Depth-based vehicle environment visualization using generative ai

Abstract

The invention discloses depth-based vehicle environment visualization using generative AI. In various examples, systems and methods related to geometry estimation and dynamic object rendering for vehicle environment visualization are disclosed. The environment around the ego-vehicle may be visualized by extracting one or more depth maps from the image data, converting the depth maps to a 3D surface topology of the surrounding environment, and / or texturing the detected 3D surface topology with the image data. Dynamic objects may be detected and masked from a first pass of texturing. Rigid body dynamic objects may be visualized by warping a respective depth map using a respective trajectory, inserting or fusing the resulting warped 3D representation of each such object into a (textured) 3D surface topology, and texturing the warped 3D representation of each object using a respective image data. The non-rigid dynamic object may be represented as a flat 2D surface and textured with corresponding image data.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application is a continuation-of U.S. Application No. 18 / 670,373, filed May 21, 2024, which claims the benefit of U.S. Provisional Application No. 63 / 566,129, filed March 15, 2024, and U.S. Provisional Application No. 63 / 565,885, filed March 15, 2024. The contents of each of the foregoing applications are incorporated herein by reference in their entirety. Background Art

[0003] Surround view systems provide the vehicle's occupants with a visual representation of the area surrounding the vehicle. Surround view systems provide the driver with the ability to see the surrounding environment (including blind spots where the driver's view is obscured by components of the driver's vehicle or other objects in the environment) without having to reposition their view (e.g., turning their head, leaving the driver's seat, leaning in a certain direction, etc.). This visualization can assist and facilitate various driving maneuvers, such as smoothly entering or exiting a parking space without colliding with vulnerable road users (such as pedestrians) or objects (such as curbs or other vehicles). An increasing number of vehicles, particularly luxury brands or newer models, are being produced equipped with surround view systems. Surround view systems in existing vehicles typically use fisheye cameras (typically mounted on the front, left, rear, and right sides of the vehicle body) to perceive the surrounding area from multiple directions. In some techniques, frames from various cameras are stitched together using camera parameter alignment and overlapping areas are combined using blending techniques to provide a top-down 360° surround view visualization.

[0004] However, existing techniques for generating visualizations of surrounding environments have various drawbacks. For example, in some existing surround view systems, two-dimensional (2D) images are used to approximate a three-dimensional (3D) visual representation of the vehicle's surroundings by modeling the geometry of the vehicle's surroundings as a virtual 3D bowl. The 3D bowl typically includes a flat, circular ground plane for the bowl's inner portion, which is connected to an outer bowl, which is represented as a curved surface rising from the ground plane to a certain height or having a slope that increases proportionally with the distance from the bowl's center. Therefore, some conventional systems project (e.g., stitch) images onto the 3D bowl, render a view of the projected image data onto the 3D bowl from the perspective of a virtual camera, and present the rendered view on a monitor visible to a vehicle occupant or operator (e.g., driver). However, the projection and / or stitching process can introduce various artifacts, including geometric distortion (e.g., size or shape misalignment), texture distortion (e.g., blurring, ghosting, object disappearance, object distortion), and color distortion. Because these artifacts may obscure or omit useful visual information and generally distract the driver, they may interfere with safe vehicle operation in some scenarios. Therefore, improved visualization techniques are needed to reduce visual artifacts, better represent useful visual information, and / or otherwise improve the visual quality of the resulting images. Summary of the Invention

[0005] Embodiments of the present disclosure relate to geometry estimation and dynamic object rendering for vehicular environment visualization. Compared to conventional systems such as those described above, the environment surrounding the ego vehicle can be visualized by texturing the detected 3D surface topology of the surrounding environment, and dynamic objects can be rendered by texturing warped 3D representations of rigid objects and / or flat 2D surfaces representing non-rigid objects.

[0006] In an embodiment, the environment around the self-machine can be visualized by extracting one or more depth maps from the image data, converting the depth maps into a 3D surface topology of the detected surrounding environment, and / or texturing the detected 3D surface topology with the image data. For example, one or more sensors (e.g., cameras) of the self-machine (e.g., a vehicle) can be used to generate sensor data frames (e.g., images) as the self-machine navigates the environment. In an example implementation, multiple frames representing the same time slice generated using different sensors (e.g., cameras) can be grouped together to represent spatial context, multiple frames representing different time slices generated using the same sensor (e.g., camera) can be grouped together to represent temporal context, and the resulting (e.g., stacked) representation of the image data can be applied to a neural network to estimate one or more depth maps (corresponding to a depth map for each image data frame). One or more depth maps can be converted into a 3D representation of the detected surface topology of the surrounding environment. For example, the geometry of the surrounding environment can be modeled as a 3D surface using a (e.g., truncated) signed distance function (SDF) that encodes the distance between each voxel in the 3D grid and the detected 3D surface topology represented by one or more depth maps. The sensor data can be back-projected onto the 3D surface representing the detected 3D surface topology using the corresponding depth value in each back-projection to generate a textured 3D surface, and a view of the textured 3D surface can be generated from the perspective of the virtual camera.

[0007] Dynamic objects (e.g., moving vehicles or pedestrians) can be detected and masked from the first pass of texturing. Rigid dynamic objects can be visualized by warping corresponding depth values ​​using corresponding trajectories, inserting or fusing the resulting warped 3D representation of each such object into a (e.g., textured) 3D surface topology, and texturing the warped 3D representation of each object using corresponding image data. Non-rigid dynamic objects can be represented as flat 2D surfaces and textured with corresponding image data. Thus, the techniques described herein can be used to more accurately visualize the environment surrounding the self-machine than the prior art. For example, estimating depth using multi-view cues and / or sensor data frames representing different time slices can provide more accurate depth estimates, and modeling the surrounding environment as the corresponding detected 3D surface topology can provide a more accurate model of depth than the prior art.

[0008] Therefore, the techniques described herein can be used to more accurately visualize the environment around a self-machine than existing techniques. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The present system and method for geometric estimation and dynamic object rendering for vehicle environment visualization are described in detail below with reference to the accompanying drawings, wherein:

[0010] Figure 1 is a schematic diagram illustrating an example environment visualization pipeline according to some embodiments of the present disclosure;

[0011] Figure 2 shows an example scenario in which depth may be estimated using frames of image data representing spatial and / or temporal context according to some embodiments of the present disclosure;

[0012] Figure 3 shows an example 3D representation of a detected 3D surface topology according to some embodiments of the present disclosure;

[0013] Figure 4 shows an example view of detected 3D surface topology according to some embodiments of the present disclosure;

[0014] Figure 5 is a flowchart illustrating a method for generating a visualization of an environment using a signed distance field according to some embodiments of the present disclosure;

[0015] Figure 6 is a flow chart illustrating a method for generating a visualization of an environment based at least on a 3D surface topology of the environment according to some embodiments of the present disclosure;

[0016] Figure 7 shows an example scenario in which a dynamic object may be detected using an image data frame according to some embodiments of the present disclosure;

[0017] Figures 8A-8B shows an example scene in which a rigid dynamic object may be inserted into the 3D surface topology of an environment according to some embodiments of the present disclosure;

[0018] Figures 9A-9B shows an example scene in which a set of non-rigid dynamic objects may be inserted into the 3D surface topology of an environment according to some embodiments of the present disclosure;

[0019] Figure 10 is a flow chart illustrating a method for generating a visualization of an environment with dynamic objects according to some embodiments of the present disclosure;

[0020] Figure 11 is a flow chart illustrating a method for generating a visualization of an environment based at least on generating graphical content for one or more 3D representations of one or more detected dynamic objects, according to some embodiments of the present disclosure;

[0021] Figure 12A is an illustration of an example autonomous vehicle according to some embodiments of the present disclosure;

[0022] Figure 12BAccording to some embodiments of the present disclosure Figure 12A Examples of camera positions and fields of view for autonomous vehicles;

[0023] Figure 12C According to some embodiments of the present disclosure Figure 12A a block diagram of an example system architecture for an example autonomous vehicle;

[0024] Figure 12D According to some embodiments of the present disclosure, a method for Figure 12A System diagram of an example of communication between autonomous vehicles;

[0025] Figure 13 is a block diagram of an example computing device suitable for implementing some embodiments of the present disclosure; and

[0026] Figure 14 is a block diagram of an example data center suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0027] Systems and methods related to geometry estimation and dynamic object rendering for vehicular environment visualization are disclosed. More specifically, the environment surrounding an ego machine can be visualized by texturing the detected 3D surface topology of the surrounding environment. Dynamic objects can be rendered by texturing warped 3D representations of rigid objects and / or planar 2D surfaces representing non-rigid objects. This technology can be used to visualize the environment surrounding an ego machine (e.g., a vehicle, robot, and / or other type of object) in systems such as parking visualization systems, surround view systems, and / or other systems.

[0028] While the present disclosure may relate to an example autonomous or semi-autonomous vehicle or machine 1200 (alternatively referred to herein as "vehicle 1200" or "ego-machine 1200"), its examples relate to Figures 12A-12DThe present disclosure may be described with respect to visualization of the environment surrounding a self-machine, but this is not intended to be limiting, and the systems and methods described herein may be used with, but are not limited to, non-autonomous vehicles or machines, semi-autonomous vehicles or machines (e.g., in one or more advanced driver assistance systems (ADAS)), autonomous vehicles or machines, manned and unmanned robots or robotic platforms, warehouse vehicles, off-road vehicles, vehicles coupled to one or more trailers, spacecraft, ships, shuttles, emergency response vehicles, motorcycles, electric or motorized bicycles, airplanes, construction vehicles, trains, submarines, remotely controlled vehicles (e.g., drones), and / or other vehicle types. Furthermore, while the present disclosure may be described with respect to visualization of the environment surrounding a self-machine, 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 field where visualization of the environment and / or dynamic objects may be used.

[0029] In some embodiments, the environment surrounding the ego machine can be visualized by extracting one or more depth maps from image data, converting the depth maps into a detected 3D surface topology of the surrounding environment, and / or texturing the detected 3D surface topology with the image data. Dynamic objects (e.g., moving vehicles or pedestrians) can be detected and masked from the first texturing pass. Rigid objects can be visualized by warping their corresponding depth values ​​using their corresponding trajectories, inserting or fusing the resulting warped 3D representation of each such object into the (e.g., textured) 3D surface topology, and texturing the warped 3D representation of each object using the corresponding image data. For non-rigid objects, warping information from previous frames is challenging because it typically requires estimating the displacement of every point on the object, not just the displacement of the object itself. Therefore, in some embodiments, non-rigid objects can be represented as flat 2D surfaces and textured with the corresponding image data. The resulting textured 3D surface topology can then be rendered to generate a visualization of the surrounding environment. This surface topology visualization can provide a more accurate view of the surrounding environment compared to environment modeling techniques such as those that model the surrounding environment as a 3D bowl. In this regard, nearby objects (e.g., mobile obstacles or their subclasses such as vehicles, pedestrians, and / or others) are less likely to be occluded or ignored in the surface topology visualization. On the other hand, distant objects are not deformed or stretched due to their geometry not matching that of the 3D bowl.

[0030] For example, one or more sensors (e.g., cameras) of a self-machine (e.g., a vehicle) can be used to generate sensor data frames (e.g., images) as the self-machine navigates the environment. Any known depth estimation technique can be used to estimate the depth of the surrounding environment using the sensor data. In some embodiments, any number of image data frames generated using any number of sensors can be (e.g., stacked into corresponding channels of a tensor and) applied to one or more neural networks to predict one or more corresponding depth maps (e.g., representing a depth estimate per pixel). In an example implementation, multiple frames generated using different sensors (e.g., cameras) and representing the same time slice can be grouped together to represent spatial context, multiple frames generated using the same sensor (e.g., camera) and representing different time slices can be grouped together to represent temporal context, and the resulting (e.g., stacked) representation of the image data can be applied to a neural network to estimate one or more depth maps (corresponding to a depth map for each image data frame). Thus, the depth of the surrounding environment can be estimated using single-view and / or multi-view cues.

[0031] In some embodiments, one or more depth maps can be converted into a 3D representation of the detected surface topology of the surrounding environment. For example, the geometry of the surrounding environment can be modeled as a 3D surface using a (e.g., truncated) signed distance function (SDF) that encodes the distance of each voxel in the 3D grid to the detected 3D surface topology represented by the one or more depth maps. In some cases, because the quality of the depth estimate may degrade beyond a certain distance from the ego machine, the distance to the detected 3D surface topology represented in the SDF can be truncated to a specified maximum value. Thus, the truncated signed distance function (TSDF) can effectively clip the detected 3D surface topology to a sphere with a specified radius. In general, a new 3D representation of the detected surface topology of the surrounding environment (e.g., a (T)SDF) can be generated for each time slice, and / or the 3D representation of the detected surface topology from the previous time slice can be updated by adding and removing deltas (e.g., scene updates). Any known technique can be used to identify and fill holes in the detected 3D surface topology of the surrounding environment (e.g., to deocclude areas that have never been observed). Thus, a 3D surface (eg, a 3D mesh) may be extracted from the detected 3D surface topology represented in the (T)SDF and may be smoothed using any known technique.

[0032] Thus, sensor data (e.g., image data) can be back-projected onto a 3D surface representing a detected 3D surface topology (e.g., a 3D surface mesh) using corresponding depth values ​​in each back-projection (e.g., represented by an unsmoothed 3D surface representation, rendered using an unstructured photometric map, etc.) to generate a textured 3D surface. This process can be understood as backward warping a texture from corresponding image data frames (e.g., from any number of cameras and / or representing multiple time slices). Areas not covered by the texture (e.g., holes or other blank areas) may not be observed, so these areas can be covered and / or represented using blur. In some embodiments, texture mapping can use image data frames representing one or more previous time slices, and uncovered or other identified old areas can be visualized by reducing color saturation, applying hues, and / or other techniques.

[0033] In some embodiments, rather than texturing the detected 3D surface topology with all image data from one or more frames, dynamic objects can be detected and masked from texturing, effectively creating a representation of the static portions of the scene. Any known technique can be applied to detect 2D and / or 3D bounding boxes or other enclosing shapes representing any desired class of dynamic objects (e.g., vehicles, vulnerable road users, etc.). Thus, detected enclosing shapes of one or more specified classes of dynamic objects can be detected (e.g., in each frame of image data), and the area inside the detected enclosing shapes can be masked from texturing in a first pass.

[0034] Some embodiments may visualize rigid objects of a specified class (e.g., a class of rigid objects such as vehicles) by tracking trajectories (e.g., 3D bounding boxes or other bounding shapes) of detected objects of a specified class (e.g., a class of rigid objects such as vehicles), identifying detected depth values ​​representing detected objects in previous time slices (e.g., based on pixel correspondences between a segmentation map and a corresponding depth map), and warping the detected depth values ​​using the corresponding detected trajectories to generate a warped 3D representation of each such object (e.g., a 3D surface representation, such as a 3D surface mesh corresponding to the detected 3D bounding box or other bounding shape). Additionally or alternatively, some embodiments may visualize non-rigid objects of a specified class (e.g., a class of non-rigid objects such as pedestrians) by modeling each such object as a flat two-dimensional surface (e.g., a piece of cardboard), inserting or fusing the flat 2D surface at a location in the (e.g., textured) 3D surface topology corresponding to the detected location of the object (e.g., the centroid of the 3D bounding box or other bounding shape), and texturing the flat 2D surface using corresponding image data (from a detected region of a corresponding image frame representing the object).

[0035] Thus, the techniques described herein can be used to more accurately visualize the environment around a self-machine than prior art techniques. For example, estimating depth using multi-view cues and / or sensor data frames representing different time slices can provide more accurate depth estimates, and modeling the surrounding environment as corresponding detected 3D surface topologies provides a more accurate depth model than prior art techniques. Thus, spatially aligning sensor data using these improved depth estimates can achieve better alignment, thereby reducing visual artifacts and improving image quality. Furthermore, by rendering dynamic objects using distorted 3D representations of rigid objects and / or texturing flat 2D surfaces representing non-rigid objects, the quality of the resulting visualization is improved compared to prior art techniques. Thus, the techniques described herein can be used to reduce visual artifacts, improve visual representations of the surrounding environment, and thereby facilitate safe operation of the self-machine.

[0036] refer to Figure 1 , Figure 1 is an example environment visualization pipeline 100 according to some embodiments of the present disclosure. It should be understood that this arrangement and other arrangements described herein are presented by way of example only. Other arrangements and elements (e.g., machines, interfaces, functions, sequences, functional groupings, etc.) may be used in addition to or in place of the arrangements and elements shown, and some elements may be omitted entirely. In addition, many of the elements described herein are functional entities that may be implemented as discrete or distributed components or in conjunction with other components and in any suitable combination and location. The various functions described herein as being performed by the entities may be performed by hardware, firmware, and / or software. For example, the various functions may be performed by a processor executing instructions stored in a memory. In some embodiments, the systems, methods, and processes described herein may be implemented using hardware, firmware, and / or software. Figures 12A-12D Example of autonomous vehicle 1200, Figure 13 The example computing device 1300 and / or Figure 14 The example data center 1400 may be performed using components, features, and / or functions similar to the components, features, and / or functions of the example data center 1400.

[0037] As a high-level overview, the environment visualization pipeline 100 can be incorporated into a self-machine, e.g. Figures 12A-12D 12. The environment visualization pipeline 100 may include any number and type of sensors 101, such as one or more cameras, that may be used to generate sensor data (e.g., image data 105) representing the surrounding environment. The environment visualization pipeline 100 may use the image data 105 to generate a visualization of the surrounding environment and / or present it on a display 180 that is visible to an occupant or operator (e.g., a driver or passenger) of the ego-object.

[0038] In an example embodiment, a self-machine (e.g., Figures 12A-12DThe autonomous vehicle 1200 in FIG. 1 is equipped with any number and type of sensors 101 (e.g., one or more cameras, such as a fisheye camera), and the sensors 101 can be used to capture overlapping sensor data frames (e.g., overlapping image data) for each time slice. In general, any suitable sensor can be used, such as Figure 12A One or more of the stereo camera 1268, wide-angle camera 1270 (e.g., fisheye camera), infrared camera 1272, surround camera 1274 (e.g., 360° camera), and / or long-range and / or medium-range camera 1298 of the vehicle 1200. Typically, different sensors have their own 3D coordinate systems. Therefore, some embodiments align the sensor data (e.g., image data 105) from the sensor 101 in a coordinate system defined relative to the self-machine, such as the vehicle equipment coordinate system. Additionally or alternatively, the environment around the self-machine can be modeled in a global 3D coordinate system (world space) and the sensor data can be aligned in the global 3D coordinate system. In an example configuration, four fisheye cameras are mounted on the front, left, rear, and right sides of the vehicle, where surrounding video is continuously captured. The vehicle's self-motion can be generated using any known technique and synchronized with the timestamps of the video frames (e.g., images). For example, absolute or relative ego-motion data (e.g., position, orientation, position and rotational velocity, position and rotational acceleration) can be determined using vehicle speed sensors, gyroscopes, accelerometers, inertial measurement units (IMUs), and / or other devices.

[0039] exist Figure 1 In the illustrated embodiment, the environment visualization pipeline 100 includes a sensor 101, an environment visualization generator 110, and a display 180. In an exemplary embodiment, the environment visualization generator 110 can generate a 3D representation of the environment surrounding the ego machine (e.g., a detected 3D surface topology) using sensor data (e.g., image data 105) from the sensor 101, texture the detected 3D surface topology by projecting the sensor data onto the detected 3D surface topology, and render a visualization of the textured 3D surface topology (or project depth from the detected 3D surface topology into a corresponding visualization and texture the visualization using the projected depth). For example, a camera of the ego machine (e.g., a car) can record video or can otherwise capture images of the environment surrounding the ego machine. The captured images can be processed by the environment visualization generator 110 to generate a textured 3D surface topology of the car's environment, and a view of the textured 3D surface topology can be presented on the display 180.

[0040] exist Figure 1In the illustrated embodiment, the environment visualization generator 110 includes: an environment modeling component 115 that generates a detected 3D surface topology of the surrounding environment; a texture mapping component 160 that textures the detected 3D surface topology using sensor data; a dynamic object processor 140 that inserts and textures 3D representations of detected dynamic objects in the (e.g., textured) detected 3D surface topology; and a view generator 170 that renders a view of the (e.g., textured) detected 3D surface topology.

[0041] In some embodiments, the environment modeling component 115 extracts depth values ​​from the sensor data from the sensor 101 and uses the extracted depth values ​​to model the surrounding environment as a detected 3D surface topology. Figure 1 In the illustrated embodiment, the environment modeling component 115 includes a depth estimator 120 (which extracts depth associated with objects within the environment), a geometry estimator 125 (which estimates the geometry of the surrounding environment), and a hole filler 130 (which fills any number of holes or other incomplete content regions, de-occluded regions, or other regions in the estimated geometry).

[0042] In some embodiments, the depth estimator 120 estimates the distance and / or direction to detected surfaces and / or objects in the environment (e.g., using any known depth estimation technique), and the geometry estimator 125 can generate a 3D surface topology representing the surrounding environment based on the distance and / or direction. In some embodiments, sensor data representing detected surfaces and / or objects (e.g., LiDAR or RADAR point clouds) can be used to determine the locations of the surfaces and objects in the environment, and the depth estimator 120 can calculate the distance (e.g., the distance to each point in the area of ​​the surrounding environment).

[0043] In some embodiments, the depth estimator 120 may apply any number of sensor data frames (e.g., image data 105) generated using any number of sensors 101 over any number of time slices to one or more neural networks to generate or predict one or more corresponding depth maps. These extracted depth maps may represent per-pixel depth estimates corresponding to each image in a set of images. The depth estimator 120 may group multiple frames representing the same time slice and generated using different sensors 101 to represent spatial context. Additionally or alternatively, the depth estimator 120 may group multiple frames representing different time slices and generated using the same sensor to represent temporal context. Any grouping of frames, image data, and / or other sensor data may be applied to one or more neural networks. Thus, the depth estimator 120 may estimate the depth corresponding to each frame of image data (e.g., using single-view and / or multi-view cues).

[0044] In some embodiments, the depth estimator 120 is implemented using a neural network (e.g., a convolutional neural network (CNN)), but this is not limiting. For example, but not limitation, the depth estimator 120 (and / or other components described herein) can include any type of multiple different networks or machine learning models, such as machine learning models using linear regression, logistic regression, decision trees, support vector machines (SVM), naive Bayes, k-nearest neighbors (Knn), K-means clustering, random forests, dimensionality reduction algorithms, gradient boosting algorithms, neural networks (e.g., autoencoders, convolutions, transformers, recurrent, perceptrons, long / short term memory (LSTM), Hopfield, Boltzmann, deep belief, deconvolution, generative adversarial, liquid machines, etc.), and / or other types of machine learning models.

[0045] Figure 2 An example scenario according to some embodiments of the present disclosure is shown, in which depth can be estimated using frames of image data representing spatial and / or temporal context. Typically, an ego machine (e.g., Figures 12A-12D The car depicted in FIG can be equipped with any number of sensors (e.g., cameras) located at any number of locations on or in the ego machine. The sensors can be used to generate sensor data frames in any number of time slices. For example, Figure 2 A car is shown equipped with a front camera 210, a left camera 215, and a right camera 220, which can be used to generate frames of image data as the car navigates an environment during a first time slice 201, a second time slice 202, and a third time slice 203. For example, the front camera 210 can be used to generate image 206, the left camera 215 can be used to generate image 205A, and the right camera 220 can be used to generate image 205B, all of which represent a common time slice (e.g., time slice 203). With one of the cameras (e.g., the front camera 210) as a reference, the same camera can be used to generate images (e.g., images 204A, 204B) representing other time slices (e.g., time slices 202, 201). Thus, a camera can be used to generate image data representing temporal context (e.g., via image data from different time slices, such as images 206, 204A, and 204B) and / or spatial context (e.g., via image data generated within the same time slice using different cameras, such as images 206, 205A, and 205B).

[0046] Thus, images representing different views and / or different time slices of an environment may be combined (e.g., stacked) and used to extract one or more depth values ​​(e.g., by Figure 1For example, image data representing a first time slice generated using a reference camera may be combined with a) image data representing the first time slice generated using other cameras and / or b) image data representing other time slices generated using the reference camera to generate a combined representation, and the combined representation of image data may be applied to a neural network to extract depth data (e.g., a depth map) corresponding to the field of view of the reference camera. Figure 2 An example depth map 207 is shown corresponding to an image 206 generated using a front-facing camera 210. Thus, depth maps corresponding to any number and type of reference sensors may be generated.

[0047] Return now Figure 1 In some embodiments, the geometry estimator 125 may convert the depth values ​​generated by the depth estimator 120 (e.g., one or more depth maps) into a 3D representation of the detected 3D surface topology of the surrounding environment. Any method for generating a 3D representation of the environment based on depth data or depth maps may be used. For example, the geometry estimator 125 may model the geometry of the surrounding environment as a 3D surface using a signed distance function (SDF) that encodes the distance of each voxel in a 3D grid to the detected 3D surface topology represented by one or more depth maps. In some embodiments, the quality of the depth estimate degrades beyond a certain distance from the ego machine, so the geometry estimator 125 may truncate the distance to the detected 3D surface topology represented in the SDF to a specified maximum value. Thus, the truncated signed distance function (TSDF) may effectively clip the detected 3D surface topology to a sphere with a specified radius. The SDF may additionally or alternatively be truncated to any specified maximum distance in any direction and / or using any other suitable geometric shape. In some embodiments, geometry estimator 125 may generate an SDF for each time slice and / or for each of one or more sensors 101 (and combine, for example, by averaging distance values ​​derived from overlapping regions of overlapping depth maps). Geometry estimator 125 may additionally or alternatively update the SDF from a previous time slice by adding and removing detected deltas (e.g., detected depth values ​​corresponding to regions of detected movement or other detected changes).

[0048] Figure 3 An example 3D representation of a detected 3D surface topology according to some embodiments of the present disclosure is shown. More specifically, Figure 33D surface 310. Grid 302 shows how a sensor 301 observing a 3D surface 310 can be used to generate an SDF represented by a grid 302 representing surface 310. For example, sensor 301 can be used to generate a range image 305 that represents the distance between sensor 301 and the portion of surface 310 represented by the corresponding pixel of range image 305. Thus, range image 305 can encode the distance between sensor 301 and surface 310. In contrast, the SDF can encode the distance between each voxel 315 in a volume 320 and its nearest point on surface 310. Grid 302 shows a cross-section of volume 320 and an extracted boundary 303 (shown as a circle) of volume 320.

[0049] In general, an SDF is a representation that can be used in 3D reconstruction and computer graphics to model the surface geometry of an object or scene. It can describe the distance from each point in 3D space (e.g., voxel 315 in volume 320) to the nearest surface of the object (e.g., surface 310), and includes a sign indicating whether the point is inside (negative distance) or outside (positive distance) the surface. For example, positive numbers outside the extracted boundary 303 in grid 302 indicate that these cells represent points outside volume 320, negative numbers inside the extracted boundary 303 indicate that these cells represent points inside surface 310, and cells with zeros represent points on surface 310.

[0050] Back to Figure 1 , the geometry estimator 125 can extract a 3D surface representation of the detected 3D surface topology, such as a 3D mesh, from the detected 3D surface topology represented in the SDF, and the geometry estimator 125 can apply any known smoothing technique to the extracted 3D mesh (and / or detected 3D surface topology). Figure 4 An example view 404 of detected 3D surface topology according to some embodiments of the present disclosure is shown. This example illustrates an implementation in which each of four cameras (e.g., front, left, right, rear) is used to generate four corresponding depth maps in each of three time slices 401, 402, 403, resulting in a total of twelve depth maps 405. Thus, the depth maps 405 can be converted (e.g., by Figure 1 The geometry estimator 125) is a 3D representation of the detected 3D surface topology (e.g., SDF, TSDF, 3D surface mesh). Figure 4 In view 404 in FIG. 4 , the detected 3D surface topology has been textured with image data (e.g., by Figure 1 A texture mapping component 160) is provided to illustrate the depth of the detected 3D surface topology.

[0051] Back to Figure 1 In some embodiments, the hole filler 130 may fill one or more detected holes or other incomplete content regions in one of the 3D representations of the detected 3D surface topology (e.g., an SDF, an extracted 3D surface mesh). Thus, depending on the implementation, the 3D surface representation may be extracted before or after the hole filler 130 has repaired or filled any holes, de-occlusions, or other regions in the detected 3D surface topology. In some embodiments, the hole filler 130 may use any known technique to identify and fill holes, de-occlusions, and / or any missing and / or corrupted content or other data in the detected 3D surface topology of the surrounding environment. For example, the hole filler 130 may identify voxels in the SDF that represent missing or inconsistent information by examining and detecting structural consistency within regions of the detected 3D surface topology, examining and detecting discontinuities in corresponding depth values ​​represented by the detected 3D surface topology, and the like. Additionally or alternatively, the hole filler 130 may use any known technique to generate graphics data to replace one or more incomplete content regions, for example, using any suitable hole filling technique (e.g., 3D inpainting) to fill identified holes or other incomplete content regions, deocclusion, and / or any missing and / or corrupted data in the detected 3D surface topology.

[0052] In some embodiments, the texture mapping component 160 can texture or otherwise apply a graphical representation of a visual element (e.g., a color, image, design, pattern, etc.) to a detected 3D surface topology (e.g., an extracted 3D surface mesh) and / or a 2D view of the detected 3D surface topology by projecting sensor data (e.g., image data 105) using depth values ​​derived from the detected 3D surface topology and / or generated by the depth estimator 120. For example, the texture mapping component 160 can map the image data 105 onto the 3D surface topology or some other 3D representation of the surrounding environment to generate a textured 3D model of the environment (e.g., a textured 3D surface topology). In some embodiments, to perform texturing, the texture mapping component 160 back-projects 2D points in pixel coordinates to 3D points using corresponding depth values ​​obtained from the detected 3D surface topology (e.g., an SDF, an extracted surface mesh), from a sensed or extracted depth map (e.g., generated using a LiDAR or RADAR sensor, generated by the depth estimator 120 using one or more neural networks, etc.), and / or otherwise. With the image data assigned to corresponding 3D points, the texture mapping component 160 can project the 3D points onto, for example, a 3D surface topology to generate a textured 3D surface topology.

[0053] Thus, in some embodiments, the texture mapping component 160 can back-project the image data from the sensor 101 onto a 3D surface representation of the detected 3D surface topology using corresponding depth values ​​(e.g., represented by an unsmoothed 3D surface representation, rendered using unstructured photometric rendering, etc.) at each back-projection to generate a textured 3D surface. In some embodiments (e.g., where depth values ​​are extracted from the current and / or previous time slices), the texture can be back-warped from corresponding image data frames from any number of sensors 101 and / or corresponding time slices. In some implementations, areas not covered by the texture can be determined to be outdated or stale. Thus, the texture mapping component 160 can cover the detected outdated or stale areas and / or can use blurring to represent such areas. Additionally or alternatively, the texture mapping component 160 can texture the 3D surface topology using image data frames representing one or more previous time slices, and can represent potentially outdated or stale areas by reducing color saturation, applying tint, and / or any other texturing technique.

[0054] Thus, the view generator 170 can render a textured view of the detected 3D surface topology (e.g., a textured 3D surface mesh), or the view generator 170 can generate an untextured view of the detected 3D surface topology (e.g., a 2D image with placeholder pixels), the texture mapping component 160 can project depth values ​​into the 2D view (e.g., generate a corresponding depth map), and the texture mapping component 160 can texture the 2D view using the corresponding (e.g., per-pixel) depth values. For example, the view generator 170 can position and orient a virtual camera in a 3D scene having a 3D surface topology and render a view of the textured 3D surface topology from the perspective of the virtual camera through a corresponding viewport. In some embodiments, the viewport may be selected based on the driving scenario (e.g., orienting the viewport in the direction of ego motion), based on a detected salient event (e.g., orienting the viewport toward the detected salient event), based on an in-cabin command (e.g., orienting the viewport in the direction indicated by a command issued by an operator or occupant of the ego machine), based on a remote command (orienting the viewport in the direction indicated by a remote command), and / or other means. Thus, the view generator 170 may output a visualization of the surrounding environment (e.g., a surround view visualization) to a display 180 (e.g., a monitor visible to an occupant or operator of the ego machine).

[0055] Now refer to Figure 5, each block of methods 500, 600, 1000, and 1100 described herein includes a computing process that can be performed using any combination of hardware, firmware, and / or software. For example, the various functions can be performed by a processor executing instructions stored in a memory. The methods can also be embodied as computer-usable instructions stored on a computer storage medium. The methods can be provided by a standalone application, a service, or a hosted service (standalone or in combination with another hosted service), or a plug-in to another product, to name a few. In addition, methods 500, 600, 1000, and 1100 are about Figure 1 The environment visualization pipeline of FIG. 5 is described by way of example. However, these methods may additionally or alternatively be performed by any system or any combination of systems, including but not limited to the systems described herein.

[0056] Figure 5 is a flow chart illustrating a method 500 for generating a visualization of an environment using a signed distance field according to some embodiments of the present disclosure. The method 500 includes, at block B510, estimating depth from image data from a set of cameras. For example, with respect to Figure 1 , the depth estimator 120 can estimate the distance and / or direction to detected surfaces and / or objects in the environment (e.g., using any known depth estimation technique). The depth estimator 120 can apply any number of sensor data frames (e.g., image data 105) generated over any number of time slices using any number of sensors 101 to one or more neural networks to generate or predict one or more corresponding depth maps. These extracted depth maps can represent per-pixel depth estimates corresponding to each image in a set of images. The depth estimator 120 can group multiple frames representing the same time slice and generated using different sensors 101 to represent spatial context. Additionally or alternatively, the depth estimator 120 can group multiple frames representing different time slices and generated using the same sensor to represent temporal context. Any grouping of frames, image data, and / or other sensor data can be applied to one or more neural networks. Thus, the depth estimator 120 can estimate the depth corresponding to each frame of image data (e.g., using single-view and / or multi-view cues). In embodiments, depth can be estimated for any number of objects in image data from a set of cameras, whether static or dynamic.

[0057] At block B520, the depth values ​​(e.g., the depth map corresponding to each of the plurality of images) are converted to an SDF (e.g., a TSDF). For example, each depth map (e.g., corresponding to each image generated using each camera in each time slice) can be converted to a corresponding SDF (e.g., a TSDF), and the SDFs can be fused together (e.g., by combining overlapping depth values, such as by averaging). For example, with respect to Figure 1 , the geometry estimator 125 can generate a 3D surface topology representing the surrounding environment based on the detected depth and / or direction to the detected objects and / or surfaces. In some embodiments, the geometry estimator 125 can convert the depth values ​​(e.g., one or more depth maps) generated by the depth estimator 120 into a 3D representation of the detected 3D surface topology of the surrounding environment. Any method for generating a 3D representation of the environment based on depth data or depth maps can be used. For example, the geometry estimator 125 can use SDFs to model the geometry of the surrounding environment as a 3D surface, and can generate (e.g., fuse) any number of SDFs to represent the detected 3D surface topology of the surrounding environment within a given time slice.

[0058] At block B530, a surface mesh representing the detected 3D surface topology is extracted from the SDF. In an embodiment, extracting the surface mesh includes traversing the SDF and identifying points where the sign of the encoded distance changes, indicating the presence of a surface. The surface points can be connected to form triangles, thereby creating an extracted mesh that approximates the underlying geometry.

[0059] At B540, the depth is projected from the surface mesh into the 2D view defined by the virtual camera. For example, Figure 1 , the view generator 170 can generate a view of the detected 3D surface topology (e.g., a 2D image with placeholder pixels for range, depth, and / or color values) from the perspective of a virtual camera positioned and oriented in the 3D scene, and the texture mapping component 160 can generate a corresponding depth map (or fill placeholder pixels) representing the distance between each pixel and the detected 3D surface topology.

[0060] At B550, the image data may be back-projected into a 2D view with the corresponding depth. Figure 1 , the texture mapping component 160 can texture the 2D view using corresponding (e.g., per-pixel) depth values ​​from the depth map. The texture mapping component 160 can texture the 2D view of the detected 3D surface topology by projecting sensor data (e.g., image data 105) using the projected depth values ​​from the depth map. In some embodiments, the depth map can be extracted from the representation of the detected 3D surface topology before texturing the depth map or generating an image using corresponding values ​​from the depth map. Additionally or alternatively, depth and color values ​​can be determined on a per-pixel basis, for example, by extracting the depth value for a given pixel and back-projecting the color of the corresponding pixel from the image data using the extracted depth value before proceeding to the next pixel. This is by way of example only, and other variations are possible (e.g., texturing the 3D surface mesh before rendering a view of the textured surface mesh, as shown in the following flow chart).

[0061] Figure 6 is a flow chart illustrating a method 600 for generating a visualization of an environment based at least on a 3D surface topology of the environment according to some embodiments of the present disclosure. The method 600 includes, at block B610, computing the 3D surface topology of the environment based at least on sensor data generated using one or more sensors of an ego machine in the environment. For example, referring to Figure 1 , the depth estimator 120 of the environment modeling component 115 can utilize sensor data (e.g., image data 105) to generate a depth map, and the geometry estimator 125 of the environment modeling component 115 can detect a 3D surface topology representing the surrounding environment based on the depth map or any other depth data generated by the depth estimator 120.

[0062] At block B620, the method includes generating a visualization of the environment based at least on applying the graphical representation to the 3D surface topology using the sensor data. Figure 1 , the texture mapping component 160 can texture a representation of the detected 3D surface topology (e.g., an extracted 3D surface mesh) by projecting sensor data (e.g., image data 105) using depth values ​​derived from the detected 3D surface topology and / or generated by the depth estimator 120. For example, the texture mapping component 160 can map the image data 105 onto the 3D surface topology or some other 3D representation of the surrounding environment to generate a textured 3D model of the environment (e.g., the textured 3D surface topology), and the view generator 170 can render a view of the textured detected 3D surface topology.

[0063] In some embodiments, rather than texturing (e.g., the detected 3D surface topology and / or corresponding 2D views) with all sensor data (e.g., image data 105) from one or more frames, dynamic objects may be detected and masked from texturing, effectively creating a representation of the static portions of the scene, and the dynamic object processor 140 may insert and texture the 3D representation of the detected dynamic objects within the (e.g., textured) detected 3D surface topology.

[0064] For example, return Figure 1In some embodiments, the texture mapping component 160 can include a masking component 165 that masks out detected dynamic objects from sensor data (e.g., image data 105) being used for texturing (e.g., detected 3D surface topology, a 2D view of the detected 3D surface topology). For example, dynamic objects can be detected from the sensor data (e.g., image data 105) (e.g., by one or more upstream components using any known techniques, such as using computer vision and / or motion detection), one or more masks or other representations of the detected dynamic objects can be generated, and the masking component 165 can omit from texturing the sensor data identified by the one or more masks representing the detected dynamic objects. More specifically, any known technique may be applied to detect 2D and / or 3D bounding boxes or other bounding shapes representing any class of dynamic objects (e.g., vehicles, road users, moving objects) (e.g., from each image frame and / or sensor data), and the detected bounding shapes may be provided to a masking component 165, which may mask one or more corresponding regions within the detected bounding shapes from texturing. These embodiments are effective for generating a textured 3D surface topology representing a detected static portion of an environment (e.g., the environment surrounding the ego machine).

[0065] Figure 7 An example scenario in which a dynamic object can be detected using an image data frame according to some embodiments of the present disclosure is shown. Figures 12A-12D The car depicted in FIG can be equipped with any number of sensors (e.g., cameras) located at any number of locations on or in the ego machine. The sensors can be used to generate sensor data frames in any number of time slices. For example, Figure 7 1 shows image data captured by five sensors of the ego machine. Images 710A, 720A, 730A, 740A, and 750A represent example images generated by the dashboard camera and the surround-view fisheye camera in a first time slice, while images 710B, 720B, 730B, 740B, and 750B represent corresponding images in a second time slice. These images and / or other sensor data can be used, for example, by Figure 1 The dynamic object processor 140 or some other component of the image processing unit 100 detects one or more dynamic objects of a specified category and generates corresponding representations (e.g., binary masks) depicting regions (e.g., pixels) of an image and / or other sensor data predicted to represent the detected dynamic objects. Figure 7 An example dynamic object mask is shown below each corresponding image. For example, a truck 712A is detected from image 710a and a pedestrian 715B is detected from image 710B to highlight a few examples.

[0066] In general, any known technique can be used to detect any suitable class of dynamic objects, such as one or more classes of rigid dynamic objects (e.g., vehicles, subclasses thereof, such as cars and trucks, etc.) and / or one or more classes of non-rigid dynamic objects (e.g., pedestrians, subclasses thereof, such as pedestrians pushing strollers, other vulnerable road users, etc.). For example, sensor data from any number and type of sensors can be stacked and applied to one or more machine learning models (e.g., one or more neural networks), which can, for example, include an output channel for each class of object to be detected. In some embodiments, the sensor data can be accumulated, self-motion compensated, projected, and / or converted to some representation accepted by the machine learning model. For example, returning Figure 1 In some embodiments, where the data from the sensor 101 includes representations of measured 3D points (e.g., a LiDAR or RADAR point cloud), the measured 3D points (and / or other data) may be accumulated (e.g., over a specified number of time slices, LiDAR or RADAR rotations, etc.), transformed to a single coordinate system (e.g., centered at the origin of the ego machine's equipment coordinate system), self-motion compensated (e.g., to the ego machine's last known position), and / or projected to form a projected image representing any suitable view (e.g., stereo, orthographic) of the 3D environment, having any number of channels (e.g., single-channel images, multi-channel images, or tensors) representing any characteristic of the data from the sensor 101 (e.g., projected positions of the measured 3D points, one or more reflectance characteristics, image data (e.g., pixel color), etc.). For example, a (cumulative, self-motion compensated) LiDAR point cloud can be projected to form a LiDAR range image having a perspective or top-down representation (e.g., a height map) of the projected 3D positions of the measured 3D points, with any number of channels (e.g., storing intensity, height, or range images, etc.). In some embodiments, one or more sensors 101 (whether of the same type or different sensors) can be used to generate sensor data of different modalities (e.g., LiDAR range images, camera images, etc.) having the same (e.g., perspective) view of the 3D environment in a common image space, and sensor data from different sensors 101 and / or sensor modalities can be stored in different channels of a multi-channel image or tensor. These are examples only, and other variations are considered within the scope of the present disclosure.

[0067] Thus, certain representations of the sensor data can be applied to a machine learning model (e.g., by one or more upstream components, by the dynamic object processor 140, etc.), and the machine learning model can extract classification data (e.g., class confidence data for any number of categories and / or corresponding channels) and / or object instance data (e.g., representations of bounding shapes) for each detected dynamic object in the 3D environment. The classification data and object instance data can be post-processed to generate class labels, 2D and / or 3D bounding boxes, closed polylines, or other bounding shapes that identify the position, geometry, and / or orientation of the detected object instances. Thus, one or more sensors 101 can be used to generate one or more sensor data frames in each time slice (e.g., at a particular frame rate, such as 30 frames per second (fps)), and the sensor data frames for each time slice can be used to detect, generate representations thereof, monitor, and / or track the positions of detected dynamic objects of one or more specified categories (e.g., rigid and / or non-rigid objects) in the surrounding environment (whether at the same frame rate as the sensor data was generated or at some other frame rate).

[0068] The dynamic object processor 140 may include a rigid object rendering component 150 that represents and textures detected rigid objects as warped 3D representations and / or a non-rigid object rendering component 145 that represents and textures detected non-rigid objects as flat 2D surfaces.

[0069] In some embodiments, the rigid object rendering component 150 (or some other component) can track the trajectories (e.g., bounding boxes or other bounding shapes) of detected objects in a specified class of rigid objects (e.g., vehicles, subclasses thereof) over time, identify detected depth values ​​representing objects detected in a previous time slice (e.g., based on pixel correspondences between a segmentation map and a corresponding depth map), and use the corresponding detected trajectories to warp the detected depth values ​​to generate a warped 3D representation (e.g., a 3D surface representation, such as a 3D surface mesh corresponding to the detected 3D bounding box or other bounding shape) for each detected object. The rigid object rendering component 150 can interpolate or fuse the warped 3D representation of each detected object into the textured or untextured detected 3D surface topology of the surrounding environment, and can texture the warped 3D representation of each detected object using corresponding image data (e.g., identified by a corresponding mask). Warping the depth data from the previous time slice can generally result in a better representation of moving objects than relying solely on depth data from the current time slice.

[0070] Figure 8A-8B An example scenario is shown in which a rigid dynamic object can be inserted into the 3D surface topology of an environment according to some embodiments of the present disclosure. For example, Figure 8A A view of the 3D surface topology of the vehicle's surroundings is shown. In this example, area 810 shows a blurry moving car. Thus, Figure 8B A view of the 3D surface topology is shown with a warped representation of the car from a previous time slice inserted into the corresponding region 820. As shown, the warped representation from the previous time slice provides a better representation of the moving car than relying solely on depth data from the current time slice.

[0071] return Figure 1 The non-rigid object rendering component 145 can visualize non-rigid objects in a specified class of non-rigid objects (e.g., pedestrians, other vulnerable road users, or subclasses thereof) by modeling each such object as a 3D representation of a flat 2D surface (e.g., a piece of cardboard or a 2D rectangle), inserting or fusing the flat 2D surface into the 3D surface topology at a location corresponding to the detected object location (e.g., the centroid of a 3D bounding box or other bounding shape), and texturing the flat 2D surface using corresponding image data (e.g., identified by a corresponding mask). Representing non-rigid objects as flat 2D surfaces generally results in a better representation of moving objects than relying solely on depth data from the current time slice.

[0072] For example, please note Figure 8B Regions 830 and 840 in FIG. 5 do not depict pedestrians present in the scene. Figures 9A-9B A corresponding scenario is shown, where a set of non-rigid dynamic objects can be inserted into a 3D surface topology. For example, Figure 9A shows views of flat 2D surfaces (e.g., surfaces 910 and 920) located at locations corresponding to locations of detected non-rigid objects (e.g., pedestrians) in the 3D surface topology, while Figure 9B A view of a 3D surface topology is shown, where a flat 2D surface has been textured using corresponding image data. Thus, regions 930 and 940 depict pedestrians present in the scene.

[0073] Figure 10 is a flow chart illustrating a method 1000 for generating a visualization of an environment with dynamic objects according to some embodiments of the present disclosure. The method 1000 includes, at block B1010, computing a 3D surface topology of the environment based at least on sensor data generated using one or more sensors of an ego machine in the environment and based at least on one or more 3D representations of one or more detected dynamic objects in the environment. For example, referring to Figure 1, the environment modeling component 115 can generate a detected 3D surface topology of the surrounding environment, and the dynamic object processor 140 can generate and interpolate 3D representations of the detected dynamic objects. For example, the dynamic object processor 140 (or some other component) can use sensor data (e.g., image data) to detect one or more specified classes of dynamic objects and generate corresponding representations (e.g., binary masks) of regions (e.g., pixels) depicting the image and / or other sensor data predicted to represent the detected dynamic objects. In general, any known technique can be used to detect any suitable class of dynamic objects, such as one or more classes of rigid dynamic objects (e.g., vehicles, subclasses thereof, such as cars and trucks, etc.) and / or one or more classes of non-rigid dynamic objects (e.g., pedestrians, subclasses thereof, such as pedestrians pushing strollers, other vulnerable road users, etc.). The dynamic object processor 140 can include: a rigid object rendering component 150 that generates and interpolates detected rigid objects as warped 3D representations; and / or a non-rigid object rendering component 145 that generates and interpolates detected non-rigid objects as flat 2D surfaces.

[0074] At block B1020, the method includes generating a visualization of the environment based at least on generating graphical content for one or more 3D representations of the 3D surface topology using sensor data. Figure 1 , the dynamic object processor 140 can texture the 3D representation of the detected dynamic object in the (e.g., textured) detected 3D surface topology. For example, the rigid object rendering component 150 can texture a distorted 3D representation of the detected rigid object, and / or the non-rigid object rendering component 145 can texture a flat 2D surface representing a detected non-rigid object using corresponding sensors (e.g., image data).

[0075] Figure 11 is a flow chart illustrating a method 1100 for generating a visualization of an environment based at least on generating graphical content for one or more 3D representations of one or more detected dynamic objects in accordance with some embodiments of the present disclosure. The method 1100 includes, at block B1110, generating a visualization of the environment based at least on generating graphical content for one or more 3D representations of one or more detected dynamic objects in a 3D surface topology of the environment, the graphical content being generated based on sensor data generated using one or more sensors of an ego machine in the environment. For example, referring to Figure 1, the environment modeling component 115 can generate a detected 3D surface topology of the surrounding environment, the dynamic object processor 140 can generate a 3D representation of the detected dynamic object and insert it into the detected 3D surface topology, the texture mapping component 160 can texture the detected 3D surface topology, the dynamic object processor 140 can texture the 3D representation of the detected dynamic object, and the view generator can render a view of the (e.g., textured) detected 3D surface topology.

[0076] The systems and methods described herein can be used by, but are not limited to, non-autonomous vehicles, semi-autonomous vehicles (e.g., in one or more adaptive driver assistance systems (ADAS)), manned and unmanned robots or robotic platforms, warehouse vehicles, off-road vehicles, vehicles connected to one or more trailers, aircraft, boats, space shuttles, emergency response vehicles, motorcycles, electric or motorized bicycles, airplanes, construction vehicles, trains, underwater vehicles, remotely operated vehicles (e.g., drones), and / or other vehicle types. In addition, the systems and methods described herein can be used for various purposes, such as, but not limited to, machine control, machine motion, machine driving, synthetic data generation, model training, perception, augmented reality, virtual reality, mixed reality, robotics, safety and surveillance, simulation and digital twins, autonomous or semi-autonomous machine applications, deep learning, environmental simulation, object or actor simulation and / or digital twins, data center processing, conversational AI, light transport simulation (e.g., ray tracing, path tracing, etc.), collaborative content creation of 3D assets, cloud computing, generative artificial intelligence, and / or any other suitable application.

[0077] The disclosed embodiments may be included in a variety of different systems, such as automotive systems (e.g., control systems for autonomous or semi-autonomous machines, perception systems for autonomous or semi-autonomous machines), systems implemented using robots, aviation systems, medical systems, boating systems, smart area monitoring systems, systems for performing deep learning operations, systems for performing simulation operations, systems for performing digital twin operations, systems implemented using edge devices, systems including one or more virtual machines (VMs), systems for performing synthetic data generation operations, systems implemented at least in part in a data center, systems for performing conversational AI operations, systems implementing one or more language models (e.g., one or more large language models (LLMs)), systems for performing light transport simulations, systems for performing collaborative content creation of 3D assets, systems implemented at least in part using cloud computing resources, and / or other types of systems.

[0078] Example autonomous vehicle

[0079] Figure 12Ais an illustration of an exemplary autonomous vehicle 1200 according to some embodiments of the present disclosure. Autonomous vehicle 1200 (alternatively referred to herein as "vehicle 1200") 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 motorized bicycle, a motorcycle, a fire truck, a police vehicle, an ambulance, a boat, a construction vehicle, an underwater vessel, a robotic vehicle, a drone, an aircraft, a vehicle coupled to a trailer (e.g., a semi-tractor-trailer truck for hauling freight), and / or another type of vehicle (e.g., a vehicle that is unmanned and / or accommodates one or more passengers). Autonomous vehicles are generally described in terms of levels of automation 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, issued on June 15, 2018, Standard No. J3016-201609, issued on September 30, 2016, and previous and future versions of such standards). Vehicle 1200 may be capable of implementing one or more of Levels 3 to 5 of autonomous driving capabilities. For example, depending on the embodiment, vehicle 1200 may be capable of driver assistance (Level 1), partial automation (Level 2), conditional automation (Level 3), high automation (Level 4), and / or full automation (Level 5). The term "autonomous" as used herein may include any and / or all types of autonomy of vehicle 1200 or other machines, such as full autonomy, high autonomy, conditional autonomy, partial autonomy, providing assisted autonomy, semi-autonomous, primarily autonomous, or other designations.

[0080] Vehicle 1200 may include components such as a chassis, a body, wheels (e.g., 2, 4, 6, 8, 18, etc.), tires, axles, and other vehicle components. Vehicle 1200 may include a propulsion system 1250, such as an internal combustion engine, a hybrid power plant, an all-electric engine, and / or another type of propulsion system. Propulsion system 1250 may be connected to a drivetrain of vehicle 1200, which may include a transmission, to achieve propulsion of vehicle 1200. Propulsion system 1250 may be controlled in response to receiving a signal from throttle / accelerator 1252.

[0081] A steering system 1254, which may include a steering wheel, may be used to steer vehicle 1200 (e.g., along a desired path or route) when propulsion system 1250 is operating (e.g., while the vehicle is in motion). Steering system 1254 may receive signals from steering actuator 1256. For fully automated (Level 5) functionality, a steering wheel may be optional.

[0082] Brake sensor system 1246 may be used to operate vehicle brakes in response to receiving signals from brake actuator 1248 and / or brake sensors.

[0083] May include one or more system on chip (SoC) 1204 ( Figure 12C ) and / or one or more GPUs can provide signals (e.g., representing commands) to one or more components and / or systems of the vehicle 1200. For example, the one or more controllers can send signals to operate the vehicle brakes via one or more brake actuators 1248, to operate the steering system 1254 via one or more steering actuators 1156, and to operate the propulsion system 1250 via one or more throttles / accelerators 1252. The one or more controllers 1236 can include one or more onboard (e.g., integrated) computing devices (e.g., supercomputers) that process sensor signals and output operational commands (e.g., signals representing commands) to enable autonomous driving and / or assist a human driver in driving the vehicle 1200. The one or more controllers 1236 can include a first controller 1236 for autonomous driving functions, a second controller 1236 for functional safety functions, a third controller 1236 for artificial intelligence functions (e.g., computer vision), a fourth controller 1236 for infotainment functions, a fifth controller 1236 for redundancy in emergency situations, and / or other controllers. In some examples, a single controller 1236 may handle two or more of the above functions, two or more controllers 1236 may handle a single function, and / or any combination thereof.

[0084] The one or more controllers 1236 may provide signals for controlling one or more components and / or systems of the vehicle 1200 in response to sensor data (eg, sensor input) received from one or more sensors. Sensor data may be received from, for example, and without limitation, global navigation satellite system (“GNSS”) sensors 1258 (e.g., global positioning system sensors), RADAR sensors 1260 , ultrasonic sensors 1262 , LiDAR sensors 1264 , inertial measurement unit (IMU) sensors 1266 (e.g., accelerometers, gyroscopes, magnetic compasses, magnetometers, etc.), microphones 1296 , stereo cameras 1268 , wide-angle cameras 1270 (e.g., fisheye cameras), infrared cameras 1272 , surround cameras 1274 (e.g., 360-degree cameras), long-range and / or mid-range cameras 1298 , speed sensors 1244 (e.g., for measuring the velocity of the vehicle 1200 ), vibration sensors 1242 , steering sensors 1240 , brake sensors (e.g., as part of a brake sensor system 1246 ), one or more occupant monitoring system (OMS) sensors 1201 (e.g., one or more interior sensors), and / or other sensor types.

[0085] One or more of the controllers 1236 may receive input (e.g., represented by input data) from the instrument cluster 1232 of the vehicle 1200 and provide output (e.g., represented by output data, display data, etc.) via a human machine interface (HMI) display 1234, an audible annunciator, a speaker, and / or via other components of the vehicle 1200. These outputs may include information such as vehicle speed, velocity, time, map data (e.g., Figure 12C The HMI display 1234 may display information regarding the presence of one or more objects (e.g., street signs, warning signs, traffic light changes, etc.) and / or information regarding driving maneuvers that the vehicle has made, is making, or will make (e.g., changing lanes now, leaving 34B in two miles, etc.).

[0086] The vehicle 1200 also includes a network interface 1224 that can communicate over one or more networks using one or more wireless antennas 1226 and / or a modem. For example, the network interface 1224 can be capable of communicating over Long Term Evolution ("LTE"), Wideband Code Division Multiple Access ("WCDMA"), Universal Mobile Telecommunications System ("UMTS"), Global System for Mobile Communications ("GSM"), IMT-CDMA Multi-Carrier ("CDMA2000"), etc. The one or more wireless antennas 1226 can also enable communication between objects in the environment (e.g., vehicles, mobile devices, etc.) using one or more local area networks such as Bluetooth, Bluetooth Low Energy ("LE"), Z-Wave, ZigBee, etc., and / or one or more low power wide area networks ("LPWAN") such as LoRaWAN, SigFox, etc.

[0087] Figure 12B For use according to some embodiments of the present disclosure Figure 12A 12. An example of camera positions and fields of view for autonomous vehicle 1200 is shown. The cameras and respective fields of view are an example embodiment and are not intended to be limiting. For example, additional and / or alternative cameras may be included and / or located at different locations on vehicle 1200.

[0088] The camera type used for the camera may include, but is not limited to, a digital camera that may be suitable for use with components and / or systems of the vehicle 1200. The camera may operate at Automotive Safety Integrity Level (ASIL) B and / or at another ASIL. The camera type may have any image capture rate, such as 60 frames per second (fps), 120fps, 240fps, and the like, depending on the embodiment. The camera may be capable of using a rolling shutter, a global shutter, another type of shutter, or a combination thereof. In some examples, the color filter array may include a red-white-white-white (RCCC) color filter array, a red-white-white-blue (RCCB) color filter array, a red-blue-green-white (RBGC) color filter array, a Foveon X3 color filter array, a Bayer sensor (RGGB) color filter array, a monochrome sensor color filter array, and / or another type of color filter array. In some embodiments, a clear pixel camera such as a camera with an RCCC, RCCB, and / or RBGC color filter array may be used in an effort to improve light sensitivity.

[0089] In some examples, one or more of the cameras can be used to perform advanced driver assistance system (ADAS) functions (e.g., as part of a redundant or fail-safe design). For example, a multi-function monocular camera can be installed to provide functions including lane departure warning, traffic sign assistance, and intelligent headlight control. One or more of the cameras (e.g., all of the cameras) can simultaneously record and provide image data (e.g., video).

[0090] One or more of the cameras can be mounted in a mounting assembly, such as a custom-designed (three-dimensional ("3D") printed) assembly, to cut out stray light and reflections from within the car (e.g., reflections from the dashboard reflected in the windshield mirror) that might interfere with the camera's ability to capture image data. With respect to the wing mirror mounting assembly, the wing mirror assembly can be custom 3D printed so that the camera mounting plate matches the shape of the wing mirror. In some examples, one or more cameras can be integrated into the wing mirror. For side-view cameras, one or more cameras can also be integrated into the four pillars at each corner of the cabin.

[0091] A camera (e.g., a front-facing camera) having a field of view that includes a portion of the environment in front of the vehicle 1200 can be used for surround vision to help identify the forward path and obstacles, as well as assist in providing information critical to generating an occupancy grid and / or determining a preferred vehicle path with the help of one or more controllers 1236 and / or control SoCs. The front-facing camera can be used to perform many of the same ADAS functions as LIDAR, including emergency braking, pedestrian detection, and collision avoidance. The front-facing camera can also be used for ADAS functions and systems, including lane departure warning ("LDW"), autonomous cruise control ("ACC"), and / or other functions such as traffic sign recognition.

[0092] A variety of cameras may be used in the front-facing configuration, including, for example, a monocular camera platform including a complementary metal oxide semiconductor ("CMOS") color imager. Another example may be a wide-angle camera 1270, which may be used to sense objects entering the field of view from the periphery (e.g., pedestrians, traffic at an intersection, or bicycles). Although Figure 12B The figure shows only one wide-angle camera, but there can be any number (including zero) of wide-angle cameras 1270 on the vehicle 1200. In addition, long-range cameras 1298 (e.g., a pair of long-view stereo cameras) can be used for depth-based object detection, especially for objects for which neural networks have not yet been trained. Long-range cameras 1298 can also be used for object detection and classification and basic object tracking.

[0093] Any number of stereo cameras 1268 may also be included in the front configuration. In at least one embodiment, one or more stereo cameras 1268 may include an integrated control unit including a scalable processing unit that may provide a multi-core microprocessor and programmable logic ("FPGA") with an integrated controller area network ("CAN") or Ethernet interface on a single chip. Such a unit may be used to generate a 3D map of the vehicle environment, including distance estimates for all points in the image. Alternative stereo cameras 1268 may include a compact stereo vision sensor that may include two camera lenses (one on the left and one on the right) and an image processing chip that may measure the distance from the vehicle to a target object and use the generated information (e.g., metadata) to activate autonomous emergency braking and lane departure warning functions. Other types of stereo cameras 1268 may be used in addition to or alternatively to those described herein.

[0094] Cameras with a field of view that includes portions of the environment to the sides of the vehicle 1200 (e.g., side view cameras) can be used for surround view, providing information used to create and update occupancy grids and generate side impact collision warnings. For example, surround cameras 1274 (e.g., Figure 12B Four surround cameras 1274 (shown in FIG) can be placed on vehicle 1200. Surround cameras 1274 can include wide-angle camera 1270, fisheye camera, 360-degree camera, and / or the like. For example, four fisheye cameras can be placed on the front, rear, and sides of the vehicle. In an alternative arrangement, the vehicle can use three surround cameras 1274 (e.g., left, right, and rear), and can utilize one or more other cameras (e.g., a forward-facing camera) as a fourth surround-view camera.

[0095] A camera having a field of view that includes a portion of the environment behind the vehicle 1200 (e.g., a rearview camera) can be used to assist with parking, surround view, rear collision warning, and creating and updating an occupancy grid. A variety of cameras can be used, including but not limited to cameras that are also suitable as front-facing cameras as described herein (e.g., long-range and / or mid-range cameras 1298, stereo cameras 1268, infrared cameras 1272, etc.).

[0096] A camera (e.g., one or more OMS sensors 1201) having a field of view encompassing portions of the interior environment within the cabin of the vehicle 1200 can be used as part of an occupant monitoring system (OMS), such as, but not limited to, a driver monitoring system (DMS). For example, an OMS sensor (e.g., OMS sensor 1201) can be used (e.g., by controller 1236) to track the gaze direction, head posture, and / or blinking of an occupant and / or driver. This gaze information can be used to determine the occupant's or driver's attention level (e.g., to detect drowsiness, fatigue, and / or distraction) and / or to take responsive action to prevent harm to the occupant or operator. In some embodiments, data from the OMS sensors can be used to implement gaze-controlled operations triggered by the driver and / or non-driver occupants, such as, but not limited to, adjusting cabin temperature and / or airflow, opening and closing windows, controlling cabin lighting, controlling the entertainment system, adjusting rearview mirrors, adjusting seat position, and / or other operations. In some embodiments, the OMS can be used for applications such as determining when an object and / or occupant has been left behind in the cabin (e.g., by detecting the occupant's presence after the driver has exited the vehicle).

[0097] Figure 12C For use according to some embodiments of the present disclosure Figure 12A 12. Block diagram of an example system architecture for an example autonomous vehicle 1200. It will be understood that this arrangement and other arrangements described herein are set forth merely as examples. Other arrangements and elements (e.g., machines, interfaces, functions, sequences, functional groupings, etc.) may be used in addition to or in place of those shown, and some elements may be omitted entirely. Further, many of the elements described herein are functional entities that may be implemented as discrete or distributed components or in conjunction with other components, and in any appropriate combination and location. The various functions described herein as being performed by entities may be implemented by hardware, firmware, and / or software. For example, the various functions may be implemented by a processor executing instructions stored in memory.

[0098] Figure 12CEach of the components, features, and systems of vehicle 1200 is illustrated as being connected via bus 1202. Bus 1202 may include a controller area network (CAN) data interface (alternatively, referred to herein as a "CAN bus"). CAN may be a network internal to vehicle 1200 that assists in controlling various features and functions of vehicle 1200, such as actuation of brakes, acceleration, braking, steering, windshield wipers, and the like. The CAN bus may be configured to have tens or even hundreds of nodes, each with its own unique identifier (e.g., CAN ID). The CAN bus may be read to find steering wheel angle, ground speed, engine revolutions per minute (RPM), button positions, and / or other vehicle status indicators. The CAN bus may be ASIL B compliant.

[0099] Although bus 1202 is described here as a CAN bus, this is not intended to be limiting. For example, FlexRay and / or Ethernet may be used in addition to or in lieu of a CAN bus. Furthermore, although bus 1202 is represented by a single line, this is not intended to be limiting. For example, there may be any number of buses 1202, which may include one or more CAN buses, one or more FlexRay buses, one or more Ethernet buses, and / or one or more other types of buses using different protocols. In some examples, two or more buses 1202 may be used to perform different functions and / or may be used for redundancy. For example, a first bus 1202 may be used for collision avoidance functionality, and a second bus 1202 may be used for drive control. In any example, each bus 1202 may communicate with any component of vehicle 1200, and two or more buses 1202 may communicate with the same component. In some examples, each SoC 1204 , each controller 1236 , and / or each computer within the vehicle may have access to the same input data (e.g., input from sensors of the vehicle 1200 ) and may be connected to a common bus such as a CAN bus.

[0100] The vehicle 1200 may include one or more controllers 1236, such as those described herein. Figure 12A Controller 1236 may be used for a variety of functions. Controller 1236 may be coupled to any of the various other components and systems of vehicle 1200 and may be used for control of vehicle 1200, artificial intelligence of vehicle 1200, infotainment for vehicle 1200, and / or the like.

[0101] The vehicle 1200 may include one or more system-on-chips (SoCs) 1204. The SoC 1204 may include a CPU 1206, a GPU 1208, a processor 1210, a cache 1212, an accelerator 1214, a data store 1216, and / or other components and features not shown. The SoC 1204 may be used to control the vehicle 1200 in a variety of platforms and systems. For example, the one or more SoCs 1204 may be combined with an HD map 1222 in a system (e.g., a system of the vehicle 1200), which may be downloaded from one or more servers (e.g., a server) via a network interface 1224. Figure 12D one or more servers 1278) to obtain map refreshes and / or updates.

[0102] The CPU 1206 may include a CPU cluster or CPU complex (alternatively, referred to herein as a "CCPLEX"). The CPU 1206 may include multiple cores and / or L2 caches. For example, in some embodiments, the CPU 1206 may include eight cores in a coherent multiprocessor configuration. In some embodiments, the CPU 1206 may include four dual-core clusters, each with a dedicated L2 cache (e.g., a 2MB L2 cache). The CPU 1206 (e.g., CCPLEX) may be configured to support simultaneous cluster operations such that any combination of CPU 1206 clusters can be active at any given time.

[0103] CPU 1206 may implement power management capabilities including one or more of the following features: each hardware block 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 due to the execution of WFI / WFE instructions; each core may be independently power gated; each core cluster may be independently clock gated when all cores are clock gated or power gated; and / or each core cluster may be independently power gated when all cores are power gated. CPU 1206 may further implement an enhanced algorithm for managing power states, in which allowed power states and expected wakeup times are specified, and hardware / microcode determines the optimal power state to enter for the core, cluster, and CCPLEX. The processing core may support a simplified power state entry sequence in software, with this work being offloaded to the microcode.

[0104] GPU 1208 may include an integrated GPU (alternatively referred to herein as an "iGPU"). GPU 1208 may be programmable and efficient for parallel workloads. In some examples, GPU 1208 may use an enhanced tensor instruction set. GPU 1208 may include one or more streaming microprocessors, each of which may include an L1 cache (e.g., an L1 cache with at least 96KB of storage capacity), and two or more of these streaming microprocessors may share an L2 cache (e.g., an L2 cache with 512KB of storage capacity). In some embodiments, GPU 1208 may include at least eight streaming microprocessors. GPU 1208 may use a computing application programming interface (API). In addition, GPU 1208 may use one or more parallel computing platforms and / or programming models (e.g., NVIDIA's CUDA).

[0105] In the case of automotive and embedded use, GPU 1208 can be power optimized to achieve optimal performance. For example, GPU 1208 can be manufactured on fin field effect transistors (FinFETs). However, this is not intended to be limiting, and GPU 1208 can be manufactured using other semiconductor manufacturing processes. Each streaming microprocessor can incorporate several mixed precision processing cores divided into multiple blocks. For example and without limitation, 64 PF32 cores and 32 PF64 cores can be divided into four processing blocks. In such an example, each processing block can be allocated 16 FP32 cores, 8 FP64 cores, 16 INT32 cores, two mixed precision NVIDIA tensor cores for deep learning matrix arithmetic, L0 instruction cache, warp scheduler, dispatch unit and / or 64KB register file. In addition, the streaming microprocessor may include independent parallel integer and floating point data paths to provide efficient execution of workloads using a mix of computation and addressing calculations. The streaming microprocessor may include independent thread scheduling capabilities to allow for finer-grained synchronization and cooperation between parallel threads. The streaming microprocessor may include a combined L1 data cache and shared memory unit to improve performance while simplifying programming.

[0106] GPU 1208 can include high bandwidth memory (HBM) and / or a 16GB HBM2 memory subsystem that provides a peak memory bandwidth of approximately 900 GB / s in some examples. In some examples, synchronous graphics random access memory (SGRAM), such as fifth generation graphics double data rate synchronous random access memory (GDDR5), can be used in addition to or in lieu of HBM memory.

[0107] The GPU 1208 may include unified memory technology that includes access counters to allow memory pages to be more accurately migrated to the processor that accesses them most frequently, thereby improving the efficiency of memory ranges shared between processors. In some examples, address translation services (ATS) support may be used to allow the GPU 1208 to directly access the CPU 1206 page tables. In such an example, when the GPU 1208 memory management unit (MMU) experiences a miss, an address translation request may be transmitted to the CPU 1206. In response, the CPU 1206 may look up the virtual-to-physical mapping for the address in its page table and transmit the translation back to the GPU 1208. In this way, unified memory technology may allow a single unified virtual address space to be used for memory of both the CPU 1206 and the GPU 1208, thereby simplifying GPU 1208 programming and porting applications to the GPU 1208.

[0108] Additionally, GPU 1208 may include access counters that can track how often GPU 1208 accesses the memory of other processors. Access counters can help ensure that memory pages are moved to the physical memory of the processor that accesses them most frequently.

[0109] SoC 1204 may include any number of caches 1212, including those described herein. For example, cache 1212 may include an L3 cache available to both CPU 1206 and GPU 1208 (e.g., connected to both CPU 1206 and GPU 1208). Cache 1212 may include a write-back cache that can track the state of lines, for example, using a cache coherence protocol (e.g., MEI, MESI, MSI, etc.). Depending on the embodiment, the L3 cache may include 4MB or more, although smaller cache sizes may also be used.

[0110] The SoC 1204 may include an arithmetic logic unit (ALU) that may be utilized in performing any of a variety of tasks or operations related to the vehicle 1200, such as processing a DNN. Furthermore, the SoC 1204 may include a floating point unit (FPU) (or other math coprocessor or digital coprocessor type) for performing mathematical operations within the system. For example, the SoC 1204 may include one or more FPUs integrated as execution units within the CPU 1206 and / or GPU 1208.

[0111] SoC 1204 may include one or more accelerators 1214 (e.g., hardware accelerators, software accelerators, or a combination thereof). For example, SoC 1204 may include a hardware accelerator cluster, which may include optimized hardware accelerators and / or large on-chip memory. The large on-chip memory (e.g., 4MB SRAM) may enable the hardware accelerator cluster to accelerate neural networks and other calculations. The hardware accelerator cluster may be used to supplement GPU 1208 and offload some tasks of GPU 1208 (e.g., freeing up more cycles of GPU 1208 to perform other tasks). As an example, accelerator 1214 may be used for targeted workloads (e.g., perception, convolutional neural networks (CNNs), etc.) that are sufficiently stable to be easily controlled for acceleration. When used herein, the term "CNN" may include all types of CNNs, including region-based or regional convolutional neural networks (RCNNs) and fast RCNNs (e.g., for object detection).

[0112] Accelerator 1214 (e.g., a hardware accelerator cluster) may include a deep learning accelerator (DLA). DLA may include one or more tensor processing units (TPUs) that can be configured to provide an additional 10 trillion operations per second for deep learning applications and reasoning. TPUs may be accelerators configured to perform image processing functions (e.g., for CNN, RCNN, etc.) and optimized for performing image processing functions. DLA may be further optimized for a specific set of neural network types and floating-point operations and reasoning. The design of DLA may provide higher performance per millimeter than a general-purpose GPU and far exceed the performance of a CPU. TPUs may perform several functions, including single-instance convolution functions, support for INT8, INT16, and FP16 data types for both features and weights, and post-processor functions.

[0113] DLA can quickly and efficiently execute neural networks, particularly CNNs, on processed or unprocessed data for any of a wide variety of functions, such as, but not limited to: CNNs for object recognition and detection using data from camera sensors; CNNs for distance estimation using data from camera sensors; CNNs for emergency vehicle detection and recognition and detection using data from microphones; CNNs for facial recognition and vehicle owner identification using data from camera sensors; and / or CNNs for safety and / or security-related events.

[0114] The DLA can perform any function of the GPU 1208, and by using an inference accelerator, for example, the designer can target any function to either the DLA or the GPU 1208. For example, the designer can focus the processing of CNNs and floating-point operations on the DLA and leave other functions to the GPU 1208 and / or other accelerators 1214.

[0115] The accelerator 1214 (e.g., a hardware accelerator cluster) may include a programmable vision accelerator (PVA), which may be referred to herein alternatively as a computer vision accelerator. The PVA may be designed and configured to accelerate computer vision algorithms for advanced driver assistance systems (ADAS), autonomous driving, and / or augmented reality (AR) and / or virtual reality (VR) applications. The PVA may provide a balance between performance and flexibility. For example, each PVA may include, for example and without limitation, any number of reduced instruction set computer (RISC) cores, direct memory access (DMA), and / or any number of vector processors.

[0116] The RISC core can interact with an image sensor (e.g., an image sensor of any camera described herein), an image signal processor, and / or the like. Each of these RISC cores can include any amount of memory. Depending on the embodiment, the RISC core can use any of a number of protocols. In some examples, the RISC core can execute a real-time operating system (RTOS). The RISC core can be implemented using one or more integrated circuit devices, application specific integrated circuits (ASICs), and / or memory devices. For example, the RISC core can include an instruction cache and / or tightly coupled RAM.

[0117] The DMA can enable components of the PVA to access system memory independently of the CPU 1206. The DMA can support any number of features used to provide optimizations for the PVA, including, but not limited to, support for multi-dimensional addressing and / or circular addressing. In some examples, the DMA can support addressing in up to six or more dimensions, which can include block width, block height, block depth, horizontal block stride, vertical block stride, and / or depth stride.

[0118] A vector processor can be a programmable processor that can be designed to efficiently and flexibly execute programming for computer vision algorithms and provide signal processing capabilities. In some examples, the PVA can include a PVA core and two vector processing subsystem partitions. The PVA core can include a processor subsystem, one or more DMA engines (e.g., two DMA engines), and / or other peripherals. The vector processing subsystem can operate as the main processing engine of the PVA and can include a vector processing unit (VPU), an instruction cache, and / or a vector memory (e.g., VMEM). The VPU core can include a digital signal processor, such as, for example, a single instruction multiple data (SIMD), a very long instruction word (VLIW) digital signal processor. The combination of SIMD and VLIW can enhance throughput and speed.

[0119] Each of the vector processors can include an instruction cache and can be coupled to dedicated memory. As a result, in some examples, each of the vector processors can be configured to execute independently of the other vector processors. In other examples, the vector processors included in a particular PVA can be configured to employ data parallelism. For example, in some embodiments, multiple vector processors included in a single PVA can execute the same computer vision algorithm, but on different regions of an image. In other examples, the vector processors included in a particular PVA can execute different computer vision algorithms simultaneously on the same image, or even execute different algorithms on sequential images or portions of images. Among other things, any number of PVAs can be included in a hardware accelerator cluster, and any number of vector processors can be included in each of these PVAs. In addition, the PVAs can include additional error correction code (ECC) memory to enhance overall system security.

[0120] The accelerator 1214 (e.g., a hardware accelerator cluster) may include an on-chip computer vision network and SRAM to provide high bandwidth, low latency SRAM for the accelerator 1214. In some examples, the on-chip memory may include at least 4MB of SRAM consisting of, for example and without limitation, eight field-configurable memory blocks that can be accessed by both the PVA and the DLA. Each pair of memory blocks may include an advanced peripheral bus (APB) interface, configuration circuitry, a controller, and a multiplexer. Any type of memory may be used. The PVA and DLA may access the memory via a backbone that provides high-speed memory access to the PVA and DLA. The backbone may include an on-chip computer vision network that interconnects the PVA and DLA to the memory (e.g., using APB).

[0121] The on-chip computer vision network can include an interface that ensures that both the PVA and DLA provide ready and valid signals before transmitting any control signals / addresses / data. Such an interface can provide separate phases and separate channels for transmitting control signals / addresses / data, as well as burst-based communication for continuous data transmission. This type of interface can comply with ISO 26262 or IEC 61508 standards, but other standards and protocols can also be used.

[0122] In some examples, SoC 1204 may include a real-time ray tracing hardware accelerator, such as that described in U.S. patent application Ser. No. 16 / 101,232, filed on Aug. 10, 2018. The real-time ray tracing hardware accelerator may be used to quickly and efficiently determine the position and extent of objects (e.g., within a world model) to generate real-time visualization simulations for RADAR signal interpretation, for sound propagation synthesis and / or analysis, for SONAR system simulations, for general wave propagation simulations, for comparison with LiDAR data for positioning and / or other functions, and / or for other uses. In some embodiments, one or more tree traversal units (TTUs) may be used to perform one or more ray tracing related operations.

[0123] The accelerator 1214 (e.g., a hardware accelerator cluster) has a wide range of uses in autonomous driving. The PVA can be a programmable vision accelerator that can be used in key processing stages in ADAS and autonomous vehicles. The capabilities of the PVA are a good match for algorithmic domains that require predictable processing, low power, and low latency. Therefore, the PVA performs well on semi-intensive or intensive rule computations, and even on small data sets that require predictable runtimes with low latency and low power. Therefore, in the context of platforms for autonomous vehicles, the PVA is designed to run classic computer vision algorithms because they are efficient at object detection and integer math operations.

[0124] For example, according to one embodiment of the technology, PVA is used to perform computer stereo vision. In some examples, a semi-global matching-based algorithm can be used, but this is not intended to be limiting. Many applications for Level 3-5 autonomous driving require on-the-fly motion estimation / stereo matching (e.g., structure from motion, pedestrian recognition, lane detection, etc.). PVA can perform computer stereo vision functions on input from two monocular cameras.

[0125] In some examples, PVA can 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 examples, PVA is used for time-of-flight depth processing, by processing raw time-of-flight data to provide processed time-of-flight data.

[0126] DLA can be used to run any type of network to enhance control and driving safety, including, for example, a neural network that outputs a confidence measure for each object detection. Such confidence values ​​can be interpreted as probabilities, or as providing a relative "weight" of each detection compared to other detections. This confidence value enables the system to make further decisions about which detections should be considered true positive detections rather than false positive detections. For example, the system can set a threshold for the confidence level and only consider detections that exceed the threshold as true positive detections. In an automatic emergency braking (AEB) system, a false positive detection would cause the vehicle to automatically perform emergency braking, which is obviously undesirable. Therefore, only the most confident detections should be considered as triggers for AEB. DLA can run a neural network for regressing confidence values. The neural network may take as its input at least some subset of parameters, such as bounding box dimensions, a ground plane estimate obtained (e.g., from another subsystem), inertial measurement unit (IMU) sensor 1266 output related to the orientation and distance of the vehicle 1200, 3D position estimates of objects obtained from the neural network and / or other sensors (e.g., LiDAR sensor 1264 or RADAR sensor 1260), etc.

[0127] SoC 1204 may include one or more data stores 1216 (e.g., memory). Data stores 1216 may be on-chip memory of SoC 1204 that may store neural networks to be executed on the GPU and / or DLA. In some examples, for redundancy and safety, data stores 1216 may be large enough to store multiple instances of the neural network. Data stores 1212 may include L2 or L3 cache 1212. References to data stores 1216 may include references to memory associated with the PVA, DLA, and / or other accelerators 1214 as described herein.

[0128] SoC 1204 may include one or more processors 1210 (e.g., embedded processors). Processor 1210 may include a boot and power management processor, which may be a dedicated processor and subsystem for handling boot power and management functions and related safety implementations. The boot and power management processor may be part of the SoC 1204 boot sequence and may provide runtime power management services. The boot power and management processor may provide clock and voltage programming, assist system low-power state transitions, SoC 1204 thermal and temperature sensor management, and / or SoC 1204 power state management. Each temperature sensor may be implemented as a ring oscillator whose output frequency is proportional to temperature, and SoC 1204 may use the ring oscillator to detect the temperature of CPU 1206, GPU 1208, and / or accelerator 1214. If the temperature is determined to exceed a threshold, the boot and power management processor may enter a temperature fault routine and place SoC 1204 in a lower power state and / or place vehicle 1200 in a driver safety parking mode (e.g., to safely park vehicle 1200).

[0129] The processor 1210 may also include a set of embedded processors that can be used as an audio processing engine. The audio processing engine can be an audio subsystem that allows for full hardware support for multi-channel audio through multiple interfaces and a wide range of flexible audio I / O interfaces. In some examples, the audio processing engine is a dedicated processor core having a digital signal processor with dedicated RAM.

[0130] The processor 1210 may also include an always-on processor engine that can provide the necessary hardware features to support low-power sensor management and wake-up use cases. The always-on processor engine may include a processor core, tightly coupled RAM, supporting peripherals (such as timers and interrupt controllers), various I / O controller peripherals, and routing logic.

[0131] Processor 1210 may also include a safety cluster engine, which includes a dedicated processor subsystem that handles safety management of automotive applications. The safety cluster engine may include two or more processor cores, tightly coupled RAM, supporting peripherals (such as timers, interrupt controllers, etc.), and / or routing logic. In safety mode, the two or more cores can operate in lockstep mode and act as a single core with comparison logic to detect any differences between their operations.

[0132] Processor 1210 may also include a real-time camera engine, which may include a dedicated processor subsystem for handling real-time camera management.

[0133] Processor 1210 may also 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.

[0134] The processor 1210 may include a video image compositer, which may be a processing block (e.g., implemented on a microprocessor) that implements the video post-processing functions required by the video playback application to produce the final image for the player window. The video image compositer may perform lens distortion correction for the wide-angle camera 1270, the surround camera 1274, and / or for the in-cab monitoring camera sensor. The in-cab monitoring camera sensor is preferably monitored by a neural network running on another instance of the advanced SoC and is configured to recognize in-cab events and respond accordingly. The in-cab system may perform lip reading to activate mobile phone service and place calls, dictate emails, change vehicle destinations, activate or change the vehicle's infotainment system and settings, or provide voice-activated web surfing. Certain functions are only available to the driver when the vehicle is operating in autonomous mode and are disabled in other circumstances.

[0135] The video image compositer can include enhanced temporal noise reduction for both spatial and temporal noise reduction. For example, in the presence of motion in the video, the noise reduction appropriately weights spatial information and downweights information provided by neighboring frames. In the case where an image or portion of an image does not include motion, the temporal noise reduction performed by the video image compositer can use information from previous images to reduce noise in the current image.

[0136] The video image compositor can also be configured to perform stereo rectification on the input stereo footage frames. The video image compositor can further be used for user interface composition when the operating system desktop is in use and the GPU 1208 does not need to continuously render new surfaces. Even when the GPU 1208 is powered on and active for 3D rendering, the video image compositor can be used to offload the GPU 1208 to improve performance and responsiveness.

[0137] The SoC 1204 may also 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 1204 may also include an input / output controller that may be controlled by software and may be used to receive I / O signals that are not assigned to a specific role.

[0138] The SoC 1204 may also include a wide range of peripheral interfaces to enable communication with peripherals, audio codecs, power management, and / or other devices. The SoC 1204 may be used to process data from cameras (connected via Gigabit multimedia serial links and Ethernet), sensors (e.g., LiDAR sensor 1264, RADAR sensor 1260, etc., which may be connected via Ethernet), data from the bus 1202 (e.g., vehicle 1200 speed, steering wheel position, etc.), and data from the GNSS sensor 1258 (connected via Ethernet or a CAN bus). The SoC 1204 may also include dedicated high-performance mass storage controllers, which may include their own DMA engines and which may be used to free the CPU 1206 from routine data management tasks.

[0139] SoC 1204 can be an end-to-end platform with a flexible architecture that spans levels 3-5 of automation, providing a comprehensive functional safety architecture that leverages and efficiently uses computer vision and ADAS technologies for diversity and redundancy, along with deep learning tools to provide a platform for a flexible and reliable driving software stack. SoC 1204 can be faster, more reliable, and even more energy- and space-efficient than conventional systems. For example, when combined with CPU 1206, GPU 1208, and data storage 1216, accelerator 1214 can provide a fast and efficient platform for Level 3-5 autonomous vehicles.

[0140] This technology therefore provides capabilities and functionality that cannot be achieved with conventional systems. For example, computer vision algorithms can be executed on CPUs, which can be configured using high-level programming languages ​​such as the C programming language to execute a wide variety of processing algorithms across a wide variety of visual data. However, CPUs often fail to meet the performance requirements of many computer vision applications, such as those related to execution time and power consumption. In particular, many CPUs are unable to execute complex object detection algorithms in real time, a requirement for in-vehicle ADAS applications and practical Level 3-5 autonomous vehicles.

[0141] In contrast to conventional systems, by providing a CPU complex, a GPU complex, and a cluster of hardware accelerators, the technology described herein allows for multiple neural networks to be executed simultaneously and / or sequentially, and the results to be combined to achieve Level 3-5 autonomous driving capabilities. For example, a CNN executed on a DLA or dGPU (e.g., GPU 1220) can include text and word recognition, allowing the supercomputer to read and understand traffic signs, including signs for which a neural network has not been specifically trained. The DLA can also include a neural network that can recognize, interpret, and provide semantic understanding of the signs, and pass that semantic understanding to a path planning module running on the CPU complex.

[0142] As another example, as required for Level 3, 4, or 5 driving, multiple neural networks can run simultaneously. For example, a warning sign consisting of "Caution: Flashing lights indicate icing conditions" along with a light can be interpreted by several neural networks, either independently or collectively. The sign itself can be identified as a traffic sign by a first neural network deployed (e.g., a trained neural network), and the text "Flashing lights indicate icing conditions" can be interpreted by a second neural network deployed, which informs the vehicle's path planning software (preferably executing on a CPU complex) that icing conditions exist when the flashing lights are detected. The flashing lights can be identified by operating a third neural network deployed over multiple frames, which informs the vehicle's path planning software of the presence (or absence) of the flashing lights. All three neural networks can run simultaneously, for example, within the DLA and / or on GPU 1208.

[0143] In some examples, a CNN for facial recognition and owner recognition can use data from a camera sensor to identify the presence of an authorized driver and / or owner of the vehicle 1200. The always-on sensor processing engine can be used to unlock the vehicle and turn on the lights when the owner approaches the driver's door, and in security mode, disable the vehicle when the owner leaves the vehicle. In this way, the SoC 1204 provides security against theft and / or carjacking.

[0144] In another example, a CNN for emergency vehicle detection and identification can use data from microphone 1296 to detect and identify emergency vehicle sirens. In contrast to conventional systems that use general classifiers to detect sirens and manually extract features, SoC 1204 uses a CNN to classify environmental and urban sounds as well as visual data. In a preferred embodiment, the CNN running on the DLA is trained to identify the relative closing speed of emergency vehicles (for example, by using the Doppler effect). The CNN can also be trained to identify emergency vehicles specific to the local area in which the vehicle is operating, as identified by the GNSS sensor 1258. Thus, for example, when operating in Europe, the CNN will seek to detect European sirens, and when in the United States, the CNN will seek to identify only North American sirens. Once an emergency vehicle is detected, with the assistance of the ultrasonic sensor 1262, the control program can be used to execute emergency vehicle safety routines, slowing the vehicle, pulling to the side of the road, stopping the vehicle, and / or idling the vehicle until the emergency vehicle passes.

[0145] The vehicle may include a CPU 1218 (e.g., a discrete CPU or dCPU) that may be coupled to the SoC 1204 via a high-speed interconnect (e.g., PCIe). The CPU 1218 may include, for example, an X86 processor. The CPU 1218 may be used to perform any of a variety of functions, including, for example, arbitrating potentially inconsistent results between ADAS sensors and the SoC 1204, and / or monitoring the status and health of the controller 1236 and / or the infotainment SoC 1230.

[0146] The vehicle 1200 may include a GPU 1220 (e.g., a discrete GPU or dGPU) that may be coupled to the SoC 1204 via a high-speed interconnect (e.g., NVIDIA's NVLINK). The GPU 1220 may provide additional artificial intelligence functionality, for example, by executing redundant and / or different neural networks, and may be used to train and / or update the neural network based at least in part on input from sensors of the vehicle 1200 (e.g., sensor data).

[0147] The vehicle 1200 may also include a network interface 1224, which may include one or more wireless antennas 1226 (e.g., one or more wireless antennas for different communication protocols, such as a cellular antenna, a Bluetooth antenna, etc.). The network interface 1224 can be used to enable wireless connections to the cloud (e.g., to a server 1278 and / or other network devices), to other vehicles, and / or to computing devices (e.g., a passenger's client device) via the Internet. To communicate with other vehicles, a direct link can be established between the two vehicles, and / or an indirect link can be established (e.g., across a network and through the Internet). The direct link can be provided using a vehicle-to-vehicle communication link. The vehicle-to-vehicle communication link can provide the vehicle 1200 with information about vehicles approaching the vehicle 1200 (e.g., vehicles in front of, to the side of, and / or behind the vehicle 1200). This functionality can be part of the cooperative adaptive cruise control functionality of the vehicle 1200.

[0148] The network interface 1224 may include a SoC that provides modulation and demodulation functionality and enables the controller 1236 to communicate over a wireless network. The network interface 1224 may include an RF front-end for up-conversion from baseband to RF and down-conversion from RF to baseband. The frequency conversion may be performed by well-known processes and / or may be performed using a super-heterodyne process. In some examples, the RF front-end functionality may be provided by a separate chip. The network interface may include wireless functionality for communicating over LTE, WCDMA, UMTS, GSM, CDMA2000, Bluetooth, Bluetooth LE, Wi-Fi, Z-wave, ZigBee, LoRaWAN, and / or other wireless protocols.

[0149] The vehicle 1200 may also include data storage 1228, which may include off-chip storage (e.g., outside the SoC 1204). The data storage 1228 may include one or more storage elements, including RAM, SRAM, DRAM, VRAM, flash memory, a hard disk, and / or other components and / or devices that can store at least one bit of data.

[0150] The vehicle 1200 may also include a GNSS sensor 1258. The GNSS sensor 1258 (e.g., GPS, assisted GPS sensor, differential GPS (DGPS) sensor, etc.) is used to assist with mapping, perception, occupancy grid generation, and / or path planning functions. Any number of GNSS sensors 1258 may be used, including, for example and without limitation, GPS using a USB connector with an Ethernet to serial (RS-232) bridge.

[0151] The vehicle 1200 may also include a RADAR sensor 1260. The RADAR sensor 1260 can be used by the vehicle 1200 for remote vehicle detection even in darkness and / or in adverse weather conditions. The RADAR functional safety level can be ASIL B. The RADAR sensor 1260 can use CAN and / or bus 1202 (e.g., to transmit data generated using the RADAR sensor 1260) for control and access to object tracking data, and in some examples access Ethernet to access raw data. A variety of RADAR sensor types can be used. For example and without limitation, the RADAR sensor 1260 can be suitable for front, rear, and side RADAR use. In some examples, a pulsed Doppler RADAR sensor is used.

[0152] The RADAR sensor 1260 can include different configurations, such as long-range with a narrow field of view, short-range with a wide field of view, short-range side coverage, and so on. In some examples, long-range RADAR can be used for adaptive cruise control functions. The long-range RADAR system can provide a wide field of view (e.g., within a range of 250m) achieved through two or more independent scans. The RADAR sensor 1260 can help distinguish between static objects and moving objects and can be used by the ADAS system for emergency braking assistance and forward collision warning. The long-range RADAR sensor may include a single-station multimode RADAR with multiple (e.g., six or more) fixed RADAR antennas and high-speed CAN and FlexRay interfaces. In the example with six antennas, the central four antennas can create a focused beam pattern that is designed to record the surroundings of the vehicle 1200 at a higher rate with minimal traffic interference from adjacent lanes. The other two antennas can expand the field of view, making it possible to quickly detect vehicles entering or leaving the lane of the vehicle 1200.

[0153] As an example, a medium-range RADAR system may include a range of up to 1260m (front) or 80m (rear) and a field of view of up to 42 degrees (front) or 1250 degrees (rear). A short-range RADAR system may include, but is not limited to, a RADAR sensor designed to be mounted on both ends of the rear bumper. When mounted on both ends of the rear bumper, such a RADAR sensor system can create two beams that continuously monitor the blind spots behind and beside the vehicle.

[0154] Short-range RADAR systems can be used in ADAS systems for blind spot detection and / or lane change assistance.

[0155] Vehicle 1200 may also include ultrasonic sensors 1262. Ultrasonic sensors 1262, which may be located on the front, rear, and / or sides of vehicle 1200, may be used for parking assistance and / or for creating and updating an occupancy grid. A variety of ultrasonic sensors 1262 may be used, and different ultrasonic sensors 1262 may have different detection ranges (e.g., 2.5 m, 4 m). Ultrasonic sensors 1262 may operate at functional safety level ASIL B.

[0156] Vehicle 1200 may include a LiDAR sensor 1264. LiDAR sensor 1264 may be used for object and pedestrian detection, emergency braking, collision avoidance, and / or other functions. LiDAR sensor 1264 may be ASIL B functional safety level. In some examples, vehicle 1200 may include multiple LiDAR sensors 1264 (e.g., two, four, six, etc.) that may use Ethernet (e.g., to provide data to a Gigabit Ethernet switch).

[0157] In some examples, the LiDAR sensor 1264 may be capable of providing a list of objects and their distances for a 360-degree field of view. Commercially available LiDAR sensors 1264 may have, for example, an advertised range of approximately 1200 meters, an accuracy of 2-3 cm, and support for 1200 Mbps Ethernet connections. In some examples, one or more non-obtrusive LiDAR sensors 1264 may be used. In such examples, the LiDAR sensor 1264 may be implemented as a small device that can be embedded in the front, back, sides, and / or corners of the vehicle 1200. In such examples, the LiDAR sensor 1264 may provide a field of view of up to 120 degrees horizontally and 35 degrees vertically, with a range of 200 meters, even for low-reflectivity objects. The front-mounted LiDAR sensor 1264 may be configured for a horizontal field of view between 45 and 135 degrees.

[0158] In some examples, LiDAR technologies such as 3D flash LiDAR may also be used. 3D flash LiDAR uses flashes of laser as an emission source to illuminate the vehicle's surroundings up to about 200 m. The flash LiDAR unit includes a receiver that records the laser pulse transmission time and the reflected light on each pixel, which in turn corresponds to the range from the vehicle to the object. Flash LiDAR can allow a highly accurate and distortion-free image of the surrounding environment to be generated with each laser flash. In some examples, four flash LiDAR sensors can be deployed, one on each side of the vehicle 1200. Available 3D flash LiDAR systems include solid-state 3D staring array LiDAR cameras (e.g., non-scanning LiDAR devices) with no moving parts other than a fan. The flash LiDAR device can use 5 nanosecond Class I (eye-safe) laser pulses per frame and can capture the reflected laser light in the form of a 3D range point cloud and co-registered intensity data. By using flash LiDAR, and because flash LiDAR is a solid-state device with no moving parts, the LiDAR sensor 1264 may be less susceptible to motion blur, vibration, and / or shock.

[0159] The vehicle may also include an IMU sensor 1266. In some examples, the IMU sensor 1266 may be located at the center of the rear axle of the vehicle 1200. The IMU sensor 1266 may include, for example and without limitation, an accelerometer, a magnetometer, a gyroscope, a magnetic compass, and / or other sensor types. In some examples, such as in a six-axis application, the IMU sensor 1266 may include an accelerometer and a gyroscope, while in a nine-axis application, the IMU sensor 1266 may include an accelerometer, a gyroscope, and a magnetometer.

[0160] In some embodiments, the IMU sensor 1266 can be implemented as a miniature, high-performance GPS-assisted inertial navigation system (GPS / INS) that combines micro-electromechanical systems (MEMS) inertial sensors, a high-sensitivity GPS receiver, and advanced Kalman filtering algorithms to provide estimates of position, velocity, and attitude. Thus, in some examples, the IMU sensor 1266 can enable the vehicle 1200 to estimate heading without input from a magnetic sensor by directly observing and correlating velocity changes from the GPS to the IMU sensor 1266. In some examples, the IMU sensor 1266 and the GNSS sensor 1258 can be combined into a single integrated unit.

[0161] The vehicle may include microphones 1296 positioned in and / or around the vehicle 1200. The microphones 1296 may be used for, among other things, emergency vehicle detection and identification.

[0162] The vehicle may also include any number of camera types, including stereo cameras 1268, wide angle cameras 1270, infrared cameras 1272, surround cameras 1274, long and / or medium range cameras 1298, and / or other camera types. These cameras may be used to capture image data around the entire periphery of the vehicle 1200. The type of camera used depends on the embodiment and the requirements of the vehicle 1200, and any combination of camera types may be used to provide the necessary coverage around the vehicle 1200. Additionally, the number of cameras may vary depending on the embodiment. For example, the vehicle may include six cameras, seven cameras, ten cameras, twelve cameras, and / or another number of cameras. As an example and not limitation, the cameras may support Gigabit Multimedia Serial Link (GMSL) and / or Gigabit Ethernet. Each of the cameras described herein may include a GMSL and / or Gigabit Ethernet network. Figure 12A and Figure 12B Described in more detail.

[0163] Vehicle 1200 may also include a vibration sensor 1242. Vibration sensor 1242 can measure vibrations of vehicle components, such as axles. For example, changes in vibration can indicate changes in the road surface. In another example, when two or more vibration sensors 1242 are used, the difference in vibration can be used to determine friction or slippage of the road surface (e.g., when there is a vibration difference between a powered drive shaft and a freely rotating shaft).

[0164] The vehicle 1200 may include an ADAS system 1238. In some examples, the ADAS system 1238 may include a SoC. The ADAS system 1238 may include autonomous / adaptive / automatic cruise control (ACC), cooperative adaptive cruise control (CACC), forward collision warning (FCW), automatic emergency braking (AEB), lane departure warning (LDW), lane keeping assist (LKA), blind spot warning (BSW), rear cross traffic warning (RCTW), collision warning system (CWS), lane centering (LC), and / or other features and functions.

[0165] The ACC system can utilize RADAR sensors 1260, LiDAR sensors 1264, and / or cameras. The ACC system can include longitudinal ACC and / or lateral ACC. Longitudinal ACC monitors and controls the distance to the vehicle immediately in front of vehicle 1200, automatically adjusting the vehicle speed to maintain a safe distance from the vehicle in front. Lateral ACC maintains distance and, when necessary, recommends that vehicle 1200 change lanes. Lateral ACC is related to other ADAS applications such as LCA and CWS.

[0166] CACC uses information from other vehicles, which can be received from other vehicles indirectly via a wireless link via the network interface 1224 and / or the wireless antenna 1226 or via a network connection (e.g., via the Internet). A direct link can be provided by a vehicle-to-vehicle (V2V) communication link, while an indirect link can be an infrastructure-to-vehicle (I2V) communication link. Typically, the V2V communication concept provides information about the vehicle immediately ahead (e.g., the vehicle immediately ahead of the vehicle 1200 and in the same lane as it), while the I2V communication concept provides information about traffic further ahead. The CACC system can include either or both of the I2V and V2V information sources. Given information about the vehicle ahead of the vehicle 1200, CACC can be more reliable, and it has the potential to improve the smoothness of traffic flow and reduce road congestion.

[0167] The FCW system is designed to alert the driver to hazards so that the driver can take corrective action. The FCW system uses a front-facing camera and / or RADAR sensor 1260 coupled to a dedicated processor, DSP, FPGA, and / or ASIC, which is electrically coupled to driver feedback such as a display, speaker, and / or vibrating component. The FCW system can provide warnings in the form of, for example, audible, visual warnings, vibrations, and / or rapid brake pulses.

[0168] The AEB system detects an impending forward collision with another vehicle or other object and can automatically apply the brakes if the driver does not take corrective action within specified time or distance parameters. The AEB system can use a front-facing camera and / or RADAR sensor 1260 coupled to a dedicated processor, DSP, FPGA and / or ASIC. When the AEB system detects a hazard, it typically first alerts the driver to take corrective action to avoid the collision, and if the driver does not take corrective action, the AEB system can automatically apply the brakes in an effort to prevent or at least mitigate the effects of the predicted collision. The AEB system can include technologies such as dynamic brake support and / or collision approach braking.

[0169] The LDW system provides visual, audible, and / or tactile warnings, such as steering wheel or seat vibrations, to alert the driver when the vehicle 1200 crosses a lane marking. When the driver indicates an intention to leave the lane by activating a turn signal, the LDW system is deactivated. The LDW system may utilize a front-facing camera coupled to a dedicated processor, DSP, FPGA, and / or ASIC that is electrically coupled to driver feedback such as a display, speaker, and / or vibration components.

[0170] The LKA system is a variation of the LDW system. If the vehicle 1200 begins to leave its lane, the LKA system provides steering input or braking to correct the vehicle 1200.

[0171] The BSW system detects and warns the driver of vehicles in the car's blind spot. The BSW system can provide visual, audible, and / or tactile alerts to indicate that 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 1260 coupled to a dedicated processor, DSP, FPGA, and / or ASIC that is electrically coupled to driver feedback such as a display, speaker, and / or vibration component.

[0172] The RCTW system can provide visual, audible, and / or tactile notifications when an object is detected outside the range of the rear-mounted camera while the vehicle 1200 is in reverse. Some RCTW systems include automatic emergency braking (AEB) to ensure that the vehicle brakes are applied to avoid a collision. The RCTW system can use one or more rear-mounted RADAR sensors 1260 coupled to a dedicated processor, DSP, FPGA, and / or ASIC that is electrically coupled to driver feedback such as a display, speaker, and / or vibration component.

[0173] Conventional ADAS systems can be prone to false positive results, which can be annoying and distracting to the driver, but are typically not catastrophic because the ADAS system alerts the driver and allows the driver to decide whether a safety condition actually exists and take action accordingly. However, in the autonomous vehicle 1200, in the event of conflicting results, the vehicle 1200 itself must decide whether to heed the results from the primary computer or the auxiliary computer (e.g., the first controller 1236 or the second controller 1236). For example, in some embodiments, the ADAS system 1238 can be a backup and / or auxiliary computer for providing perception information to the backup computer rationality module. The backup computer rationality monitor can run redundant and diverse software on hardware components to detect failures in perception and dynamic driving tasks. The output from the ADAS system 1238 can be provided to the supervisory MCU. If the outputs from the primary and auxiliary computers conflict, the supervisory MCU must determine how to reconcile the conflict to ensure safe operation.

[0174] In some examples, the primary computer can be configured to provide a confidence score to the supervisory MCU, indicating the primary computer's confidence in the selected result. If the confidence score exceeds a threshold, the supervisory MCU can follow the primary computer's direction, regardless of whether the secondary computer provides conflicting or inconsistent results. In the event that the confidence score does not meet the threshold and the primary and secondary computers indicate different results (e.g., a conflict), the supervisory MCU can arbitrate between these computers to determine the appropriate result.

[0175] The supervisory MCU can be configured to run a neural network that is trained and configured to determine conditions under which the secondary computer provides a false alarm based, at least in part, 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. For example, when the secondary computer is a RADAR-based FCW system, the neural network in the supervisory MCU can learn when the FCW system is identifying a metal object that is not actually a danger, such as a drain 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 disregard the LDW when a cyclist or pedestrian is present and a lane departure is actually the safest strategy. In embodiments that include a neural network running on the supervisory MCU, the supervisory MCU can include at least one of a DLA or a GPU suitable for running the neural network with associated memory. In a preferred embodiment, the supervisory MCU can include and / or be included as a component of the SoC 1204.

[0176] In other examples, the ADAS system 1238 can include an auxiliary computer that uses traditional computer vision rules to perform ADAS functions. In this way, the auxiliary computer can use classic computer vision rules (if-then), and the presence of a neural network in the supervisory MCU can improve reliability, safety, and performance. For example, diverse implementations and intentional non-identity make the entire system more fault-tolerant, especially with respect to failures caused by software (or software-hardware interface) functions. For example, if there is a software vulnerability or bug in the software running on the main computer and the non-identical software code running on the auxiliary computer provides the same overall result, the supervisory MCU can be more confident that the overall result is correct and that the vulnerability in the software or hardware on the main computer did not cause a substantial error.

[0177] In some examples, the output of the ADAS system 1238 can be fed into the primary computer's perception block and / or the primary computer's dynamic driving task block. For example, if the ADAS system 1238 indicates a forward collision warning due to an object immediately ahead, the perception block can use this information when identifying the object. In other examples, the secondary computer can have its own neural network that is trained and thus reduces the risk of false positives as described herein.

[0178] The vehicle 1200 may also include an infotainment SoC 1230 (e.g., an in-vehicle infotainment system (IVI)). Although illustrated and described as an SoC, the infotainment system may not be an SoC and may include two or more separate components. The infotainment SoC 1230 may include a combination of hardware and software that can be used to provide audio (e.g., music, personal digital assistant, navigation instructions, news, radio, etc.), video (e.g., TV, movies, streaming, etc.), phone (e.g., hands-free calling), network connectivity (e.g., LTE, WiFi, etc.), and / or information services (e.g., a navigation system, rear parking assistance, radio data system, vehicle-related information such as fuel level, total distance covered, brake fuel level, oil level, door open / closed, air filter information, etc.) to the vehicle 1200. For example, the infotainment SoC 1230 may include a radio, a disc player, a navigation system, a video player, USB and Bluetooth connectivity, an onboard computer, in-car entertainment, WiFi, steering wheel audio controls, hands-free voice controls, a head-up display (HUD), an HMI display 1234, 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 1230 may further be used to provide information (e.g., visual and / or auditory) to a user of the vehicle, such as information from an ADAS system 1238, autonomous driving information such as planned vehicle maneuvers, trajectories, surrounding environment information (e.g., intersection information, vehicle information, road information, etc.), and / or other information.

[0179] The infotainment SoC 1230 may include GPU functionality. The infotainment SoC 1230 may communicate with other devices, systems, and / or components of the vehicle 1200 via a bus 1202 (e.g., a CAN bus, Ethernet, etc.). In some examples, the infotainment SoC 1230 may be coupled to a supervisory MCU so that in the event of a failure of a primary controller 1236 (e.g., a primary and / or backup computer of the vehicle 1200), the infotainment system's GPU may perform some self-driving functions. In such an example, the infotainment SoC 1230 may place the vehicle 1200 in a driver-safe parking mode as described herein.

[0180] The vehicle 1200 may also include an instrument cluster 1232 (e.g., a digital instrument panel, an electronic instrument cluster, a digital instrument panel, etc.). The instrument cluster 1232 may include a controller and / or a supercomputer (e.g., a separate controller or a supercomputer). The instrument cluster 1232 may include a set of instruments, such as a speedometer, fuel level, oil pressure, a tachometer, an odometer, a turn indicator, a gear position indicator, a seat belt warning light, a parking brake warning light, an engine check light, airbag (SRS) system information, lighting controls, safety system controls, navigation information, etc. In some examples, information may be displayed and / or shared between the infotainment SoC 1230 and the instrument cluster 1232. Thus, the instrument cluster 1232 may be included as part of the infotainment SoC 1230, or vice versa.

[0181] Figure 12D For cloud-based servers and Figure 12A 12. System diagram of communication between an example autonomous vehicle 1200. System 1276 may include a server 1278, a network 1290, and a vehicle including vehicle 1200. Server 1278 may include multiple GPUs 1284(A)-1284(H) (collectively referred to herein as GPUs 1284), PCIe switches 1282(A)-1282(D) (collectively referred to herein as PCIe switches 1282), and / or CPUs 1280(A)-1280(B) (collectively referred to herein as CPUs 1280). GPUs 1284, CPUs 1280, and PCIe switches may be interconnected with a high-speed interconnect such as, for example and without limitation, NVLink interface 1288 developed by NVIDIA and / or PCIe connection 1286. In some examples, GPUs 1284 are connected via NVLink and / or NVSwitch SoCs, and GPUs 1284 and PCIe switches 1282 are connected via PCIe interconnects. Although eight GPUs 1284, two CPUs 1280, and two PCIe switches are shown, this is not intended to be limiting. Depending on the embodiment, each of the servers 1278 may include any number of GPUs 1284, CPUs 1280, and / or PCIe switches. For example, each of the servers 1278 may include eight, sixteen, thirty-two, and / or more GPUs 1284.

[0182] Server 1278 can receive image data from a vehicle via network 1290, the image data representing images showing unexpected or changed road conditions, such as recently begun road construction. Server 1278 can transmit neural network 1292, updated neural network 1292, and / or map information 1294, including information about traffic and road conditions, via network 1290 and to the vehicle. Updates to map information 1294 can include updates to HD map 1222, such as information about construction sites, potholes, curves, flooding, or other obstacles. In some examples, neural network 1292, updated neural network 1292, and / or map information 1294 can be generated from new training and / or data received from any number of vehicles in the environment and / or based on experience from training performed at a data center (e.g., using server 1278 and / or other servers).

[0183] Server 1278 can be used to train a machine learning model (e.g., a neural network) based on training data. The training data can be generated using the vehicle and / or can be generated in simulation (e.g., using a game engine). In some examples, the training data is labeled (e.g., in cases where the neural network benefits from supervised learning) and / or undergoes other preprocessing, while in other examples, the training data is not labeled and / or preprocessed (e.g., in cases where the neural network does not require supervised learning). Training can be performed according to any one or more categories of machine learning techniques, including but not limited to the following categories: supervised training, semi-supervised training, unsupervised training, self-learning, reinforcement learning, joint learning, transfer learning, feature learning (including principal component and cluster analysis), multilinear subspace learning, manifold learning, representation learning (including alternative dictionary learning), rule-based machine learning, anomaly detection, and any variants or combinations thereof. Once the machine learning model is trained, the machine learning model can be used by the vehicle (e.g., transmitted to the vehicle via network 1290), and / or the machine learning model can be used by server 1278 to remotely monitor the vehicle.

[0184] In some examples, server 1278 can receive data from the vehicle and apply the data to the latest real-time neural network for real-time intelligent reasoning. Server 1278 can include a deep learning supercomputer and / or a dedicated AI computer powered by GPU 1284, such as the DGX and DGX Station machines developed by NVIDIA. However, in some examples, server 1278 can include the deep learning infrastructure of a data center using only CPU power.

[0185] The deep learning infrastructure of server 1278 may be capable of rapid real-time inference and may use this capability to assess and verify the health of the processors, software, and / or associated hardware in vehicle 1200. For example, the deep learning infrastructure may receive periodic updates from vehicle 1200, such as an image sequence and / or objects located in the image sequence that vehicle 1200 has located (e.g., via computer vision and / or other machine learning object classification techniques). The deep learning infrastructure may run its own neural networks to identify objects and compare them to the objects identified by vehicle 1200, and if the results do not match and the infrastructure concludes that the AI ​​in vehicle 1200 has malfunctioned, server 1278 may transmit a signal to vehicle 1200 instructing the vehicle's 1200 fail-safe computer to take control, notify passengers, and complete a safe parking maneuver.

[0186] For inference, server 1278 may include a GPU 1284 and one or more programmable inference accelerators (e.g., NVIDIA's TensorRT). The combination of GPU-powered servers and inference acceleration can enable real-time responses. In other examples, such as where performance is less important, CPU, FPGA, and other processor-powered servers can be used for inference.

[0187] Example computing device

[0188] Figure 13 13 is a block diagram of an example computing device 1300 suitable for implementing some embodiments of the present disclosure. Computing device 1300 may include an interconnect system 1302 that directly or indirectly couples the following devices: memory 1304, one or more central processing units (CPUs) 1306, one or more graphics processing units (GPUs) 1308, a communication interface 1310, input / output (I / O) ports 1312, I / O components 1314, a power supply 1316, one or more presentation components 1318 (e.g., display(s)), and one or more logic units 1320. In at least one embodiment, computing device(s) 1300 may include one or more virtual machines (VMs), and / or any of its components may include virtual components (e.g., virtual hardware components). For non-limiting examples, one or more of GPUs 1308 may include one or more vGPUs, one or more of CPUs 1306 may include one or more vCPUs, and / or one or more of logic units 1320 may include one or more virtual logic units. As such, computing device(s) 1300 may include discrete components (e.g., a full GPU dedicated to computing device 1300), virtual components (e.g., a portion of a GPU dedicated to computing device 1300), or a combination thereof.

[0189] although Figure 13 The various blocks of are shown as being connected via interconnect system 1302 using wires, but this is not intended to be limiting and is provided for clarity only. For example, in some embodiments, presentation component 1318 (such as a display device) may be considered to be I / O component 1314 (e.g., if the display is a touch screen). As another example, CPU 1306 and / or GPU 1308 may include memory (e.g., memory 1304 may represent a storage device in addition to the memory of GPU 1308, CPU 1306, and / or other components). Thus, Figure 13 The computing devices are illustrative only. No distinction is made between such categories as "workstation," "server," "laptop," "desktop," "tablet," "client device," "mobile device," "handheld device," "game console," "electronic control unit (ECU)," "virtual reality system," and / or other device or system types, as all are considered Figure 13 within the range of computing devices.

[0190] Interconnect system 1302 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 1302 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, there is a direct connection between components. As an example, CPU 1306 may be directly connected to memory 1304. Further, CPU 1306 may be directly connected to GPU 1308. In the case where there is a direct or point-to-point connection between components, interconnect system 1302 may include a PCIe link to perform the connection. In these examples, the PCI bus need not be included in computing device 1300.

[0191] Memory 1304 may include any of a variety of computer-readable media. Computer-readable media can be any available media that can be accessed by computing device 1300. Computer-readable media can include volatile and non-volatile media, as well as removable and non-removable media. By way of example and not limitation, computer-readable media can include computer storage media and communication media.

[0192] Computer storage media may include 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 1304 may store computer-readable instructions (e.g., representing (one or more) programs and / or (one or more) 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 disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage devices or other magnetic storage devices, or any other medium that can be used to store the desired information and that can be accessed by computing device 1300. As used herein, computer storage media does not include signals themselves.

[0193] Computer storage media can 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 includes 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 can 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.

[0194] The CPU 1306 may be configured to execute at least some of the computer-readable instructions to control one or more components of the computing device 1300 to perform one or more of the methods and / or processes described herein. The CPUs 1306 may each include one or more cores (e.g., one, two, four, eight, twenty-eight, seventy-two, etc.) capable of handling numerous software threads simultaneously. The CPU 1306 may include any type of processor and may include different types of processors depending on the type of computing device 1300 being 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 1300, the processor may be an Advanced RISC Machine (ARM) processor implemented using Reduced Instruction Set Computing (RISC) or an x86 processor implemented using Complex Instruction Set Computing (CISC). The computing device 1300 may also include one or more CPUs 1306 in addition to one or more microprocessors or supplemental coprocessors (such as a math coprocessor).

[0195] In addition to or in lieu of CPU(s) 1306, GPU(s) 1308 may be configured to execute at least some of the computer-readable instructions to control one or more components of computing device 1300 to perform one or more of the methods and / or processes described herein. One or more of GPUs 1308 may be integrated GPUs (e.g., with one or more of CPUs 1306) and / or one or more of GPUs 1308 may be discrete GPUs. In embodiments, one or more of GPUs 1308 may be coprocessors for one or more of CPUs 1306. GPU 1308 may be used by computing device 1300 to render graphics (e.g., 3D graphics) or perform general-purpose computations. For example, GPU 1308 may be used for general-purpose computing on a GPU (GPGPU). GPU 1308 may include hundreds or thousands of cores capable of handling hundreds or thousands of software threads simultaneously. GPU 1308 may generate pixel data for an output image in response to rendering commands (e.g., rendering commands received from CPU 1306 via a host interface). GPU 1308 may include graphics memory (e.g., display memory) for storing pixel data or any other suitable data (e.g., GPGPU data). Display memory may be included as part of memory 1304. GPU 1308 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 through a switch (e.g., using NVSwitch). When combined, each GPU 1308 may generate pixel data or GPGPU data for a different portion of an output or for different outputs (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.

[0196] In addition to or in lieu of the CPU 1306 and / or GPU 1308, the logic unit 1320 may be configured to execute at least some of the computer-readable instructions to control one or more components of the computing device 1300 to perform one or more of the methods and / or processes described herein. In embodiments, the CPU(s) 1306, the GPU(s) 1308, and / or the logic unit(s) 1320 may execute any combination of methods, processes, and / or portions thereof, either discretely or jointly. One or more of the logic units 1320 may be part of and / or integrated into one or more of the CPU 1306 and / or GPU 1308, and / or one or more of the logic units 1320 may be discrete components or otherwise external to the CPU 1306 and / or GPU 1308. In embodiments, one or more of logic units 1320 may be a co-processor to one or more of CPUs 1306 and / or one or more of GPUs 1308 .

[0197] Examples of logic unit 1320 include one or more processing cores and / or components thereof, such as a data processing unit (DPU), a tensor core (TC), a tensor processing unit (TPU), a pixel vision core (PVC), a vision processing unit (VPU), a graphics processing cluster (GPC), a texture processing cluster (TPC), a streaming multiprocessor (SM), a tree transverse 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 a peripheral component interconnect express (PCIe) element, etc.

[0198] The communication interface 1310 may include one or more receivers, transmitters, and / or transceivers that enable the computing device 1300 to communicate with other computing devices via an electronic communication network (including wired and / or wireless communications). The communication interface 1310 may include components and functionality that implement communication over any of a number of different networks, such as a wireless network (e.g., Wi-Fi, Z-Wave, Bluetooth, Bluetooth LE, ZigBee, etc.), a wired network (e.g., via Ethernet or Wi-Fi), a low-power wide-area network (e.g., LoRaWAN, SigFox, etc.), and / or the Internet. In one or more embodiments, the logic unit 1320 and / or the communication interface 1310 may include one or more data processing units (DPUs) for transmitting data received over the network and / or over the interconnect system 1302 directly to (e.g., memory of) one or more GPUs 1308.

[0199] I / O ports 1312 can enable computing device 1300 to be logically coupled to other devices including I / O components 1314, presentation component(s) 1318, and / or other components, some of which may be built into (e.g., integrated into) computing device 1300. Illustrative I / O components 1314 include a microphone, mouse, keyboard, joystick, game pad, game controller, satellite dish, scanner, printer, wireless device, and the like. I / O components 1314 can provide a natural user interface (NUI) that processes air gestures, voice, or other physiological input generated by the user. In some cases, the input can be transmitted to an appropriate network element for further processing. The NUI can implement any combination of voice recognition, stylus recognition, facial recognition, biometric recognition, gesture recognition on and near the screen, air gestures, head and eye tracking, and touch recognition (as described in more detail below) associated with the display of computing device 1300. Computing device 1300 may include a depth camera for gesture detection and recognition, such as a stereo camera system, an infrared camera system, an RGB camera system, touch screen technology, and combinations thereof. Additionally, computing device 1300 may include an accelerometer or gyroscope that enables detection of motion (e.g., as part of an inertial measurement unit (IMU)). In some examples, computing device 1300 may use the output of the accelerometer or gyroscope to render immersive augmented or virtual reality.

[0200] The power supply 1316 may include a hardwired power supply, a battery power supply, or a combination thereof. The power supply 1316 may provide power to the computing device 1300 to enable the components of the computing device 1300 to operate.

[0201] The presentation component 1318 may include a display (e.g., a monitor, a touch screen, a television screen, a head-up display (HUD), other display types, or a combination thereof), speakers, and / or other presentation components. The presentation component 1318 may receive data from other components (e.g., the GPU 1308, the CPU 1306, the DPU, etc.) and output the data (e.g., as images, video, sound, etc.).

[0202] Sample Data Center

[0203] Figure 14 An example data center 1400 that may be used in at least one embodiment of the present disclosure is shown. The data center 1400 may include a data center infrastructure layer 1410, a framework layer 1420, a software layer 1430, and / or an application layer 1440.

[0204] like Figure 14 As shown, the data center infrastructure layer 1410 may include a resource coordinator 1412, grouped computing resources 1414, and node computing resources ("node CRs") 1416(1)-1416(N), where "N" represents any complete positive integer. In at least one embodiment, the node CRs 1416(1)-1416(N) may include, but is not limited to, any number of central processing units (CPUs) or other processors (including DPUs, accelerators, field programmable gate arrays (FPGAs), graphics processors or graphics processing units (GPUs), etc.), memory devices (e.g., dynamic read-only memories), 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 node CRs from the node CRs 1416(1)-1416(N) may correspond to a server having one or more of the above-mentioned computing resources. Furthermore, in some embodiments, node CRs 1416(1)-14161(N) may include one or more virtual components, such as vGPUs, vCPUs, etc., and / or one or more of node CRs 1416(1)-1416(N) may correspond to a virtual machine (VM).

[0205] In at least one embodiment, the grouped computing resources 1414 may include separate groups of node CRs 1416 housed in one or more racks (not shown), or multiple racks housed in data centers at different geographical locations (also not shown). Separate groups of node CRs 1416 within the grouped computing resources 1414 may include grouped computing, network, memory, or storage resources that can be configured or allocated to support one or more workloads. In at least one embodiment, several node CRs 1416 including CPUs, GPUs, DPUs, and / or other processors may be grouped in one or more racks to provide computing resources to support one or more workloads. One or more racks may also include any number of power modules, cooling modules, and / or network switches in any combination.

[0206] Resource coordinator 1412 may configure or otherwise control one or more node CRs 1416(1)-1416(N) and / or grouped computing resources 1414. In at least one embodiment, resource coordinator 1412 may comprise a software design infrastructure (SDI) management entity for data center 1400. Resource coordinator 1412 may comprise hardware, software, or some combination thereof.

[0207] In at least one embodiment, Figure 14 As shown, the framework layer 1420 may include a job scheduler 1433, a configuration manager 1434, a resource manager 1436 and / or a distributed file system 1438. The framework layer 1420 may include a framework that supports software 1432 of the software layer 1430 and / or one or more applications 1442 of the application layer 1440. The software 1432 or the application 1442 may include network-based service software or applications, such as those provided by Amazon Web Services, Google Cloud, and Microsoft Azure. The framework layer 1420 may be, but is not limited to, a free and open source software network application framework (such as Apache Spark) that can utilize the distributed file system 1438 for large-scale data processing (e.g., "big data"). TM(hereinafter referred to as "Spark"). In at least one embodiment, the job scheduler 1433 may include a Spark driver to facilitate scheduling workloads supported by the different layers of the data center 1400. The configuration manager 1434 may be capable of configuring the different layers, such as the software layer 1430 and the framework layer 1420 (which includes Spark and a distributed file system 1438 for supporting large-scale data processing). The resource manager 1436 may be capable of managing clustered or grouped computing resources that are mapped to the distributed file system 1438 and the job scheduler 1433 or allocated to support the distributed file system 1438 and the job scheduler 1433. In at least one embodiment, the clustered or grouped computing resources may include the grouped computing resources 1414 at the data center infrastructure layer 1410. The resource manager 1436 may coordinate with the resource coordinator 1412 to manage these mapped or allocated computing resources.

[0208] In at least one embodiment, the software 1432 included in the software layer 1430 may include software used by at least a portion of the node CRs 1416(1)-1416(N), the grouped computing resources 1414, and / or the distributed file system 1438 of the framework layer 1420. The one or more types of software may include, but are not limited to, Internet web search software, email virus scanning software, database software, and streaming video content software.

[0209] In at least one embodiment, the applications 1442 included in the application layer 1440 may include one or more types of applications used by at least a portion of the node CRs 1416(1)-1416(N), the grouped computing resources 1414, and / or the distributed file system 1438 of the framework layer 1420. The one or more types of applications may include, but are not limited to, any number of genomic 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.

[0210] In at least one embodiment, any of configuration manager 1434, resource manager 1436, and resource coordinator 1412 can implement any number and type of self-modification actions based on any amount and type of data acquired in any technically feasible manner. The self-modification actions can save a data center operator of data center 1400 from making potentially poor configuration decisions and potentially avoiding underutilized and / or poorly performing portions of the data center.

[0211] According to one or more embodiments described herein, data center 1400 may include tools, services, software, or other resources to train one or more machine learning models or to predict or infer information using one or more machine learning models. For example, the machine learning model(s) may be trained by computing weight parameters according to a neural network architecture using the software and / or computing resources described above with respect to data center 1400. 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 1400 using weight parameters computed using one or more training techniques, such as, but not limited to, those described herein.

[0212] In at least one embodiment, data center 1400 may use a CPU, an application-specific integrated circuit (ASIC), a GPU, an FPGA, and / or other hardware (or virtual computing resources corresponding thereto) to perform training and / or inference using the aforementioned resources. In addition, one or more of the software and / or hardware resources described above may be configured to allow a user to train or perform inference services on information, such as image recognition, speech recognition, or other artificial intelligence services.

[0213] Sample network environment

[0214] A network environment suitable for implementing embodiments of the present disclosure may include one or more client devices, servers, network attached storage (NAS), other backend devices, and / or other device types. The client devices, servers, and / or other device types (e.g., each device) may be configured to: Figure 13 The backend devices 1300 may be implemented on one or more instances of the computing device(s) 1300 - for example, each device may include similar components, features, and / or functionality of the computing device(s) 1300. In addition, in the case of implementing backend devices (e.g., servers, NAS, etc.), the backend devices may be included as part of the data center 1400, examples of which are discussed herein with respect to FIG. Figure 14 Describe in more detail.

[0215] The components of the network environment can communicate with each other via a network, which can be wired, wireless, or both. The network can include multiple networks or one of multiple networks. For example, the network can 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. In the case where the network includes a wireless telecommunications network, components such as base stations, communication towers, or even access points (and other components) can provide wireless connectivity.

[0216] Compatible network environments may include one or more peer-to-peer network environments (in which case the server may not be included in the network environment) and one or more client-server network environments (in which case one or more servers may be included in the network environment). In a peer-to-peer network environment, the functionality described herein for the server may be implemented on any number of client devices.

[0217] In at least one embodiment, the network environment may include one or more cloud-based network environments, distributed computing environments, combinations thereof, and the like. 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 servers, which may include one or more core network servers and / or edge servers. The framework layer may include software supporting the software layer and / or a framework for one or more applications at the application layer. The software or application may include network-based service software or applications, respectively. In an embodiment, one or more client devices may use network-based service software or applications (e.g., by accessing the service software and / or application via one or more application programming interfaces (APIs)). The framework layer may be, but is not limited to, a free and open source software network application framework that can use a distributed file system for large-scale data processing (e.g., "big data").

[0218] A cloud-based network environment can provide cloud computing and / or cloud storage that performs any combination of the computing and / or data storage functions described herein (or one or more portions thereof). Any of these different functions can be distributed across multiple locations from a central or core server (e.g., one or more data centers that can be distributed across a state, region, country, global, etc.). If the connection to the user (e.g., client device) is relatively close to an edge server, the core server can assign at least a portion of the functionality to the edge server. A cloud-based network environment can be private (e.g., limited to a single organization), public (e.g., available to many organizations), and / or a combination thereof (e.g., a hybrid cloud environment).

[0219] The client device(s) may include the Figure 13At least some of the components, features, and functionality of the described example computing device(s) 1300. By way of example and not limitation, the client device may be implemented as a personal computer (PC), a laptop computer, a mobile device, a smartphone, a tablet computer, a smartwatch, 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 camera, a surveillance device or system, a vehicle, a boat, a spacecraft, a virtual machine, a drone, a robot, a handheld communication device, a hospital device, a gaming device or system, an entertainment system, a vehicle computer system, an embedded system controller, a remote control, an appliance, a consumer electronic device, a workstation, an edge device, any combination of the depicted devices, or any other suitable device.

[0220] The present disclosure can be described in the general context of machine-usable instructions or computer code executed by a computer or other machine such as a personal digital assistant or other handheld device, including computer-executable instructions such as program modules. Generally, program modules including routines, programs, objects, components, data structures, etc. refer to code that performs a specific task or implements a specific abstract data type. The present disclosure can be practiced in a variety of system configurations, including handheld devices, consumer electronics, general-purpose computers, more specialized computing devices, etc. The present disclosure can also be practiced in a distributed computing environment where tasks are performed by remote processing devices linked through a communication network.

[0221] As used herein, the phrase "and / or" with respect to two or more elements should be interpreted as referring to only one element or combination of elements. For example, "element A, element B, and / or element C" may include only element A, only element B, only element C, element A and element B, element A and element C, element B and element C, or elements A, B, and C. In addition, "at least one of element A or element B" may include at least one of element A, at least one of element B, or at least one of element A and at least one of element B. Further, "at least one of element A and element B" may include at least one of element A, at least one of element B, or at least one of element A and at least one of element B.

[0222] The subject matter of the present disclosure is described in detail herein to meet statutory requirements. However, the description itself is not intended to limit the scope of the present disclosure. On the contrary, the present inventors have contemplated that the claimed subject matter may also be embodied in other ways to include steps that are different from the steps described herein in conjunction with other current or future technologies, or combinations of similar steps. Moreover, although the terms "step" and / or "block" may be used herein to imply different elements of the method employed, these terms should not be interpreted as implying any particular order among or between the various steps disclosed herein, unless the order of the steps is explicitly described.

[0223] Sample text support

[0224] In an example embodiment, one or more processors include: processing circuitry for: generating a three-dimensional (3D) surface topology of an environment based at least on sensor data generated by one or more sensors of an ego machine in the environment; and generating a visualization of the environment based at least on texturing the detected 3D surface topology using the sensor data.

[0225] In some embodiments, the processing circuit is further configured to: estimate depth data based on at least a stack of images representing a common time slice from different perspectives; and generate the detected 3D surface topology based on at least the depth data.

[0226] In some embodiments, the processing circuit is further configured to: estimate depth data based at least on stacked images representing a common perspective in different time slices from different perspectives; and generate the detected 3D surface topology based at least on the depth data.

[0227] In some embodiments, the processing circuit is further configured to represent the detected 3D surface topology of the environment using a 3D signed distance function.

[0228] In some embodiments, the processing circuit is further configured to represent the detected 3D surface topology of the environment using a 3D signed distance function truncated to a sphere having a specified radius.

[0229] In some embodiments, the processing circuit is further configured to generate the detected 3D surface topology of the environment based at least on a plurality of depth maps representing overlapping views of the environment.

[0230] In some embodiments, the processing circuit is further configured to detect and fill one or more holes in the detected 3D surface topology of the environment.

[0231] In some embodiments, the processing circuit is further configured to apply smoothing to the detected 3D surface topology.

[0232] In some embodiments, the processing circuit is further configured to: dewarp the sensor data using depth data represented by the detected 3D surface topology of the environment.

[0233] In some embodiments, the processing circuit is further configured to texturize the detected one or more white space regions of the 3D surface topology based at least on applying blur.

[0234] In some embodiments, the one or more processors are included in at least one of: a control system for an autonomous or semi-autonomous machine; a perception system for an autonomous or semi-autonomous machine; a system for performing simulation operations; a system for performing digital twin operations; a system for performing light transport simulations; a system for performing collaborative content creation of 3D assets; a system for performing deep learning operations; a system for performing real-time streaming; a system for generating or presenting one or more of augmented reality content, virtual reality content, or mixed reality content; a system implemented using an edge device; a system implemented using a robot; a system for performing conversational AI operations; a system for generating synthetic data; a system for generating synthetic data using AI; a system comprising one or more virtual machines (VMs); a system implemented at least in part in a data center; or a system implemented at least in part using cloud computing resources.

[0235] In some embodiments, a system includes one or more processors configured to generate a visualization of an environment based at least on texturing a detected three-dimensional (3D) surface topology of the environment using sensor data generated by one or more sensors of a self-machine in the environment.

[0236] In an example embodiment, the one or more processors are further configured to represent the detected 3D surface topology of the environment using a 3D signed distance function.

[0237] In an example embodiment, the one or more processors are further configured to generate the detected 3D surface topology of the environment based on at least a plurality of depth maps representing overlapping views of the environment.

[0238] In an example embodiment, the one or more processors are further configured to detect and fill one or more holes in the detected 3D surface topology of the environment.

[0239] In an example embodiment, the one or more processors are further configured to apply smoothing to the detected 3D surface topology.

[0240] In an example embodiment, the one or more processors are further configured to dewarp the sensor data using depth data represented by the detected 3D surface topology of the environment.

[0241] In an example embodiment, the system is included in at least one of: a control system for an autonomous or semi-autonomous machine; a perception system for an autonomous or semi-autonomous machine; a system for performing simulation operations; a system for performing digital twin operations; a system for performing light transport simulations; a system for performing collaborative content creation of 3D assets; a system for performing deep learning operations; a system for performing real-time streaming; a system for generating or presenting one or more of augmented reality content, virtual reality content, or mixed reality content; a system implemented using an edge device; a system implemented using a robot; a system for performing conversational AI operations; a system for generating synthetic data; a system for generating synthetic data using AI; a system comprising one or more virtual machines (VMs); a system implemented at least in part in a data center; or a system implemented at least in part using cloud computing resources.

[0242] In an embodiment, a method includes: generating a detected three-dimensional (3D) surface topology of an environment based at least on image data generated using one or more cameras of an ego machine in the environment; and generating a visualization of the environment based at least on projecting the image data into the detected 3D surface topology.

[0243] In an embodiment, the method is performed by at least one of: a control system for an autonomous or semi-autonomous machine; a perception system for an autonomous or semi-autonomous machine; a system for performing simulation operations; a system for performing digital twin operations; a system for performing light transport simulations; a system for performing collaborative content creation of 3D assets; a system for performing deep learning operations; a system for performing real-time streaming; a system for generating or presenting one or more of augmented reality content, virtual reality content, or mixed reality content; a system implemented using an edge device; a system implemented using a robot; a system for performing conversational AI operations; a system for generating synthetic data; a system for generating synthetic data using AI; a system comprising one or more virtual machines (VMs); a system implemented at least in part in a data center; or a system implemented at least in part using cloud computing resources.

[0244] In an example embodiment, one or more processors include: processing circuitry for: generating a detected 3D surface topology of the environment based at least on sensor data generated using one or more sensors of a self-machine in the environment, and based at least on one or more three-dimensional (3D) representations of one or more detected dynamic objects in the environment; and generating a visualization of the environment based at least on texturing the one or more 3D representations of the one or more detected dynamic objects in the detected 3D surface topology using the sensor data.

[0245] In an example embodiment, the processing circuit is further configured to mask the one or more detected dynamic objects during the first pass of texturing the detected 3D surface topology.

[0246] In an example embodiment, the processing circuit is further used to: generate a first detected 3D surface topology of the environment representing a static portion of the environment; and update the first detected 3D surface topology based at least on inserting the one or more 3D representations of the one or more detected dynamic objects into the first detected 3D surface topology.

[0247] In an example embodiment, the processing circuit is further configured to generate one or more 3D representations of one or more detected dynamic objects in one or more rigid object classes based at least on warping one or more detected depth values ​​of the one or more detected dynamic objects using one or more detected trajectories of the one or more detected dynamic objects.

[0248] In an example embodiment, the processing circuit is further used to: fuse at least a first 3D representation of at least a first detected dynamic object among the one or more detected dynamic objects generated based at least on the following steps into the detected 3D surface topology: tracking a trajectory of the first detected dynamic object; identifying one or more detected depth values ​​representing the first detected object in a previous time slice; and warping the one or more detected depth values ​​using the trajectory.

[0249] In an example embodiment, the processing circuit is further configured to generate one or more 3D representations of one or more detected dynamic objects of one or more non-rigid object classes based at least on inserting a 3D representation of a two-dimensional (2D) surface at a location in the detected 3D surface topology corresponding to a detected location of the first detected object in the environment for at least a first detected dynamic object of the one or more detected dynamic objects.

[0250] In an example embodiment, the processing circuit is further configured to fuse at least a first 3D representation of a flat surface at a location corresponding to a detected centroid of a corresponding one of the one or more detected dynamic objects into the detected 3D surface topology.

[0251] In an example embodiment, wherein texturing the one or more 3D representations of the one or more detected dynamic objects uses a set of segmented sensor data classified as belonging to the one or more detected dynamic objects.

[0252] In an example embodiment, the one or more processors are included in at least one of: a control system for an autonomous or semi-autonomous machine; a perception system for an autonomous or semi-autonomous machine; a system for performing simulation operations; a system for performing digital twin operations; a system for performing light transport simulations; a system for performing collaborative content creation of 3D assets; a system for performing deep learning operations; a system for performing real-time streaming; a system for generating or presenting one or more of augmented reality content, virtual reality content, or mixed reality content; a system implemented using an edge device; a system implemented using a robot; a system for performing conversational AI operations; a system for generating synthetic data; a system for generating synthetic data using AI; a system comprising one or more virtual machines (VMs); a system implemented at least in part in a data center; or a system implemented at least in part using cloud computing resources.

[0253] In an embodiment, a system includes one or more processors configured to generate a visualization of an environment by texturing one or more 3D representations of one or more detected dynamic objects in a detected three-dimensional (3D) surface topology based at least on sensor data generated by one or more sensors of a self-machine in the environment.

[0254] In an embodiment, the one or more processors are further configured to mask the one or more detected dynamic objects during the first pass of texturing the detected 3D surface topology.

[0255] In an embodiment, the one or more processors are further configured to: generate a first detected 3D surface topology of the environment representing a static portion of the environment; and update the first detected 3D surface topology based at least on inserting the one or more 3D representations of the one or more detected dynamic objects into the first detected 3D surface topology.

[0256] In an embodiment, the one or more processors are further configured to generate one or more 3D representations of the one or more detected dynamic objects based at least on warping one or more detected depth values ​​of the one or more detected dynamic objects using one or more detected trajectories of the one or more detected dynamic objects in one or more rigid object classes.

[0257] In an embodiment, the one or more processors are further configured to: fuse at least a first 3D representation of at least a first detected dynamic object among the one or more detected dynamic objects generated based at least on the following steps into the detected 3D surface topology: tracking a trajectory of the first detected dynamic object; identifying one or more detected depth values ​​representing the first detected object in a previous time slice; and warping the one or more detected depth values ​​using the trajectory.

[0258] In an embodiment, the one or more processors are further configured to generate one or more 3D representations of the one or more detected dynamic objects based at least on inserting a 3D representation of a two-dimensional (2D) surface of at least a first detected object of one or more detected dynamic objects of one or more non-rigid object classes at a position in the detected 3D surface topology corresponding to a detected position of the first detected object in the environment.

[0259] In an embodiment, the one or more processors are further configured to fuse at least a first 3D representation of a flat surface at a location corresponding to a detected centroid of a corresponding one of the one or more detected dynamic objects into the detected 3D surface topology.

[0260] In an embodiment, wherein texturing the one or more 3D representations of the one or more detected dynamic objects uses a set of segmented sensor data classified as belonging to the one or more detected dynamic objects.

[0261] In an embodiment, the system is included in at least one of: a control system for an autonomous or semi-autonomous machine; a perception system for an autonomous or semi-autonomous machine; a system for performing simulation operations; a system for performing digital twin operations; a system for performing light transport simulations; a system for performing collaborative content creation of 3D assets; a system for performing deep learning operations; a system for performing real-time streaming; a system for generating or presenting one or more of augmented reality content, virtual reality content, or mixed reality content; a system implemented using an edge device; a system implemented using a robot; a system for performing conversational AI operations; a system for generating synthetic data; a system for generating synthetic data using AI; a system comprising one or more virtual machines (VMs); a system implemented at least in part in a data center; or a system implemented at least in part using cloud computing resources.

[0262] In an embodiment, a method comprises: generating a detected 3D surface topology of an environment based at least on image data generated using one or more cameras of an ego machine in the environment, and based at least on one or more three-dimensional (3D) representations of one or more detected dynamic objects in the environment; and generating a visualization of the environment based at least on projecting the image data onto the one or more 3D representations of the one or more detected dynamic objects in the detected 3D surface topology.

[0263] In an embodiment, the method is performed by at least one of: a control system for an autonomous or semi-autonomous machine; a perception system for an autonomous or semi-autonomous machine; a system for performing simulation operations; a system for performing digital twin operations; a system for performing light transport simulations; a system for performing collaborative content creation of 3D assets; a system for performing deep learning operations; a system for performing real-time streaming; a system for generating or presenting one or more of augmented reality content, virtual reality content, or mixed reality content; a system implemented using an edge device; a system implemented using a robot; a system for performing conversational AI operations; a system for generating synthetic data; a system for generating synthetic data using AI; a system comprising one or more virtual machines (VMs); a system implemented at least in part in a data center; or a system implemented at least in part using cloud computing resources.

Claims

1. One or more processors, including: A processing circuit, the processing circuit being configured to: computing a 3D surface topology of the environment based at least on sensor data generated using one or more sensors of the ego machine in the environment and at least on one or more three-dimensional (3D) representations of one or more detected dynamic objects in the environment; as well as A visualization of the environment is generated based at least on generating graphical content for the one or more 3D representations in the 3D surface topology using the sensor data. 2 . The one or more processors of claim 1 , the processing circuitry further configured to mask the one or more detected dynamic objects during a first pass of generating the visualization.

3. The one or more processors of claim 1 , wherein the processing circuit is further configured to: computing a first 3D surface topology of the environment representing a static portion of the environment; and The first 3D surface topology is updated based at least on inserting the one or more 3D representations of the one or more detected dynamic objects into the first 3D surface topology.

4. The one or more processors of claim 1 , wherein the one or more detected dynamic objects comprise at least one rigid object from one or more rigid object classes, wherein the processing circuitry is further configured to generate one or more 3D representations of the at least one rigid object based at least on warping one or more detected depth values ​​corresponding to the at least one rigid object using one or more detected trajectories corresponding to the at least one rigid object.

5. One or more processors according to claim 1, the processing circuit is further used to: fuse at least a first 3D representation of at least a first detected dynamic object among the one or more detected dynamic objects generated based at least on the following operations into the 3D surface topology: tracking a trajectory of the first detected dynamic object; identifying one or more detected depth values ​​representing the first detected object in a previous time slice; and warping the one or more detected depth values ​​using the trajectory.

6. The one or more processors of claim 1 , wherein the one or more detected dynamic objects comprise at least one non-rigid object of one or more non-rigid object classes, and wherein the processing circuitry is further configured to generate one or more 3D representations of the at least one non-rigid object based at least on inserting a 3D representation of a two-dimensional (2D) surface for the at least one non-rigid object at a location in the 3D surface topology corresponding to a detected location of the at least one non-rigid object in the environment.

7. One or more processors according to claim 1, the processing circuit is further used to: fuse at least a first 3D representation of a flat surface at a position corresponding to the detected center of mass of a corresponding detected dynamic object in the one or more detected dynamic objects into the 3D surface topology.

8. The one or more processors of claim 1, wherein generating graphical content for the one or more 3D representations is based at least on a set of segmented sensor data classified as corresponding to the one or more detected dynamic objects.

9. The one or more processors of claim 1 , wherein the one or more processors are included in at least one of: control systems for autonomous or semi-autonomous machines; Perception systems for autonomous or semi-autonomous machines; a system for performing simulation operations; Systems for performing digital twin operations; a system for performing light transport simulations; A system for performing collaborative content creation of 3D assets; Systems for performing deep learning operations; A system for performing real-time streaming; Systems for generating or presenting one or more of augmented reality content, virtual reality content, or mixed reality content; Systems implemented using edge devices; Systems implemented using robots; Systems for performing conversational AI operations; Systems for generating synthetic data; Systems for generating synthetic data using AI; A system comprising one or more virtual machines VM; A system implemented at least in part in a data center; or A system implemented at least in part using cloud computing resources.

10. A system comprising one or more processors configured to generate a visualization of an environment based at least on generating graphical content for one or more 3D representations of one or more detected dynamic objects in a three-dimensional (3D) surface topology of the environment, the graphical content being generated based on sensor data generated using one or more sensors of an ego machine in the environment. 11 . The system of claim 10 , the one or more processors further configured to mask the one or more detected dynamic objects during a first pass of texturing the detected 3D surface topology.

12. The system of claim 10, wherein the one or more processors are further configured to: generating a first 3D surface topology of the environment representing a static portion of the environment; and The first 3D surface topology is updated based at least on inserting the one or more 3D representations of the one or more detected dynamic objects into the first 3D surface topology.

13. The system of claim 10 , wherein the one or more detected dynamic objects comprise at least one rigid object from one or more rigid object classes, the one or more processors being further configured to generate one or more 3D representations of the at least one rigid object based at least on warping one or more detected depth values ​​of the at least one rigid object using one or more detected trajectories corresponding to the at least one rigid object.

14. The system according to claim 10, wherein the one or more processors are further configured to: fuse at least a first 3D representation of at least a first detected dynamic object among the one or more detected dynamic objects generated at least based on the following operations into the 3D surface topology: tracking a trajectory of the first detected dynamic object; identifying one or more detected depth values ​​representing the first detected object in a previous time slice; and warping the one or more detected depth values ​​using the trajectory.

15. The system of claim 10 , wherein the one or more detected dynamic objects comprise at least one non-rigid object from one or more non-rigid object classes, the one or more processors further configured to generate one or more 3D representations of the at least one non-rigid object based at least on inserting a 3D representation of a two-dimensional (2D) surface for the at least one non-rigid object at a location in the 3D surface topology corresponding to a detected location of the at least one non-rigid object in the environment.

16. The system of claim 10, wherein the one or more processors are further configured to fuse at least a first 3D representation of a flat surface at a location corresponding to a detected center of mass of a corresponding one of the one or more detected dynamic objects into the 3D surface topology.

17. The system of claim 10, wherein generating graphical content for the one or more 3D representations of the one or more detected dynamic objects is based at least on a set of segmented sensor data classified as corresponding to the one or more detected dynamic objects.

18. The system of claim 10, wherein the system is included in at least one of the following: control systems for autonomous or semi-autonomous machines; Perception systems for autonomous or semi-autonomous machines; a system for performing simulation operations; Systems for performing digital twin operations; a system for performing light transport simulations; A system for performing collaborative content creation of 3D assets; Systems for performing deep learning operations; A system for performing real-time streaming; Systems for generating or presenting one or more of augmented reality content, virtual reality content, or mixed reality content; Systems implemented using edge devices; Systems implemented using robots; Systems for performing conversational AI operations; Systems for generating synthetic data; Systems for generating synthetic data using AI; A system comprising one or more virtual machines VM; A system implemented at least in part in a data center; or A system implemented at least in part using cloud computing resources.

19. A method comprising: computing a 3D surface topology of the environment based at least on image data generated using one or more cameras of the ego machine in the environment and at least on one or more three-dimensional (3D) representations of one or more detected dynamic objects in the environment; as well as A visualization of the environment is generated based at least on projecting the image data onto the one or more 3D representations of the one or more detected dynamic objects in the 3D surface topology.

20. The method of claim 19, wherein the method is performed by at least one of: control systems for autonomous or semi-autonomous machines; Perception systems for autonomous or semi-autonomous machines; a system for performing simulation operations; Systems for performing digital twin operations; a system for performing light transport simulations; A system for performing collaborative content creation of 3D assets; Systems for performing deep learning operations; A system for performing real-time streaming; Systems for generating or presenting one or more of augmented reality content, virtual reality content, or mixed reality content; Systems implemented using edge devices; Systems implemented using robots; Systems for performing conversational AI operations; Systems for generating synthetic data; Systems for generating synthetic data using AI; A system comprising one or more virtual machines VM; A system implemented at least in part in a data center; or A system implemented at least in part using cloud computing resources.

Images