. [Figure 8D] FIG. 8B is a system diagram of communication between a cloud-based server and the example autonomous vehicle of FIG. 8A, according to some embodiments of the present disclosure. [Figure 9] FIG. 1 is a block diagram of an exemplary computing device suitable for use in implementing some embodiments of the present disclosure. [Figure 10] FIG. 1 is a block diagram of an exemplary data center suitable for use in implementing some embodiments of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0007] Systems and methods are disclosed for belief propagation for range image mapping in autonomous machine applications. The present disclosure may be described, without limitation, with respect to an exemplary autonomous vehicle 800 (alternatively referred to herein as a “vehicle 800” or “ego-vehicle 800,” examples of which are described with respect to FIGS. 8A-8D ). For example, but not by way of limitation, the systems and methods described herein may be used by non-autonomous vehicles, semi-autonomous vehicles (e.g., in one or more adaptive driver assistance systems (ADAS)), manned and unmanned robots or robotics platforms, warehouse vehicles, off-road vehicles, vehicles coupled to one or more trailers, airships, boats, shuttles, emergency response vehicles, motorcycles, electric or mopeds, aircraft, construction vehicles, submarines, drones, and / or other vehicle or autonomous machine types. Additionally, while the present disclosure may be described with respect to generating scene flows for autonomous driving, this is not intended to be limiting, and the systems and methods described herein may be used in augmented reality, virtual reality, mixed reality, robotics, security and surveillance, autonomous or semi-autonomous machine applications, and / or any other technology space in which scene flows may be used.

[0008] Embodiments of the present disclosure relate to scene flow determination using depth sensors, such as LiDAR sensors used on autonomous vehicles, by creating and analyzing range images generated from LiDAR range (or other projection) images. For example, LiDAR sensors may generate 3D LiDAR data representing their field of view or sensory field, and the LiDAR data may represent a standard spherical projection or other projection type. The LiDAR sensors may continuously generate LiDAR data such that multiple sets of LiDAR data are generated, each set representing a single sweep or rotation of the LiDAR sensor. The sets of LiDAR data may be separated by time intervals, for example, corresponding to the time interval required to complete one rotation, and embodiments of the present disclosure may use a first (e.g., current) set of LiDAR data and compare it with a second, previous (e.g., immediately preceding) set of LiDAR data to determine various information from the data, as described herein. In some examples, two consecutive sets of LiDAR data may be analyzed, although this is not limiting and the analyzed sets of LiDAR data may correspond to different configurations of data other than consecutive frames.

[0009] Complex 3D LiDAR data (e.g., representing a 3D point cloud) can be simplified into 2D LiDAR range images (or other projection image types) and encoded with additional information (e.g., depth flow) to generate 2.5D depth flow images (alternatively referred to herein as “depth flow images”) that enable faster analysis. In some embodiments, depth flow images can also be reduced in resolution using pyramidal or multi-scale techniques to help identify optical flow information more quickly, especially when greater motion occurs between frames. A depth flow image can include a set of pixels, each with a two-dimensional (2D) coordinate and a depth flow value (indicating change in depth or distance across frames). In some embodiments, depth flow can be simplified as a set of labels (e.g., when a label-based algorithm, such as belief propagation, is employed) indicating motion toward the sensor, motion away from the sensor, and / or no motion toward or away (e.g., stationary). For example, these depth flow values ​​may be represented as -1 (towards), +1 (away from), and 0 (stationary), respectively. In some embodiments, depth flow may be more granular and may represent fast movement towards the sensor, slow movement towards the sensor, no movement towards the sensor, slow movement away from the sensor, and / or fast movement away from the sensor. This simplified depth flow information may enable the analysis described herein.

[0010] When comparing two, e.g., consecutive, range images, one range image (e.g., a previous image) may be transformed or rectified into the coordinate system of the subsequent or current range image (or vice versa) to compensate for egomotion. By compensating for egomotion, subsequent analysis need only consider the motion of the object relative to the ego vehicle, since the motion of the ego vehicle has been excluded, which may allow alignment of the two range images in the same coordinate system.

[0011] In some embodiments, two or more range images are compared to determine a 3D motion vector associated with each pixel therein. The 3D motion vectors are generated by analysis of the two or more depth images via a message-passing belief propagation algorithm. For example, pixel data may be propagated to neighboring pixels (e.g., pixel nodes represented in a matrix, table, grid, etc.) via the belief propagation algorithm, and the data received at the neighboring pixels (e.g., neighboring nodes of the pixel representation) is analyzed and used to determine updated values ​​for the pixels. Additional iterations are then performed to refine the results.

[0012] In general, belief propagation is a message-passing algorithm for inferring information from graphical models. While belief propagation can be exact in some cases, belief propagation is often used for approximations. In a belief propagation algorithm, two constraints may need to be defined. The first constraint determines a data term, and the second constraint determines a smoothness term. The data term may be used to identify similar or identical pixels across two range images so that the motion of the object represented by the pixel can be determined. The smoothness term may be used to adjust values ​​across neighboring pixels so that identified objects move together, or so that boundaries between objects or surfaces are treated as such. Messages may be sent between pixels of the same depth flow image (e.g., pixel nodes corresponding to pixels) and carry information about both the data term and the smoothness term. In particular, the messages may include depth flow information (to help identify the general in and out directions for identifying the correct pixels between images), 2D pixel flow information (motion of pixels between frames), and / or cost (to help identify how much influence pixels should have on their neighbors).

[0013] The data terms may identify which pixels in two or more range images correspond to the same object (or part of the same object). Data terms may be useful in embodiments because the physical object and / or depth sensor is moving between images. Variables such as depth flow, reflectance, color value, intensity, time of flight (ToF), texture, return behavior, and / or other information (such as information represented by LiDAR data or information represented in LiDAR range images) may be analyzed in the data terms.

[0014] Messages may not be passed between pixels (e.g., pixel nodes) that correspond to dissimilar objects. For example, adjacent pixels that are calculated not to correspond to the same physical object in the physical environment (such as because the detected difference in depth or reflectance between the two pixels exceeds a certain threshold) do not pass messages between each other. This is because these pixels should not influence their neighboring pixels because they correspond to different physical objects. Therefore, the disparity between these neighboring pixels may not be unduly affected by the belief propagation algorithm.

[0015] The smoothness term may identify which values ​​for depth flow are approximately correct for pixels corresponding to the same physical object. Generally, physical objects moving in space move as a single unit. Therefore, incomplete and inaccurate data for some pixels can be adjusted and influenced by neighboring pixels to take this information into account. For example, a cost may be assigned to each pixel based on how reliable the data associated with it is. For pixels with incomplete data (e.g., no return to the LiDAR sensor), a lower cost may be assigned so that the incomplete data is more likely to be influenced by those neighboring pixels by the belief propagation algorithm. This may make it possible to create a dense motion field despite missing data from the original LiDAR scan.

[0016] By passing and analyzing these messages between pixels of the same depth flow image, the accuracy of various data at each pixel can be improved to refine 2D (or 2.5D) motion vectors. Missing or incomplete data for other pixels can also be estimated and refined.

[0017] The 2D (or 2.5D) vectors can then be transformed back into 3D space using the known correspondence between 2D image space and 3D world space positions. Thus, 3D motion vectors can be determined for points in the LiDAR point cloud, representing the difference (or movement) of the point, and therefore the object or obstacle, across frames of LiDAR data. In this way, a message-passing belief propagation algorithm allows this complex analysis to be simplified and completed before returning to 3D space, thereby reducing the system's computational requirements and execution time by avoiding analysis in a 3D coordinate system. Ultimately, the scene flow can be used to detect and track dynamic objects around the machine, showing the movement of various physical objects around the depth sensor and allowing various appropriate actions to be taken by the autonomous or semi-autonomous machine to account for the objects.

[0018] Referring to FIG. 1A, FIG. 1A illustrates an exemplary scene flow generation system 100 according to some embodiments of the present disclosure. It should be understood that this and other configurations described herein are provided by way of example only. Other configurations and elements (e.g., machines, interfaces, functions, sequences, groupings of functions, etc.) may be used in addition to or instead of those illustrated, and some elements may be omitted entirely. Furthermore, many of the elements described herein are functional entities that may be implemented as individual or distributed components or in conjunction with other components, and in any suitable combination and location. Various functions described herein as being performed by entities may be implemented by hardware, firmware, and / or software. For example, various functions may be implemented by a processor executing instructions stored in a memory. In some embodiments, system 100 may include components, features, and / or functionality similar to those of vehicle 800 described herein with respect to FIGS. 8A-8D, the exemplary computing device 900 of FIG. 9, and / or the exemplary data center 1000 of FIG. 10.

[0019] As shown in FIG. 1A , the scene flow system 100 may include one or more depth perception sensors 102 (e.g., LiDAR sensors, RADAR sensors, ultrasonic sensors, etc.). The depth perception sensors 102 may generally include a transmitter and a receiver, may include any suitable field of view or sensory field, such as a wide field of view (e.g., from 180 degrees to 360 degrees), and in embodiments, may move (e.g., rotate) relative to the depth perception sensor 102 to obtain a wider area of ​​view. The LiDAR, for example, signals may reflect from objects in the vicinity of the depth perception sensor 102. The objects may be moving relative to the depth perception sensor 102, and the depth perception sensor 102 may be moving relative to an underlying surface (e.g., a road over which an autonomous vehicle is traveling). A receiver may receive (directly or indirectly) indications of these various reflected signals, which may be stored and / or transmitted as data for later analysis.

[0020] The depth perception sensor 102 may have a sensor controller 104 that may be used to control the operation of and interpret the results of the depth perception sensor 102. For example, the sensor controller 104 or other processor may receive the sensor data 106 and process, analyze, or otherwise perform calculations related to the sensor data 106. In some embodiments, the depth perception sensor 102 and / or the sensor controller 104 may be similar to the LiDAR sensor 864 described with respect to FIGS. 8A-8C or may be another type of depth perception sensor 102.

[0021] The sensor controller 104 may output the sensor data 106 to a computing system 108, such as a computing system executing within the vehicle 800 and / or the example computing device 900 of FIG. 9 . The sensor data 106 may be in any of a variety of forms, such as, but not limited to, a 3D LiDAR point cloud. The sensor data 106 may be analyzed to perform various functions related thereto and may be used with other sensor data 106 (such as from the various sensors shown in FIGS. 8A-8C and described herein). In embodiments in which a LiDAR sensor is used, the sensor data 106 may be referred to as LiDAR data. However, in other embodiments of the present disclosure, the sensor data 106 may be another type of depth data (e.g., from RADAR, ultrasound, etc.).

[0022] The system 100 may include a coordinate converter 110 that may convert the sensor data 106 into different formats, frames of reference (e.g., from 3D to 2D or 2.5D, etc.), and / or types (e.g., from point cloud to projection or range image). As a non-limiting example, the coordinate converter 110 may convert LiDAR data representing a 3D point cloud into 2D range images or other projection image types, and a depth flow generator may analyze the depth information between frames to generate “2.5D” depth flow images using a depth flow image generator 112. In some embodiments, the coordinate converter 110 may convert one or more range images (based on known or tracked motion (e.g., egomotion) of the vehicle 800) into the same coordinate system, and after analysis for optical flow, message passing, etc. in the 2.5D space, the coordinate converter 110 may convert the resulting 2.5D depth flow images into a 3D scene flow representation.

[0023] The system 100 may further include a depth flow generator 116 that may create a depth flow image using one or more sets of sensor data 106. As shown in FIG. 5, a range image may be generated from LiDAR data using one or more LiDAR sensors. A first set of LiDAR data 502 captured at time T1 and a second set of LiDAR data 504 captured at time T2 are shown in FIG. 5. For example, the sets of LiDAR data may represent a spherical projection or other projection type extending from the depth perception sensor 102. The LiDAR sensor may generate LiDAR data continuously such that multiple sets of LiDAR data are generated. Each such set may represent a single sweep, rotation, or other capture of the LiDAR sensor, and the sets of LiDAR data may be separated by a time interval. In some examples, to generate an optical flow image or a depth flow image as described herein, the depth flow image generator 112 may use a second (e.g., current or most recent) set of LiDAR data 504 and compare it to a previous set of LiDAR data 502 to determine various information from the data. In some examples, two consecutive sets of LiDAR data may be analyzed, although this is not limiting and the analyzed sets of LiDAR data may correspond to different configurations other than consecutive frames.

[0024] In some embodiments, the optical flow image or depth flow image may be a simplified "2.5D" rendering of 3D information, where pixels (e.g., each pixel) have coordinates (e.g., (x, y) coordinates indicating, for example, azimuth and altitude, respectively) and one or more associated variables, such as depth and / or depth flow. By reducing the 3D representation to a 2D or 2.5D representation, computational and execution times may be reduced.

[0025] In some examples, depth flow image generator 112 may generate a first range image based at least in part on first LiDAR data generated using one or more LiDAR sensors at time T1. The first range image may be generated such that pixels of the image include depth flow values ​​indicative of changes in depth values ​​across frames, for example. Similarly, depth flow image generator 112 may generate a second range image based at least in part on second LiDAR data generated using one or more LiDAR sensors at time T2, after T1. The first range image and second range image may then be stored, analyzed, and / or compared to each other (and to a third depth flow image, a fourth depth flow image, etc.), as described herein.

[0026] In embodiments of the present disclosure, the range image may be generated using a LiDAR range image generated using data representing one or more LiDAR point clouds. The LiDAR point clouds may correspond to raw data output by a LiDAR sensor (and / or a LiDAR controller), as described herein. The LiDAR data may represent any combination of various variables associated with the transmitted and received signals, such as, but not limited to, reflectance information, texture information, time-of-flight (ToF) information, color information, and / or intensity information.

[0027] Complex 3D LiDAR data (e.g., representing a 3D point cloud) can be simplified into a 2D LiDAR range image (or other projection image type) and encoded with additional information (such as depth flow) to generate a 2.5D depth flow image that allows for rapid analysis. In some embodiments, the depth flow image can also be reduced in resolution using pyramidal or multi-scale techniques to help identify optical flow information more quickly, especially when there is greater motion between frames (such as would occur for a depth perception sensor 102 on an autonomous vehicle operating at highway speeds). A depth flow image can include a set of pixels, each with a set of 2D coordinates and a depth flow. The depth flow, in some embodiments, can be stored as a label on the pixel.

[0028] The simplified depth flow representation can be used in processing for scene flow instead of more complex 3D LiDAR point cloud data. The simplified computations may enable the analysis described herein to be performed in real time, and the results can be back-projected into 3D space so that accurate 3D scene flow information is generated with less computational and processing time.

[0029] System 100 may include a belief propagator 114 that may identify or assign pixels with one or more characteristics, such as shown in Figure 2. For example, belief propagator 114 may pass messages between pixels of a depth flow image (e.g., pixel nodes of a representation of pixels, such as a matrix, graph, table, etc.) to refine and analyze information therein, as shown in Figure 3. In some embodiments, costs may be assigned to pixels, as shown in Figure 4.

[0030] Thus, embodiments of the present disclosure may use a belief propagation algorithm to pass messages between adjacent pixels (e.g., pixel nodes) in a depth flow image. In general, belief propagation may include a message-passing algorithm for inferring information from a graphical model. While belief propagation can be exact in some cases, belief propagation is often used for approximation. For example, the exact information in LiDAR data is simplified into a depth flow image, message passing approximates the values ​​of neighboring pixels, and the approximated values ​​are returned to the coordinate space of the generated LiDAR data. The simplification, message passing, and return may be performed more quickly than a full analysis in full 3D, while still allowing for significant definiteness in obstacle detection, identification, and / or tracking.

[0031] To aid the belief propagation algorithm in passing messages between adjacent pixels (e.g., pixel nodes in a pixel representation), labels may be associated with pixels that contain information that may be relevant to the adjacent pixels. It should be appreciated that "pixel," as used herein, may refer to any subdivision of a depth flow image and may correspond to a node representing a pixel or a subdivision thereof in a representation of the depth flow image (e.g., a matrix, table, graph, etc.). The resolution of individual pixels may be reduced to allow for faster processing. Similarly, groups of pixels may be grouped together, e.g., messages may be passed between groups of adjacent pixels rather than from pixel to pixel in an embodiment.

[0032] Referring to Figure 2, Figure 2 shows exemplary labels that may be associated with pixels. For example, a belief propagation label may correspond to a 2.5D rectangular block. The illustrated labels may represent image displacement and depth flow, and the label volume may correspond to movement in the Z direction (e.g., toward and away from the depth perception sensor 102).

[0033] In some embodiments, depth flow may be simplified into a set of labels indicating movement toward the sensor, movement away from the sensor, and no movement toward or away (e.g., stationary). For example, these depth flow values ​​(Z in FIG. 2 ) may be represented as −1, +1, and 0, respectively. In some non-limiting embodiments, these simplified data flow labels may be known as ternary labels using balanced ternary logic. In some embodiments, depth flow may be a simplified set of labels indicating fast movement toward the sensor, slow movement toward the sensor, no movement relative to the sensor, slow movement away from the sensor, and fast movement away from the sensor. For example, depth flow values ​​may be represented as −2, −1, 0, +1, and +2, respectively (not shown). However, these labels are for illustrative purposes only, and in other embodiments, different label types may be used to represent depth flow or depth change between frames.

[0034] Referring to FIG. 3, FIG. 3 illustrates different label types that can be associated with a pixel. For example, FIG. 3 illustrates an exemplary graph showing the calculation of a cost function for a depth flow data term. The x-axis of the graph in FIG. 3 corresponds to the detected depth difference for that pixel, and the y-axis of the graph in FIG. 3 corresponds to the associated cost. As a first example, if the depth difference is 0.0 and the depth flow value is 0, a cost of 0 is assigned, indicating that the cost is extremely low when the corresponding physical object has not moved and two adjacent pixels are the same distance apart. As a second example, if the depth difference is 0.125 (at the first tick mark on FIG. 3), the lowest cost depth flow label is associated with Z=1, the second lowest cost is associated with Z=0, and the highest cost is associated with Z=-1. The cost is selected as the y-value at the appropriate depth flow value for that pixel and the corresponding depth flow for that pixel. It should be appreciated that the y values ​​shown on the graphs, as well as the corresponding slopes of the plots, are for illustrative purposes only, and that actual cost values ​​and depth label functions may vary based on any number of factors.

[0035] Once assigned a label, the pixels may pass messages among themselves via a belief propagation algorithm, as shown in FIG. 4 ; for example, a node corresponding to a pixel may pass a message to a neighboring node corresponding to an adjacent or nearby pixel. For example, belief propagator 114 may pass one or more messages between pixels of a second set of pixels that at least partially indicate the pixel's respective depth flow value and the pixel's respective cost. In some embodiments, one or more messages are passed using the belief propagation algorithm. Generally, a message at least partially indicating a label described herein may be sent to a neighboring pixel. The received message may then be analyzed; in particular, the image displacement and depth flow with associated cost may be compared to the corresponding values ​​for that pixel. A new message with the refined value may then be sent and analyzed to further refine the label value. After a certain threshold has passed, the iteration of messages and analysis may terminate. In some instances, the threshold may be a certain number of iterations, the passage of a certain time interval, achieving a certain threshold confidence level, a threshold associated with processor and / or memory capacity, and / or another threshold.

[0036] Use of the belief propagation algorithm may satisfy various constraints. For example, the belief propagation algorithm may identify which adjacent pixels correspond to the same physical object (e.g., data term), and the belief propagation algorithm may "smooth" the adjacent values ​​so that adjacent pixels corresponding to the same physical object initially have similar combined values ​​(e.g., smoothness term). Thus, the belief propagation algorithm may determine which adjacent pixels correspond to the same physical object through the data term, and may determine an approximate value for the pixel associated with that physical object through the smoothness term. Multiple iterations of message passing and analysis over the same optical flow or depth flow image may determine the physical objects and an approximate value for their motion. Multiple iterations over multiple consecutive depth flow images may identify new physical objects (such as physical objects that were previously occluded or that enter the range of the depth perception sensor 102), and changes (such as changes in direction or velocity) of previously identified physical objects may be determined and tracked.

[0037] In some embodiments, messages may be sent between pixels (e.g., nodes corresponding to pixels) of the same depth flow image and may carry information about a data term and a smoothness term. In some embodiments, the data term may be determined during a first set of iterations, and the smoothness term may be determined during a subsequent iteration of the same depth flow image. In other embodiments, the data term and the smoothness term may be determined simultaneously. In particular, messages may include depth flow information (to help identify general in- and out-directions for identifying the correct pixel between images) and / or costs (to help identify how much influence pixels should have on their neighbors).

[0038] An example of message passing is shown in FIG. 4. An exemplary target pixel "X" is shown in the center of FIG. 4. The target pixel is adjacent to neighboring pixels in the horizontal and vertical directions. In some embodiments, the target pixel may also be adjacent to neighboring pixels in the horizontal direction or other designations (not shown). In FIG. 4, neighboring pixels in the vertical direction are labeled v+ (vertically above the target pixel) and v− (vertically below the target pixel). Neighboring pixels in the horizontal direction are labeled h+ (horizontally to the right of the target pixel) and h− (horizontally to the left of the target pixel). It should be appreciated that any pixel in a depth flow image may be designated as a target pixel with corresponding neighboring pixels. For example, pixel h+ in FIG. 4 may be designated as a target pixel that would have a horizontal neighbor as pixel "x," as well as another pixel further to the right (not shown in FIG. 4). Furthermore, although referred to as pixels, message passing may be between nodes or other representations of pixels in a graph, matrix, table, or other representation of an image.

[0039] Messages may be passed between respective pixels. For example, a pixel of interest may send messages to its neighboring pixels, and the messages may each include at least one label, or a portion thereof. In an embodiment, the messages may correspond to simple packets of information about one or more labels associated with the respective pixel. As described herein, by sharing information with neighboring pixels, various aspects of the data may be determined.

[0040] An outgoing message may be labeled with a lowercase lambda and a subscript of the pixel from which the outgoing message originates, as in Figure 4, with the recipient of the message added in parentheses (this convention is purely for illustrative purposes). For example, an outgoing message from target pixel x to its vertically upper neighbor is λ x (v+). The target pixel can then be specified as λ x (v+), λ x (v-), λ x (h+), and λ x(h-) may be created and sent. As described herein, in embodiments of the present disclosure, other messages (such as messages to diagonally adjacent pixels) may also be sent. The outgoing messages may provide information about one or more current labels associated with the target pixel. In some instances, the one or more labels associated with the target pixel are all equivalent, so that the outgoing messages from the target pixel (e.g., the node corresponding to the target pixel) to each neighboring pixel (e.g., the node corresponding to the neighboring pixel) may be similar or equivalent. In other embodiments, the content of at least one outgoing message may be selected based on one or more characteristics of the neighboring pixels.

[0041] Incoming messages are labeled in Figure 4 with a lowercase pi and a subscript of the pixel where the message is destined, with the sender of the message added in parentheses (this convention is purely for illustrative purposes). For example, an incoming message to a target pixel x from its horizontal right neighbor is pi x (h+). The target pixel can then be specified as π x (h+), π x (h-), π x (v+), and π x (v-). As described herein, in embodiments of the present disclosure, other messages may also be received (such as messages from diagonally adjacent pixels). The incoming messages provide information about one or more current labels associated with neighboring pixels. In some embodiments, the one or more labels associated with neighboring pixels are independent, and therefore the content of one or more of the outgoing messages is distinct.

[0042] It should also be appreciated that a single message may be specified differently based on the viewpoint of the pixels involved. For example, an outgoing message λ x (v+) is also the incoming message from the point of view of pixel v+. Similarly, the incoming message π x (h+) is also the outgoing message from the point of view of pixel h+.x (h+) may also be similar or equivalent to other messages sent by pixel h+ to its neighboring pixels.

[0043] As described herein, the data term identifies which adjacent pixels correspond to the same physical object, and the smoothness term adjusts adjacent values ​​so that adjacent pixels corresponding to the same physical object initially have similar composite values. In an embodiment, belief propagator 114 may calculate one or more data terms indicating one or more of the first set of pixels in the first range image that correspond to one or more of the second set of pixels in the second range image based on one or more messages. Data terms may be used because the physical object and / or depth sensor is moving between images. Variables such as depth flow, reflectivity, color value, intensity, time of flight (ToF), texture, return behavior, and / or other information (such as information represented by LiDAR data or information represented in a LiDAR range image) may be analyzed in the data terms. In an example, a physical object may be relatively constant with respect to one or more of these variables. For example, because the physical object moves as a single unit, the physical object may move with a similar depth flow. As another example, reflectance may vary based on the physical object (eg, a painted vehicle may have a higher reflectance value than a pedestrian).

[0044] In an embodiment, the belief propagator 114 calculates one or more smoothness terms that indicate revised depth flow values ​​for a set of pixels in the depth flow image based at least in part on the one or more messages. The smoothness terms may identify which values ​​for depth flow are approximately correct for pixels that correspond to the same physical object. Generally, physical objects moving in space move as a single unit. Therefore, incomplete or inaccurate data for some pixels can be adjusted and influenced by neighboring pixels to take this information into account. For example, a cost is assigned to each pixel based on how reliable the data in each pixel is. For pixels with incomplete data (e.g., no return to the LiDAR sensor), a lower cost may be assigned so that the incomplete data is more likely to be influenced by those neighboring pixels by the belief propagation algorithm. This allows for the creation of a dense motion field despite missing data from the original LiDAR scan.

[0045] In some embodiments, belief propagator 114 may prevent the passing of messages across boundaries corresponding to disparate physical objects. Whether any two adjacent pixels correspond to the same object may be calculated by comparing the relative depth and / or depth flow associated with the two respective pixels. Adjacent pixels that are calculated not to correspond to the same physical object in the physical environment may not pass messages between each other (e.g., because a detected difference in depth or reflectance between the two pixels exceeds a certain threshold). (For example, messages may not be passed between nodes corresponding to pixels in a representation of an image, such as a matrix, table, or graph.) This is because these pixels should not affect their neighboring pixels because they correspond to different physical objects. Therefore, disparities between these neighboring pixels may not be affected by the belief propagation algorithm.

[0046] In some embodiments, belief propagator 114 may determine, based at least in part on depth data from the second LiDAR data, that the target pixel corresponds to a target object that is distinct from neighboring objects corresponding to neighboring pixels. Based on this determination, passing of messages between the target pixel and neighboring pixels may be stopped or prevented. For example, preventing message passing may be performed by determining (before sending a message) whether the pixels involved correspond to the same physical object. In other embodiments, preventing message passing may be performed by placing a virtual barrier around pixels corresponding to the same object so that the determination does not need to be performed again during subsequent iterations of the process. In yet other embodiments, preventing message passing may be performed by adding a label (such as an object reference number or other indicator) to at least one pixel that indicates the physical object. In this embodiment, if the label does not correspond to the same object to which the target pixel corresponds, the message is passed, but the received message may be ignored.

[0047] The system 100 may include a scene flow generator 116 that may convert the 2.5D optical flow information or depth flow information back to 3D space (e.g., after message passing), e.g., using a coordinate transformer 110, to generate a scene flow. For example, by passing and analyzing these messages between pixels of the same depth flow image (e.g., nodes corresponding to pixels in a representation of the depth flow image), the accuracy of various data at each pixel is improved relative to a refined 2D motion vector. The refined 2D motion vector may indicate optical flow changes and / or depth flow changes for a physical object corresponding to the target pixel. Missing or incomplete data for other pixels may also be estimated and refined based at least in part on the refined 2D motion vector.

[0048] The scene flow generator 116 may calculate one or more motion vectors corresponding to the second set of pixels based at least in part on the one or more data terms and the one or more smoothness terms, and the one or more motion vectors may represent relative pixel positions between the first range image and the second range image. The relative pixel positions may indicate movement of physical objects in the environment near the depth perception sensor 102.

[0049] In embodiments of the present disclosure, the scene flow generator 116 may calculate or compute one or more motion vectors in 2D or 2.5D space and then transform the one or more motion vectors into 3D space to generate one or more 3D motion vectors. In embodiments, the processor transforms the determined 2D motion vectors into 3D space based at least in part on information in the 2D vectors and the corresponding LiDAR data for the 2D vectors. The 2D vectors, sometimes referred to as 2.5D motion vectors because they include depth information, can be transformed back into 3D space using known correspondences between 2D image space and 3D world space positions (e.g., using intrinsic and / or extrinsic sensor parameters). Thus, 3D motion vectors can be calculated for points in a LiDAR point cloud, where the 3D motion vectors represent various information about the displacement of 3D points of the point cloud, for example, indicating the displacement of physical objects across frames of LiDAR data. In this way, a message-passing belief propagation algorithm allows this complex analysis to be simplified and completed, then converted back to 3D space to generate 3D scene flow information.

[0050] Thus, the scene flow generator 116 may generate a scene flow based at least in part on one or more 3D motion vectors representing the scene flow between the first and second range images. The scene flow may be used to detect, identify, and / or track various physical objects within the environment of the vehicle 800, for example, within the field of view or sensory field of the depth perception sensor 102.

[0051] 6A-6B, an example scene flow 602 is presented along with a corresponding distance image 604. For example, FIG. 6A shows the example scene flow 602 from a first orientation, and FIG. 6B shows the example scene flow 602 from FIG. 6A from a second orientation (e.g., zoomed in). The example scene flow 602 shows a detected object 606 and an indication of the corresponding movement of the detected object. For example, the scene flow 602 is sparse and includes a central region 608 that corresponds to the location of the depth perception sensor 102 (not shown in FIG. 6A or 6B) and a radius extending therefrom. At the edge of the central region 608 are a series of concentric circles 610 that may indicate that information is available for this area, but that no obstacles are detected. If an obstacle 606 is detected, there may be an obstacle in the concentric circles 610, and there may be an occluded region 612 (indicated by a blank space) behind the obstacle 606 (from the perspective of the depth perception sensor 102). During additional iterations of the process, the occluded object may become visible as the obstacle and / or the perception sensor move relative to one another, thereby removing the occlusion.

[0052] 1A , system 100 may include a vehicle controller 118 that may analyze information in the scene flow to determine obstacles in the physical environment surrounding depth perception sensor 102. Vehicle controller 118 may then command one or more vehicle actions based on the determined obstacles, such as applying the brakes or turning the vehicle. For example, vehicle controller 118 may execute an autonomous driving software stack that may include a perception layer, a world model management layer, a planning layer, a control layer, an actuation layer, an obstacle avoidance layer, and / or one or more other layers.

[0053] 1B , which illustrates a time flow diagram illustrating an exemplary embodiment of system 100 over a time interval. For example, as shown in FIG. 1B , during operation of system 100, at time T1, LiDAR data 150 is collected. The LiDAR data 150 is aggregated into a 3D LiDAR point cloud 152, and the 3D LiDAR point cloud is projected to generate a range image. An exemplary range image is shown in FIG. 5 . At time T2 (after time T1), LiDAR data 150 is again collected and aggregated into a second 3D LiDAR point cloud 152, the LiDAR point cloud is projected to generate a range image, and one or more range images are used to generate a second 2.5D depth flow image 154. In this 2.5D depth flow image 154, various pixels (e.g., nodes corresponding to pixels) may pass messages between them via a message passing algorithm 156. When the message is passed during the first iteration, the depth flow values ​​(and possibly other labels and values) may be refined and compared to the range image to produce refined values ​​158. After a number of iterations, a 2.5D depth flow image 160 and a 3D scene flow 162 may be produced. To further refine the 3D scene flow over time, LiDAR data collected at time T3 (not shown) is further processed and compared to data from time T2.

[0054] Referring now to FIG. 7, each block of method 700 described herein includes computational processes that may be implemented using any combination of hardware, firmware, and / or software. For example, various functions may be implemented by a processor executing instructions stored in memory. Method 700 may also be implemented as computer-usable instructions stored on a computer storage medium. Method 700 may be provided by a standalone 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. Furthermore, by way of illustration, method 700 is described with respect to scene flow generation system 100 of FIG. 1A. However, this method may additionally or alternatively be performed by any one system or any combination of systems, including, but not limited to, those described herein.

[0055] 7 is a flow diagram illustrating a method 700 for generating a scene flow according to some embodiments of the present disclosure. The method 700 includes, at block B702, generating at least a first range image and a second range image. For example, a first range image corresponding to a first set of LiDAR data generated at a first time and a second range image corresponding to a second set of LiDAR data generated at a second time after the first time may be generated. The range images may encode various information, such as depth, intensity, Time of Flight, etc.

[0056] At block B704, the method 700 includes generating a depth flow image using the first range image and the second range image. For example, at least the depth values ​​corresponding to the first range image and the second range image may be compared to determine a depth flow value indicating a change in depth for a particular (e.g., matched) pixel or point in the range image over time. To compare the two range images, one of the range images may be transformed into the coordinate space of the other range image, and the transformation may be performed using the tracked egomotion of the vehicle 800 between the capture of the first and second LiDAR data used to generate the range image.

[0057] The method 700 includes passing messages between pixels of the depth flow image at block B706. For example, messages may be passed via a belief propagation algorithm between nodes corresponding to pixels of a representation (e.g., a matrix, graph, table, etc.) of the image to update values ​​in the depth flow image and / or to fill in missing information about missing or incomplete data.

[0058] The method 700 includes, at block B708, calculating data terms indicative of which pixels correspond to which physical objects, which may include comparison with the first depth flow image.

[0059] The method 700 includes, at block B710, calculating a smoothness term that adjusts the value of the label based on information received from neighboring pixels.

[0060] Method 700 includes calculating and / or refining a motion vector based at least in part on the data term and the smoothness term at block B712. In some embodiments, method 700 may include multiple iterations of the loop of blocks B706-B712, in which the smoothness term adjusts value and the data term identifies an object.

[0061] After a certain threshold has passed, the method 700 may include generating a scene flow based on the motion vectors at block B714. For example, the scene flow may be generated from 3D motion vectors determined by converting the 2.5D motion vectors into 3D space.

[0062] Method 700 includes, at block B716, identifying an obstacle based at least in part on the scene flow. For example, an obstacle may include a physical object that may directly or indirectly affect autonomous vehicle 800. Thus, an obstacle may be moving, such as another vehicle or pedestrian, or stationary, such as a building or tree. An obstacle may have a position relative to the vehicle, a 3D motion vector (which may include acceleration, rotation, or other motion indication), and a relative size (based on how many pixels correspond to the obstacle). In some embodiments, based at least in part on the information about the obstacle, system 100 may determine a likelihood that the obstacle will affect the vehicle, and the system may determine one or more corrective actions to be taken to avoid the obstacle. For example, vehicle controller 118 may determine that the vehicle should brake, turn, or accelerate to avoid the obstacle. If the obstacle cannot be avoided, the system 100 may determine that one or more corrective actions should be taken, such as minimizing damage or activating other safety features.

[0063] The method 700 includes controlling the vehicle to avoid the identified obstacle by taking the identified one or more corrective actions at block 718. Controlling the vehicle may include sending commands to any of a number of vehicle systems, such as the vehicle systems described with respect to Figures 8A-8D.

[0064] (Exemplary Autonomous Vehicle) 8A is a diagram of an example autonomous vehicle 800 according to some embodiments of the present disclosure. Autonomous vehicle 800 (alternatively referred to herein as "vehicle 800") may include, but is not limited to, a passenger vehicle, such as a car, a truck, a bus, a first responder vehicle, a shuttle, an electric or moped, a motorcycle, a fire engine, a police vehicle, an ambulance, a boat, a construction vehicle, a submarine, a drone, a vehicle coupled to a trailer, and / or another type of vehicle (e.g., unmanned and / or carrying one or more passengers). Autonomous vehicles are generally described in terms of automation levels as 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, published June 15, 2018; Standard No. J3016-201609, published September 30, 2016; and previous and future versions of this standard). Mobile vehicle 800 may be capable of functions according to one or more of levels 3 through 5 of autonomous driving. For example, vehicle 800 may be capable of conditional automation (Level 3), high automation (Level 4), and / or full automation (Level 5), depending on the embodiment.

[0065] Mobile vehicle 800 may include components such as a vehicle chassis, body, wheels (e.g., 2, 4, 6, 8, 18, etc.), tires, axles, and other components. Mobile vehicle 800 may include a propulsion system 850, such as an internal combustion engine, a hybrid power plant, a fully electric engine, and / or another propulsion system type. Propulsion system 850 may be connected to a drive train of mobile vehicle 800, which may include a transmission, to enable propulsion of mobile vehicle 800. Propulsion system 850 may be controlled in response to receiving a signal from a throttle / accelerator 852.

[0066] A steering system 854, which may include a steering wheel, may be used to steer the vehicle 800 (e.g., along a desired course or route) when the propulsion system 850 is operating (e.g., when the vehicle is moving). The steering system 854 may receive signals from a steering actuator 856. A steering wheel may be optional for fully automated (Level 5) functionality.

[0067] Brake sensor system 846 may be used to operate vehicle brakes in response to receiving signals from brake actuators 848 and / or brake sensors.

[0068] A controller 836, which may include one or more system-on-chip (SoC) 804 (FIG. 8C) and / or a GPU, can provide signals (e.g., representations of commands) to one or more components and / or systems of the vehicle 800. For example, the controller can send signals to operate vehicle brakes via one or more brake actuators 848, to operate a steering system 854 via one or more steering actuators 856, and to operate a propulsion system 850 via one or more throttle / accelerators 852. The controller 836 may include one or more on-board (e.g., integrated) computing devices (e.g., supercomputers) that process sensor signals and output operational commands (e.g., signals representing commands) to enable rhythmic driving and / or assist a driver in operating the vehicle 800. The controllers 836 may include a first controller 836 for autonomous driving functions, a second controller 836 for functional safety functions, a third controller 836 for artificial intelligence functions (e.g., computer vision), a fourth controller 836 for infotainment functions, a fifth controller 836 for redundancy in emergency situations, and / or other controllers. In some instances, a single controller 836 may handle two or more of the foregoing functions, and two or more controllers 836 may handle a single function and / or any combination thereof.

[0069] Controller 836 may provide signals to control one or more components and / or systems of vehicle 800 in response to sensor data (e.g., sensor inputs) received from one or more sensors. Sensor data may be received from, for example, and without limitation, global navigation satellite system sensors 858 (e.g., global positioning system sensors), RADAR sensors 860, ultrasonic sensors 862, LIDAR sensors 864, inertial measurement unit (IMU) sensors 866 (e.g., accelerometers, gyroscopes, magnetic compasses, magnetometers, etc.), microphones 896, stereo cameras 868, wide-view cameras 870 (e.g., fisheye cameras), infrared cameras 872, surround cameras 874 (e.g., 360-degree cameras), long-range and / or medium-range cameras 898, speed sensors 844 (e.g., for measuring the speed of the moving vehicle 800), vibration sensors 842, steering sensors 840, brake sensors (e.g., as part of a brake sensor system 846), and / or other sensor types.

[0070] One or more of the controllers 836 may receive input (e.g., represented by input data) from the instrument cluster 832 of the vehicle 800 and provide output (e.g., represented by output data, display data, etc.) via a human-machine interface (HMI) display 834, an audible annunciator, a loudspeaker, and / or other components of the vehicle 800. The output may include information such as vehicle velocity, speed, time, map data (e.g., HD map 822 of FIG. 8C ), position data (e.g., the location of vehicle 800 on a map, etc.), direction, the locations of other vehicles (e.g., occupancy grid), information about objects and object situations as known by the controller 836, etc. For example, the HMI display 834 may display information regarding the presence of one or more objects (e.g., road signs, warning signs, traffic light changes, etc.) and / or a driving maneuver that the moving vehicle has performed, is performing, or will perform (e.g., changing lanes now, taking exit 34B in 3.22 km (2 miles), etc.).

[0071] The mobile vehicle 800 further includes a network interface 824 that can communicate over one or more networks using one or more wireless antennas 826 and / or a modem. For example, the network interface 824 can be capable of communication over LTE, WCDMA, UMTS, GSM, CDMA2000, etc. The wireless antenna 826 can also enable communication between objects in the environment (e.g., mobile vehicles, mobile devices, etc.) using local area networks such as Bluetooth, Bluetooth LE, Z-Wave, ZigBee, etc., and / or low power wide-area networks (LPWANs) such as LoRaWAN, SigFox, etc.

[0072] 8B is an illustration of camera positions and fields of view of the exemplary autonomous vehicle 800 of FIG. 8A, according to some embodiments of the present disclosure. The cameras and their respective fields of view are one illustrative example and are not intended to be limiting. For example, additional and / or alternative cameras may be included and / or the cameras may be located at different positions on the vehicle 800.

[0073] The camera type may include, but is not limited to, a digital camera adapted for use with components and / or systems of the mobile vehicle 800. The camera may be capable of operating at Automotive Safety Integrity Level (ASIL) B and / or at another ASIL. The camera type may be capable of any image capture rate, such as 60 frames per second (fps), 120 fps, 240 fps, 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 instances, the color filter array may include a red clear clear clear (RCCC) color filter array, a red clear clear blue (RCCB) color filter array, a red blue green clear (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, clear pixel cameras, such as cameras with RCCC, RCCB, and / or RBGC color filter arrays, may be used in an effort to increase light sensitivity.

[0074] In some instances, one or more of the cameras may 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 mono camera may be installed to provide functions including lane departure warning, traffic sign assist, and intelligent headlamp control. One or more of the cameras (e.g., all cameras) may simultaneously record and provide image data (e.g., video).

[0075] One or more of the cameras may be mounted in a mounting part, such as a custom-designed (e.g., 3D printed) part, to filter out stray light and reflections from within the vehicle (e.g., reflections from the dashboard reflected in the windshield mirror) that may interfere with the camera's image data capture ability. Referring to a side mirror mounting part, the side mirror part may be custom 3D printed so that the camera mounting plate fits the shape of the side mirror. In some instances, the camera may be integrated into the side mirror. For side view cameras, the camera may also be integrated into four posts at each corner of the cabin.

[0076] A camera (e.g., a forward-facing camera) with a field of view that includes a portion of the environment in front of the vehicle 800 may be used for surround view to aid in identifying a forward path and obstacles and, with the assistance of one or more controllers 836 and / or control SoCs, to provide information essential for generating an occupancy grid and / or determining a preferred vehicle path. Forward-facing cameras may be used to perform many of the same ADAS functions as LIDAR, including emergency braking, pedestrian detection, and collision avoidance. Forward-facing cameras may also be used for ADAS functions and systems, including other functions such as lane departure warning (LDW), autonomous cruise control (ACC), and / or traffic sign recognition.

[0077] Various cameras can be used in a forward-facing configuration, including, for example, a monocular camera platform including a complementary metal oxide semiconductor (CMOS) color imager. Another example can be a wide-view camera 870 that can be used to understand objects that come into view from the periphery (e.g., pedestrians, crossing traffic, or bicycles). While only one wide-view camera is shown in FIG. 8B, any number of wide-view cameras 870 can be present in the vehicle 800. Additionally, a long-range camera 898 (e.g., a long-view stereo camera pair) can be used for depth-based object detection, particularly for objects for which a neural network has not yet been trained. The long-range camera 898 can also be used for object detection and classification, as well as basic object tracking.

[0078] One or more stereo cameras 868 may also be included in the forward-facing configuration. The stereo camera 868 may include an integrated control unit with an extensible processing unit, which may provide programmable logic (FPGA) and a multi-core microprocessor with a CAN or Ethernet interface integrated on a single chip. Such a unit may be used to generate a 3D map of the vehicle's environment, including distance estimates for all points in the image. An alternative stereo camera 868 may include a compact stereo vision sensor, which may include two camera lenses (one on the left and one on the right) and an image processing chip that can measure the distance from the vehicle to objects of interest and use the generated information (e.g., metadata) to activate autonomous emergency braking and lane departure warning functions. Other types of stereo cameras 868 may be used in addition to or instead of those described herein.

[0079] Cameras having a field of view that includes portions of the environment to the sides of the mobile vehicle 800 (e.g., side-view cameras) may be used for surround view, providing information used to create and update the occupancy grid and generate side-impact collision warnings. For example, surround cameras 874 (e.g., four surround cameras 874 as shown in FIG. 8B ) may be positioned on the mobile vehicle 800. The surround cameras 874 may include wide-view cameras 870, fisheye cameras, 360-degree cameras, and / or the like. For example, four fisheye cameras may be positioned on the front, rear, and sides of the mobile vehicle. In an alternative arrangement, the mobile vehicle may use three surround cameras 874 (e.g., left, right, and rear) and utilize one or more other cameras (e.g., a forward-facing camera) as a fourth surround-view camera.

[0080] A camera having a field of view that includes the portion of the environment behind the moving vehicle 800 (e.g., a rearview camera) may be used for parking assistance, surround view, rear collision warning, and creating and updating an occupancy grid. As described herein, a wide variety of cameras may be used, including, but not limited to, cameras that are also suitable as forward-facing cameras (e.g., long-range and / or medium-range camera 898, stereo camera 868, infrared camera 872, etc.).

[0081] FIG. 8C is a block diagram of an example system architecture for the example autonomous vehicle 800 of FIG. 8A , in accordance with some embodiments of the present disclosure. It should be understood that this and other arrangements described herein are merely illustrative. Other arrangements and elements (e.g., machines, interfaces, functions, sequences, groupings of functions, etc.) may be used in addition to or instead of those shown, and some elements may be omitted altogether. Furthermore, many of the elements described herein are functional entities that may be implemented as separate or distributed components or in conjunction with other components, and in any suitable combination and location. Various functions described herein as being performed by entities may be implemented by hardware, firmware, and / or software. For example, various functions may be implemented by a processor executing instructions stored in a memory.

[0082] Each of the components, features, and systems of the mobile vehicle 800 in FIG. 8C is shown connected via a bus 802. The bus 802 may include a controller area network (CAN) data interface (alternatively referred to as a "CAN bus"). The CAN may be a network within the mobile vehicle 800 used to help control various features and functions of the mobile vehicle 800, such as braking, acceleration, braking, steering, windshield wiper operation, etc. The CAN bus may be configured to have dozens or even hundreds of nodes, each with its own unique identifier (e.g., CAN ID). The CAN bus may be read to determine steering angle, ground speed, engine revolutions per minute (RPM), button position, and / or other vehicle status indicators. The CAN bus may be ASIL B compliant.

[0083] Although the bus 802 is described herein as being a CAN bus, this is not intended to be limiting. For example, FlexRay and / or Ethernet may be used in addition to or as an alternative to a CAN bus. Additionally, although a single line is used to represent the bus 802, this is not intended to be limiting. There may be any number of buses 802, which may include, for example, 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 instances, two or more buses 802 may be used to perform different functions and / or for redundancy. For example, a first bus 802 may be used for collision avoidance functions, and a second bus 802 may be used for actuation control. In any instance, each bus 802 may communicate with any of the components of the vehicle 800, and two or more buses 802 may communicate with the same component. In some instances, each SoC 804, each controller 836, and / or each computer in the vehicle may have access to the same input data (e.g., input from sensors in the vehicle 800) and may be connected to a common bus, such as a CAN bus.

[0084] Mobile vehicle 800 may include one or more controllers 836, such as those described herein with respect to FIG. 8A. Controller 836 may be used for a variety of functions. Controller 836 may be coupled to any of various other components and systems of mobile vehicle 800 and may be used for control of mobile vehicle 800, artificial intelligence of mobile vehicle 800, infotainment for mobile vehicle 800, and / or the like.

[0085] Mobile vehicle 800 may include a system-on-chip (SoC) 804. SoC 804 may include a CPU 806, a GPU 808, a processor 810, a cache 812, an accelerator 814, a data store 816, and / or other components and features not shown. SoC 804 may be used to control mobile vehicle 800 in a variety of platforms and systems. For example, SoC 804 may be coupled in a system (e.g., that of mobile vehicle 800) with an HD map 822 that can obtain map refreshes and / or updates via a network interface 824 from one or more servers (e.g., server 878 of FIG. 8D ).

[0086] The CPU 806 may include a CPU cluster or CPU complex (alternatively referred to as a "CCPLEX"). The CPU 806 may include multiple cores and / or L2 caches. For example, in some embodiments, the CPU 806 may include eight cores in a coherent multiprocessor configuration. In some embodiments, the CPU 806 may include four dual-core clusters, each with its own dedicated L2 cache (e.g., a 2M L2 cache). The CPU 806 (e.g., a CCPLEX) may be configured to support simultaneous cluster operation, allowing any combination of clusters of CPUs 806 to be active at any given time.

[0087] The CPU 806 may implement power management capabilities including one or more of the following features: individual hardware blocks may be automatically clock gated when idle to conserve dynamic power; each core clock may be gated when the core is not actively executing instructions by executing a WFI / WFE instruction; 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. The CPU 806 may further implement an enhanced algorithm for managing power states, where allowable power states and expected wake-up times are specified and hardware / microcode determines the best power state for entering the cores, clusters, and CCPLEX. The processing cores may support simplified power state entry sequences in software with work offloaded to microcode.

[0088] GPU808 may include an integrated GPU (alternatively referred to herein as an "iGPU"). GPU808 may be programmable and efficient for parallel workloads. In some instances, GPU808 may use an enhanced tensor instruction set. GPU808 may include one or more streaming microprocessors, where each streaming microprocessor may include an L1 cache (e.g., an L1 cache with at least 96 KB of storage capacity) and two or more of the streaming microprocessors may share a cache (e.g., an L2 cache with 512 KB of storage capacity). In some embodiments, GPU808 may include at least eight streaming microprocessors. GPU808 may use a compute application programming interface (API). Additionally, GPU808 may use one or more parallel computing platforms and / or programming models (e.g., NVIDIA's CUDA).

[0089] The GPU 808 may be power-optimized for best performance in automotive and embedded use cases. For example, the GPU 808 may be fabricated on FinFET (Fin field-effect transistor) chips. However, this is not intended to be limiting, and the GPU 808 may be fabricated using other semiconductor manufacturing processes. Each streaming microprocessor may incorporate several mixed-precision processing cores partitioned into multiple blocks. For example, and without limitation, 64 PF32 cores and 32 PF64 cores may be partitioned into four processing blocks. In such an example, each processing block may be assigned 16 FP32 cores, 8 FP64 cores, 16 INT32 cores, two mixed-precision NVIDIA tensor cores for deep learning matrix operations, an L0 instruction cache, a warp scheduler, a dispatch unit, and / or a 64KB register file. Additionally, the streaming microprocessor may include independent parallel integer and floating-point data paths to provide efficient execution of workloads with a mix of computational and addressing operations. Streaming microprocessors may include independent thread scheduling capabilities to allow finer-grained synchronization and coordination among concurrent threads. Streaming microprocessors may include a combined L1 data cache and shared memory unit to improve performance while simplifying programming.

[0090] The GPU 808 may, in some instances, include a high bandwidth memory (HBM) and / or 16 GB HBM2 memory subsystem to provide up to 900 GB / s of peak memory bandwidth. In some instances, synchronous graphics random-access memory (SGRAM), such as graphics double data rate type five synchronous random-access memory (GDDR5), may be used in addition to or in place of the HBM memory.

[0091] The GPU 808 may include unified memory technology that includes access counters to enable more accurate movement of memory pages to the processors that access them most frequently, thereby improving the efficiency of storage areas shared between processors. In some instances, address translation service (ATS) support may be used to enable the GPU 808 to directly access the CPU 806 page tables. In such instances, when the GPU 808 memory management unit (MMU) experiences a miss, an address translation request may be sent to the CPU 806. In response, the CPU 806 may consult its page table for a virtual-to-real mapping of addresses and send the translation back to the GPU 808. As such, unified memory technology may enable a single unified virtual address space for both CPU 806 and GPU 808 memory, thereby simplifying GPU 808 programming and porting of applications to the GPU 808.

[0092] Additionally, GPU 808 may include access counters that can record the frequency of GPU 808's accesses to the memory of other processors. The access counters can help ensure that memory pages are moved to the physical memory of the processors that are accessing the pages most frequently.

[0093] The SoC 804 may include any number of caches 812, including those described herein. For example, the cache 812 may include an L3 cache available to both the CPU 806 and the GPU 808 (e.g., connected to both the CPU 806 and the GPU 808). The cache 812 may include a write-back cache that can record line state, such as by using a cache coherence protocol (e.g., MEI, MESI, MSI, etc.). The L3 cache may include 4 MB or more, depending on the implementation, although smaller cache sizes may also be used.

[0094] The SoC 804 may include an arithmetic logic unit (ALU) that may be utilized in performing processing for any of various tasks or operations (e.g., processing DNNs) of the vehicle 800. Additionally, the SoC 804 may include a floating point unit (FPU) (or other math co-processor or math co-processor type) for performing mathematical operations within the system. For example, the SoC 104 may include one or more FPUs integrated as execution units within the CPU 806 and / or GPU 808.

[0095] The SoC 804 may include one or more accelerators 814 (e.g., hardware accelerators, software accelerators, or a combination thereof). For example, the SoC 804 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., 4 MB of SRAM) may enable the hardware acceleration cluster to accelerate neural networks and other operations. The hardware acceleration cluster may be used to complement the GPU 808 and to offload some of the GPU 808's tasks (e.g., to free up more cycles for the GPU 808 to perform other tasks). As an example, the accelerator 814 may be used for target workloads that are sufficiently stable to be suitable for acceleration (e.g., perception, convolutional neural networks (CNNs), etc.). As used herein, the term "CNN" may include all types of CNNs, including region-based or regional convolutional neural networks (RCNNs) and Faster RCNNs (e.g., as used for object detection).

[0096] The accelerator 814 (e.g., a hardware acceleration cluster) may include a deep learning accelerator (DLA). The DLA may include one or more tensor processing units (TPUs), which can be configured to provide an additional 10 trillion operations per second for deep learning applications and inference. The TPU may be an accelerator configured and optimized to perform image processing functions (e.g., CNN, RCNN, etc.). The DLA may also be optimized for a specific set of neural network types and floating-point operations, as well as inference. The DLA design can provide more performance per millimeter than a general-purpose GPU, significantly exceeding the performance of a CPU. The TPU can perform several functions, including, for example, single-instance convolution functions, supporting INT8, INT16, and FP16 data types for both features and weights, and post-processor functions.

[0097] The DLA can quickly and efficiently run neural networks, particularly CNNs, on processed or unprocessed data for any of a variety of functions, including, but not limited to: CNNs for object identification and detection using data from camera sensors, CNNs for distance estimation using data from camera sensors, CNNs for emergency vehicle detection and identification using data from microphones, CNNs for face recognition and moving vehicle owner identification using data from camera sensors, and / or CNNs for security and / or safety related events.

[0098] The DLA can perform any function of the GPU 808, and by using an inference accelerator, for example, a designer can target either the DLA or the GPU 808 for any function. For example, a designer can focus on processing CNNs and floating-point operations on the DLA, and offload other functions to the GPU 808 and / or other accelerators 814.

[0099] The accelerator 814 (e.g., a hardware acceleration cluster) may include a programmable vision accelerator (PVA), which may alternatively be 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, but is not limited to, any number of reduced instruction set computer (RISC) cores, direct memory access (DMA), and / or any number of vector processors.

[0100] The RISC cores may interact with an image sensor (e.g., an image sensor in any of the cameras described herein), an image signal processor, and / or the like. Each RISC core may include any amount of memory. The RISC cores may use any of several protocols, depending on the embodiment. In some instances, the RISC cores may execute a real-time operating system (RTOS). The RISC cores may be implemented using one or more integrated circuit devices, application specific integrated circuits (ASICs), and / or memory devices. For example, the RISC cores may include an instruction cache and / or tightly coupled RAM.

[0101] The DMA may allow components of the PVA to access system memory independent of the CPU 806. The DMA may support any number of features used to provide optimizations to the PVA, including, but not limited to, supporting multi-dimensional addressing and / or circular addressing. In some instances, the DMA may support up to six or more dimensions of addressing, which may include block width, block height, block depth, horizontal block stepping, vertical block stepping, and / or depth stepping.

[0102] A vector processor may be a programmable processor that can be designed to efficiently and flexibly execute computer vision algorithm programming and provide signal processing capabilities. In some instances, a PVA may include a PVA core and two vector processing subsystem partitions. The PVA core may include a processor subsystem, a DMA engine (e.g., two DMA engines), and / or other peripherals. The vector processing subsystem may act as the PVA's primary processing engine and may include a vector processing unit (VPU), an instruction cache, and / or a vector memory (e.g., VMEM). The VPU core may include a digital signal processor, such as a single instruction, multiple data (SIMD), or very long instruction word (VLIW) digital signal processor. The combination of SIMD and VLIW can increase throughput and speed.

[0103] Each vector processor may include an instruction cache and may be coupled to dedicated memory. As a result, in some instances, each vector processor may be configured to execute independently of other vector processors. In other instances, 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 instances, 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 sequential images or portions of an image. In particular, any number of PVAs may be included in a hardware-accelerated cluster, and any number of vector processors may be included in each PVA. Additionally, the PVA may include additional error correcting code (ECC) memory to enhance overall system security.

[0104] The accelerator 814 (e.g., a hardware acceleration cluster) may include a computer vision network-on-chip and SRAM to provide high-bandwidth, low-latency SRAM for the accelerator 814. In some instances, the on-chip memory may include, for example, and without limitation, at least 4 MB of SRAM consisting of eight field-configurable memory blocks that may be accessible by both the PVA and 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 can access the memory through a backbone that provides the PVA and DLA with high-speed access to the memory. The backbone may include a computer vision network-on-chip that interconnects the PVA and DLA to the memory (e.g., using the APB).

[0105] The computer vision network-on-chip may include an interface that determines, prior to the transmission of any control signals, addresses, or data, that both the PVA and DLA provide ready and valid signals. Such an interface may provide separate phases and separate channels for transmitting control signals, addresses, and data, as well as burst-type communication for continuous data transfer. This type of interface may conform to the ISO 26262 or IEC 61508 standards, although other standards and protocols may also be used.

[0106] In some instances, SoC 804 may include a real-time ray tracing hardware accelerator, such as that described in U.S. Patent Application No. 16 / 101,232, filed August 10, 2018. The real-time ray tracing hardware accelerator may be used to quickly and efficiently determine the location and scale of objects (e.g., within a world model) to generate real-time visualization simulations for RADAR signal interpretation, for acoustic propagation synthesis and / or analysis, for SONAR system simulation, for general wave propagation simulation, for comparison to LIDAR data for localization and / or other functions, and / or other uses. In some embodiments, one or more tree traversal units (TTUs) may be used to perform one or more ray tracing-related operations.

[0107] The accelerator 814 (e.g., a hardware accelerator cluster) has diverse applications for autonomous driving. The PVA may be a programmable vision accelerator that can be used for critical processing stages in ADAS and autonomous vehicles. The capabilities of the PVA make it well suited to algorithmic domains that require predictable processing at low power and low latency. In other words, the PVA works well for semi-dense or dense regular computations on small data sets that require predictable execution times along with low latency and low power. Therefore, because the PVA is efficient at object detection and operating on integer computations, in the context of a platform for autonomous vehicles, the PVA is designed to run classic computer vision algorithms.

[0108] For example, according to one embodiment of the present technology, PVA is used to perform computer stereo vision. A semi-global matching-based algorithm may be used in some instances, but this is not intended to be limiting. Many applications for Level 3-5 autonomous driving require motion estimation / stereo matching on the fly (e.g., structure from motion, pedestrian recognition, lane detection, etc.). PVA can perform computer stereo vision functions with input from two monocular cameras.

[0109] In some instances, PVA may be used to perform dense optical flow by processing raw RADAR data (e.g., using a 4D Fast Fourier Transform) to provide processed RADAR. In other instances, PVA is used in time of flight depth processing, for example, by processing raw time of flight data to provide processed time of flight data.

[0110] DLA can be used to implement 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 the relative "weight" of each detection compared to other detections. This confidence value allows the system to make further decisions regarding which detections should be considered true positives rather than false positives. For example, the system can set a confidence threshold and consider only detections above the threshold as true positives. In an automatic emergency braking (AEB) system, a false positive detection would cause a moving vehicle to automatically apply emergency braking, which is clearly undesirable. Therefore, only the most confident detections should be considered to trigger AEB. DLA can implement a neural network that regresses the confidence value. The neural network may receive as its input at least some subset of parameters, such as bounding box dimensions, a ground plane estimate obtained (e.g., from another subsystem), an inertial measurement unit (IMU) sensor 866 output that correlates with the vehicle 800 orientation, range, and 3D position estimate of the object obtained from the neural network and / or other sensors (e.g., a LIDAR sensor 864 or a RADAR sensor 860), and others.

[0111] The SoC 804 may include a data store 816 (e.g., memory). The data store 816 may be on-chip memory of the SoC 804 and may store neural networks to be executed by the GPU and / or DLA. In some instances, the data store 816 may have a capacity large enough to store multiple instances of the neural network for redundancy and safety. The data store 816 may comprise an L2 or L3 cache 812. References to the data store 816 may include references to memory associated with the GPU, DLA, and / or other accelerators 814, as described herein.

[0112] The SoC 804 may include one or more processors 810 (e.g., embedded processors). The processors 810 may include a boot and power management processor, which may be a dedicated processor and subsystem for handling boot power and management capabilities and related security enforcement. The boot and power management processor may be part of the SoC 804 boot sequence and may provide run-time power management services. The boot power and management processor may provide clock and voltage programming, assist with system low-power state transitions, manage the SoC 804 thermal and temperature sensors, and / or manage the SoC 804 power state. Each temperature sensor may be implemented as a ring oscillator whose output frequency is proportional to temperature, and the SoC 804 may use the ring oscillator to detect the temperature of the CPU 806, GPU 808, and / or accelerator 814. If the temperature is determined to exceed a threshold, the boot and power management processor may enter a temperature fault routine, place the SoC 804 in a lower power state, and / or place the vehicle 800 in a Chauffeur safe stop mode (e.g., bring the vehicle 800 to a safe stop).

[0113] The processor 810 may further include a set of embedded processors that can perform the functions of an audio processing engine. The audio processing engine may be an audio subsystem that allows full hardware support for multi-channel audio through multiple interfaces and a wide and flexible range of audio I / O interfaces. In some instances, the audio processing engine is a dedicated processor core that includes a digital signal processor with dedicated RAM.

[0114] The processor 810 may further include an always-on processor engine that can provide the necessary hardware features to support low-power sensor management and wake use cases. The always-on processor engine may include a processor core, tightly coupled RAM, support peripherals (e.g., timers and interrupt controllers), various I / O controller peripherals, and routing logic.

[0115] The processor 810 may further include a safety cluster engine that includes a processor subsystem dedicated to handling safety management for 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 function as a single core with comparison logic to detect any differences between their operations.

[0116] The processor 810 may further include a real-time camera engine, which may include a dedicated processor subsystem for handling real-time camera management.

[0117] The processor 810 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.

[0118] The processor 810 may include a video image compositor, which may be a processing block (e.g., implemented in a microprocessor) that implements video post-processing functions required by the video playback application to produce the final image for the player window. The video image compositor may perform lens distortion correction on the wide-view camera 870, the surround camera 874, and / or the in-cabin surveillance camera sensor. The in-cabin surveillance camera sensor is preferably monitored by a neural network running on a separate instance of the advanced SoC, configured to identify in-cabin events and respond appropriately. The in-cabin system may perform lip reading to activate cellular service and make phone calls, dictate emails, change the vehicle's destination, activate or change the vehicle's infotainment system and settings, or provide voice-activated web surfing. Certain features are available to the driver only when operating in autonomous mode and are disabled otherwise.

[0119] The video image combiner may include enhanced temporal noise reduction for both spatial and temporal noise reduction. For example, when motion occurs in the video, the noise reduction reduces the weight of information provided by adjacent frames and appropriately weights spatial information. When an image or portion of an image does not contain motion, the temporal noise reduction performed by the video image combiner can use information from previous images to reduce noise in the current image.

[0120] The video image compositor may also be configured to perform stereo rectification on the input stereo lens frames. The video image compositor may further be used for user interface compositing when the operating system desktop is in use, and the GPU 808 is not required to continuously render new surfaces. Even when the GPU 808 is powered on and actively performing 3D rendering, the video image compositor may be used to offload the GPU 808 to improve performance and responsiveness.

[0121] The SoC 804 may further include a mobile industry processor interface (MIPI) camera serial interface, a high-speed interface for receiving video and input from a camera, and / or a video input block that may be used for camera and related pixel input functions. The SoC 804 may further include an input / output controller that may be controlled by software and that may be used to receive I / O signals that are not committed to a specific role.

[0122] The SoC 804 may further include a wide range of peripheral interfaces to enable communication with peripherals, audio codecs, power management, and / or other devices. The SoC 804 may be used to process data from cameras (e.g., connected via gigabit multimedia serial links and Ethernet), sensors (e.g., LIDAR sensors 864, RADAR sensors 860, etc., which may be connected via Ethernet), data from the bus 802 (e.g., vehicle 800 speed, steering wheel position, etc.), and GNSS sensors 858 (e.g., connected via Ethernet or CAN bus). The SoC 804 may further include a dedicated high-performance mass storage controller, which may include its own DMA engine and may be used to offload routine data management tasks from the CPU 806.

[0123] The SoC 804 may be an end-to-end platform with a flexible architecture spanning levels 3-5 of automation, thereby providing a comprehensive functional safety architecture that leverages and efficiently uses computer vision and ADAS techniques for diversity and redundancy, and provides a platform for a flexible, reliable driving software stack along with deep learning tools. The SoC 804 may be faster, more reliable, and more energy- and space-efficient than conventional systems. For example, when the accelerator 814 is combined with the CPU 806, GPU 808, and data store 816, it can provide a fast and efficient platform for levels 3-5 of autonomous vehicles.

[0124] This technology therefore offers capabilities and functionality not achievable by conventional systems. For example, computer vision algorithms can be implemented on a central processing unit (CPU), which can be configured using a high-level programming language, such as the C programming language, to execute a wide variety of processing algorithms across a wide variety of visual data. However, CPUs often cannot meet the performance requirements of many computer vision applications, including those related to execution time and power consumption. Specifically, many CPUs cannot execute complex object detection algorithms in real time, a requirement for in-vehicle ADAS applications and practical Level 3-5 autonomous vehicles.

[0125] In contrast to conventional systems, by providing a CPU complex, a GPU complex, and a hardware acceleration cluster, the technology described herein allows multiple neural networks to run simultaneously and / or serially and the results to be combined to enable Level 3-5 autonomous driving capabilities. For example, a CNN running on the DLA or dGPU (e.g., GPU820) can include text and word recognition, enabling the supercomputer to read and understand traffic signs, including signs for which the neural network was not specifically trained. The DLA can further include a neural network that can identify, interpret, and provide a semantic understanding of the signs and pass the semantic understanding to a route planning module running on the CPU complex.

[0126] As another example, multiple neural networks may be run simultaneously, as required for Level 3, 4, or 5 operation. For example, a warning sign consisting of "Caution: Flashing lights indicate icy conditions" along with a lightning flash may be interpreted independently or collectively by several neural networks. The sign itself may be identified as a traffic sign by a first deployed neural network (e.g., a trained neural network), and the text "Flashing lights indicate icy conditions" may be interpreted by a second deployed neural network that notifies the vehicle's route planning software (preferably running on a CPU complex) that icy conditions exist when the flashing light is detected. The flashing light may be identified by running a third deployed neural network over multiple frames, informing the vehicle's route planning software of the presence (or absence) of the flashing light. All three neural networks may run simultaneously, such as within the DLA and / or on the GPU 808.

[0127] In some instances, a CNN for facial recognition and vehicle owner identification can use data from the camera sensor to identify the presence of a legitimate driver and / or owner of the vehicle 800. An always-on sensor processing engine can be used to unlock the vehicle and turn on the lights when the owner approaches the driver's side door, and in security mode, to disable operation of the vehicle when the owner leaves the vehicle. In this manner, the SoC 804 provides security against theft and / or vehicle hijacking.

[0128] In another example, a CNN for emergency vehicle detection and identification can detect and identify emergency vehicle sirens using data from microphone 896. In contrast to conventional systems that use general classifiers to detect sirens and manually extract features, SoC 804 uses CNNs for environmental and urban sound classification, as well as visual data classification. In a preferred embodiment, the CNN running on the DLA is trained to identify the relative terminal velocity 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 mobile vehicle is operating, as identified by GNSS sensor 858. Thus, for example, when operating in Europe, the CNN would attempt to detect European sirens, and when in the United States, the CNN would attempt to identify only North American sirens. After an emergency vehicle is detected, a control program can be used to perform emergency vehicle safety routines, such as slowing down the mobile vehicle, stopping it at the side of the road, parking it, and / or idling it, with the assistance of ultrasonic sensor 862, until the emergency vehicle has passed.

[0129] The vehicle may include a CPU 818 (e.g., a discrete CPU or dCPU) that may be coupled to the SoC 804 via a high-speed interconnect (e.g., PCIe). The CPU 818 may include, for example, an X86 processor. The CPU 818 may be used to perform any of a variety of functions, including, for example, reconciling potentially inconsistent results between the ADAS sensors and the SoC 804 and / or monitoring the status and health of the controller 836 and / or the infotainment SoC 830.

[0130] Vehicle 800 may include a GPU 820 (e.g., a discrete GPU or dGPU) that may be coupled to SoC 804 via a high-speed interconnect (e.g., NVIDIA's NVLINK). GPU 820 may provide additional artificial intelligence functionality, such as by running redundant and / or different neural networks, and may be used to train and / or update neural networks based on input (e.g., sensor data) from sensors in vehicle 800.

[0131] The mobile vehicle 800 may further include a network interface 824, which may include one or more wireless antennas 826 (e.g., one or more wireless antennas for different communication protocols, such as a cellular antenna, a Bluetooth antenna, etc.). The network interface 824 may be used to enable wireless connections with the cloud via the Internet (e.g., with the server 878 and / or other network devices), with other mobile vehicles, and / or with computing devices (e.g., passenger client devices). To communicate with other mobile vehicles, a direct link may be established between the two mobile vehicles and / or an indirect link may be established (e.g., through a network and via the Internet). A direct link may be provided using a vehicle-to-vehicle communication link. The vehicle-to-vehicle communication link may provide the mobile vehicle 800 with information about mobile vehicles in its vicinity (e.g., vehicles in front of, beside, and / or behind the mobile vehicle 800). This functionality may be part of a cooperative adaptive cruise control function of the mobile vehicle 800.

[0132] The network interface 824 may include an SoC that provides modulation and demodulation functions and enables the controller 836 to communicate over a wireless network. The network interface 824 may include a radio frequency front end for upconversion from baseband to radio frequency and downconversion from radio frequency to baseband. The frequency conversion may be performed through well-known processes and / or may be performed using a superheterodyne process. In some instances, the radio frequency front end functionality may be provided by a separate chip. The network interface may include wireless functionality for communicating via LTE, WCDMA, UMTS, GSM, CDMA2000, Bluetooth, Bluetooth LE, Wi-Fi, Z-Wave, ZigBee, LoRaWAN, and / or other wireless protocols.

[0133] Mobile vehicle 800 may further include a data store 828, which may include off-chip (e.g., off-SoC 804) storage. Data store 828 may include one or more memory elements, including RAM, SRAM, DRAM, VRAM, flash, hard disk, and / or other components and / or devices capable of storing at least one bit of data.

[0134] The vehicle 800 may further include a GNSS sensor 858. The GNSS sensor 858 (e.g., a GPS, an aided GPS sensor, a differential GPS (DGPS) sensor, etc.) aids in mapping, perception, occupancy grid generation, and / or route planning functions. Any number of GNSS sensors 858 may be used, including, for example, but not limited to, a GPS using a USB connector with an Ethernet to serial (RS-232) bridge.

[0135] The mobile vehicle 800 may further include a RADAR sensor 860. The RADAR sensor 860 may be used by the mobile vehicle 800 for long-range mobile vehicle detection, even in darkness and / or severe weather conditions. The RADAR functional safety level may be ASIL B. In some instances, the RADAR sensor 860 may use the CAN and / or bus 802 for control and to access object tracking data (e.g., to transmit data generated by the RADAR sensor 860), with access to Ethernet for accessing raw data. A wide variety of RADAR sensor types may be used. For example, and without limitation, the RADAR sensor 860 may be suitable for front, rear, and side RADAR use. In some instances, a pulse-Doppler RADAR sensor is used.

[0136] The RADAR sensor 860 may include different configurations, such as long-range with a narrow field of view, short-range with a wide field of view, and short-range side coverage. In some instances, long-range RADAR may be used for adaptive cruise control functions. Long-range RADAR systems may provide a wide field of view achieved by two or more independent scans, such as within a 250-meter range. The RADAR sensor 860 may help distinguish between static and moving objects and may be used by ADAS systems for emergency brake assist and forward collision warning. Long-range RADAR sensors may include monostatic multimodal RADAR with multiple (e.g., six or more) fixed RADAR antennas and high-speed CAN and FlexRay interfaces. In one example with six antennas, the center four antennas may create a focused beam pattern designed to record the surroundings of the moving vehicle 800 at high speeds with minimal interference from traffic in adjacent lanes. The other two antennas may widen the field of view, allowing for rapid detection of moving vehicles entering or leaving the moving vehicle's lane.

[0137] As an example, a medium-range RADAR system may include a range of up to 860 meters (front) or 80 meters (rear) and a field of view of up to 42 degrees (front) or 850 degrees (rear). A short-range RADAR system may include, but is not limited to, a RADAR sensor designed to be mounted on either end of a rear bumper. When mounted on either end of a rear bumper, such a RADAR sensor system can create two beams that constantly monitor the blind spots behind and adjacent to a moving vehicle.

[0138] Short-range RADAR systems may be used in ADAS systems for blind spot detection and / or lane change assist.

[0139] The mobile vehicle 800 may further include ultrasonic sensors 862. The ultrasonic sensors 862, which may be positioned on the front, rear, and / or sides of the mobile vehicle 800, may be used for parking assistance and / or for creating and updating an occupancy grid. A variety of ultrasonic sensors 862 may be used, and different ultrasonic sensors 862 may be used for different ranges of detection (e.g., 2.5 m, 4 m). The ultrasonic sensors 862 may operate at an ASIL B functional safety level.

[0140] The mobile vehicle 800 may include a LIDAR sensor 864. The LIDAR sensor 864 may be used for object and pedestrian detection, emergency braking, collision avoidance, and / or other functions. The LIDAR sensor 864 may be functional safety level ASIL B. In some instances, the mobile vehicle 800 may include multiple (e.g., two, four, six, etc.) LIDAR sensors 864 that can use Ethernet (e.g., to provide data to a Gigabit Ethernet switch).

[0141] In some instances, the LIDAR sensor 864 may be capable of providing a list of objects and their distances in a 360-degree field of view. Commercially available LIDAR sensors 864 may have an advertised range of approximately 800 m, with an accuracy of 2 cm to 3 cm, and support for an 800 Mbps Ethernet connection, for example. In some instances, one or more non-protruding LIDAR sensors 864 may be used. In such instances, the LIDAR sensor 864 may be implemented as a small device that may be integrated into the front, rear, sides, and / or corners of the vehicle 800. In such instances, the LIDAR sensor 864 may have a range of 200 m, even for low-reflecting objects, and provide up to a 120-degree horizontal and 35-degree vertical field of view. A front-mounted LIDAR sensor 864 may be configured for a horizontal field of view between 45 and 135 degrees.

[0142] In some instances, LIDAR technology such as 3D flash LIDAR may also be used. 3D flash LIDAR uses a laser flash as a transmitter to illuminate the surroundings of the vehicle up to approximately 200 meters. The flash LIDAR unit includes a receptor that records the laser pulse transit time and the reflected light at each pixel, which in turn corresponds to the range from the vehicle to the object. Flash LIDAR may enable a highly accurate and distortion-free image of the surroundings to be generated with every laser flash. In some instances, four flash LIDAR sensors may be deployed, one on each side of the vehicle 800. Available 3D flash LIDAR systems include solid-state 3D steering array LIDAR cameras (e.g., non-scanning LIDAR devices) with no moving parts other than the blower. Flash LIDAR devices may use 5 nanosecond Class I (eye-safe) laser pulses per frame and may capture reflected laser light in the form of a 3D range point cloud and coregistered intensity data. By using flash LIDAR, and because flash LIDAR is a solid-state device with no moving parts, the LIDAR sensor 864 may be less susceptible to motion blur, vibration, and / or shock.

[0143] The mobile vehicle may further include an IMU sensor 866. In some instances, the IMU sensor 866 may be positioned at the center of the rear axle of the mobile vehicle 800. The IMU sensor 866 may include, for example, but not limited to, an accelerometer, a magnetometer, a gyroscope, a magnetic compass, and / or other sensor types. In some instances, such as in a six-axis application, the IMU sensor 866 may include an accelerometer and a gyroscope, while in a nine-axis application, the IMU sensor 866 may include an accelerometer, a gyroscope, and a magnetometer.

[0144] In some embodiments, the IMU sensor 866 may be implemented as a miniature, high-performance GPS-Aided Inertial Navigation System (GPS / INS) that combines micro-electro-mechanical system (MEMS) inertial sensors, a highly sensitive GPS receiver, and advanced Kalman filtering algorithms to provide estimates of position, velocity, and attitude. As such, in some instances, the IMU sensor 866 may enable the vehicle 800 to estimate heading without requiring input from a magnetic sensor by directly observing and correlating changes in velocity from the GPS to the IMU sensor 866. In some instances, the IMU sensor 866 and the GNSS sensor 858 may be combined in a single integrated unit.

[0145] The mobile vehicle may include a microphone 896 placed in and / or around the mobile vehicle 800. The microphone 896 may be used for emergency vehicle detection and identification, among other things.

[0146] The vehicle may further include any number of camera types, including stereo cameras 868, wide-view cameras 870, infrared cameras 872, surround cameras 874, long-range and / or mid-range cameras 898, and / or other camera types. The cameras may be used to capture image data around the entire exterior of the vehicle 800. The types of cameras used depend on the implementation and requirements of the vehicle 800, and any combination of camera types may be used to achieve the desired coverage around the vehicle 800. Additionally, the number of cameras may vary depending on the implementation. For example, the vehicle may include six cameras, seven cameras, ten cameras, twelve cameras, and / or another number of cameras. The cameras may support, by way of example only, Gigabit Multimedia Serial Link (GMSL) and / or Gigabit Ethernet. Each camera is described in further detail herein with reference to FIGS. 8A and 8B.

[0147] The vehicle 800 may further include a vibration sensor 842. The vibration sensor 842 may measure vibrations of vehicle components, such as an axle. For example, a change in vibration may indicate a change in the road surface. In another example, when two or more vibration sensors 842 are used, the difference in vibration may be used to determine friction or slippage of the road surface (e.g., when the difference in vibration is between a powered axle and a free-spinning axle).

[0148] The mobile vehicle 800 may include an ADAS system 838. In some instances, the ADAS system 838 may include an SoC. The ADAS system 838 may include autonomous / adaptive / automatic cruise control (ACC), cooperative adaptive cruise control (CACC), forward crash warning (FCW), automatic emergency braking (AEB), lane departure warning (LDW), lane keep assist (LKA), blind spot warning (BSW), rear cross-traffic warning (RCTW), collision warning system (CWS), lane centering (LC), and / or other features and functions.

[0149] The ACC system may use a RADAR sensor 860, a LIDAR sensor 864, and / or a camera. The ACC system may include longitudinal ACC and / or lateral ACC. The longitudinal ACC monitors and controls the distance to the vehicle directly ahead of the vehicle 800 and automatically adjusts the vehicle speed to maintain a safe distance from the vehicle ahead. The lateral ACC performs distance maintenance and advises the vehicle 800 to change lanes when necessary. The lateral ACC is related to other ADAS applications such as LCA and CWS.

[0150] CACC uses information from other moving vehicles, which may be received from other moving vehicles via a wireless link via the network interface 824 and / or wireless antenna 826, or indirectly via a network connection (e.g., via the Internet). A direct link may be provided by a vehicle-to-vehicle (V2V) communication link, while an indirect link may be an infrastructure-to-vehicle (I2V) communication link. Generally, V2V communication concepts provide information about the immediately preceding moving vehicle (e.g., the moving vehicle directly ahead of the moving vehicle 800 that is in the same lane as the moving vehicle 800), while I2V communication concepts provide information about traffic further ahead. A CACC system may include either or both I2V and V2V information sources. Given information about moving vehicles ahead of the moving vehicle 800, CACC may be more reliable, potentially allowing for smoother traffic flow and reducing road congestion.

[0151] The FCW system is designed to warn the driver of hazards so that the driver can take corrective action. The FCW system uses a forward-facing camera and / or RADAR sensor 860 coupled to a dedicated processor, DSP, FPGA, and / or ASIC, electrically coupled to driver feedback such as a display, speaker, and / or vibration components. The FCW system can provide warnings in the form of an audio or visual alarm, vibration, and / or a quick brake pulse.

[0152] An AEB system can detect an imminent forward collision with another moving vehicle or other object and automatically apply the brakes if the driver does not take corrective action within specified time or distance parameters. The AEB system can use a forward-facing camera and / or RADAR sensor 860 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; if the driver does not take corrective action, the AEB system can automatically apply the brakes as part of an effort to prevent, or at least mitigate, the effects of the predicted collision. The AEB system may include techniques such as dynamic brake support and / or collision imminent braking.

[0153] The LDW system provides visual, audible, and / or tactile warnings, such as vibration of the steering wheel or seat, to alert the driver when the mobile vehicle 800 crosses a lane marking. The LDW system does not activate when the driver indicates an intentional lane departure by activating a turn signal. The LDW system may use a forward-facing camera coupled to a dedicated processor, DSP, FPGA, and / or ASIC electrically coupled to driver feedback, such as a display, speaker, and / or vibration components.

[0154] The LKA system is a modification of the LDW system, which provides steering input or braking to correct the vehicle 800 if it begins to drift out of its lane.

[0155] The BSW system detects and warns the driver of a moving vehicle in the vehicle's blind spot. The BSW system can provide visual, audible, and / or tactile warnings to indicate that merging or changing lanes is unsafe. The system can provide additional warnings when the driver uses a turn signal. The BSW system can use a rear-facing camera and / or RADAR sensor 860 coupled to a dedicated processor, DSP, FPGA, and / or ASIC, electrically coupled to driver feedback, e.g., a display, speaker, and / or vibration component.

[0156] The RCTW system can provide visual, audible, and / or tactile notifications when an object is detected outside the range of the rear camera when the vehicle 800 is backing up. Some RCTW systems include AEB to ensure vehicle brakes are applied to avoid a collision. The RCTW system can use one or more rear-facing RADAR sensors 860 coupled to a dedicated processor, DSP, FPGA, and / or ASIC, electrically coupled to driver feedback, e.g., a display, speaker, and / or vibration components.

[0157] Because conventional ADAS systems alert the driver and allow the driver to determine whether a safety condition truly exists and act accordingly, conventional ADAS systems can be prone to producing false positives that, while not usually catastrophic, can be annoying and distracting to the driver. However, in an autonomous vehicle 800, when results conflict, the vehicle 800 itself must decide whether to heed results from a primary computer or a secondary computer (e.g., the first controller 836 or the second controller 836). For example, in some embodiments, the ADAS system 838 may be a backup and / or secondary computer that provides perception information to a backup computer rationality module. The backup computer rationality monitor can run redundant software on hardware components to detect failures in perception and dynamic driving tasks. Output from the ADAS system 838 may be provided to a supervisory MCU. When outputs from the primary and secondary computers conflict, the supervisory MCU must determine how to reconcile the conflict to ensure safe operation.

[0158] In some instances, the primary computer may be configured to provide a reliability score to the supervising MCU indicating the reliability of the primary computer in a selected outcome. If the reliability score exceeds a threshold, the supervising MCU may follow the primary computer's instructions regardless of whether the secondary computers provide conflicting or inconsistent results. If the reliability score does not meet the threshold, and the primary and secondary computers provide different (e.g., conflicting) results, the supervising MCU may arbitrate between the computers to determine the appropriate outcome.

[0159] The supervisory MCU may be configured to execute a neural network trained and configured to determine conditions under which the secondary computer will provide a false alarm based on outputs from the primary and secondary computers. Thus, the neural network in the supervisory MCU can learn when the output of the secondary computer can be trusted and when it cannot be trusted. For example, when the secondary computer is a RADAR-based FCW system, the neural network in the supervisory MCU can learn when the FCW identifies a metal object that is not actually dangerous, such as a sewer grate or manhole cover, which triggers an alarm. Similarly, when the secondary computer is a camera-based LDW system, the neural network in the supervisory MCU can learn to ignore the LDW when a bicyclist or pedestrian is present and lane departure is, in fact, the safest maneuver. In embodiments including a neural network running on the supervisory MCU, the supervisory MCU may include at least one of a DLA or a GPU suitable for executing the neural network with associated memory. In a preferred embodiment, the supervising MCU may comprise and / or be included as a component of the SoC 804 .

[0160] In other instances, the ADAS system 838 may include a secondary computer that performs ADAS functions using traditional rules of computer vision. As such, the secondary computer may use classical computer vision rules (if-then), and the presence of a neural network in the supervisory MCU may improve reliability, safety, and performance. For example, diverse implementations and intentional non-identity may make the overall system more fault-tolerant, particularly to failures caused by software (or software-hardware interface) functions. For example, if a software bug or error exists in software running on the primary computer and non-identical software code running on the secondary computer provides the same overall result, the supervisory MCU may have greater confidence that the overall result is correct and that a bug in the software or hardware on the primary computer has not caused a critical error.

[0161] In some instances, the output of the ADAS system 838 can be fed to the perception block of the primary computer and / or the dynamic driving task block of the primary computer. For example, if the ADAS system 838 indicates a forward collision warning due to an object directly ahead, the perception block can use this information when identifying the object. In other instances, the secondary computer can have its own neural network that is trained as described herein, thus reducing the risk of false positives.

[0162] The mobile vehicle 800 may further include an infotainment SoC 830 (e.g., an in-vehicle infotainment system (IVI)). Although shown and described as an SoC, the infotainment system need not be an SoC and may include two or more separate components. The infotainment SoC 830 may include a combination of hardware and software that may be used to provide audio (e.g., music, personal digital assistants, navigation instructions, news, radio, etc.), video (e.g., TV, movies, streaming, etc.), telephony (e.g., hands-free calling), network connectivity (e.g., LTE, Wi-Fi, etc.), and / or information services (e.g., navigation system, reverse parking assist, wireless data system, vehicle-related information such as fuel level, total distance traveled, brake fuel level, oil level, door opening / closing, air filter information, etc.) to the mobile vehicle 800. For example, the infotainment SoC 830 may be a radio, a disc player, a navigation system, a video player, USB and Bluetooth connectivity, a car computer, in-car entertainment, Wi-Fi, steering wheel audio controls, hands-free voice control, a heads-up display (HUD), an HMI display 834, 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 830 may further be used to provide information (e.g., visual and / or audible) to a user of the vehicle, such as information from an ADAS system 838, autonomous driving information such as planned vehicle maneuvers, trajectory, surrounding environment information (e.g., intersection information, vehicle information, road information, etc.), and / or other information.

[0163] The infotainment SoC 830 may include GPU functionality. The infotainment SoC 830 may communicate with other devices, systems, and / or components of the mobile vehicle 800 via the bus 802 (e.g., CAN bus, Ethernet, etc.). In some instances, the infotainment SoC 830 may be coupled to the supervisory MCU so that the infotainment system's GPU can perform some self-driving functions in the event of a failure of the primary controller 836 (e.g., the primary and / or backup computer of the mobile vehicle 800). In such instances, the infotainment SoC 830 may place the mobile vehicle 800 in a Chauffeur safe stop mode, as described herein.

[0164] The mobile vehicle 800 may further include an instrument cluster 832 (e.g., a digital dash, an electronic instrument cluster, a digital instrument panel, etc.). The instrument cluster 832 may include a controller and / or a supercomputer (e.g., a separate controller or supercomputer). The instrument cluster 832 may include a set of instruments such as a speedometer, fuel level, oil pressure, a tachometer, an odometer, turn signals, a gear shift position indicator, a seat belt warning light, a parking brake warning light, an engine malfunction light, airbag (SRS) system information, lighting controls, safety system controls, navigation information, etc. In some instances, information may be displayed and / or shared between the infotainment SoC 830 and the instrument cluster 832. In other words, the instrument cluster 832 may be included as part of the infotainment SoC 830, or vice versa.

[0165] 8D is a system diagram of communication between the cloud-based server and example autonomous vehicle 800 of FIG. 8A in accordance with some embodiments of the present disclosure. System 876 may include server 878, network 890, and vehicle 800. Server 878 may include multiple GPUs 884(A)-884(H) (collectively referred to herein as GPUs 884), PCIe switches 882(A)-882(H) (collectively referred to herein as PCIe switches 882), and / or CPUs 880(A)-880(B) (collectively referred to herein as CPUs 880). GPUs 884, CPUs 880, and PCIe switches may be interconnected with a high-speed interconnect, such as, but not limited to, an NVLink interface 888 developed by NVIDIA and / or a PCIe connection 886. In some instances, the GPUs 884 are connected via an NVLink and / or NVSwitch SoC, and the GPUs 884 and PCIe switch 882 are connected via a PCIe interconnect. While eight GPUs 884, two CPUs 880, and two PCIe switches are illustrated, this is not intended to be limiting. Depending on the embodiment, each server 878 may include any number of GPUs 884, CPUs 880, and / or PCIe switches. For example, the servers 878 may each include 8, 16, 32, and / or more GPUs 884.

[0166] Server 878 can receive image data from the mobile vehicles over network 890, representing images showing unexpected or changed road conditions, such as recently started road construction. Server 878 can transmit neural network 892, updated neural network 892, and / or map information 894, including information about traffic and road conditions, to the mobile vehicles over network 890. Updates to map information 894 can include updates to HD map 822, such as information about construction sites, potholes, detours, flooding, and / or other obstacles. In some instances, neural network 892, updated neural network 892, and / or map information 894 may result from new training and / or experience represented in data received from any number of mobile vehicles in the environment and / or based on training performed at a data center (e.g., using server 878 and / or other servers).

[0167] The server 878 may be used to train a machine learning model (e.g., a neural network) based on training data. The training data may be generated by a mobile vehicle and / or generated in a simulation (e.g., using a game engine). In some instances, the training data is tagged (e.g., if the neural network benefits from supervised learning) and / or undergoes other preprocessing, while in other instances, the training data is not tagged and / or preprocessed (e.g., if the neural network does not require supervised learning). The training may 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, federated learning, transfer learning, feature learning (including principal component and cluster analysis), multilinear subspace learning, manifold learning, representation learning (including preliminary dictionary learning), rule-based machine learning, anomaly detection, and variations or combinations thereof. After the machine-learned model is traced, it may be used by the vehicle (e.g., transmitted to the vehicle via network 890) and / or it may be used by server 878 to remotely monitor the vehicle.

[0168] In some instances, server 878 can receive data from mobile vehicles and apply the data to state-of-the-art real-time neural networks for real-time intelligent inference. Server 878 can include deep learning supercomputers and / or dedicated AI computers powered by GPUs 884, such as the DGX and DGX Station machines developed by NVIDIA. However, in some instances, server 878 can include deep learning infrastructure that uses only CPU-powered data centers.

[0169] The deep learning infrastructure of server 878 may be capable of rapid real-time inference and may use that capability to evaluate and verify the health of the processor, software, and / or associated hardware within mobile vehicle 800. For example, the deep learning infrastructure may receive periodic updates from mobile vehicle 800 (e.g., via computer vision and / or other machine learning object classification techniques), such as a sequence of images and / or objects where mobile vehicle 800 was located within the sequence of images. The deep learning infrastructure may run its own neural network to identify objects and compare them to objects identified by mobile vehicle 800; if the results are inconsistent and the infrastructure concludes that the AI ​​within mobile vehicle 800 is not functioning properly, server 878 may send a signal to mobile vehicle 800 instructing its failsafe computer to take control, notify passengers, and complete a safe parking maneuver.

[0170] For inference, server 878 may include a GPU 884 and one or more programmable inference accelerators (e.g., NVIDIA's TensorRT). The combination of a GPU-powered server and inference acceleration can enable real-time responsiveness. In other instances, such as when less performance is required, servers powered by CPUs, FPGAs, and other processors may be used for inference.

[0171] Exemplary Computing Devices 9 is a block diagram of an example computing device 900 suitable for use in implementing some embodiments of the present disclosure. The computing device 900 may include an interconnection system 902 that indirectly or directly couples the following devices: memory 904, one or more central processing units (CPUs) 906, one or more graphics processing units (GPUs) 909, a communication interface 910, input / output (I / O) ports 912, input / output components 914, a power supply 916, one or more presentation components 918 (e.g., displays), and one or more logic units 920. In at least one embodiment, the computing device 900 may include one or more virtual machines (VMs), and / or any of its components may include virtual components (e.g., virtual hardware components). As non-limiting examples, one or more of GPUs 908 may include one or more vGPUs, one or more of CPUs 906 may include one or more vCPUs, and / or one or more of logical units 920 may include one or more virtual logical units. As such, computing device 900 may include discrete components (e.g., an entire GPU dedicated to computing device 900), virtual components (e.g., a portion of a GPU dedicated to computing device 900), or a combination thereof.

[0172] While the various blocks in FIG. 9 are depicted as connected via interconnection system 902 with lines, this is not intended to be limiting and is merely for clarity. For example, in some embodiments, a presentation component 918, such as a display device, may be considered an I / O component 914 (e.g., if the display is a touch screen). As another example, CPU 906 and / or GPU 908 may include memory (e.g., memory 904 may represent a storage device in addition to the memory of GPU 908, CPU 906, and / or other components). In other words, the computing devices in FIG. 9 are merely exemplary. Categories such as “workstation,” “server,” “laptop,” “desktop,” “tablet,” “client device,” “mobile device,” “handheld device,” “gaming console,” “electronic control unit (ECU),” “virtual reality system,” and / or other device or system types are all intended to be within the scope of the computing devices in FIG. 9 and therefore will not be distinguished from one another.

[0173] Interconnect system 902 may represent one or more links or buses, such as an address bus, a data bus, a control bus, or a combination thereof. Interconnect system 902 may include one or more bus or link 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 or link. In some embodiments, direct connections exist between components. As an example, CPU 906 may be directly connected to memory 904. Further, CPU 906 may be directly connected to GPU 908. When direct or point-to-point connections exist between components, interconnect system 902 may include a PCIe link to implement the connections. In these examples, a PCI bus need not be included in computing device 900.

[0174] Memory 904 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 900. Computer-readable media may include both volatile and nonvolatile media, and removable and non-removable media. By way of example, and not limitation, computer-readable media may comprise computer storage media and communication media.

[0175] Computer storage media may include both volatile and nonvolatile media, and / or removable and non-removable media implemented in any method or technology for storage of information, such as computer-readable instructions, data structures, program modules, and / or other data types. For example, memory 904 may store computer-readable instructions (e.g., representing programs and / or program elements), such as an operating system. Computer storage media may include, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and that can be accessed by computing device 900. As used herein, computer storage media does not include the signals themselves.

[0176] Computer storage media may embody computer-readable instructions, data structures, program modules, and / or other data types in a modulated data signal such as a carrier wave or other transport mechanism and include any information delivery media. The term "modulated data signal" may refer to a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal. By way of example, and not limitation, computer storage media may include wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, RF, infrared and other wireless media. Combinations of any of the above should also be included within the scope of computer-readable media.

[0177] The CPU 906 may be configured to execute at least some of the computer-readable instructions to control one or more components of the computing device 900 to perform one or more of the methods and / or processes described herein. The CPU 906 may include one or more (e.g., 1, 2, 4, 8, 28, 72, etc.) cores, each capable of simultaneously processing multiple software threads. The CPU 906 may include any type of processor, and may include different types of processors depending on the type of computing device 900 implemented (e.g., a processor with fewer cores for a mobile device and a processor with more cores for a server). For example, depending on the type of computing device 900, the processor may be an Advanced RISC Machines (ARM) processor implemented using Reduced Instruction Set Computing (RISC) or an x86 processor implemented using Complex Instruction Set Computing (CISC). The computing device 900 may include one or more CPUs 906 within one or more microprocessors or auxiliary coprocessors, such as computational coprocessors.

[0178] In addition to or instead of CPU 906, GPU 908 may be configured to execute at least some of the computer-readable instructions to control one or more components of computing device 900 to perform one or more of the methods and / or processes described herein. One or more of GPUs 908 may be integrated GPUs (e.g., with one or more of CPUs 906) and / or one or more of GPUs 908 may be discrete GPUs. In an embodiment, one or more of GPUs 908 may be coprocessors of one or more of CPUs 906. GPU 908 may be used by computing device 900 to render graphics (e.g., 3D graphics) or perform general-purpose computing. For example, GPU 908 may be used with GPGPU (General-Purpose Computing on a GPU) The GPU 908 may be used for graphics processing (GPU). The GPU 908 may include hundreds or thousands of cores capable of processing hundreds or thousands of software threads simultaneously. The GPU 908 may generate pixel data for an output image in response to rendering commands (e.g., rendering commands from the CPU 906 received via a host interface). The GPU 908 may include graphics memory, e.g., display memory, for storing pixel data or any other suitable data, e.g., GPGPU data. The display memory may be included as part of the memory 904. GPU 908 may include two or more GPUs operating in parallel (e.g., via a link). The link may connect the GPUs directly (e.g., using NVLINK) or may connect the GPUs via a switch (e.g., using NVSwitch). When coupled together, each GPU 908 may generate pixel data or GPGPU data for a different portion of the output or for a different output (e.g., a first GPU for a first image and a second GPU for a second image). Each GPU may include its own memory or may share memory with other GPUs.

[0179] In addition to or instead of CPU 906 and / or GPU 908, logic unit 920 may be configured to execute at least some of the computer-readable instructions to control one or more of computing devices 900 to perform one or more of the methods and / or processes described herein. In an embodiment, CPU 906, GPU 908, and / or logic unit 920 may discretely or jointly execute any combination of methods, processes, and / or portions thereof. One or more of logic units 920 may be part of and / or integrated with one or more of CPU 906 and / or GPU 908, and / or one or more of logic units 920 may be discrete components to or otherwise external to CPU 906 and / or GPU 908. In an embodiment, one or more of logic units 920 may be a coprocessor of one or more of CPU 906 and / or GPU 908.

[0180] Examples of logic unit 920 include one or more processing cores and / or components thereof, such as a tensor core (TC), a tensor processing unit (TPU), a pixel visual core (PVC), a vision processing unit (VPU), a graphics processing cluster (GPC), a texture processing cluster (TPC), a streaming multiprocessor (SM), a tree traversal unit (TTU), an artificial intelligence accelerator (AIA), a deep learning accelerator (DLA), an arithmetic logic unit (ALU), an application specific integrated circuit (ASIC), a floating point unit (FPU), an input / output (I / O) element, a peripheral component interconnect (PCI) or peripheral component interconnect express (PCIe) element, and / or the like.

[0181] The communications interface 910 may include one or more receivers, transmitters, and / or transceivers that enable the computing device 900 to communicate with other computing devices over electronic communications networks, including wired and / or wireless communications. The communications interface 910 may include components and functionality to enable communication over any of several 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 or InfiniBand), a low-power wide area network (e.g., LoRaWAN, SigFox, etc.), and / or the Internet.

[0182] The I / O ports 912 may enable the computing device 900 to be logically coupled to other devices, including I / O components 914, presentation components 918, and / or other components, some of which may be built into (e.g., integrated with) the computing device 900. Exemplary I / O components 914 include a microphone, mouse, keyboard, joystick, game pad, game controller, satellite dish, scanner, printer, wireless device, etc. The I / O components 914 may provide a natural user interface (NUI) that processes air gestures, voice, or other physiological input generated by the user. In some cases, the input may be sent to an appropriate network element for further processing. The NUI may implement any combination of voice recognition, stylus recognition, facial recognition, biometric recognition, on-screen and adjacent-screen gesture recognition, air gestures, head and eye tracking, and touch recognition in connection with the display of the computing device 900 (as described in more detail below). The computing device 900 may include a depth camera, such as a stereoscopic camera system, an infrared camera system, an RGB camera system, touch screen technology, and combinations thereof, for gesture detection and recognition. Additionally, the computing device 900 may include an accelerometer or gyroscope (e.g., as part of an inertia measurement unit (IMU)) to enable detection of movement. In some instances, the output of the accelerometer or gyroscope may be used by the computing device 900 to render immersive augmented or virtual reality.

[0183] The power supply 916 may include a hardwired power supply, a battery power supply, or a combination thereof. The power supply 916 may provide power to the computing device 900 to enable the components of the computing device 900 to operate.

[0184] The presentation component 918 may include a display (e.g., a monitor, a touch screen, a television screen, a heads-up display (HUD), other display types, or a combination thereof), speakers, and / or other presentation components. The presentation component 918 can receive data from other components (e.g., GPU 908, CPU 906, etc.) and output data (e.g., as images, video, sound, etc.).

[0185] (Example Data Center) 10 illustrates an example data center 1000 that may be used in at least one embodiment of the present disclosure. The data center 1000 may include a data center infrastructure layer 1010, a framework layer 1020, a software layer 1030, and / or an application layer 1040.

[0186] 10 , data center infrastructure layer 1010 may include a resource orchestrator 1012, grouped computational resources 1014, and node computational resources (“node CRs”) 1016(1) through 1016(N), where “N” represents any integer, natural number. In at least one embodiment, node CRs 1016(1) through 1016(N) may include, but are not limited to, any number of central processing units (“CPUs”) or other processors (including accelerators, field programmable gate arrays (FPGAs), graphics processors or graphics processing units (GPUs), etc.), memory devices (e.g., dynamic read-only memory), storage devices (e.g., solid-state or disk drives), network input / output (“NW I / O”) devices, network switches, virtual machines (“VMs”), power modules, and / or cooling modules, etc. In some embodiments, one or more of the nodes CR 1016(1)-1016(N) may correspond to a server having one or more of the aforementioned computational resources. Additionally, in some embodiments, the nodes CR 1016(1)-1016(N) may include one or more virtual components, such as a vGPU, a vCPU, and / or the like, and / or one or more of the nodes CR 1016(1)-1016(N) may correspond to a virtual machine (VM).

[0187] In at least one embodiment, the grouped computing resources 1014 may include distinct groups of nodes CR 1016 housed within one or more racks (not shown), or multiple racks housed in data centers in various geographic locations (also not shown). The distinct groups of nodes CR 1016 within the grouped computing resources 1014 may include grouped computing, network, memory, or storage resources that can be configured or assigned to support one or more workloads. In at least one embodiment, several nodes CR 1016, including CPUs, GPUs, and / or other processors, may be grouped within one or more racks to provide computing resources to support one or more workloads. The one or more racks may also include any number of power modules, cooling modules, and / or network switches, in any combination.

[0188] The resource orchestrator 1012 can configure or otherwise control one or more nodes CR 1016(1)-1016(N) and / or grouped computational resources 1014. In at least one embodiment, the resource orchestrator 1012 can include a software design infrastructure (“SDI”) management entity of the data center 1000. The resource orchestrator 1012 can include hardware, software, or some combination thereof.

[0189] In at least one embodiment, as shown in FIG. 10 , framework layer 1020 may include a job scheduler 1032, a configuration manager 1034, a resource manager 1036, and / or a distributed file system 1038. Framework layer 1020 may include frameworks to support software 1032 in software layer 1030 and / or one or more applications 1042 in application layer 1040. Software 1032 or applications 1042 may include web-based service software or applications, such as those offered by Amazon Web Services, Google Cloud, and Microsoft Azure, respectively. Framework layer 1020 may be, but is not limited to, a type of free and open source software web application framework, such as Apache Spark™ (hereinafter “Spark”), which may use distributed file system 1038 for large-scale data processing (e.g., “big data”). In at least one embodiment, the job scheduler 1032 may include a Spark driver to facilitate scheduling of workloads supported by the various tiers of the data center 1000. The configuration manager 1034 may be capable of configuring different tiers, such as the software layer 1030 and the framework layer 1020, which includes Spark and a distributed file system 1038 to support large-scale data processing. The resource manager 1036 may be capable of managing clustered or grouped computing resources that are mapped or assigned to support the distributed file system 1038 and the job scheduler 1032. In at least one embodiment, the clustered or grouped computing resources may include the computing resources 1014 grouped in the data center infrastructure layer 1010. The resource manager 1036 may coordinate with the resource orchestrator 1012 to manage these mapped or assigned computing resources.

[0190] In at least one embodiment, software 1032 included in software layer 1030 may include software used by at least a portion of nodes CR 1016(1)-1016(N), grouped computational resources 1014, and / or distributed file system 1038 of framework layer 1020. The one or more types of software may include, but are not limited to, internet web page searching software, email virus scanning software, database software, and streaming video content software.

[0191] In at least one embodiment, the applications 1042 included in the application layer 1040 may include one or more types of applications used by at least a portion of the nodes CR 1016(1)-1016(N), the grouped computational resources 1014, and / or the distributed file system 1038 of the framework layer 1020. The one or more types of applications may include, but are not limited to, any number of genomics applications, cognitive computing, and machine learning applications, including training or inference software, machine learning framework software (e.g., PyTorch, TensorFlow, Caffe, etc.), and / or other machine learning applications used in conjunction with one or more embodiments.

[0192] In at least one embodiment, any of configuration manager 1034, resource manager 1036, and resource orchestrator 1012 can implement any number and type of self-modifying actions based on any amount and type of data obtained in any technically possible manner. The self-modifying actions can free data center operators of data center 1000 from making potentially poor configuration decisions and possibly avoiding underutilized and / or underperforming portions of the data center.

[0193] Data center 1000 may include tools, services, software, or other resources for training one or more machine learning models or for predicting or inferring information using one or more machine learning models according to one or more embodiments described herein. For example, a machine learning model may be trained by calculating weight parameters according to a neural network architecture using the software and / or computing resources described above with respect to data center 1000. In at least one embodiment, a trained or deployed machine learning model corresponding to one or more neural networks may be used to infer or predict information using the resources described above with respect to data center 1000, for example, by using weight parameters calculated via one or more training techniques, including but not limited to those described herein.

[0194] In at least one embodiment, data center 1000 may use CPUs, application specific integrated circuits (ASICs), GPUs, FPGAs, and / or other hardware (or corresponding virtual computing resources) for training and / or performing inference using such resources. Additionally, one or more of such software and / or hardware resources may be configured as services, such as image recognition, speech recognition, or other artificial intelligence services, to enable users to train or perform inference on information.

[0195] (Example network environment) A network environment suitable for use in implementing embodiments of the present disclosure may include one or more client devices, servers, network attached storage (NAS), other back-end devices, and / or other device types. The client devices, servers, and / or other device types (e.g., each device) may be implemented with one or more instances of computing device 900 of FIG. 9 , e.g., each device may include similar components, features, and / or functionality of computing device 900. Additionally, if a back-end device (e.g., server, NAS, etc.) is implemented, the back-end device may be included as part of data center 1000, examples of which are further detailed herein with respect to FIG. 10.

[0196] Components of a network environment may communicate with each other via a network, which may be wired, wireless, or both. A network may include multiple networks or a network of networks. Illustratively, a network may include one or more wide area networks (WANs), one or more local area networks (LANs), one or more public networks, such as the Internet and / or the Public Switched Telephone Network (PSTN), and / or one or more private networks. When a network includes a wireless telecommunications network, components such as base stations, communication towers, or access points (as well as other components) may provide wireless connectivity.

[0197] Compatible network environments may include one or more peer-to-peer network environments (wherein a server may not be included in the network environment) and one or more client-server network environments (wherein a server or servers may be included in the network environment). In a peer-to-peer network environment, functionality described herein with respect to a server may be implemented in any number of client devices.

[0198] In at least one embodiment, the network environment may include one or more cloud-based network environments, distributed computing environments, combinations thereof, etc. The cloud-based network environment may include a framework layer, a job scheduler, a resource manager, and a distributed file system implemented on one or more of the servers, which may include one or more core network servers and / or edge servers. The framework layer may include a framework to support software in the software layer and / or one or more applications in the application layer. The software or applications may each include web-based service software or applications. In an embodiment, one or more of the client devices may use the web-based service software or applications (e.g., by accessing the service software and / or applications via one or more application programming interfaces (APIs)). The framework layer may be, but is not limited to, a type of free and open source software web application framework that may use a distributed file system for large-scale data processing (e.g., “big data”).

[0199] A cloud-based network environment may provide cloud computing and / or cloud storage that implements any combination of the computing and / or data storage functions (or one or more portions thereof) described herein. Any of these various functions may be distributed across multiple locations from a central or core server (e.g., one or more data centers that may be distributed across a state, region, country, or the world). When a user (e.g., a client device) is connected relatively close to an edge server, the core server may delegate at least a portion of its functionality to the edge server. A cloud-based network environment may be private (e.g., limited to a single organization), public (e.g., available to multiple organizations), and / or a combination thereof (e.g., a hybrid cloud environment).

[0200] A client device may include at least some of the components, features, and functionality of the exemplary computing device 900 described herein with respect to Figure 9. By way of illustration, and not limitation, a client device may be embodied as a personal computer (PC), a laptop computer, a mobile device, a smartphone, a tablet computer, a smart watch, a wearable computer, a personal digital assistant (PDA), an MP3 player, a virtual reality headset, a global positioning system (GPS) or device, a video player, a video camera, a surveillance device or system, a vehicle, a boat, an airship, a virtual machine, a drone, a robot, a handheld communications device, a hospital device, a gaming device or system, an entertainment system, a vehicle computing system, an embedded system controller, a remote control, an appliance, a consumer electronic device, a workstation, an edge device, any combination of these depicted devices, or any other suitable device.

[0201] The present disclosure may be described in the general context of computer code or machine-usable instructions, including computer-executable instructions, such as program modules, being executed by a computer or other machine, such as a personal digital assistant or other handheld device. Generally, program modules, including routines, programs, objects, components, data structures, etc., refer to code that performs particular tasks or implements particular abstract data types. The present disclosure may be implemented in a variety of configurations, including handheld devices, consumer electronics, general-purpose computers, more specialized computing devices, etc. The present disclosure may also be implemented in distributed computing environments where tasks are performed by remote processing devices linked through a communications network.

[0202] As used herein, the term "and / or" in reference to two or more elements should be interpreted to mean one element only or a combination of elements. For example, "element A, element B, and / or element C" may include element A only, element B only, element C only, elements A and B, elements A and C, elements B and C, or elements A, B, and C. Additionally, "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. Furthermore, "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.

[0203] The subject matter of the present disclosure has been described with specificity to meet statutory requirements. However, that description itself is not intended to limit the scope of the disclosure. Rather, the inventors contemplate that the claimed subject matter may be implemented in other ways, including different steps or combinations of steps similar to those described herein, in conjunction with other current or future technologies. Furthermore, although the terms "step" and / or "block" may be used herein to connote different elements of the method used, these terms should not be construed as implying any particular order among the various steps disclosed herein unless and when the order of individual steps is explicitly described.

Claims

1. generating a depth flow image representing one or more changes in depth values ​​between corresponding pixels of at least two successive range images; passing data corresponding to one or more neighboring pixels of the depth flow image using a belief propagation algorithm to update one or more values ​​associated with the one or more neighboring pixels of the depth flow image; calculating, based at least in part on the one or more updated values, one or more motion vectors in image space corresponding to movement of the one or more adjacent pixels in the depth flow image between the at least two consecutive distance images; generating a scene flow representation by transforming the one or more motion vectors into corresponding one or more motion vectors in world space; One or more circuits for performing A processor comprising:

2. the processor: Control systems for autonomous or semi-autonomous machines; Autonomous or semi-autonomous machine perception systems, a system for performing a simulation operation; a system for performing deep learning operations; a system implemented using edge devices; Systems implemented using robots, a system incorporating one or more Virtual Machines (VMs); a system that is at least partially implemented in a data center; or 10. The processor of claim 1, wherein the processor is included in at least one of: a system implemented at least in part using cloud computing resources.

3. determining one or more positions of one or more objects based at least in part on the scene flow representation; performing one or more actions of an autonomous or semi-autonomous machine based at least in part on the one or more positions of the one or more objects; The processor of claim 1 further comprising processing circuitry for performing:

4. The processor of claim 1 , further comprising processing circuitry for performing one or more actions of an autonomous or semi-autonomous machine based at least in part on the scene flow.

5. The processor of claim 1 , wherein the at least two successive range images are generated using corresponding Light Detection and Ranging (LiDAR) point clouds.

6. 2. The processor of claim 1, wherein when another pair of one or more adjacent pixels is determined not to correspond to the same object, data corresponding to the other pair of one or more adjacent pixels is prevented from being passed.

7. A step in which one or more circuits included in a processor generate a depth flow image based at least in part on a first LiDAR distance image generated using one or more LiDAR sensors at a first time and a second LiDAR distance image generated using the one or more LiDAR sensors at a second time after the first time, wherein the depth flow image includes one or more pixels labeled with one or more depth flow values; passing data corresponding to one or more neighboring pixels of the depth flow image to the one or more circuits to update one or more values ​​associated with the one or more neighboring pixels of the depth flow image to generate an updated depth flow image, the passed data including a depth flow value for each of the one or more neighboring pixels and a cost for each of the one or more neighboring pixels, the cost being calculated using a cost function; the one or more circuits calculating, based at least in part on the updated depth flow image, one or more motion vectors representing relative pixel positions between the first LiDAR range image and the second LiDAR range image; A method comprising:

8. The method of claim 7 , wherein the first LiDAR range image and the second LiDAR range image are generated using data representing one or more LiDAR point clouds.

9. The method of claim 7 , wherein the first LiDAR range image is transformed into the same coordinate system as the second LiDAR range image based at least in part on the calculated egomotion.

10. 8. The method of claim 7, wherein the first LiDAR range image and the second LiDAR range image each represent at least one of reflectivity information, intensity information, depth information, time of flight (ToF) information, return information, or classification information.

11. 8. The method of claim 7, wherein the one or more motion vectors are calculated in two-dimensional (2D) space, the method further comprising the one or more circuits transforming the one or more motion vectors to three-dimensional (3D) space to generate one or more 3D motion vectors.

12. The method of claim 11 , wherein the one or more 3D motion vectors represent a scene flow representation between the first LiDAR range image and the second LiDAR range image.

13. The method of claim 7 , wherein the data corresponding to the one or more neighboring pixels is passed using a belief propagation algorithm.

14. The depth flow image includes a first pixel adjacent to a second pixel, and the method includes: determining, by the one or more circuits based at least in part on depth data from second LiDAR data, that the first pixel corresponds to a second object distinct from the first object corresponding to the second pixel; preventing the one or more circuits from passing data corresponding to the first pixel based at least in part on the determining; 14. The method of claim 13, further comprising:

15. one or more processing units; one or more memory units for storing instructions; wherein the instructions, when executed by the one or more processing units, cause the one or more processing units to: generating a depth flow image representing one or more changes in one or more depth values ​​between corresponding pixels of at least two successive range images; passing data corresponding to the one or more neighboring pixels of the depth flow image using a belief propagation algorithm to update one or more values ​​associated with a pixel of the depth flow image; calculating, based at least in part on the one or more updated values, one or more image-space motion vectors corresponding to movement of the one or more adjacent pixels in the depth flow image between the at least two consecutive distance images; converting the image space motion vectors to world space motion vectors to generate a scene flow representation; A system for causing an operation including

16. The system comprises: Control systems for autonomous or semi-autonomous machines; Autonomous or semi-autonomous machine perception systems, a system for performing a simulation operation; a system for performing deep learning operations; a system implemented using edge devices; Systems implemented using robots, a system incorporating one or more virtual machines (VMs); a system that is at least partially implemented in a data center; or 16. The system of claim 15, wherein the system is included in at least one of: a system implemented at least in part using cloud computing resources.

17. The operation is determining one or more positions of one or more objects based at least in part on the scene flow; performing one or more actions of an autonomous or semi-autonomous machine based at least in part on the one or more positions of the one or more objects; The system of claim 15 further comprising:

18. 16. The system of claim 15, wherein the actions further comprise performing one or more actions of an autonomous or semi-autonomous machine based at least in part on the scene flow representation.

19. The system of claim 15 , wherein the at least two successive range images are generated using corresponding LiDAR point clouds.

20. 16. The system of claim 15, wherein at least one message is prevented from being passed between one or more other pairs of adjacent pixels when the one or more other pairs of adjacent pixels are determined not to correspond to the same object.

Images