PERCEPTIONAL DIVERSITY FOR OBJECT IDENTIFICATION IN ROBOTIC SYSTEMS AND APPLICATIONS

Abstract

The present disclosure relates to the detection of objects in detection zones using multiple analysis techniques. These multiple analysis techniques can be used to analyze sensor data corresponding to the detection zones. The multiple analysis techniques can be selected based on at least two of the analysis techniques exhibiting computational diversity by performing different types of computational analyses on the sensor data in relation to each other, and at least two of the analysis techniques exhibiting implementation diversity by being implemented on different types of computing platforms in relation to each other.

Description

BACKGROUND

[0001] Object detection is a crucial task for many robotics and autonomous systems. For example, object detection systems can be used to detect objects (such as people, animals, substances, stationary objects, dynamic objects, and / or other objects), and they can be specifically tasked with detecting objects within defined areas, which may be referred to as object detection zones ("detection zones"). For instance, detection systems can be used to determine whether an object is within a certain proximity of a machine's operating area, where the operating area may encompass a range within a certain distance of the machine and thus be designated as a detection zone.

[0002] Object detection can be used in conjunction with certain safety measures to help prevent machines from exhibiting unsafe and / or harmful behaviors. For example, areas or spaces around or within a robot's operating area, designated as potentially hazardous areas, may be included in detection zones where object detection is critical. For instance, in response to an object detection action corresponding to a detection zone, a detection system may cause the machine to operate in a specific way—such as slowing down or halting operation. Similarly, the machine may stop in response to the system performing the detection action corresponding to the detection zone.

[0003] Some current approaches to performing detections that correspond to detection zones involve using a two-dimensional (2D) planar laser rangefinder or lidar to scan within detection zones. For example, the planar 2D laser rangefinder or lidar can be used to scan a circular plane extending from 270 to 360 degrees around the machine. In these cases, detections can be performed when an object intersects the circular plane. For example, the planar 2D laser rangefinder or lidar can detect an object within the circular plane. However, such an approach is limited to detections in 2D space and may not be applicable in some cases. For example, such an approach cannot properly perform detection in three-dimensional (3D) planes. For example, the object may be located above the circular plane being scanned, such as...Flying objects and / or hanging objects. The planar 2D laser rangefinder or lidar cannot properly detect flying objects and / or hanging objects.

[0004] Furthermore, some current approaches may be limited due to the use of a specific detection technique. For example, a particular detection system may use a specific detection technique to analyze input data obtained using a sensor to determine the detections. This specific detection technique may involve an algorithm used to analyze the input data. In some cases, the algorithm may not be suitable for a particular detection zone or specifications of the example system. For example, the algorithm may need to be preconfigured to compute a specific set of disparities, which may prevent inference time matching.Such algorithms may not be suitable for detection zone requirements that may require flexibility in utilizing trade-offs between latency and depth quantization at inference time. SUMMARY

[0005] According to one or more embodiments of the present disclosure, one or more systems and / or methods can be configured to detect objects in detection zones using multiple analysis techniques. For example, in some embodiments, sensor data corresponding to the detection zones, obtained using one or more sensors, can be analyzed using the multiple analysis techniques. In some embodiments, the multiple analysis techniques can be selected based on at least two of the multiple analysis techniques exhibiting computational diversity by performing different types of computational analyses on the sensor data in relation to one another.In some embodiments, to further increase robustness and diversity, at least two of the multiple analysis techniques can be implemented on different types of computing or hardware platforms, devices or components.

[0006] One or more embodiments of the present disclosure can help to detect the presence of objects in detection zones more accurately. Additionally or alternatively, the detection of the objects can be carried out in a variety of ways, which can enable increased robustness of the detection. For example, according to one or more embodiments, the variety of detection methods can increase the accuracy of detecting objects in the respective detection zones under varying detection conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] The present systems and methods for detecting objects in detection zones and applications are described in detail below with reference to the accompanying figures, wherein: Fig. 1A-1B illustrate an exemplary system configured to determine the detection of an object in detection zones according to one or more embodiments of the present disclosure; Fig. 1C an exemplary diagram 130 illustrates a trade-off between safety and productivity when combining one or more detection algorithms according to one or more embodiments of the present disclosure; Fig. 1D illustrates an exemplary detection zone according to some embodiments of the present disclosure; Fig. 2A-2B illustrate an exemplary system configured to determine the detection of an object in a detection zone according to one or more embodiments of the present disclosure; Fig. 3 is an exemplary flowchart illustrating a method for detecting objects in detection zones of a machine according to one or more embodiments of the present disclosure; Fig. 4A is an illustration of an exemplary autonomous vehicle according to one or more embodiments of the present disclosure; Fig. 4B is an example of camera positions and fields of view for the exemplary autonomous vehicle of Fig. 4A according to one or more embodiments of the present disclosure; Fig. 4C a block diagram of an exemplary system architecture for the exemplary autonomous vehicle of Fig. 4A according to one or more embodiments of the present disclosure; Fig. 4D, a system diagram for communication between one or more cloud-based servers and the example autonomous vehicle from Fig. 4A according to one or more embodiments of the present disclosure; Fig. 5 is a block diagram of an exemplary computing device suitable for use in implementing one or more embodiments of the present disclosure; and Fig. 6 is a block diagram of an exemplary data center suitable for use in implementing one or more embodiments of the present disclosure. DETAILED DESCRIPTION

[0008] Systems and methods relating to the detection of objects in specific areas, referred to as "detection zones," are disclosed in the present disclosure. In some embodiments, the detection zones may include any area of ​​interest surrounding or near a machine. For example, the detection zones may include operating areas of a vehicle or machine (generally referred to as a "machine"). For example, the operating areas of the machine may include any area in which a section of the machine can move or operate. For example, an operating area corresponding to a robot arm may include any area in which or toward which the robot arm can move. Additionally or alternatively, the operating areas may include any area in which an operation by the machine can introduce something into the area.For example, the operating areas may include an area where electromagnetic waves, radiation, etc. (e.g., X-rays, laser beams, etc.) may be present due to emissions from the machine.

[0009] Additionally or alternatively, the detection zones may include hazardous areas or restricted areas (generally referred to as "restricted areas"). For example, restricted areas may include areas that, due to confidentiality, contamination, or safety considerations, may require restricted access with respect to persons, animals, substances, or objects. In some cases, restricted areas may include areas that may be restricted with respect to access and / or objects or substances (e.g., dust, radiation, certain gases, etc.). In this disclosure, a general reference to an object or objects may include persons, substances, or any other applicable type of thing that may enter the detection zones.

[0010] In some embodiments, the detection zones may include areas related to ego-machines. For example, the detection zones may include areas in front of, behind, and / or around an ego-machine. Additionally or alternatively, the detection zones may include areas above or below the ego-machine. In these or other embodiments, the detection zones may move along with the ego-machine as it moves. For example, a particular detection zone, as illustrated graphically, may be represented as a boundary shape surrounding the ego-machine. The boundary shape may be defined as a specific size and shape depending on the ego-machine and the machine's detection capabilities or requirements. For example, the boundary shape may be defined to cover areas that should be kept clear for the ego-machine to avoid collisions.In these and other embodiments, the boundary shape can move along with the ego-machine as the ego-machine moves. For example, the boundary shape can maintain a position and / or orientation relative to the current location and / or pose of the ego-machine.

[0011] In some embodiments, object detection can be based on at least detection data generated using one or more sensors. For example, in some embodiments, the detection system can include one or more analysis algorithms that can be applied to the detection data to determine whether an object is detected in the detection zone.

[0012] One or more of the embodiments disclosed herein may relate to the detection of objects in detection zones assigned to one or more Ego-machines, which may include any applicable machine or system capable of performing one or more autonomous or semi-autonomous operations. Exemplary Ego-machines may include, but are not limited to, vehicles (land, sea, space, and / or air), robots, robotic platforms, etc. For example, the Ego-machine computing applications may include one or more applications that can be executed by an autonomous or semi-autonomous vehicle, such as an exemplary Autonomous Vehicle 400 (hereafter referred to alternatively as "Vehicle 400" or "Ego-Machine 400"), which, with respect to the Fig. As described in sections 4A-4D. In the present disclosure, reference to an “autonomous vehicle” or “semi-autonomous vehicle” can include any vehicle that can be configured to perform one or more autonomous or semi-autonomous navigation or driving operations. Thus, such vehicles can also include vehicles in which an operator is required or in which an operator can also perform such operations.

[0013] The systems and procedures described herein may be used by, among others, 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, controlled and uncontrolled robots or robotic platforms, warehouse vehicles, all-terrain vehicles, vehicles coupled to one or more trailers, flying ships, boats, shuttles, emergency vehicles, motorcycles, electric or motorized bicycles, aircraft, construction vehicles, underwater vehicles, drones and / or other types of vehicles.Furthermore, the systems and methods described herein can be used for a variety of purposes, including but not limited to machine control, machine locomotion, machine driving, synthetic data generation, model training, perception, augmented reality, virtual reality, mixed reality, robotics, security 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 (such as by using one or more language models, such as one or more large language models (LLMs)), light transport simulation (e.g., ray tracing, path tracking, etc.), collaborative content creation for 3D assets, cloud computing, and / or other suitable applications.

[0014] Disclosed embodiments may be included in a wide variety of different systems, such as automotive systems (e.g., a control system for an autonomous or semi-autonomous machine, a perception system for an autonomous or semi-autonomous machine), systems implemented using a robot, aviation systems, medical systems, boat systems, intelligent area monitoring systems, systems for performing deep learning operations, systems for performing simulation operations, systems for performing digital twin operations, systems implemented using an edge device, systems containing one or more virtual machines (VMs), systems for performing synthetic data generation operations, systems implemented at least partially in a data center, systems for performing conversational AI operations (e.g.,Systems that implement one or more language models, such as Large Language Models (LLMs)), systems for hosting real-time streaming applications, systems for displaying one or more virtual reality content, augmented reality content or mixed reality content, systems for performing light transport simulation, systems for performing collaborative content creation for 3D assets, systems that are implemented at least partially using cloud computing resources, and / or other types of systems.

[0015] The embodiments of the present disclosure can help to improve the accuracy of object detection in detection zones. For example, in one or more embodiments of the present disclosure, the detection system can perform object detection in a variety of ways, which can improve the accuracy of the detection system. For example, according to one or more embodiments, the detection system can use different techniques that can be used for object detection in the detection zones. In some embodiments, the different techniques can be selected based on computational diversity among the different techniques. Additionally or alternatively, the different techniques can be selected based on implementation diversity among the different techniques.

[0016] One or more embodiments of the present disclosure can help to improve the accuracy compared to some conventional approaches to object detection. For example, some conventional approaches to object detection may involve the use of a two-dimensional (2D) planar laser rangefinder or a lidar. For example, the planar 2D laser rangefinder or the lidar may be used to scan a circular plane extending from 270 to 360 degrees around the detection system. In these cases, detection can be performed when an object intersects the circular plane. However, such an approach may be limited to 2D spaces and / or may be limited in height (based on the type of lidar sensor used) and may not be applicable in certain cases. For example, such an approach may not be effective if the detection zones include vertical planes.For example, the detection zones may include areas where the ground is not flat and / or where objects may be located on different vertical planes. With a planar 2D laser rangefinder or lidar, uneven ground can lead to false positive detections due to some parts of the ground intersecting the circular plane.

[0017] Furthermore, some conventional approaches may involve the use of a specific computing platform. For example, the specific computing platform may include a specific hardware platform and / or a specific software platform capable of receiving and processing sensor data to determine detection. However, the use of a specific computing platform can introduce the risk of common failures. For example, a computational error on the specific computing platform may cause false positives and / or missed detections.

[0018] The embodiments of the present disclosure are explained with reference to the accompanying figures. It is understood that the figures are illustrative and schematic representations of such exemplary embodiments and are not limiting and are not necessarily to scale. In the figures, features with the same numbers indicate the same structure and function, unless otherwise described.

[0019] In relation to the figures, they illustrate Fig. 1A-1B is an exemplary system 100 configured to determine the detection of an object in a detection zone according to one or more embodiments of the present disclosure. In some embodiments, the system 100 may be implemented on a machine. For example, the system 100 may be implemented on the vehicle 400 of the Fig. 4A-4D can be implemented. For example, system 100 can be configured to detect objects located within a detection zone of vehicle 400. In some cases, vehicle 400 can use system 100 to detect objects while the vehicle 400 is moving. In another example, system 100 can be implemented on a machine with moving parts (e.g., robot arms). In these cases, system 100 can be configured to detect objects within operating areas of the machine (e.g., areas that the robot arms can reach and / or where the robot arms must operate safely).

[0020] As described in detail herein, the system 100 can generally include a first computing platform (“first platform”) 104, a second computing platform (“second platform”) 106, and a third computing platform (“third platform”). In some embodiments, the system 100 can include additional computing platforms. The first platform 104, the second platform 106, and the third platform 108 can be configured to determine a first detection 116a, a second detection 116b, and a third detection 116c, respectively, based on image data 102 and / or other sensor data (e.g., lidar, radar, ultrasound, etc.). Thus, when reference is made herein to image data 102, this can also include other types or modalities of sensor data.

[0021] The image data 102 can include any type of data representative of the detection zone. For example, the image data 102 can include an image depicting the detection zone. For example, in some embodiments, the image data 102 can be obtained using one or more sensors. The one or more sensors can include any suitable system, device, or apparatus that can be used to obtain the image data 102, such as laser scanners, camera systems, and / or other suitable optoelectronic devices. In the present disclosure, a reference to an “image” can also refer to the image data corresponding to it. Likewise, an image can include a representation of the image data and other sensor data representations (e.g., area images, point clouds, etc.).), which are assigned to other sensor data modalities, may also be used herein without deviating from the scope of the present disclosure.

[0022] In some embodiments, the image data 102 obtained using the one or more sensors can contain a specific number of pixels. For example, the specific number of pixels can depend on the size of the image data 102, the specifications of the one or more sensors, the specifications of computing platforms, etc. For example, one or more of the sensors can be relatively high-resolution sensors and can accordingly be used to generate image data with a larger number of pixels than one or more sensors that may have a lower resolution.

[0023] In some embodiments, the number of pixels and / or other properties of the image data 102 can be transformed. For example, the image data 102 obtained using the one or more sensors can be preprocessed using one or more preprocessing functions to be more suitable for processing by the first platform 104, the second platform 106, and / or the third platform 108. For example, the image data 102 can be transformed into a specific size (e.g., number of pixels) and a specific format. The one or more preprocessing functions can include any suitable functions to convert raw image data into a more suitable format for processing.For example, one or more preprocessing functions may include format conversion, noise reduction, tone mapping, multiple exposure merging, distortion correction, cropping, downscaling, and rectification.

[0024] In some embodiments, one or more objects can be represented in the image data 102. Furthermore, in the present disclosure, objects represented in the image data 102 and represented by corresponding images can also be referred to as "present in" or "in" the image data 102 or the corresponding image. Pixels of the image data 102 containing information corresponding to objects can also be referred to as "occupied by" objects. Likewise, a pixel representing an object or a portion of the object can be referred to as "containing" the object, or the object can be referred to as "present in" or "in" the pixel. Additionally, the area and / or objects represented or depicted by the image data 102 and / or a pixel can be referred to as "corresponding" to the image data 102 and / or the pixel.Furthermore, in the present disclosure, a reference to a pixel may include a reference to the one or more objects or areas corresponding to the pixel.

[0025] In some embodiments, the image data can contain information about which pixels are occupied by the one or more objects. For example, a particular pixel can be considered occupied if at least part of that pixel contains at least part of the one or more objects. Conversely, the particular pixel can be considered unoccupied if no part of the one or more objects is present in that pixel. For example, the particular pixel can correspond to an area of ​​the detection zone devoid of any objects. For example, the particular pixel can correspond to an area that includes the background, sky, etc.

[0026] In some embodiments, one or more sensors may include a stereo camera module. In these and other embodiments, the stereo camera module may include two or more cameras that can be used to generate a three-dimensional (3D) representation of an area. In some embodiments, the area represented by the stereo camera module may correspond to the detection zone. In some embodiments, the two or more cameras may be configured to obtain images of the area from different positions. For example, the stereo module may include a first camera and a second camera, where the first camera may be configured to obtain a first image of the area and the second camera may be configured to obtain a second image of the area.In some cases, the first and second cameras may be offset from each other, so that they have different perspectives. In these and other embodiments, the first and second images may exhibit a disparity due to the different perspectives. For example, the one or more objects found in the first and second images may occupy different sets of pixels. In these cases, the disparity may indicate a displacement (e.g., a distance) between the different sets of pixels in the first and second images. In some embodiments, the presence of disparity may indicate the presence of depth in the area, and the disparity may be used to estimate depth. In some embodiments, the image data 102 may include at least the first and second images.Additionally or alternatively, the image data 102 can include the disparity between the first image and the second image.

[0027] In some embodiments, the first platform 104, the second platform 106, and the third platform 108 may each include a first depth module 110a, a second depth module 110b, and a third depth module 110c (hereinafter collectively referred to as "depth modules 110"). In some embodiments, one or more of the depth modules 110 may be implemented using hardware that includes one or more processors, central processing units (CPUs), graphics processing units (GPUs), data processing units (DPUs), parallel processing units (PPUs), microprocessors (e.g., for performing or controlling the performance of one or more operations), field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), accelerators (e.g.,Deep learning accelerators (DLAs), optical flow accelerators (OFAs), programmable vision accelerators (including one or more direct memory address (DMA) systems and / or vector processing units (VPUs)), and / or other processor types are included. In some other cases, one or more of these modules may be implemented using a combination of hardware and software. In the present disclosure, operations described as being performed by a respective module may include operations that the respective module can instruct a corresponding computing system to perform. In these or other embodiments, one or more of these modules may be implemented by one or more computing devices, such as the one described in relation to the... Fig. 4A-4D, Fig. 5 and / or Fig. Section 6 is described in more detail. In some embodiments, the first depth module 110a, the second depth module 110b, and the third depth module 110c can be implemented using different hardware relative to each other. For example, in some embodiments, the first depth module 110a can be implemented using an OFA, the second depth module 110b can be implemented using a DLA, and the third depth module 110c can be implemented using a GPU.

[0028] In some embodiments, the one or more depth modules 110 can be configured to obtain the image data 102. The one or more depth modules 110 can further be configured to identify the one or more objects represented by the image data 102. For example, the one or more depth modules 110 can be configured to determine a number of objects and / or locations of the objects present in the image data 102. Additionally or alternatively, the one or more depth modules 110 can be configured to determine a depth corresponding to the one or more objects identified in the image data 102.For example, the first depth module 110a can be configured to determine a first depth 112a, the second depth module 110b can be configured to determine a second depth 112b, and the third depth module 110c can be configured to determine a third depth 112c. The first depth 112a, the second depth 112b, and the third depth 112c can generally or collectively be referred to as "the depths 112" or "the depth 112".

[0029] In some embodiments, one or more of the depths 112 can include a relationship between the one or more objects and a reference point. For example, in some embodiments, one or more of the depths can be defined as a continuous depth that may refer to a distance of the one or more objects from the reference point—e.g., an origin or rig coordinate corresponding to the machine or ego machine, such as the center of a vehicle's rear axle or a point along a robot arm.

[0030] Additionally or alternatively, one or more depths 112 can correspond to a disparity of one or more objects from the reference point. For example, objects farther from the reference point can have a greater depth than objects closer to the reference point. In these and other embodiments, the distance corresponding to one or more of the depths 112 can be represented using any suitable metrics of measurement. For example, the distance can be represented in meters, centimeter meters, inches, etc.

[0031] In some embodiments, the reference point can correspond to the position of one or more sensors. For example, the reference point can be determined as the position of the stereo camera module. In these cases, one or more of the depths 112 of the one or more objects can be determined as the distance from the stereo camera module to the one or more objects. In some embodiments, the reference point can be determined as any other suitable position with respect to the system 100 and the detection zone. For example, in some embodiments, the reference point can be located in the middle, at the edge, in front, at the rear, or at any other part of the machine implementing the system 100.

[0032] Additionally or alternatively, in some embodiments, one or more of the depths can be represented in a binary classification format. For example, the depth can refer to whether the one or more objects are closer to or farther from a specific depth level. For example, there can be one or more depth levels, each located at a different distance from the reference point. In these and other embodiments, the positions of one or more objects can be compared to the specified depth level to determine whether the one or more objects are closer to or farther from the reference point than the specified depth level. In some embodiments, one or more objects can be compared to multiple depth levels, which can provide finer depth quantization.

[0033] For example, the position of a particular object can be compared to a first depth level and a second depth level. The particular object can be determined to be farther from the reference point than the first depth level, but closer than the second depth level. In these cases, the particular object may be located between the first and second depth levels. In these cases, comparing one or more objects to multiple depth levels can lead to a more detailed depth quantization of the one or more objects. In these and other embodiments, one or more of the depths can accordingly specify the relative depth relationships between the particular object and the depth levels.

[0034] In some embodiments, the distances between one or more depth levels can be configured according to a specified depth quantization. For example, the distances between one or more depth levels can be narrower if greater depth quantization is required, and the distances between one or more depth levels can be larger if less depth quantization is required.

[0035] In some embodiments, one or more of the depths 112 can be determined on a pixel-wise basis for the pixels corresponding to the one or more objects. In some embodiments, different parts of the one or more objects can have different depths. In these and other embodiments, the depths of the one or more objects can be determined by determining respective depths for respective sections of the one or more objects corresponding to the pixels. For example, individual pixel depths can be determined for the pixels occupied by the one or more objects. For example, a distance between the reference point and each of the pixels occupied by the one or more objects can be determined.In these and other embodiments, the individual pixel depths can be used to determine one or more of the depths 112 in the continuous depth format.

[0036] Additionally or alternatively, individual pixel depths for the pixels occupied by the one or more objects can be determined in a binary classification format. Such an approach can provide more detailed depth quantization than determining the depths of the one or more objects as a whole. For example, part of an object represented by a particular pixel may be closer than a certain depth level to the reference point, while the rest of the object may be farther than that depth level. Such an approach can indicate how only a small portion of the object is located at a given depth.

[0037] In some embodiments, one or more of the depths 112 can be represented using a depth map. The depth map can include a 3D representation of the depths of some (e.g., all) of the pixels present in the image data 102. For example, the one or more objects can be projected onto a 3D space based on their individual pixel depths. For example, each of the pixels corresponding to the one or more objects can be placed at specific positions in the 3D space according to the reference point and the individual pixel depths.

[0038] In some embodiments, one or more depth modules 110 can be configured to determine the depth of one or more objects using diverse approaches. For example, the first depth module 110a can determine the depth in the continuous depth format, while the second depth module 110b can determine the depth in the binary classification format. Additionally or alternatively, the diverse approaches can include the use of machine learning (ML)-based techniques versus the use of one or more machine vision-based techniques. The diverse approaches can also include different ways in which data can be analyzed or aggregated. For example, some ML-based techniques can implement 2D or 3D convolution techniques as part of the training and / or depth determination.Additionally or alternatively, other ML-based techniques can implement iterative refinement techniques, with each iteration improving on estimates from a previous step as part of the training and / or depth determination.

[0039] In some embodiments, the diverse approaches may include differences in the ability to manipulate one or more parameters corresponding to the approaches. In these and other embodiments, some approaches may allow the manipulation of one or more parameters during runtime. Additionally or alternatively, some of the diverse approaches may include a flexible range, which may include the ability to define a specific range at or near runtime, such that the approach is restricted to estimating depth only for objects within the defined range.

[0040] In some embodiments, the one or more depth modules 110 can be implemented on different computing platforms. For example, in some embodiments, the first platform 104, the second platform 106, and the third platform 108 can comprise hardware and / or software platforms, wherein at least two of the first platform 104, the second platform 106, and the third platform 108 differ with respect to each other. In some embodiments, the different computing platforms can cause a difference in the results of the one or more depth modules 110. The different computing platforms are defined here as differing at least with respect to Fig. 2A-2B discussed in more detail.

[0041] Additionally or alternatively, the one or more depth modules 110 can be configured to determine depth confidences. For example, the first depth 112a can include a first depth confidence. In some embodiments, the first depth confidence can indicate how confident the first depth module 110a is with the depth of the one or more objects determined by the first depth module 110a. For example, the first depth confidence can indicate how confident the first depth module 110a is with the depth determined for each of the one or more objects. In some embodiments, the first depth confidence can include how confident the first depth module 110a is with the depth determined for the pixels occupied by the one or more objects.

[0042] In some embodiments, the depth determined by one or more depth modules 110 can be represented in the same format as the image data 102. For example, the image data 102 can be transformed into a 16-bit image format for the first depth module 110a. In these and other cases, the first depth 112 can also be represented in the 16-bit image format.

[0043] In some embodiments, the first platform 104, the second platform 106, and the third platform 108 may each include a first detection module 114a, a second detection module 114b, and a third detection module 114c (collectively referred to herein as "detection modules 114"). In some embodiments, the detection modules 114 may be implemented using hardware that includes one or more processors, central processing units (CPUs), graphics processing units (GPUs), data processing units (DPUs), parallel processing units (PPUs), microprocessors (e.g., for performing or controlling the performance of one or more operations), field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), accelerators (e.g.,Deep learning accelerators (DLAs), programmable vision accelerators (including one or more direct memory address (DMA) systems and / or vector processing units (VPUs)), and / or other processor types. In some other cases, one or more of these modules may be implemented using a combination of hardware and software. In the present disclosure, operations described as being performed by a respective module may include operations that the respective module can instruct a corresponding computing system to perform. In these or other embodiments, one or more of these modules may be implemented by one or more computing devices, such as the one described in relation to the... Fig. 4A-4D, Fig. 5 and / or Fig. Section 6 is described in more detail.

[0044] In some embodiments, the one or more detection modules 114 can be configured to obtain the one or more depths 112. The one or more detection modules 114 can further be configured to determine whether a depth of any object present in the one or more depths 112 lies within a specific area of ​​the detection zone. For example, the first detection module 114a can be configured to determine a first detection 116a, the second detection module 114b can be configured to determine a second detection 116b, and the third detection module 114c can be configured to determine a third detection 116c. The first detection 116a, the second detection 116b and the third detection 116c may generally or collectively be referred to as “the detections 116” or “the detection 116”.

[0045] In some embodiments, one or more of the detections 116 can indicate whether detections of objects within the detection zone are performed. In some embodiments, the detections can be represented using one or more voxels. For example, the detection zone can be subdivided into one or more voxels or 3D cubes. The one or more voxels can generate a 3D grid of cubes that cover the detection zone. In some embodiments, the one or more voxels can be of the same size. In some embodiments, the size of the one or more voxels can be determined based on the format of the one or more depths 112. For example, one or more of the depths 112 can be represented in the 16-bit image format. In these cases, the size of the one or more voxels can be determined to subdivide the detection zone according to the 16-bit image format.In some embodiments, the first detection 116a, the second detection 116b, and the third detection 116c can be configured to be of the same format. For example, the one or more voxels for the first detection 116a, the second detection 116b, and the third detection 116c can be arranged and / or dimensioned identically. In other embodiments, however, the one or more voxels for the first detection 116a, the second detection 116b, and the third detection 116c may not be arranged and / or dimensioned identically.

[0046] In some embodiments, one or more of the detection modules 114 can be configured to project the depth map, representing one or more of the depths 112, onto the one or more voxels. For example, the pixels corresponding to different parts of the one or more objects can be placed on the 3D grid of cubes according to the individual pixel depths. In these and other embodiments, one or more of the detection modules 114 can be configured to determine which of the one or more voxels contain the depth of any one of the one or more objects.

[0047] Additionally or alternatively, one or more of the detection modules 114 can be configured to determine whether the one or more objects detected within the detection zone are located in specific areas of interest within the detection zone. For example, one or more of the detections 116 can indicate whether the depths of any one or more objects lie within the specific areas of interest. For example, the one or more voxels containing at least a portion of the one or more existing objects can be examined to determine where the one or more objects are located with respect to the detection zone. In some embodiments, the specific areas of interest can include a portion of the detection zone.For example, the specific areas of interest can include the portion of the detection zone located within a certain proximity of the machine implementing System 100. For instance, areas in the immediate vicinity of the machine might be of particular interest for safety reasons. In these cases, the one or more voxels within a certain distance from the machine can be designated as an area of ​​interest. In these cases, the detections can indicate whether any of the one or more voxels within the area of ​​interest contains any part of the one or more objects.

[0048] In some embodiments, the detection zone can include one or more areas of interest. For example, the detection zone can be subdivided into one or more sub-detection zones, each covering an area of ​​particular interest. In some embodiments, the sub-detection zones can be determined based on the machine's operating characteristics. For example, the machine can include a moving part (such as a robot arm) that can operate within a specific operating range.In these and other embodiments, a first underdetection zone can include any area that the moving part can reach during operation, a second underdetection zone can include any area that should be kept clear to prevent interference from the operation, and a third underdetection zone can include areas that the moving part might not reach. Examples of different underdetection zones can be found in relation to... Fig. 1D of the present revelation will be discussed in more detail.

[0049] Additionally or alternatively, one or more of the detection modules 114 can be configured to determine detection confidence levels. For example, the first detection 116a can include a first detection confidence level, the second detection 116b can include a second detection confidence level, and the third detection 116c can include a third detection confidence level. In some embodiments, the first detection confidence level can indicate how confident the first detection module 114a is in detecting the one or more objects determined by the first detection module 114a. For example, the first detection confidence level can indicate how confident the first detection module 114a is in the detection determined for each of the one or more objects.In some embodiments, the first detection confidence can indicate how reliably the first detection module 114a detects the one or more objects in each of the one or more voxels. In some embodiments, the first detection confidence can be represented as a number within a confidence range (e.g., 0–100). In some embodiments, the first, second, and third detection confidences can be represented on the same confidence range.

[0050] In some embodiments, the system 100 may include an aggregation module 118. In some embodiments, the aggregation module 118 may be configured to obtain the first detection 116a, the second detection 116b, and the third detection 116c. The aggregation module 118 may further be configured to determine a final detection 120 based on at least the first detection 116a, the second detection 116b, and / or the third detection 116c.

[0051] In some embodiments, the aggregation module 118 can be implemented using hardware that includes one or more processors, central processing units (CPUs), graphics processing units (GPUs), data processing units (DPUs), parallel processing units (PPUs), microprocessors (e.g., for performing or controlling the power of one or more operations), field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), accelerators (e.g., deep learning accelerators (DLAs), optical flow accelerators (OFAs)), programmable vision accelerators (including one or more direct memory address (DMA) systems and / or vector processing units (VPUs)), and / or other processor types. In some other cases, the aggregation module 118 can be implemented using a combination of hardware and software.In the present disclosure, operations described as being performed by a respective module may include operations that the respective module can instruct a corresponding computing system to perform. In these or other embodiments, the aggregation module 118 may be implemented by one or more computing devices, such as the one described in relation to the... Fig. 6A-6D, Fig. 7 and / or Fig. 8 is described in more detail.

[0052] In some embodiments, the final detection 120 can indicate whether a detection of one or more objects in the detection zone is performed. For example, the final detection 120 can indicate whether the one or more objects are located in the image data 102. Additionally or alternatively, the final detection 120 can indicate where in the detection zone the one or more objects are detected. In some embodiments, the final detection 120 can be a result of combining the first detection 116a, the second detection 116b, and the third detection 116c. By combining two or more of the detections 116, the final detection 120 can represent the detection of the one or more objects in the detection zone more accurately than using only one of the detections 116.

[0053] In some embodiments, the aggregation module 118 can be configured to aggregate the detections 116. For example, the first detection 116a, the second detection 116b, and the third detection 116c can be aggregated to determine the final detection 120. In some embodiments, the final detection 120 can be represented in the same format as the detections 116. For example, the final detection 120 can be represented using a final set of voxels, which can be arranged and dimensioned identically to the one or more voxels of the detections 116. In these and other embodiments, each voxel of the final set of voxels can have a corresponding voxel in each of the detections 116. In these and other embodiments, the final detection 120 can be determined on a voxel-wise basis.For example, each of the final set of voxels of final detection 120 can be determined by aggregating corresponding voxels from detections 116. In these and other cases, each voxel of the final set of voxels of final detection 120 can indicate whether an object is present in the corresponding voxel.

[0054] In some embodiments, the aggregation of the corresponding voxels can be based on one or more combination algorithms. In the present disclosure, the combination algorithms can refer to any suitable method for combining the corresponding voxels from the detections 116 to determine the final set of voxels of the final detection 120.

[0055] In some embodiments, a particular combination algorithm may include a tuning algorithm that may involve combining the detections 116 based at least on the detection confidences contained in the detections 116. For example, the aggregation module 118 for the final set of voxels (e.g., each voxel) of the final detection 120 may analyze corresponding voxels assigned to each of the detections 116, together with the detection confidences assigned to the corresponding voxels.

[0056] For example, the first detection 116a can include the first detection confidence level, the second detection 116b can include the second detection confidence level, and the third detection 116c can include the third detection confidence level. In these and other embodiments, the aggregation module 118 can give more weight to a detection with a higher detection confidence level. For example, in cases where there is a conflict between the detections 116, a detection with the higher confidence level can prevail over one with a lower confidence level.

[0057] For example, for a given voxel of the final set of voxels, a corresponding first voxel from the first detection 116a may indicate a detection (e.g., an object is present in the corresponding first voxel), while a corresponding second voxel from the second detection 116b and a corresponding third voxel from the third detection 116c may indicate a non-detection (e.g., no object is present). In such a case, the aggregation module 118 can observe the detection confidences of the first detection 116a, the second detection 116b, and the third detection 116c, and a detection with a higher confidence can be given more weight than detections with lower confidences when determining the final detection 120. For example, the first detection 116a may have complete confidence (e.g.,100) of the detection for the specified voxel, while the second detection 116b and the third detection 116c may each indicate a lower confidence level (e.g., 30) of non-detection for the specified voxel. In some of these cases, the detection of the first detection 116a may predominate over the non-detections of the second detection 116b and the third detection 116c, and despite only one of the three detections indicating a detection, the specified voxel may be determined by the final detection 120 to have a detection.

[0058] In such cases, where there is a conflict between the detections 116, the aggregation module 118 can be configured to compare an overall confidence for a detection with an overall confidence for a non-detection. For example, in the example above, the overall confidence for detection might be 100. In contrast, the overall confidence for non-detection might be 60 (e.g., 30 + 30). Since the overall confidence for detection is greater than the overall confidence for non-detection, the aggregation module 118 can determine that a detection is present in the given voxel.

[0059] In some embodiments, the combination algorithm may involve determining a trade-off between safety and productivity. For example, illustrates Fig. Figure 1C shows an exemplary diagram 130 that visualizes the trade-off. In some embodiments, the diagram 130 may include a moving scale 132 indicating a preference between safety and productivity, where an indicator 134 may be positioned. In these and other embodiments, as the indicator 134 moves toward a first side 136 of the moving scale 132, it may indicate a preference for safety over productivity. Conversely, as the indicator 134 moves toward a second end 138 of the moving scale 132, the preference for productivity may increase and the preference for safety may decrease. In some embodiments, the aggregation module 118 may define a threshold requirement based at least on where the indicator 134 is located on the moving scale 132.In these and other embodiments, the threshold requirement can define a number of detections required by the detections 116 to determine a detection for the final detection 120. For example, a specific voxel of the final detection 120 can be determined to represent the detection of one or more objects in cases where the threshold requirement number of corresponding voxels from the detections 116 indicates a detection.

[0060] In some embodiments, reliability may be prioritized over productivity. For example, indicator 134 may be positioned closer to the first side 136 than to the second end 138 of the sliding scale 132, as illustrated by a first indicator 134a. In these cases, the threshold requirement may be set as a relatively low number, making it easier to meet the requirements for the final detection 120 and therefore easier to determine. For example, the threshold requirement may be set as a number indicating a slight minority or a minority of the detections 116. For example, the slight minority may indicate a number of less than half of the detections 116.For example, from the first detection 116a, the second detection 116b and the third detection 116c, only one corresponding voxel indicating a detection may be required to find a detection in the specific voxel of the final detection 120.

[0061] In some embodiments, productivity may be considered more important than safety. For example, indicator 134 may be positioned slightly more towards the second end 138 of the sliding scale 132 than towards the first side 136, as illustrated by a second indicator 134b. In these and other embodiments, the threshold requirement may be set as a number representing at least a majority of the detections 116. For example, in some embodiments, the threshold requirement may be set as a number representing a slight majority of the detections 116. For example, the slight majority may indicate a number of more than half of the detections 116.In a case with three detections, the threshold requirement can be set to two, whereby at least two of the three detections may be required to indicate a detection for the specific voxel for the final detection 120, in order to find a detection for the specific voxel. In another case with five detections, at least three of the five detections may be required to indicate a detection.

[0062] In some embodiments, if the focus shifts from safety to productivity, the threshold requirement may also increase. For example, four detections may be required instead of three of the five to indicate a detection. In some embodiments, the highest level of productivity may be sought. For example, indicator 134 may be positioned at one end of the second end 138 of the sliding scale 132, as illustrated by a third indicator 134c. In these and other embodiments, a detection by all of the detections 116 may be required to find a detection for the specific voxel for the final detection 120. In some embodiments, indicator 134 may be moved along the sliding scale 132 to adjust the threshold requirement according to specific machine requirements.In some embodiments, depending on the specific security criteria, the threshold for security purposes may decrease - e.g., if a detection is identified, the system may determine that a detection is present.

[0063] With renewed reference to Fig. In some embodiments, after determining detection results with respect to each of the final set of voxels, the final detection 120 can be reduced to one or more general representations indicating whether an object was detected in one or more of the sub-detection zones. For example, the final detection 120 can include one or more binary values, each indicating a detection of an object in one or more of the sub-detection zones. The one or more binary values ​​(e.g., 0 for non-detection, 1 for detection) can simply indicate whether an object was detected in each of the sub-detection zones. In these and other embodiments, the one or more general representations can be determined at least based on the detection results corresponding to the final set of voxels.For example, one or more general representations may be the result of combining the detection results of the final set of voxels located within corresponding sub-detection zones.

[0064] In some embodiments, combining the detection results of the final set of voxels may involve determining the number of voxels in the final set that corresponds to a detection. For example, the number of voxels in the final set that indicates a detection within a specific sub-detection zone may be obtained. In these and other embodiments, this number may be compared to a reduction threshold to determine the general representation for the specific sub-detection zone. The reduction threshold may define the number of voxels required to indicate a detection, thus finding a detection for the general representation of the specific sub-detection zone. In some embodiments, the reduction threshold may be any number of voxels suitable for the machine implementing System 100.For example, a particular machine might require an operating area free of objects or particles. In these cases, the reduction threshold can be set as low as a single voxel. For instance, in response to finding a detection in any of the final set of voxels, the general representation might be determined to indicate a detection. In another example, the particular machine might only be interested in detecting larger objects. In these cases, the reduction threshold can be higher.

[0065] In some embodiments, the system 100 may include a control module 122. In some embodiments, the control module 122 may be configured to obtain the final detection 120. In some embodiments, the control module 122 may be implemented using hardware that includes one or more processors, central processing units (CPUs), graphics processing units (GPUs), data processing units (DPUs), parallel processing units (PPUs), microprocessors (e.g., for performing or controlling the performance of one or more operations), field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), accelerators (e.g., deep learning accelerators (DLAs), optical flow accelerators (OFAs)), programmable vision accelerators (including one or more direct memory address (DMA) systems and / or vector processing units (VPUs)), and / or other processor types.In some other cases, the control module 122 can be implemented using a combination of hardware and software. In the present disclosure, operations described as being performed by a respective module may include operations that the respective module can instruct a corresponding computing system to perform. In these or other embodiments, the control module 122 can be implemented by one or more computing devices, such as the one described in relation to the... Fig. 6A-6D, Fig. 7 and / or Fig. 8 is described in more detail.

[0066] In some embodiments, the control module 122 can further be configured to instruct the machine implementing the system 100 to perform or stop one or more operations based on at least the final detection 120. For example, in some embodiments, the control module 122 can instruct the machine to perform one or more attenuation operations in response to the final detection 120 indicating a detection. In some embodiments, the one or more attenuation operations can include any operations that can be performed to maintain and / or improve safety. For example, in some embodiments, the attenuation operations can include a notification regarding a detection. For example, the notification can include alerting security personnel, a warning alarm (e.g.,visual and / or audible) or any other suitable type of notification.

[0067] In these and other embodiments, the mitigation operations can involve causing a change in one or more machine operations corresponding to the underdetection zone in which the detection occurred. For example, the mitigation operations can involve causing the machine to enter a protective stop, limiting the machine's speed, and / or triggering warning signals. The protective stop can involve halting one or more machine operations, such as any moving or traveling parts. In some embodiments, the control module 122 can select different mitigation operations depending on the context and / or circumstances of the detection. For example, the control module 122 can select the mitigation operations depending on the underdetection zone in which the detection occurred.

[0068] For example, it illustrates Fig. 1D an exemplary detection zone 150 according to some embodiments of the present disclosure. In some embodiments, the detection zone 150 may refer to a machine 152. For example, in some embodiments, the machine 152 may be a vehicle, such as the vehicle 400 of Fig. 4. In some embodiments, the detection zone 150 can include one or more sub-detection zones. For example, the detection zone 150 can include a first sub-detection zone 154, a second sub-detection zone 156, and a third sub-detection zone 158. Although illustrated with three sub-detection zones, the detection zone can include any number of suitable sub-detection zones. Additionally or alternatively, although illustrated as rectangular boxes, the sub-detection zones can be of any shape and / or format suitable for the machine 152. For example, the machine 152 may be interested in an area surrounding the machine in a circular shape. Additionally or alternatively, the one or more sub-detection zones can apply equally in all directions.For example, one or more under-detection zones may include areas behind, around, above and / or below machine 152.

[0069] In some embodiments, the detection zone 150 can be subdivided into one or more sub-detection zones based on at least one distance from a reference point. For example, the first sub-detection zone 154 can include an area within a first distance from the reference point, the second sub-detection zone 156 can include an area within a second distance from the reference point, excluding the area included in the first sub-detection zone, and the third sub-detection zone 158 can include an area within a third distance from the reference point, excluding the area covered by the first and second sub-detection zones. In some embodiments, the reference point can be located at any position within or near the machine 152. For example, the reference point can be located at a front end of the machine 152.In another case, the reference point can be located in the center of machine 152.

[0070] In some embodiments, one or more sub-detection zones can be defined according to a degree of restriction of access and / or infiltration by objects. For example, the first sub-detection zone 154 can cover an area in the immediate vicinity of the machine 152 and can include the highest degree of restriction. For example, the first sub-detection zone 154 can include physical operating areas of the machine 152. In these and other embodiments, the first sub-detection zone 152 can be designated as an exclusion zone. The exclusion zone can correspond to an area that must always be free of objects for operation.

[0071] Additionally or alternatively, the second sub-detection zone 156 can be designated as a protection zone. In some embodiments, the protection zone can include areas that contain operating areas and / or that are located around them and that may be potentially hazardous. For example, the second sub-detection zone 156 can include areas that the machine 152 may encounter while in motion, or it can include areas that are immediately adjacent to an exclusion zone—so that any object in these zones could soon enter the exclusion zones, and thus protective measures can be taken in advance.

[0072] Additionally or alternatively, the third under-detection zone 158 can be designated as a warning zone. In some embodiments, the warning zone may include areas that might not affect the operation of the machine 152, but could without appropriate precautions.

[0073] In some embodiments, different attenuation processes can be operated by the machine 152 based on at least one or more sub-detection zones in which a detection of one or more objects is carried out. For example, a detection in the first sub-detection zone, the second sub-detection zone, and the third sub-detection zone can lead to one or more different attenuation processes.

[0074] In some embodiments, the one or more attenuation processes corresponding to the first sub-detection zone 154 can include a protective stop and / or a speed limiting of the machine 152. For example, one or more operations (e.g., any moving or traveling parts) of the machine 152 can be stopped and / or the speed of the one or more operations can be reduced. For example, the one or more operations can be stopped and / or reduced based on at least one position of the one or more objects detected in the first sub-detection zone 154. For example, in response to the detection of the one or more objects that are closer to the machine 152, the one or more operations of the machine 152 can be stopped.

[0075] In some embodiments, the one or more attenuation processes corresponding to the second sub-detection zone 156 may include a protective stop. In this and other cases, the one or more objects detected in the second sub-detection zone 156 may not be close enough to enter the first sub-detection zone 154. In these cases, the movement of the machine 152 may be stopped to avoid a collision with the one or more objects.

[0076] Additionally or alternatively, the attenuation processes in response to the detection of one or more objects in the third sub-detection zone 158 may include operating the machine 152 at a limited speed and / or activating warning signals. In some embodiments, the warning signals may include any visual and / or audible signals that can announce the presence of the machine 152. For example, the warning signals may include an alarm sound that can alert workers in the third sub-detection zone to the presence of the machine 152.

[0077] Fig. Figures 2A-2B illustrate an exemplary System 200 configured to determine the detection of an object in a detection zone according to one or more embodiments of the present disclosure. System 200 can be an exemplary implementation of System 100 from Fig. They can be 1A-1B. For example, System 200 can replace System 100. Fig. 1A-1B illustrate with specific types of detectors that can be used as the depth modules 110, which are implemented on specific types of computing platforms.

[0078] For example, the system 200 can include a semi-global matching (SGM) detector 206a implemented on an optical flow accelerator (OFA) 204a, a Bi3D detector 206b implemented on a deep learning accelerator (DLA) 204b, and an efficient semi-supervised (ESS) detector 206c implemented on a graphics processing unit (GPU) 204c. In some embodiments, the SGM detector 206a, the Bi3D detector 206b, and the ESS detector 206c can be collectively referred to as "depth detectors 206". In these and other embodiments, a reference to a detector may include the type of detector (e.g., SGM, Bi3D, ESS) used and / or the hardware platform (e.g., OFA, DLA, GPU) that implements the detector. For example, a reference to the SGM detector 206a may include a reference to the OFA 204a, which implements the SGM detector 206a.

[0079] Although the depth detectors 206 are illustrated with specific examples, the depth detector 206 can include any other suitable type of detector. For example, in some embodiments, the depth detectors 206 can include, among others, one or more of SGM, Bi3D, ESS, block matching (BM), block matching with dynamic programming (BMDP), belief matching (BM), gradient feature matching (GF), and histogram of oriented gradient (HOG). Additionally or alternatively, the depth detectors 206 can be implemented on any other suitable hardware platform, although they are illustrated as being implemented on specific hardware platforms (e.g., OFA, DLA, and GPU).For example, the depth detectors 206 can be implemented using, among other things, one or more central processing units (CPUs), GPUs, OFAs, DLAs, programmable vision accelerators (PVAs), video input (VI) controllers, image signal processors (ISPs), and video image composition (VIC) engines. Where different hardware is used for each detection, the hardware in embodiments can be free from common failures. As such, if one of the hardware elements were to have a problem or defect and thus produce inappropriate results, these problems or defects would not propagate to the other hardware elements—thus enabling accurate detection to continue even if one or more of the detectors or detection modules were taken offline or malfunctioning.

[0080] In these and other embodiments, the depth detectors 206 can be configured to determine depths based on at least image data 202. In some embodiments, the image data 202 can include one or more images representing the detection zone. For example, in some embodiments, the image data 202 can be obtained using one or more sensors. In some embodiments, the image data 202 can be added to the image data 102 of Fig. 1A corresponds.

[0081] In some embodiments, the depth detectors 206 can receive the image data 202 and determine the depth of one or more objects present in the image data 202. In these and other embodiments, the determined depths can differ due to differences between one or more of the depth detectors 206. For example, the SGM detector 206a can be configured to determine an SGM depth 208a, the Bi3D detector 206b can determine a Bi3D depth 208b, and the ESS detector 206c can determine an ESS depth 208c. In the present disclosure, the SGM depth 208a, the Bi3D depth 208b, and the ESS depth 208c can be collectively referred to as "depths 208". In these and other embodiments, one or more of the depths 208 can differ with respect to one another.For example, in some embodiments, one or more of the depth detectors 206 can correspond to different parameters that characterize the depth detectors 206.

[0082] In some embodiments, the SGM detector 206a can be a computer vision-based detector configured to determine the SGM depth 208a based on at least the image data 202. Additionally or alternatively, the SGM depth 208a can correspond to a disparity of one or more objects from a reference point, wherein objects farther from the reference point may have a greater depth than objects closer to the reference point.

[0083] In some embodiments, the SGM detector 206a can be configured to combine a local pixel-matching cost function with global power minimization constraints. For example, the SGM detector 206a can receive image data 202, comprising a first image and a second image, from a set of stereo cameras. For one or more pixels corresponding to the one or more objects in the first image, the SGM detector 206a can determine corresponding pixels in the second image. In some embodiments, the SGM detector 206a can apply a census transformation to determine the corresponding pixels in the second image. For example, the SGM detector 206a can be configured to receive the image data 202 as a grayscale image represented in the Luma format. For example, the image data 202 can be represented in a 16-bit Luma format and / or an 8-bit Luma format.In these and other embodiments, the census transformation can compare one or more pixels with a number of neighboring pixels. For example, the intensity (e.g., a luma value) of individual pixels can be compared with the intensity of neighboring pixels. The results of the comparison can be represented as a number that can be used to determine corresponding pixels between the first and second images.

[0084] In some embodiments, the SGM detector 206a can perform regression-based classification. For example, disparity values ​​can be assigned to one or more pixels of the image data 202. In these and other embodiments, a reference to a pixel can include a reference to the one or more objects or areas corresponding to the pixel. The disparity values ​​can be representative of distances from the one or more pixels in the first image to the corresponding pixels in the second image. In these and other embodiments, the disparity values ​​can be used to determine distances of the one or more pixels from a specific reference point, with the distances representing depths of the one or more pixels.

[0085] In some embodiments, the SGM detector 206a can be configured such that one or more of its specifications can be set during runtime. For example, the SGM detector 206a can be configured to specify the number of iterations and / or the accuracy of the depth estimation. For instance, the SGM detector 206a can be iterated one or more times, with the iterations refining the cost output of the SGM detector 206a. Additionally or alternatively, the granularity of the output grid size can be configured. For example, the output grid size can be configured according to a desired quality of depth estimation. For instance, a smaller grid size can provide a more detailed depth estimate than a larger grid size.

[0086] In some embodiments, the Bi3D detector 206b can be a machine learning-based detector configured to use a set of binary classifications to determine the depths of one or more objects in the image data 202. For example, instead of testing whether the one or more pixels are at a certain depth (e.g., by assigning a disparity value to each pixel), the Bi3D detector 206b can classify the one or more pixels corresponding to the one or more objects as being closer to or farther from a given depth level. Thus, the Bi3D detector 206b can implement the 2D convolution technique, which aggregates features within the same depth level. In some embodiments, the Bi3D detector 206b can determine the depth in the binary classification format, as shown in Fig. 1A-1B described in more detail.

[0087] In some embodiments, the Bi3D detector 206b can include a flexible range. For example, the Bi3D detector 206b can estimate depth for objects in a specific area (e.g., the area in the immediate vicinity of the machine). In some cases, the specific area can be determined by specifying depth levels to be used with the binary classifications. Additionally or alternatively, the Bi3D detector 206b can allow parameter manipulation during runtime. For example, classifying one or more pixels as closer to or farther from the given depth level can be iterated for multiple depth levels. In some embodiments, finer depth quantization can be obtained by iterating the classification.

[0088] In some embodiments, the Bi3D detector 206b can be configured to receive image data 202 represented in RGB format. For example, the image data 202 can be represented in RGB8 format, where one or more pixels present in the image data are represented by three colors (e.g., red, green, and blue), and each color is represented by an 8-bit integer. In some embodiments, the Bi3D detector 206b can be configured to receive image data in any other suitable format.

[0089] In some embodiments, the ESS detector 206c can be configured to determine the ESS depth 208c using a machine learning-based algorithm. In some embodiments, the ESS detector 206c can receive the image data 202 in RGB format similar to the Bi3D detector 206b. In these and other embodiments, the ESS detector 206c can be configured to apply a regression-based classification to the image data 202 to determine the ESS depth 208c based on at least the image data 202.

[0090] In these and other embodiments, while the ESS detector 206c may be similar to the Bi3D detector 206b in that both are ML-based detectors, the ESS detector 206c and the Bi3D detector 206b may employ different approaches to representing depth. For example, while the Bi3D detector 206b represents depth in binary classifications using different depth levels, the ESS detector 206c may represent depth in a continuous depth format. For example, the ESS detector 206c represents depth as a distance from a given reference point, rather than whether a given pixel corresponding to an object is closer to or farther from a given depth level. In some embodiments, the continuous depth format adopted by the ESS detector 206c may be used with respect to Fig. Sections 1A-1B of the present revelation will be described in more detail.

[0091] In some embodiments, the ESS detector 206c can be configured to handle all depth levels simultaneously in order to determine a probability distribution over all possible depth levels for the one or more objects in the image data 202. For example, the ESS detector 206c can implement 3D convolution to aggregate data over multiple depth levels. For example, the ESS detector 206c can extract features (e.g., the one or more objects) from the image data 202 and arrange the extracted features in a tensor that spans the multiple depth levels. In these and other embodiments, the tensor can be constructed and processed in a way that yields a probability distribution over multiple depth levels for the one or more pixels corresponding to the extracted features of the image data 202.For example, the probability distribution can specify the probabilities that the one or more pixels corresponding to the extracted features are located at different depth levels. In some embodiments, the depths of the extracted features can be determined by finding the maximum value under the probability distribution. For example, it can be determined that the one or more pixels corresponding to the extracted features are located at a depth level with the highest probability.

[0092] In some embodiments, the depth detectors 206 can be implemented on different hardware platforms. For example, one or more of the depth detectors 206 can be implemented on different hardware platforms. In these and other embodiments, different hardware platforms can be selected based on the parameters that characterize the one or more depth detectors 206. For example, the OFA 204a can be suitable for the SGM detector 206a because the OFA 204a can be dedicated to optical flow and / or stereovision calculations. For example, the OFA 204a can be configured to generate an optical flow and / or a disparity map from image data. The DLA 204b can be suitable for the Bi3D detector 206b because the DLA 204b is used to run neural networks.The GPU 204c can be suitable for the ESS detector 206c insofar as the GPU 204c can be used to provide high-performance parallel computing, which can be used by the ESS detector 206c to determine the depth of one or more objects across multiple depth levels simultaneously.

[0093] In these and other embodiments, the depth detectors 206 can have different properties and / or parameters suitable for the hardware platform. In some embodiments, the different properties and / or parameters can improve the robustness of the system 200. For example, the different properties and / or parameters of the depth detectors 206 and / or the hardware platforms can reduce the probability of an error in the detection of one or more objects. For example, differences between the depth detectors 206 can allow the system 200 to incorporate different approaches for detecting one or more objects. The different properties and / or parameters between the depth detectors, as discussed herein according to some embodiments, are illustrated and summarized in Table 1 below. Table 1 Tiefendete ktor ML / CV Eingabefor mat Flexibl er Bereic h? 3D-vs. 2D-Faltu ng (R)egressio n vs. (K)lassifika tion Inferenzzeitpara meter? SGM CV Luma N - R J Bi3D ML RGB J 2D K J ESS ML RGB N 3D R N

[0094] In some embodiments, the system 200 can include a first detection module 212a, a second detection module 212b and a third detection module 212c (collectively referred to as "detection modules 212"), which are equivalent to the detection modules 114 from Fig. 1A. In some embodiments, the detection modules 212 can be configured to obtain the depths 208 in order to determine one or more detections. For example, the first detection module 212a can be configured to determine an SGM detection 214a based at least on the SGM depth 208a, the second detection module 212b can be configured to determine a Bi3D detection 214b based at least on the Bi3D depth 208b, and the third detection module 212c can be configured to determine an ESS detection 214c based at least on the ESS depth 208c. In some embodiments, the SGM detection 214a, the Bi3D detection 214b and the ESS detection 214c can be collectively referred to as the “detections 214”.

[0095] In these and other embodiments, one or more of the detections 214 can indicate whether detections of objects are being made within the detection zone. Additionally or alternatively, one or more of the detection modules 212 can be configured to determine whether the one or more objects detected within the detection zone are located in certain areas of interest within the detection zone. In some embodiments, the determination by one or more of the detections 214 can involve one or more operations that are related to Fig. 1B are described, which correspond to determining the detections 116.

[0096] In some embodiments, the detection modules 212 can be implemented on one or more CPUs. For example, the first detection module 212a can be implemented on a first CPU 210a, the second detection module 212b can be implemented on a second CPU, and the third detection module 212b can be implemented on a third CPU 210c. In some embodiments, the first CPU 210a, the second CPU 210b, and the third CPU 210c can refer to the same CPU; for example, a single CPU can implement all detection modules 212. In some embodiments, one or more of the first CPU 210a, the second CPU 210b, and the third CPU 210c can be different from each other.

[0097] In some embodiments, the system 200 may include an aggregation module 218 implemented on a fourth CPU 216. In some embodiments, the fourth CPU 216 may be the same CPU as one or more of the first CPU 210a, the second CPU 210b, and the third CPU 210c. In some embodiments, the fourth CPU 216 may be different from one or more of the other CPUs. In some embodiments, the aggregation module 218 may be configured to obtain the SGM detection 214a, the Bi3D detection 214b, and the ESS detection 214c. The aggregation module 218 may further be configured to determine a final detection 220 based on at least the SGM detection 214a, the Bi3D detection 214b, and / or the ESS detection 214c.

[0098] In some embodiments, the final detection 220 can indicate whether one or more objects are detected in the detection zone. For example, the final detection 220 can indicate whether the one or more objects are located in the image data 202. Additionally or alternatively, the final detection 220 can indicate where in the detection zone the one or more objects are detected. In some embodiments, the final detection 220 can be a result of combining the SGM detection 214a, the Bi3D detection 214b, and the ESS detection 214c. By combining two or more of the detections 214, the final detection 220 can represent the detection of one or more objects in the detection zone more accurately than using only one of the detections 214. In some embodiments, the aggregation module 218 can determine the final detection 220 by aggregating the detections 214.In some embodiments, the aggregation of one or more of the detections 214 may include one or more processes that are related to . Fig. 1B are described, which correspond to the aggregation of the detections 116. In some embodiments, one or more control operations can be performed based on the final detection 220, as described in relation to Fig. 1B described.

[0099] Fig. Figure 3 is a flowchart illustrating a method 300 for detecting objects in detection zones of a machine according to one or more embodiments of the present disclosure. In some embodiments, one or more of the operations of method 300 may be performed with respect to system 100. Fig. 1A-1B. One or more operations of procedure 300 can be performed by any suitable system, device, or apparatus, such as system 100, the one(s) described in relation to the Fig. 4A-4D described autonomous vehicle system(s), which in relation to Fig. 5 described computing device(s) and / or the(s) in relation to Fig. 6 data system(s) described in the present disclosure.

[0100] Procedure 300 can include one or more blocks. Although illustrated with discrete blocks, the operations associated with one or more of the blocks of Procedure 300 can be subdivided into additional blocks, combined into fewer blocks, or eliminated, depending on the specific implementation.

[0101] In some embodiments, the method may include Block 302. In Block 302, sensor data corresponding to an operating range of a machine may be analyzed using multiple analysis techniques. In some embodiments, the multiple analysis techniques may be used to detect objects within the operating range. In these and other embodiments, the multiple analysis techniques may include one or more depth techniques, which may vary in how depth is determined from a reference point. For example, the one or more depth techniques may obtain the sensor data and determine the depths of one or more objects present in the sensor data. In some embodiments, the sensor data may include stereoscopic image data. For example, a stereo camera may be used to obtain the sensor data corresponding to the operating range of the machine.

[0102] In some embodiments, at least two of the multiple analysis techniques can exhibit computational diversity. For example, at least two of the analysis techniques can be configured to perform different types of computational analyses on the sensor data in relation to each other. In some embodiments, the computational diversity among the multiple analysis techniques can provide robust object detection. For example, a single analysis technique may involve certain computational errors and / or limitations. By implementing different analysis techniques, these certain computational errors and / or limitations can be reduced. In some embodiments, examples of the multiple analysis techniques in the present disclosure can be found as in relation to Fig. 1A-1B will be discussed and illustrated further.

[0103] In some embodiments, at least two of the analysis techniques from the set of techniques may exhibit implementation diversity. For example, at least two of the analysis techniques may be implemented on different types of computing platforms in relation to each other. For example, a particular analysis technique may be implemented on specific computing hardware that is most suitable for that particular analysis technique. In some embodiments, examples of the different computing platforms and corresponding analysis techniques may be given in the present disclosure as in relation to Fig. 1A-1C will be discussed and illustrated further.

[0104] In Block 304, the detection of an object within the operating range can be determined based on at least the analysis. For example, it can be determined, based at least on the results of several analysis techniques, whether one or more objects are present in the sensor data. For example, individual detection results corresponding to the respective analysis techniques of the several analysis techniques can be determined. In these and other cases, an overall detection result can be determined based at least on the individual detection results. In some embodiments, the determination of the overall detection result may involve combining the individual detection results.In some cases, the individual detection results can be combined based on one or more combination algorithms to determine the overall detection result, which represents the presence and / or position of one or more objects within the machine's operating range. In some embodiments, determining the overall detection result may involve one or more operations related to... Fig. 1A-1C are described, which correspond to determining the final detection 120.

[0105] In block 306, one or more operations of the machine can be controlled based on at least the detection of the object. For example, one or more attenuation operations can be performed by the machine in response to the detection of the object. In some embodiments, the one or more attenuation operations can vary depending on where the detection of the object is made relative to the machine. For example, the one or more attenuation operations can vary depending on at least the distance of the object from the machine. In some embodiments, examples of varying attenuation operations can be found in the present disclosure, such as with respect to Fig. 1A-1B and Fig. 1D will be discussed and illustrated further.

[0106] Modifications, additions, or omissions may be made to Method 300 without altering the scope of this disclosure. For example, the operations of Method 300 may be implemented in a different order. Additionally or alternatively, two or more operations may be performed simultaneously. Furthermore, the outlined operations and actions are provided only as examples, and some of the operations and actions may be optional, combined into fewer operations and actions, or extended into additional operations and actions without affecting the essence of the described embodiments. Exemplary autonomous vehicle

[0107] Fig. Figure 4A illustrates an exemplary autonomous vehicle 400 according to some embodiments of the present disclosure. The autonomous vehicle 400 (hereinafter referred to alternatively as "vehicle 400") may include, among other things, a passenger vehicle such as a car, truck, bus, first responder vehicle, shuttle, electric or motorized bicycle, motorcycle, fire engine, police vehicle, ambulance, boat, construction vehicle, underwater vehicle, drone and / or another type of vehicle (which, for example, is unmanned and / or carries one or more passengers).Autonomous vehicles are generally described in terms of automation levels defined by the National Highway Traffic Safety Administration (NHTSA), a division of the U.S. Department of Transportation, and the Society of Automotive Engineers (SAE) in the "Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles" (Standard No. J3016-201806, published June 15, 2018, Standard No. J3016-201609, published September 30, 2016, and previous and future versions of this standard). The Vehicle 400 may be capable of functionality corresponding to one or more of Levels 3 through 5 of autonomous driving levels. The Vehicle 400 may be capable of functionality corresponding to one or more of Levels 1 through 5 of autonomous driving levels.For example, the Vehicle 400 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), depending on the embodiment. The term "autonomous," as used herein, may include any and / or all types of autonomy for the Vehicle 400 or any other machine, such as fully autonomous, highly autonomous, conditionally autonomous, partially autonomous, providing assistive autonomy, semi-autonomous, primarily autonomous, or any other designation.

[0108] The vehicle 400 can include components such as a chassis, a vehicle body, wheels (e.g., 2, 4, 6, 8, 18, etc.), tires, axles, and other vehicle components. The vehicle 400 can include a propulsion system 450, such as an internal combustion engine, a hybrid electric power plant, a pure electric motor, and / or another type of propulsion system. The propulsion system 450 can be connected to a drivetrain of the vehicle 400, which may include a transmission to enable the propulsion of the vehicle 400. The propulsion system 450 can be controlled in response to signals received from the throttle / accelerator pedal 452.

[0109] A steering system 454, which may include a steering wheel, can be used to steer the vehicle 400 (e.g., along a desired path or route) when the drive system 450 is in operation (e.g., when the vehicle is in motion). The steering system 454 can receive signals from a steering actuator 456. The steering wheel may be optional for full automation (Level 5) functionality.

[0110] The brake sensor system 446 can be used to actuate the vehicle brakes in response to receiving signals from the brake actuators 448 and / or brake sensors.

[0111] The controller(s) 436, which controls one or more CPU(s), System-on-Chips (SoCs) 404 ( Fig. 4C) and / or GPU(s) can provide signals (e.g., representing commands) to one or more components and / or systems of the vehicle 400. For example, the controller(s) can send signals to actuate the vehicle brakes via one or more brake actuators 448, to actuate the steering system 454 via one or more steering actuators 456, and / or to actuate the propulsion system 450 via one or more throttle / accelerator pedals 452. The controller(s) 436 can include one or more onboard (e.g., integrated) computing devices (e.g., supercomputers) that process sensor signals and issue operating commands (e.g., signals representing commands) to enable autonomous driving and / or to assist a human driver in driving the vehicle 400.The control(s) 436 can include a first control 436 for autonomous driving functions, a second control 436 for functional safety functions, a third control 436 for artificial intelligence functionality (e.g., computer vision), a fourth control 436 for infotainment functionality, a fifth control 436 for redundancy in emergency conditions, and / or other controls. In some examples, a single control 436 can handle two or more of the above functionalities; two or more control 436s can handle a single functionality and / or any combination thereof.

[0112] The controller(s) 436 can provide the signals for controlling one or more components and / or systems of the vehicle 400 in response to sensor data received from one or more sensors (e.g., sensor inputs). The sensor data can be received, for example, and without limitation, from one or more sensors 458 for global navigation satellite systems (e.g., sensors for the global positioning system), one or more radar sensors 460, one or more ultrasonic sensors 462, one or more lidar sensors 464, one or more sensors 466 for inertial measurement units (IMUs) (e.g., accelerometers, gyroscopes, magnetic compasses, magnetometers, etc.), one or more microphones 496, one or more stereo cameras 468, one or more wide-angle cameras 470 (e.g., fisheye cameras), one or more infrared cameras 472, one or more surround cameras 474 (e.g.,360-degree cameras), one or more long-range and / or medium-range cameras 498, one or more speed sensors 444 (e.g., for measuring the speed of the vehicle 400), one or more vibration sensors 442, one or more steering sensors 440, one or more brake sensors 446 (e.g., as part of the brake sensor system 446), and / or other sensor types.

[0113] One or more of the controller(s) 436 can receive inputs (e.g., represented by input data) from an instrument cluster 432 of the vehicle 400 and provide outputs (e.g., represented by output data, display data, etc.) via a display 434, a human-machine interface (HMI), an audible indicator, a loudspeaker, and / or via other components of the vehicle 400. The outputs can include information such as vehicle speed, velocity, time, map data (e.g., the HD map 422 from Fig. 4C), location data (e.g., the location of vehicle 400, as on a map), direction, location of other vehicles (e.g., an occupancy grid), information about objects and the status of objects as perceived by the control(s) 436, etc. For example, the HMI display 434 can show information about the presence of one or more objects (e.g., a road sign, warning sign, traffic light change, etc.) and / or information about driving maneuvers that the vehicle has performed, is performing, or will perform (e.g., changing lanes now, entering exit 34B in two miles, etc.).

[0114] The vehicle 400 further includes a network interface 424, which can use one or more wireless antennas 426 and / or one or more modems to communicate over one or more networks. For example, the network interface 424 can be capable of communication over LTE, WCDMA, UMTS, GSM, CDMA2000, etc. The wireless antenna(s) 426 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 LE, Z-Wave, ZigBee, etc., and / or one or more low-power wide area networks (LPWANs), such as LoRaWAN, SigFox, etc.

[0115] Fig. 4B is an example of camera positions and fields of view for the exemplary autonomous vehicle 400 from Fig. 4A according to some embodiments of the present disclosure. The cameras and respective fields of view are an exemplary embodiment and are not intended to be limiting. For example, additional and / or alternative cameras may be included and / or the cameras may be located at different positions on the vehicle 400.

[0116] The camera types may include, but are not limited to, digital cameras that can be adapted for use with the components and / or systems of the Vehicle 400. The camera(s) may operate at Automotive Safety Integrity Level (ASIL) B and / or another ASIL. The camera types may be capable of any frame rate, such as 60 frames per second (fps), 120 fps, 240 fps, etc., depending on the configuration. The cameras may be capable of using rolling apertures, global apertures, another type of aperture, or a combination thereof. In some examples, the color filter array may include a Red Clear Clear Clear (RCCC) color filter array, a Red Clear Clear Blue (RCCB) color filter array, a Red Blue Green Clear (RBGC) color filter array, a Foveon X3 color filter array, a Bayer Sensors (RGGB) color filter array, a Monochrom Sensor color filter array and / or another type of color filter array.In some embodiments, clear pixel cameras, such as cameras with an RCCC, an RCCB and / or an RBGC color filter array, can be used in an effort to increase light sensitivity.

[0117] In some examples, one or more of the camera(s) 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 mono camera can be installed to provide functions that include lane departure warning, traffic sign recognition, and intelligent headlight control. One or more of the camera(s) (e.g., all cameras) can simultaneously record and provide image data (e.g., video).

[0118] One or more of the cameras can be mounted in an assembly, such as a custom-designed (3D-printed) assembly, to filter out stray light and reflections from inside the car (e.g., reflections from the dashboard that are reflected in the windshield mirrors) that can interfere with the camera's image acquisition capabilities. Referring to vane mirror mounting assemblies, the vane mirror assemblies can be custom 3D-printed so that the camera mounting plate matches the shape of the vane mirror. In some examples, the camera(s) can be integrated into the vane mirror. For side-view cameras, the camera(s) can also be integrated into the four pillars at each corner of the cabin.

[0119] Cameras with a field of view that includes portions of the environment in front of the vehicle (e.g., forward-facing cameras) can be used for surround view to help identify forward paths and obstacles, and to help provide, with the aid of one or more controllers and / or control SoCs, information critical for generating an occupancy grid and / or determining preferred vehicle paths. Forward-facing cameras can be used to perform many of the same ADAS functions as LiDAR, including emergency braking, pedestrian detection, and collision avoidance. Forward-facing cameras can also be used for ADAS functions and systems that include lane departure warnings (LDW), autonomous cruise control (ACC), and / or other functions, such as traffic sign recognition.

[0120] A variety of cameras can be used in a forward-facing configuration, including, for example, a monocular camera platform incorporating a CMOS (Complementary Metal Oxide Semiconductor) color imager. Another example could be a 470 long-range camera, which can be used to detect objects coming into view from the periphery (e.g., pedestrians, crossing traffic, or bicycles). Although in Fig. While Figure 4B illustrates only one long-range camera, there can be any number of long-range cameras 470 on the vehicle 400. Additionally, a long-range camera 498 (e.g., a pair of long-range stereo cameras) can be used for depth-based object detection, especially for objects for which a neural network has not yet been trained. The long-range camera 498 can also be used for object detection and classification, as well as basic object tracking.

[0121] One or more Stereo Cameras 468 can also be included in a forward-facing configuration. The Stereo Camera(s) 468 can include an integrated control unit comprising a scalable processing unit that can provide programmable logic (FPGA) and a multi-core microprocessor with an integrated CAN or Ethernet interface on a single chip. Such a unit can be used to generate a 3D map of the vehicle's surroundings, including a distance estimate for all points in the image. An alternative Stereo Camera(s) 468 can include one or more compact stereo vision sensors, which can include two camera lenses (one each on the left and right) and an image processing chip that can measure the distance from the vehicle to the target object and use the generated information (e.g., metadata) to activate the autonomous emergency braking and lane departure warning functions.Other types of stereo camera(s) 468 can be used in addition to or as an alternative to those described here.

[0122] Cameras with a field of view that includes sections of the environment to the sides of the vehicle 400 (e.g., side-view cameras) can be used for surround view, providing information that is used to generate and update the occupancy grid, as well as to generate side-impact collision warnings. For example, the surround-view camera(s) 474 (e.g., four surround-view cameras 474, as in Fig. (Figure 4B illustrates) is positioned on the vehicle 400. The surround-view camera(s) 474 may include one or more wide-angle cameras 470, one or more fisheye cameras, one or more 360-degree cameras, and / or the like. For example, four fisheye cameras may be positioned at the front, rear, and sides of the vehicle. In an alternative arrangement, the vehicle may use three surround-view camera(s) 474 (e.g., left, right, and rear) and may use one or more other cameras (e.g., a forward-facing camera) as a fourth surround-view camera.

[0123] Cameras with a field of view that includes portions of the area behind the vehicle 400 (e.g., rear-view cameras) can be used for parking assistance, surround view, rear collision warnings, and generating and updating the occupancy grid. A wide variety of cameras can be used, including, but not limited to, cameras that are also suitable as one or more forward-facing cameras (e.g., long-range and / or medium-range camera(s) 498, stereo camera(s) 468, infrared camera(s) 472, etc.), as described herein.

[0124] Fig. 4C is a block diagram of an exemplary system architecture for the exemplary autonomous vehicle 400 from Fig. 4A according to some embodiments of the present disclosure. It is understood that these and other arrangements described herein are presented only as examples. Other arrangements and elements (e.g., machines, interfaces, functions, sequences, groupings of functions, etc.) may be used in addition to or instead of those shown, and some elements may be omitted entirely. Furthermore, 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 position. Various functions described herein as being performed by entities may be executed by hardware, firmware, and / or software. For example, various functions may be performed by a processor executing instructions stored in memory.

[0125] Each of the components, features and systems of the 400 vehicle in Fig. 4C is illustrated as being connected via bus 402. Bus 402 may include a Controller Area Network (CAN) data interface (referred to herein alternatively as the "CAN bus"). A CAN can be a network within the vehicle 400 that is used to assist in controlling various features and functionality of the vehicle 400, such as actuation of brakes, acceleration, braking, steering, windshield wipers, etc. A CAN bus may be configured to have dozens or even hundreds of nodes, each with its own unique identifier (e.g., a CAN ID). The CAN bus can be read to obtain 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.

[0126] Although the 402 bus is described herein as a CAN bus, this is not intended to be restrictive. For example, FlexRay and / or Ethernet may be used in addition to or as an alternative to the CAN bus. Although a single line is used to represent the 402 bus, this is not intended to be restrictive. For example, there may be any number of 402 buses, 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 a different protocol. In some examples, two or more 402 buses may be used to perform different functions and / or for redundancy. For example, a first 402 bus may be used for collision avoidance functionality, and a second 402 bus may be used for actuation control.In any given example, each bus 402 can communicate with any of the vehicle 400 components, and two or more buses 402 can communicate with the same components. In some examples, each SoC 404, each controller 436, and / or each computer within the vehicle can have access to the same input data (e.g., inputs from vehicle 400 sensors) and can be connected to a common bus, such as the CAN bus.

[0127] The vehicle 400 can include one or more control units 436, such as those described herein in relation to Fig. 4A described. The controller(s) 436 can be used for a variety of functions. The controller(s) 436 can be coupled with any of the various other components and systems of the vehicle 400 and can be used to control the vehicle 400, the vehicle 400's artificial intelligence, the vehicle 400's infotainment system, and / or the like.

[0128] The vehicle 400 can include one or more system-on-a-chip (SoC) 404. The SoC 404 can include CPU(s) 406, GPU(s) 408, processor(s) 410, cache(s) 412, accelerator 414, data storage 416, and / or other components and features not illustrated. The SoC(s) 404 can be used to control the vehicle 400 in a variety of platforms and systems. For example, the SoC(s) 404 can be combined in a system (e.g., the system of the vehicle 400) with an HD card 422, which performs map updates and / or updates via a network interface 424 from one or more servers (e.g., server 478). Fig. 4D) can be obtained.

[0129] The CPU(s) 406 may include a CPU cluster or CPU complex (hereinafter referred to as "CCPLEX"). The CPU(s) 406 may include multiple cores and / or L2 caches. For example, in some embodiments, the CPU(s) 406 may include eight cores in a coherent multiprocessor configuration. In some embodiments, the CPU(s) 406 may include four dual-core clusters, each cluster having a dedicated L2 cache (e.g., a 2 MB L2 cache). The CPU(s) 406 (e.g., the CCPLEX) may be configured to support concurrent cluster operation, allowing any combination of CPU(s) 406 clusters to be active at any given time.

[0130] The CPU(s) 406 can implement power management capabilities that include one or more of the following features: Individual hardware blocks can be automatically clocked while idle to save dynamic power; Each core can be clocked when the core is not actively executing instructions due to the execution of WFI / WFE instructions; Each core can be clocked independently; Each core cluster can be clocked independently when all cores are clocked; and / or Each core cluster can be clocked independently when all cores are clocked. The CPU(s) 406 can further implement an improved power state management algorithm, specifying acceptable power states and expected wake-up times, with the hardware / microcode determining the best power state to input for the core, cluster, and CCPLEX.The processing kernels can support simplified performance state entry sequences in software, offloading the work to microcode.

[0131] The GPU(s) 408 may include an integrated GPU (referred to herein alternatively as an "iGPU"). The GPU(s) 408 may be programmable and efficient for parallel workloads. The GPU(s) 408 may, in some examples, use an extended tensor instruction set. The GPU(s) 408 may include one or more streaming microprocessors, each of which may include an L1 cache (e.g., an L1 cache with a minimum storage capacity of 96 KB), and two or more of the streaming microprocessors may share an L2 cache (e.g., an L2 cache with a storage capacity of 512 KB). In some embodiments, the GPU(s) 408 may include at least eight streaming microprocessors. The GPU(s) 408 can use one or more computing application programming interfaces (APIs).Additionally, the GPU(s) 408 can use one or more parallel computing platforms and / or programming models (e.g., NVIDIA's CUDA).

[0132] The GPU(s) 408 can be performance-optimized for best performance in automotive and embedded applications. For example, the GPU(s) 408 can be manufactured using a Fin field-effect transistor (FinFET). However, this is not a limitation, and the GPU(s) 408 can be manufactured using other semiconductor manufacturing processes. Each streaming microprocessor can contain a number of mixed-precision processing cores partitioned into multiple blocks. For example, and without limitation, 64 PF32 cores and 32 PF64 cores can be partitioned into four processing blocks. In such an example, each processing block could be allocated 16 FP32 cores, 8 FP64 cores, 16 INT32 cores, two mixed-precision NVIDIA TENSOR COREs for deep learning matrix arithmetic, a low-instruction cache, a warp scheduler, a dispatch unit, and / or a 64 KB register file.Additionally, streaming microprocessors can include independent parallel integer and floating-point data paths to provide efficient execution of workloads with a mix of computational and addressing operations. Streaming microprocessors can include independent thread scheduling capabilities to enable finer synchronization and collaboration between parallel threads. Streaming microprocessors can also include a combined L1 data cache and shared memory to improve performance while simplifying programming.

[0133] The GPU(s) 408 can include High Bandwidth Memory (HBM) and / or a 16 GB HBM2 memory subsystem to provide a peak memory bandwidth of approximately 900 GB / second in some examples. In some examples, in addition to or as an alternative to HBM memory, synchronous graphics random access memory (SGRAM), such as double-data-rate type five synchronous graphics random access memory (GDDR5), can be used.

[0134] The GPU(s) 408 can incorporate a unified memory technology that includes access counters to enable more accurate migration of memory pages to the processor that accesses them most frequently, thereby improving the efficiency of memory areas shared between processors. In some examples, support for Address Translation Services (ATS) can be used to allow the GPU(s) 408 to directly access the page tables of the CPU(s) 406. In such examples, if the memory management unit (MMU) of the GPU(s) 408 experiences a failure, an address translation request can be passed to the CPU(s) 406. In response, the CPU(s) 406 can search its page tables for the virtual-to-physical mapping for the address and pass the translation back to the GPU(s) 408.Thus, the unified memory technology can enable a single unified virtual address space for memory of both the CPU(s) 406 and the GPU(s) 408, thereby simplifying the programming and porting of applications from GPU(s) 408 to GPU(s) 408.

[0135] Additionally, the GPU(s) 408 can include an access counter that tracks the frequency of GPU 408 access to the memory of other processors. This access counter can help ensure that memory pages are moved to the physical memory of the processor that accesses them most frequently.

[0136] The SoC(s) 404 can include any number of Cache(s) 412, including those described herein. For example, the Cache(s) 412 can include an L3 cache that is available to both the CPU(s) 406 and the GPU(s) 408 (e.g., connected to both the CPU(s) 406 and the GPU(s) 408). The Cache(s) 412 can include a write-back cache capable of tracking row states, such as using a cache coherence protocol (e.g., MEI, MESI, MSI, etc.). The L3 cache can be 4 MB or larger, depending on the implementation, although smaller cache sizes are possible.

[0137] The SoC(s) 404 may include an arithmetic logic unit (ALU) that can be used to perform processing related to any of the many tasks or operations of the Vehicle 400, such as processing deep neural networks (DNNs). Additionally, the SoC(s) 404 may include a floating-point unit (FPU) or other mathematical or numeric coprocessor types for performing mathematical operations within the system. For example, the SoC(s) 404 may include one or more FPUs integrated as execution units within a CPU(s) 406 and / or GPU(s) 408.

[0138] The SoC(s) 404 can include one or more accelerators 414 (e.g., hardware accelerators, software accelerators, or a combination thereof). For example, the SoC(s) 404 can include a hardware acceleration cluster, which may contain optimized hardware accelerators and / or large on-chip memory. The large on-chip memory (e.g., 4 MB SRAM) can enable the hardware acceleration cluster to accelerate neural networks and other computations. The hardware acceleration cluster can be used to complement the GPU(s) 408 and offload some of the GPU's tasks (e.g., to free up more GPU cycles for other tasks). As an example, the Accelerator 414 can be used for targeted workloads (e.g., perception, convolutional neural networks (CNNs), etc.) that are stable enough to be accelerable.The term “CNN”, as used herein, can include all types of CNNs, including region-based or regional convolutional neural networks (RCNNs) and fast RCNNs (e.g., such as those used for object detection).

[0139] The Accelerator 414 (e.g., the hardware acceleration cluster) can include one or more Deep Learning Accelerators (DLAs). The DLA(s) can include one or more Tensor Processing Units (TPUs), which can be configured to provide an additional ten trillion operations per second for deep learning applications and inference. The TPUs can be accelerators configured and optimized to perform image processing functions (e.g., for CNNs, RCNNs, etc.). The DLA(s) can further be optimized for a specific set of neural network types and floating-point operations, as well as inference. The design of the DLA(s) can provide more performance per millimeter than a general-purpose GPU and significantly exceeds the performance of a CPU.The TPU(s) can perform several functions, including a single-instance folding function that supports, for example, INT8, INT16 and FP16 data types for both features and weights, as well as post-processor functions.

[0140] The DLA(s) can quickly and efficiently execute neural networks, especially CNNs, on processed or unprocessed data for any of a wide variety of functions, including, but not limited to: a CNN for object identification and detection using data from camera sensors; a CNN for distance estimation using data from camera sensors; a CNN for emergency vehicle detection and identification using data from microphones; a CNN for facial recognition and vehicle owner identification using data from camera sensors; and / or a CNN for safety and / or security-related events.

[0141] The DLA(s) can perform any function of the GPU(s) 408, and by using an inference accelerator, a designer can, for example, target either the DLA(s) or the GPU(s) 408 for any given function. For example, the designer can concentrate the processing of CNNs and floating-point operations on the DLA(s) and leave other functions to the GPU(s) 408 and / or the other accelerator(s) 414.

[0142] The Accelerator 414 (e.g., the hardware acceleration cluster) may include one or more Programmable Vision Accelerators (PVAs), which may alternatively be referred to herein as a computer vision accelerator. The PVA(s) 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(s) may provide a balance between performance and flexibility. For example, each PVA may, without limitation, include any number of Reduced Instruction Set Computer (RISC) cores, Direct Memory Access (DMA), and / or any number of vector processors.

[0143] The RISC cores can interact with image sensors (e.g., the image sensors of any of the cameras described herein), image signal processor(s), and / or the like. Each RISC core can include any amount of memory. Depending on the embodiment, the RISC cores can use any of a number of protocols. In some examples, the RISC cores can run a real-time operating system (RTOS). The RISC cores can be implemented using one or more integrated circuit devices, application-specific integrated circuits (ASICs), and / or memory devices. For example, the RISC cores can include an instruction cache and / or tightly coupled RAM.

[0144] The DMA can enable components of the PVA(s) to access system memory independently of the CPU(s) 406. The DMA can support any number of features used to provide optimization for the PVA, including, but not limited to, supporting multidimensional addressing and / or circular addressing. In some examples, the DMA can support up to six or more dimensions of addressing, which may include block width, block height, block depth, horizontal block step size, vertical block step size, and / or depth step size.

[0145] Vector processors can be programmable processors designed to efficiently and flexibly execute programming for computer vision algorithms and provide signal processing capabilities. In some examples, the PVA might include a PVA core and two vector processing subsystem partitions. The PVA core may include a processor subsystem, DMA engine(s) (e.g., two DMA engines), and / or other peripherals. The vector processing subsystem can act as the primary processing engine of the PVA and may include a vector processing unit (VPU), an instruction cache, and / or vector memory (e.g., VMEM). A VPU core might include a digital signal processor, such as a single-instruction, multiple-data (SIMD) digital signal processor and a very long instruction word (VLIW). The combination of SIMD and VLIW can improve throughput and speed.

[0146] Each vector processor can include an instruction cache and can be coupled to dedicated memory. Consequently, in some examples, each vector processor can be configured to run independently of the others. In other examples, the vector processors included in a particular PVA can be configured to employ data parallelism. For example, in some embodiments, the multitude of vector processors included in a single PVA can execute the same computer viewing algorithm, but on different regions of an image. In other examples, the vector processors included in a particular PVA can simultaneously execute different computer viewing algorithms on the same image, or even different algorithms on successive images or sections of an image.Among other things, any number of PVAs can be included in the hardware acceleration cluster, and any number of vector processors can be included in each of the PVAs. Additionally, the PVA(s) can include additional error correction code (ECC) memory to improve overall system security.

[0147] The Accelerator 414 (e.g., the hardware acceleration cluster) may include an on-chip and SRAM computer view network to provide high-bandwidth, low-latency SRAM for the Accelerator 414. In some examples, the on-chip memory may include at least 4 MB of SRAM, consisting, for example, and without limitation, of eight field-configurable memory blocks accessible to 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 the DLA can access the memory via a backbone that provides high-speed memory access to the PVA and the DLA. The backbone may include an on-chip computer view network that connects the PVA and the DLA to the memory (e.g., using the APB).

[0148] The on-chip computer network can include an interface that, prior to the transmission of a control signal / address / data, determines that both the PVA and the DLA provide ready and valid signals. Such an interface can provide separate phases and channels for transmitting control signals / addresses / data, as well as burst-type communication for continuous data transmission. This type of interface can conform to the ISO 26262 or IEC 61508 standards, although other standards and protocols can be used.

[0149] In some examples, the SoC(s) 404 may include a real-time ray-tracing hardware accelerator, as described in U.S. Patent Application No. 16 / 101,232, filed on August 10, 2018. The real-time ray-tracing hardware accelerator can be used to quickly and efficiently determine the positions and dimensions of objects (e.g., within a world model) to generate real-time visualization simulations for radar signal interpretation, sound propagation synthesis and / or analysis, SONAR system simulation, general wave propagation simulation, comparison with lidar data for localization purposes, and / or other functions and / or other uses. In some embodiments, one or more tree traversal units (TTUs) may be used to perform one or more ray-tracing-related operations.

[0150] The Accelerator 414 (e.g., the hardware accelerator cluster) has a wide range of uses for autonomous driving. The PVA can be a programmable vision accelerator used for key processing stages in ADAS and autonomous vehicles. The PVA's capabilities are a good match for algorithmic domains that require predictable processing with low power consumption and low latency. In other words, the PVA performs well on semi-dense or dense regular computation, even on small datasets, that require predictable runtimes with low latency and low power consumption. Thus, in the context of autonomous vehicle platforms, PVAs are designed to execute classical computer vision algorithms because they are efficient at object detection and operate on integer mathematics.

[0151] For example, according to one embodiment of the technology, the PVA is used to perform computer stereo vision. In some examples, a semi-global matching-based algorithm can be used, although this is not intended to be restrictive. Many Level 3-5 autonomous driving applications require on-the-fly motion estimation / stereo matching (e.g., structure of motion, pedestrian detection, lane detection, etc.). The PVA can perform computer stereo vision on input from two monocular cameras.

[0152] In some examples, the PVA can be used to perform dense optical flow processing. For instance, raw RADAR data is processed (e.g., using a fast 4D Fourier transform) to process the RADAR data. In other examples, the PVA is used for runtime depth processing, for example, by processing raw runtime data to provide processed runtime data.

[0153] The DLA can be used to run any type of network to improve control and driving safety, including, for example, a neural network that outputs a confidence score for each object detection. Such a confidence score can be interpreted as a probability or as providing a relative "weight" for each detection compared to other detections. This confidence score allows the system to make further decisions about which detections should be considered true positives rather than false positives. For example, the system can set a confidence threshold and consider only those detections that exceed the threshold as true positives. In an automatic emergency braking (AEB) system, false positives would cause the vehicle to automatically perform emergency braking, which is obviously undesirable.Therefore, only the most reliable detections should be considered as triggers for AEB. The DLA can run a neural network to reduce the confidence level. The neural network can take as its input at least a subset of parameters, such as boundary box dimensions, ground level estimation obtained (e.g., from another subsystem), output from sensor 466 of the inertial measurement unit (IMU) correlated with the vehicle's orientation 400, distance, 3D position estimates of the object obtained from the neural network and / or other sensors (e.g., LIDAR sensor(s) 464 or RADAR sensor(s) 460), among others.

[0154] The SoC(s) 404 may include data storage 416 (e.g., memory). The data storage 416 may be on-chip memory of the SoC(s) 404 capable of storing neural networks to be executed on the GPU and / or the DLA. In some examples, the data storage 416 may be large enough to store multiple instances of neural networks for redundancy and security. The data storage 416 may include L2 or L3 cache(s) 412. A reference to the data storage 416 may include a reference to the memory allocated to the PVA, the DLA, and / or the other accelerator(s) 414, as described herein.

[0155] The SoC(s) 404 may include one or more Processor(s) 410 (e.g., embedded processors). The Processor(s) 410 may include a Boot and Power Management Processor, which may be a dedicated processor and subsystem for handling boot power and management functions and associated security enforcement. The Boot and Power Management Processor may be part of the SoC(s) 404's boot sequence and may provide runtime power management services. The Boot and Power Management Processor may provide clock and voltage programming, assistance with system low-power state transitions, management of SoC(s) 404's thermal and temperature sensors, and / or management of SoC(s) 404's power states.Each temperature sensor can be implemented as a ring oscillator whose output frequency is proportional to the temperature, and the SoC(s) 404 can use the ring oscillators to detect temperatures of the CPU(s) 406, GPU(s) 408, and / or accelerator(s) 414. If temperatures are determined to exceed a threshold, the boot and power management processor can enter a temperature fault routine and put the SoC(s) 404 into a lower power state and / or put the vehicle 400 into a chauffeur-in-safe-stop mode (e.g., bring the vehicle 400 to a safe stop).

[0156] The 410 processor(s) may further include a set of embedded processors that can serve as an audio processing engine. The audio processing engine can be an audio subsystem that provides full hardware support for multi-channel audio across multiple interfaces and a wide and flexible range of audio I / O interfaces. In some examples, the audio processing engine is a dedicated processor core with a digital signal processor and dedicated RAM.

[0157] The 410 processor(s) 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 control peripherals, and routing logic.

[0158] The 410 processor(s) may also include a security cluster engine, which comprises a dedicated processor subsystem for handling security management for automotive applications. The security cluster engine may include two or more processor cores, tightly coupled RAM, supporting peripherals (such as timers, interrupt control, etc.), and / or routing logic. In a security mode, the two or more cores may operate in a lock-step mode, functioning as a single core with comparison logic to detect differences between their operations.

[0159] The processor(s) 410 may also include a real-time camera engine, which may include a dedicated processor subsystem for handling real-time camera management.

[0160] The 410 processor(s) 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.

[0161] The processor(s) 410 may include a video image composer, which may be a processing block (e.g., implemented on a microprocessor) that implements video post-processing functions required by a video playback application to generate the final image for the playback window. The video image composer may perform lens distortion correction on the wide-view camera(s) 470, the omnidirectional camera(s) 474, and / or on in-cabin monitoring camera sensors. An in-cabin monitoring camera sensor is preferably monitored by a neural network running on a separate instance of the Advanced SoC, configured to identify and respond to events in the cabin.The in-cabin system can lip-read to activate cellular service and make phone calls, dictate emails, change the vehicle's destination, activate or change the infotainment system and vehicle settings, or provide voice-activated web browsing. Certain functions are only available to the driver when the vehicle is operating in autonomous mode and are otherwise disabled.

[0162] The video image composer can include enhanced temporal noise reduction for both spatial and temporal noise reduction. For example, if there is motion in a video, the noise reduction appropriately weights spatial information, thus reducing the impact of information provided by adjacent frames. If a frame or portion of a frame does not contain motion, the temporal noise reduction performed by the video image composer can use information from the previous frame to reduce noise in the current frame.

[0163] The video image composer can also be configured to perform stereo rectification on input stereo lens images. Furthermore, the video image composer can be used for user interface composition when the operating system desktop is in use and the GPU(s) 408 are not required to continuously render new surfaces. Even when the GPU(s) 408 are powered on and actively performing 3D rendering, the video image composer can be used to offload the GPU(s) 408 to improve performance and responsiveness.

[0164] The SoC(s) 404 may further include a Mobile Industrial Processor Interface (MIPI) camera serial interface for receiving video and camera input, a high-speed interface, and / or a video input block that can be used for camera and related pixel input functions. The SoC(s) 404 may also include software-controlled input / output control(s) that can be used to receive I / O signals not assigned to a specific role.

[0165] The SoC(s) 404 can also include a wide range of peripheral interfaces to enable communication with peripheral devices, audio codecs, power management, and / or other devices. The SoC(s) 404 can be used to process data from cameras (e.g., connected via Gigabit Multimedia Serial Link and Ethernet), sensors (e.g., LiDAR sensor(s) 464, RADAR sensor(s) 460, etc., which may be connected via Ethernet), data from bus 402 (e.g., vehicle speed 400, steering wheel position, etc.), and data from GNSS sensor(s) 458 (e.g., connected via Ethernet or CAN bus). The SoC(s) 404 may also include dedicated high-performance mass storage controllers, which may include their own DMA engines and which can be used to offload routine data management tasks from the CPU(s) 406.

[0166] The SoC(s) 404 can be an end-to-end platform with a flexible architecture spanning automation levels 3-5, providing a comprehensive functional safety architecture that leverages and efficiently utilizes computer vision and ADAS techniques for diversity and redundancy, and provides a platform for a flexible, reliable driving software stack along with deep learning tools. The SoC(s) 404 can be faster, more reliable, and even more energy-efficient and compact than conventional systems. For example, the accelerator(s) 414, when combined with the CPU(s) 406, the GPU(s) 408, and the data storage(s) 416, can provide a fast, efficient platform for autonomous vehicles at levels 3-5.

[0167] This technology thus provides capabilities and functionality that cannot be achieved by conventional systems. For example, computer vision algorithms can be run on CPUs that can be configured using a high-level programming language, such as C, to execute a wide variety of processing algorithms over a wide variety of visual data. However, CPUs are often unable 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, which is a requirement for in-vehicle ADAS applications and for practical Level 3-5 autonomous vehicles.

[0168] Unlike conventional systems, the technology described herein, by providing a CPU complex, a GPU complex, and a hardware acceleration cluster, enables multiple neural networks to run simultaneously and / or sequentially and to combine the results to enable autonomous driving functionality at levels 3-5. For example, a CNN running on the DLA or the dGPU (e.g., the GPU(s) 420) can include text and word recognition, allowing the supercomputer to read and understand traffic signs, including signs for which the neural network has not been specifically trained.The DLA can further include a neural network that is able to identify and interpret the character, provide a semantic understanding of the character, and forward this semantic understanding to the path planning modules running on the CPU complex.

[0169] As another example, multiple neural networks can run simultaneously, as required for driving at levels 3, 4, or 5. For instance, a warning sign consisting of "Warning: Flashing lights indicate icy conditions" can be interpreted independently or jointly by multiple neural networks, along with an electric warning light. The sign itself can be identified as a traffic sign by a first neural network (e.g., a trained neural network), while the text "Flashing lights indicate icy conditions" can be interpreted by a second neural network, which informs the vehicle's path planning software (preferably running on the CPU) that the detection of flashing lights indicates icy conditions.The flashing light can be identified across multiple images by operating a third, deployed neural network, which informs the vehicle's path planning software about the presence (or absence) of flashing lights. All three neural networks can run simultaneously, for example, within the DLA and / or on the GPU(s) 408.

[0170] In some examples, a CNN for facial recognition and vehicle owner identification can use data from camera sensors to identify the presence of an authorized driver and / or owner of the vehicle 400. The always-on sensor processing engine can be used to unlock the vehicle when the owner approaches the driver's door and turn on the lights, and in security mode, to disable the vehicle when the owner leaves. In this way, the SoC(s) 404 provide security against theft and / or carjacking.

[0171] In another example, a CNN for emergency vehicle detection and identification can use data from microphones 496 to detect and identify emergency vehicle sirens. Unlike conventional systems that use general classifiers to detect sirens and manually extract features, the SoC(s) 404 use the CNN to classify ambient and urban noise as well as visual data. In a preferred embodiment, the CNN running on the DLA is trained to identify the relative closing speed of the emergency vehicle (e.g., by using the Doppler effect). The CNN can also be trained to identify emergency vehicles specific to the local area in which the vehicle operates, as identified by the GNSS sensor(s) 458.Thus, for example, when operating in Europe, the CNN attempts to detect European sirens, and when operating in the United States, it attempts to identify only North American sirens. Once an emergency vehicle is detected, a control program can be used to execute an emergency vehicle safety routine, slowing the vehicle down, pulling it to the side of the road, parking the vehicle, and / or putting the vehicle into neutral with the assistance of ultrasonic sensors 462 until the emergency vehicle(s) pass by.

[0172] The vehicle may include a CPU(s) 418 (e.g., discrete CPU(s) or dCPU(s)) which may be coupled to the SoC(s) 404 via a high-speed connection (e.g., PCIe). The CPU(s) 418 may, for example, include an x86 processor. The CPU(s) 418 may be used to perform any one of a variety of functions, including, for example, arbitrating potentially inconsistent results between ADAS sensors and the SoC(s) 404 and / or monitoring the status and health of the control unit(s) 436 and / or the infotainment SoC 430.

[0173] The Vehicle 400 can include a GPU(s) 420 (e.g., discrete GPU(s) or dGPU(s)) that can be coupled to the SoC(s) 404 via a high-speed connection (e.g., NVIDIA's NVLINK). The GPU(s) 420 can provide additional artificial intelligence functionality, such as running redundant and / or different neural networks, and can be used to train and / or update neural networks based on inputs (e.g., sensor data) from the Vehicle 400's sensors.

[0174] The vehicle 400 may also include the network interface 424, which may include one or more wireless antennas 426 (e.g., one or more wireless antennas for various communication protocols, such as a cellular antenna, a Bluetooth antenna, etc.). The network interface 424 can be used to enable wireless connectivity via the internet to the cloud (e.g., to the server(s) 478 and / or other network devices), to other vehicles, and / or to computing devices (e.g., passenger client devices). To communicate with other vehicles, a direct connection between the two vehicles and / or an indirect connection (e.g., via networks and the internet) can be established. Direct connections can be provided using a vehicle-to-vehicle communication link.The vehicle-to-vehicle communication link can provide the vehicle 400 with information about vehicles in its vicinity (e.g., vehicles in front of, to the side of, and / or behind the vehicle 400). This functionality can be part of a cooperative adaptive cruise control functionality of the vehicle 400.

[0175] The network interface 424 can include a system-on-a-chip (SoC) that provides modulation and demodulation functionality, enabling the controller(s) 436 to communicate over wireless networks. The network interface 424 can include a high-frequency (RF) front-end for upconversion from baseband to RF and downconversion from RF to baseband. The frequency conversions can be performed using well-known processes and / or superheterodyne processes. In some examples, the RF front-end functionality can be provided by a separate chip. The network interface can include wireless functionality for communication over LTE, WCDMA, UMTS, GSM, CDMA2000, Bluetooth, Bluetooth LE, Wi-Fi, Z-Wave, ZigBee, LoRaWAN, and / or other wireless protocols.

[0176] The vehicle 400 may further include one or more data storage devices 428, which may include off-chip memory (e.g., outside the SoC(s) 404). The data storage device(s) 428 may include one or more memory elements, including RAM, SRAM, DRAM, VRAM, flash, hard disks, and / or other components and / or devices capable of storing at least one data bit.

[0177] The vehicle 400 can further include one or more GNSS sensor(s) 458. The GNSS sensor(s) 458 (e.g., GPS, supported GPS sensors, differential GPS sensors (DGPS sensors), etc.) are used to assist with mapping, perception, occupancy grid generation, and / or path planning functions. Any number of GNSS sensor(s) 458 can be used, including, for example, and without limitation, a GPS using a USB connector with an Ethernet-to-serial (RS-232) bridge.

[0178] The vehicle 400 may also include one or more RADAR sensor(s) 460. The RADAR sensor(s) 460 may be used by the vehicle 400 for long-range vehicle detection, even in darkness and / or severe weather conditions. RADAR functional safety levels can be ASIL B. The RADAR sensor(s) 460 may, in some examples, use CAN and / or bus 402 (e.g., to transmit data generated by the RADAR sensor(s) 460) for control and to access object tracking data, with Ethernet access for accessing raw data. A wide variety of RADAR sensor types may be used. For example, and without limitation, the RADAR sensor(s) 460 may be suitable for front, rear, and side radar applications. In some examples, one or more pulse Doppler radar sensors are used.

[0179] The RADAR sensor(s) 460 can incorporate various configurations, such as a long-range sensor with a narrow field of view, a short-range sensor with a wide field of view, short-range side coverage, etc. In some examples, a long-range RADAR can be used for adaptive cruise control functionality. The long-range RADAR systems can provide a wide field of view, achieved through two or more independent scans, such as within a range of 250 m. The RADAR sensor(s) 460 can help distinguish between stationary and moving objects and can be used by ADAS systems for emergency braking assistance and forward collision warning. Long-range RADAR sensors can incorporate monostatic multimodal RADAR with multiple (e.g., six or more) fixed RADAR antennas and a high-speed CAN and FlexRay interface.In an example with six antennas, the middle four antennas can generate a focused beam pattern designed to record the area around vehicle 400 at higher speeds with minimal interference from traffic in adjacent lanes. The other two antennas can expand the field of view, making it possible to quickly detect vehicles entering or exiting vehicle 400's lane.

[0180] Medium-range radar systems, for example, can have a range of up to 160 m (front) or 80 m (rear) and a field of view of up to 42 degrees (front) or 150 degrees (rear). Short-range radar systems can, without restriction, include radar sensors designed to be installed at both ends of the rear bumper. When installed at both ends of the rear bumper, such radar sensor systems can generate two beams that continuously monitor the blind spot at the rear and to the sides of the vehicle.

[0181] Short-range radar systems can be used in an ADAS system for blind spot detection and / or lane change assistance.

[0182] The vehicle 400 may also include one or more ultrasonic sensor(s) 462. The ultrasonic sensor(s) 462, which may be positioned at the front, rear, and / or sides of the vehicle 400, may be used for parking assistance and / or for generating and updating an occupancy grid. A wide variety of ultrasonic sensor(s) 462 may be used, and different ultrasonic sensor(s) 462 may be used for different detection ranges (e.g., 2.5 m, 4 m). The ultrasonic sensor(s) 462 may operate at functional safety levels of ASIL B.

[0183] The vehicle 400 can include one or more LiDAR sensor(s) 464. The LiDAR sensor(s) 464 can be used for object and pedestrian detection, emergency braking, collision avoidance, and / or other functions. The LiDAR sensor(s) 464 can be functional safety level ASIL B. In some examples, the vehicle 400 can include multiple LiDAR sensors 464 (e.g., two, four, six, etc.) that can use Ethernet (e.g., to provide data to a Gigabit Ethernet switch).

[0184] In some examples, the LIDAR sensor(s) 464 may be capable of providing a list of objects and their distances for a 360-degree field of view. The commercially available LIDAR sensor(s) 464 may have an advertised range of approximately 100 m with an accuracy of 2-3 cm and support for, for example, a 100 Mbps Ethernet connection. In some examples, one or more non-protruding LIDAR sensors 464 may be used. In such examples, the LIDAR sensor(s) 464 may be implemented as a small device that can be embedded in the front, rear, sides, and / or corners of the vehicle 400. The LIDAR sensor(s) 464 can provide up to a horizontal 120-degree and vertical 35-degree field of view with a 200-m range in such examples, even for objects with low reflectivity.The front-mounted LIDAR sensor(s) 464 can be configured for a horizontal field of view between 45 degrees and 135 degrees.

[0185] In some examples, LiDAR technologies, such as 3D flash LiDAR, can also be used. 3D flash LiDAR uses a laser pulse as a transmission source to illuminate the vehicle's surroundings up to approximately 200 m. A flash LiDAR unit includes a receiver that records the laser pulse's travel time and the reflected light at each pixel, which corresponds to the distance from the vehicle to objects. Flash LiDAR can enable the creation of highly accurate and distortion-free images of the environment with each laser pulse. In some examples, four flash LiDAR sensors can be used, one on each side of the vehicle. Available 3D flash LiDAR systems include a solid-state 3D rigid-array LiDAR camera with no moving parts except for a fan (e.g., a non-scanning LiDAR device).The flash LIDAR device can use a 5-nanosecond Class I (eye-safe) laser pulse per image and can capture the reflected laser light in the form of 3D area point clouds and co-recorded intensity data. By using flash LIDAR, and because flash LIDAR is a solid-state device with no moving parts, the LIDAR sensor(s) 464 can be less susceptible to motion blur, vibration, and / or shock.

[0186] The vehicle may further include one or more IMU sensor(s) 466. The IMU sensor(s) 466 may be located in the center of the rear axle of the vehicle 400 in some examples. The IMU sensor(s) 466 may include, for example, without limitation, one or more accelerometers, one or more magnetometers, one or more gyroscopes, one or more magnetic compasses, and / or other sensor types. In some examples, such as in six-axis applications, the IMU sensor(s) 466 may include accelerometers and gyroscopes, while in nine-axis applications, the IMU sensor(s) 466 may include accelerometers, gyroscopes, and magnetometers.

[0187] In some embodiments, the IMU sensor(s) 466 can be implemented as a miniature high-performance GPS-based inertial navigation system (GPS / INS) that combines micro-electromechanical system (MEMS) inertial sensors, a high-sensitivity GPS receiver, and advanced Kalman filter algorithms to provide estimates of position, velocity, and orientation. Thus, in some examples, the IMU sensor(s) 466 can enable the vehicle 400 to estimate direction without requiring input from a magnetic sensor by directly observing changes in velocity from the GPS and correlating them with the IMU sensor(s) 466. In some examples, the IMU sensor(s) 466 and the GNSS sensor(s) 458 can be combined in a single integrated unit.

[0188] The vehicle may contain one or more microphone(s) 496, which are placed in and / or around the vehicle 400. The microphone(s) 496 may be used, among other things, for emergency vehicle detection and identification.

[0189] The vehicle may also include any number of camera types, including stereo camera(s) 468, wide-angle camera(s) 470, infrared camera(s) 472, surround-view camera(s) 474, long-range and / or medium-range camera(s) 498, and / or other camera types. The cameras may be used to capture image data around the entire periphery of the vehicle 400. The camera types used depend on the embodiment and requirements of the vehicle 400, and any combination of camera types may be used to provide the necessary coverage around the vehicle 400. 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 without limitation, the cameras may support Gigabit Multimedia Serial Link (GMSL) and / or Gigabit Ethernet.Each of the camera(s) is described herein in relation to . Fig. 4A and Fig. 4B is described in more detail.

[0190] The vehicle 400 may also include one or more vibration sensor(s) 442. The vibration sensor(s) 442 may measure vibrations of vehicle components, such as the axle(s). For example, changes in vibration may indicate a change in the road surface. In another example, if two or more vibration sensors 442 are used, the differences between the vibrations may be used to determine friction or slippage of the road surface (e.g., if the difference in vibration is between a power-driven axle and a freely rotating axle).

[0191] The vehicle 400 may include an ADAS system 438. In some examples, the ADAS system 438 may include a system of compute (SoC). The ADAS system 438 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 alert (RCTW), collision warning system (CWS), lane centering (LC), and / or other features and functionality.

[0192] The ACC systems can use one or more radar sensors 460, one or more lidar sensors 464, and / or one or more cameras. The ACC systems can include longitudinal ACC and / or lateral ACC. Longitudinal ACC monitors and controls the distance to the vehicle immediately in front of the vehicle 400 and automatically adjusts the vehicle speed to maintain a safe distance from vehicles ahead. Lateral ACC maintains a safe distance and notifies the vehicle 400 to change lanes if necessary. Lateral ACC relates to other ADAS applications, such as LCA and CWS.

[0193] CACC uses information from other vehicles, which can be received via the network interface 424 and / or the one or more wireless antenna(s) 426 from other vehicles via a wireless connection or indirectly via a network connection (e.g., via the Internet). Direct connections can be provided through a vehicle-to-vehicle (V2V) communication link, while indirect connections can be an infrastructure-to-vehicle (I2V) communication link. In general, the V2V communication concept provides information about the vehicles immediately ahead (e.g., vehicles immediately in front of and in the same lane as the vehicle 400), while the I2V communication concept provides information about traffic further ahead of the vehicle. CACC systems can incorporate one or both of the I2V and V2V information sources.Given the information from the vehicles ahead of vehicle 400, CACC can be more reliable and has the potential to improve the smoothness of traffic flow and reduce congestion on the road.

[0194] FCW systems are designed to warn the driver of a hazard so that the driver can take corrective action. FCW systems use a forward-facing camera and / or one or more RADAR sensor(s) 460 coupled with a dedicated processor, DSP, FPGA, and / or ASIC, which is / are electrically coupled to driver feedback, such as a display, speaker, and / or vibration component. FCW systems can provide a warning, such as an audible signal, a visual warning, a vibration, and / or a rapid braking pulse.

[0195] AEB systems detect an impending forward collision with another vehicle or object and can automatically apply the brakes if the driver does not take corrective action within a specified time or distance parameter. AEB systems can use one or more forward-facing camera(s) and / or one or more radar sensor(s) coupled with a dedicated processor, DSP, FPGA, and / or ASIC. When the AEB system detects a hazard, it typically first warns the driver to take corrective action to avoid the collision. If the driver fails to take corrective action, the AEB system can automatically apply the brakes to prevent or at least mitigate the effects of the predicted collision. AEB systems may incorporate techniques such as dynamic brake assist and / or anticipatory braking.

[0196] Lane Departure Warning (LDW) systems provide visual, audible, and / or tactile warnings, such as steering wheel or seat vibrations, to alert the driver if the vehicle crosses lane markings. An LDW system will not activate if the driver indicates an intentional lane departure by activating a turn signal. LDW systems may use forward-facing cameras coupled to a dedicated processor, DSP, FPGA, and / or ASIC, which is electrically coupled to driver feedback, such as a display, speaker, and / or vibration component.

[0197] Lane Keeping Assist (LKA) systems are a variation of Lane Departure Warning (LDW) systems. LKA systems provide steering or braking input to correct the vehicle if it begins to leave its lane. Blind Spot Warning (BSW) systems detect and warn the driver of vehicles in a vehicle's blind spot. BSW systems can provide visual, audible, and / or tactile warnings to indicate that merging or changing lanes is unsafe. The system can provide an additional warning if the driver uses a turn signal. BSW systems can use one or more rear-facing cameras and / or one or more radar sensors.

[0198] RCTW systems can provide visual, audible, and / or tactile alerts when an object is detected outside the range of the rear camera while the vehicle is reversing. Some RCTW systems include AEB to ensure the vehicle's brakes are applied to avoid a collision. RCTW systems can use one or more rear-facing radar sensors coupled to a dedicated processor, DSP, FPGA, and / or ASIC, which is electrically coupled to driver feedback, such as a display, speaker, and / or vibration component.

[0199] Conventional ADAS systems can be prone to false positives, which can be annoying and distracting for a driver, but are typically not catastrophic because the ADAS systems warn the driver and allow the driver to decide whether a safety condition actually exists and act accordingly. However, in an autonomous vehicle 400, the vehicle 400 itself must decide, even in the case of conflicting results, whether the result should be heeded by a primary computer or a secondary computer (e.g., a first controller 436 or a second controller 436). For example, in some embodiments, the ADAS system 438 can be a backup and / or secondary computer for providing perceptual information to a backup computer rationality module.The backup computer rationality monitor can run redundant, diverse software on hardware components to detect errors in perception and dynamic driving tasks. Outputs from the ADAS system 438 can be provided to a monitoring MCU. If outputs from the primary and secondary computers conflict, the monitoring MCU must determine how to resolve the conflict to ensure safe operation.

[0200] In some examples, the primary computer can be configured to provide the monitoring MCU with a confidence score indicating its confidence in the chosen outcome. If the confidence score exceeds a threshold, the monitoring MCU can follow the direction of the primary computer, regardless of whether the secondary computer provides a conflicting or inconsistent result. If the confidence score does not meet the threshold, and if the primary and secondary computers report different results (e.g., conflict), the monitoring MCU can arbitrate between the computers to determine the appropriate outcome.

[0201] The monitoring MCU can be configured to run a neural network(s) trained and configured to determine, based on outputs from the primary and secondary computers, the conditions under which the secondary computer provides false alarms. Thus, the neural network(s) in the monitoring MCU can learn when the secondary computer's output can be trusted and when it cannot. For example, if the secondary computer is a radar-based FCW system, the neural network(s) in the monitoring MCU can learn when the FCW system identifies metallic objects that are not actually hazards, such as a drainage grate or manhole cover, triggering an alarm.Similarly, if the secondary computer is a camera-based lane departure warning (LDW) system, a neural network in the monitoring MCU can learn to override the LDW when cyclists or pedestrians are present and leaving the lane is indeed the safest maneuver. In embodiments that include a neural network(s) running on the monitoring MCU, the monitoring MCU may include a data lathe (DLA) and / or a GPU suitable for running the neural network(s) with associated memory. In preferred embodiments, the monitoring MCU may include the SoC(s) 404 and / or be included as a component thereof.

[0202] In other examples, the ADAS system 438 can include a secondary computer that performs ADAS functionality using conventional computer vision rules. Thus, the secondary computer can use classic computer vision rules (if-then), and the presence of a neural network or neural networks in the monitoring MCU can improve reliability, safety, and performance. For example, the diverse implementation and intended non-identity make the overall system more fault-tolerant, especially against errors caused by software (or software-hardware interface) functionality.For example, if there is a software bug or error in the software running on the primary computer, and the non-identical software code running on the secondary computer provides the same overall result, the monitoring MCU can have greater confidence that the overall result is correct, and the bug in the software or hardware on the primary computer does not cause a material error.

[0203] In some examples, the output of the ADAS system 438 can be fed into the perception block and / or the dynamic driving task block of the primary computer. For example, if the ADAS system 438 displays a forward collision warning due to an object immediately ahead, the perception block can use this information in object identification. In other examples, the secondary computer may have its own trained neural network, thus reducing the risk of false positives, as described herein.

[0204] The Vehicle 400 may also include the Infotainment SoC 430 (e.g., an in-vehicle infotainment system (IVI)). Although illustrated and described as a single SoC, the infotainment system may not be a SoC and may include two or more discrete components. The Infotainment SoC 430 may include a combination of hardware and software that can be used to provide the Vehicle 400 with audio (e.g., music, a personal digital assistant, navigation instructions, news, radio, etc.), video (e.g., television, movies, streaming, etc.), telephone (e.g., hands-free calling), network connectivity (e.g., LTE, Wi-Fi, etc.), and / or information services (e.g., navigation systems, rear parking assistance, a radio data system, vehicle-related information such as fuel level, total distance traveled, brake fluid level, oil level, door opening / closing, air filter information, etc.).For example, the Infotainment SoC 430 can include radios, record players, navigation systems, video game players, USB and Bluetooth connectivity, car putters, in-vehicle entertainment, Wi-Fi, steering wheel audio controls, hands-free voice control, a head-up display (HUD), an HMI display 434, 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 430 can also be used to provide information (e.g., visual and / or audible) to the vehicle's user(s), such as information from the ADAS system 438, autonomous driving information such as planned vehicle maneuvers, trajectories, information about the surrounding environment (e.g., intersection information, vehicle information, road information, etc.), and / or other information.

[0205] The infotainment SoC 430 may include GPU functionality. The infotainment SoC 430 can communicate with other devices, systems, and / or components of the vehicle 400 via bus 402 (e.g., CAN bus, Ethernet, etc.). In some examples, the infotainment SoC 430 may be coupled with a monitoring MCU so that the infotainment system's GPU can perform some self-driving functions in the event of a failure of the primary controller(s) 436 (e.g., the primary and / or backup computer of the vehicle 400). In such an example, the infotainment SoC 430 may place the vehicle 400 into a chauffeur-in-safe-stop mode, as described herein.

[0206] The vehicle 400 may further include an instrument cluster 432 (e.g., a digital instrument cluster, an electronic instrument cluster, a digital dashboard, etc.). The instrument cluster 432 may include a controller and / or a supercomputer (e.g., a discrete controller or a supercomputer). The instrument cluster 432 may include a set of gauges, such as a speedometer, fuel level, oil pressure, tachometer, odometer, turn signals, gear shift position indicator, seat belt warning light(s), parking brake warning light(s), engine failure light(s), 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 430 and the instrument cluster 432. In other words, the instrument cluster 432 may be included as part of the infotainment SoC 430, or vice versa.

[0207] Fig. 4D is a system diagram for communication between one or more cloud-based servers and the exemplary autonomous vehicle 400 from Fig. 4A according to some embodiments of the present disclosure. The system 476 may include one or more servers 478, one or more networks 490, and vehicles, including the vehicle 400. The server(s) 478 may include a plurality of GPUs 484(A)-484(H) (hereinafter collectively referred to as GPUs 484), PCIe switches 482(A)-482(H) (hereinafter collectively referred to as PCIe switches 482), and / or CPUs 480(A)-480(B) (hereinafter collectively referred to as CPUs 480). The GPUs 484, the CPUs 480, and the PCIe switches may be connected by high-speed links, such as, but not limited to, NVLink interfaces 488 developed by NVIDIA and / or PCIe links 486. In some examples, the GPUs 484 are connected via NVLink and / or NVSwitch SoC, and the GPUs 484 and the PCIe switches 482 are connected via PCIe links.Although eight GPUs 484, two CPUs 480, and two PCIe switches are illustrated, this is not intended to be restrictive. Depending on the implementation, each Server 478 can contain any number of GPUs 484, CPUs 480, and / or PCIe switches. For example, the Server 478 can contain eight, sixteen, thirty-two, and / or more GPUs 484.

[0208] The server(s) 478 can receive image data from the network(s) 490 and the vehicles, representing images that show unexpected or changed road conditions, such as recently started roadworks. The server(s) 478 can transmit neural networks 492, updated neural networks 492, and / or map information 494, including information relating to traffic and road conditions, to the vehicles via the network(s) 490 and the vehicles. The map information updates 494 may include updates to the HD map 422, such as information relating to construction sites, potholes, detours, flooding, and / or other obstacles.In some examples, the neural networks 492, the updated neural networks 492 and / or the map information 494 may result from new training and / or experience represented in data received from any number of vehicles in the environment, and / or based on training performed in a data center (e.g. using the server(s) 478 and / or other servers).

[0209] Server 478 can be used to train machine learning models (e.g., neural networks) based on training data. The training data can be generated by the vehicles and / or created in a simulation (e.g., using a game engine). In some examples, the training data is tagged (e.g., where the neural network benefits from supervised learning) and / or subjected to other preprocessing, while in other examples, the training data is not tagged and / or preprocessed (e.g., where the neural network does not require supervised learning).Training can be performed according to any one or more classes of machine learning techniques, including, without limitation, classes such as: supervised training, semi-supervised training, unsupervised training, self-learning, reinforcement learning, federated learning, transfer learning, feature learning (including principal component and cluster analysis), multilinear subspace learning, diverse learning, representation learning (including free dictionary learning), rule-based machine learning, anomaly detection, and any variants or combinations thereof. Once the machine learning models are trained, they can be used by the vehicles (e.g., transmitted to the vehicles via network(s) 490), and / or the machine learning models can be used by server(s) 478 to remotely monitor the vehicles.

[0210] In some examples, the Server 478 can receive data from the vehicles and apply that data to current real-time neural networks for intelligent real-time inference. The Server 478 can include deep learning supercomputers and / or dedicated AI computers powered by GPU(s) 484, such as DGX and DGX Station machines developed by NVIDIA. However, in some examples, the Server 478 can include deep learning infrastructure that uses only CPU-based data centers.

[0211] The deep learning infrastructure of the server(s) 478 is capable of rapid real-time inference and can use this capability to assess and verify the health of the processors, software, and / or associated hardware in the vehicle 400. For example, the deep learning infrastructure can receive periodic updates from the vehicle 400, such as a sequence of images and / or objects indicating where the vehicle 400 was located within that sequence of images (e.g., via computer vision and / or other machine learning object classification techniques).The deep learning infrastructure can run its own neural network to identify the objects and compare them with the objects identified by the vehicle 400, and if the results do not match and the infrastructure concludes that the AI ​​in the vehicle 400 is malfunctioning, the server(s) 478 can transmit a signal to the vehicle 400 that instructs a fail-safe computer of the vehicle 400 to take over control, notify the passengers and complete a safe parking maneuver.

[0212] For inference, the server(s) 478 can include the GPU(s) 484 and one or more programmable inference accelerators (e.g., NVIDIA's TensorRT). Combining servers with GPU power and inference acceleration can enable real-time responsiveness. In other scenarios, such as when performance is less critical, servers powered by CPUs, FPGAs, and other processors can be used for inference. Example computing device

[0213] Fig. Figure 5 is a block diagram of one or more exemplary computing devices 500 suitable for use in implementing some embodiments of the present disclosure. The computing device 500 may include a connection system 502 that directly or indirectly couples the following devices: memory 504, one or more central processing units (CPUs) 506, one or more graphics processing units (GPUs) 508, a communication interface 510, input / output (I / O) ports 512, input / output components 514, a power supply 516, one or more presentation components 518 (e.g., display(s)), and one or more logic units 520. In at least one embodiment, the computing device(s) 500 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-restrictive examples, one or more of the GPUs 508 can comprise one or more vGPUs, one or more of the CPUs 506 can comprise one or more vCPUs, and / or one or more of the logic units 520 can comprise one or more virtual logic units. Accordingly, a computing device 500 can include discrete components (e.g., a complete GPU dedicated to the computing device 500), virtual components (e.g., a portion of a GPU dedicated to the computing device 500), or a combination thereof.

[0214] Although the various blocks made Fig. Where components 5 are shown connected via lines through the connection system 502, this is not intended to be restrictive and is merely for clarity. For example, in some embodiments, a presentation component 518, such as a display device, may be considered an I / O component 514 (e.g., if the display is a touchscreen). As another example, the CPUs 506 and / or GPUs 508 may include memory (e.g., the memory 504 may be representative of a memory device in addition to the memory of the GPUs 508, the CPUs 506, and / or other components). In other words, the computing device is made up of Fig. 5 is for illustrative purposes only. No distinction is made between categories such 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 fall within the scope of the computing device. Fig. 5 should be considered.

[0215] The 502 interconnect system can represent one or more connections or buses, such as an address bus, a data bus, a control bus, or a combination thereof. The 502 interconnect system can include one or more bus or connection 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 connection. In some embodiments, there are direct connections between components. For example, the CPU 506 can be directly connected to the memory 504. Furthermore, the CPU 506 can be directly connected to the GPU 508. When there is a direct or point-to-point connection between components, the 502 interconnect system can include a PCIe connection to implement the connection.In these examples, a PCI bus does not need to be included in the Computing Device 500.

[0216] The memory 504 can contain any of a variety of computer-readable media. Computer-readable media can be any available media accessible to the computing device 500. Computer-readable media can include both volatile and non-volatile media, as well as removable and non-removable media. For example, and without limitation, computer-readable media can include computer storage media and communication media.

[0217] Computer storage media can include both volatile and non-volatile media, and / or removable and non-removable media, implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, and / or other data types. For example, memory can store 504 computer-readable instructions (which represent, for example, a program or programs and / or a program element or program elements), such as an operating system.Computer storage media may include, but are not limited to, RAM, ROM, EEPROM, flash memory or other storage technology, CD-ROM, Digital Versatile Discs (DVD) or other optical disk storage, magnetic cartridges, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and that can be accessed by the Computing Device 500. As used herein, computer storage media do not include signals per se.

[0218] 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 include any information transmission media. The term "modulated data signal" can refer to a signal in which one or more of its properties are set or modified in such a way as to encode information in the signal. For example, and without limitation, computer storage media can include wired media, such as a wired network or a directly wired connection, and wireless media, such as acoustic, RF, infrared, and other wireless media. Combinations of any of the foregoing should also be included in the scope of computer-readable media.

[0219] The CPU(s) 506 can be configured to execute at least some of the computer-readable instructions to control one or more components of the Computing Device 500 to perform one or more of the procedures and / or processes described herein. The CPU(s) 506 can each include one or more cores (e.g., one, two, four, eight, twenty-eight, seventy-two, etc.) capable of handling a plurality of software threads concurrently. The CPU(s) 506 can include any type of processor and can include different types of processors depending on the type of Computing Device 500 implemented (e.g., processors with fewer cores for mobile devices and processors with more cores for servers).For example, depending on the type of computing device 500, the processor can be an Advanced RISC Machines (ARM) processor implemented using Reduced Instruction Set Computing (RISC), or an x86 processor implemented using Complex Instruction Set Computing (CISC). The computing device 500 can include one or more CPUs 506 in addition to one or more microprocessors or supplementary coprocessors, such as mathematical coprocessors.

[0220] In addition to or alternatively to the CPU(s) 506, the GPU(s) 508 may be configured to execute at least some of the computer-readable instructions to control one or more components of the computing device 500 to perform one or more of the procedures and / or processes described herein. One or more of the GPU(s) 508 can be an integrated GPU (e.g., with one or more of the CPU(s) 506) and / or one or more of the GPU(s) 508 can be a discrete GPU. In embodiments, one or more of the GPU(s) 508 can be a coprocessor of one or more of the CPU(s) 506. The GPU(s) 508 can be used by the computing device 500 to render graphics (e.g., 3D graphics) or to perform general-purpose computing. For example, the GPU(s) 508 can be used for general-purpose computing on GPUs (GPGPU).The GPU(s) 508 can include hundreds or thousands of cores capable of handling hundreds or thousands of software threads simultaneously. The GPU(s) 508 can generate pixel data for output images in response to rendering commands (e.g., rendering commands from the CPU(s) 506 received via a host interface). The GPU(s) 508 can include graphics memory, such as display memory, for storing pixel data or any other suitable data, such as GPGPU data. The display memory can be included as part of the memory 504. The GPU(s) 508 can include two or more GPUs operating in parallel (e.g., via a link). The link can connect the GPUs directly (e.g., using NVLINK) or connect them via a switch (e.g., using NVSwitch).When combined, each GPU can generate 508 pixel data or GPGPU data for different sections 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 can have its own dedicated memory or share memory with other GPUs.

[0221] In addition to or as an alternative to the CPU(s) 506 and / or the GPU(s) 508, the logic unit(s) 520 may be configured to execute at least some of the computer-readable instructions to control one or more components of the computing device 500 to perform one or more of the methods and / or processes described herein. In embodiments, the CPU(s) 506, the GPU(s) 508, and / or the logic unit(s) 520 may discretely or jointly perform any combination of the methods, processes, and / or sections thereof. One or more of the logic units 520 may be part of and / or integrated within one or more of the CPU(s) 506 and / or the GPU(s) 508, and / or one or more of the logic units 520 may be discrete components or otherwise separate from the CPU(s) 506 and / or the GPU(s) 508.In embodiments, one or more of the logic units 520 may be a coprocessor of one or more of the CPU(s) 506 and / or one or more of the GPU(s) 508.

[0222] Examples of the logic unit(s) 520 include one or more processing cores and / or components thereof, such as data processing units (DPUs), tensor cores (TCs), tensor processing units (TPUs), visual pixel cores (PVCs), vision processing units (VPUs), graphics processing clusters (GPCs), texture processing clusters (TPCs), streaming multiprocessors (SMs), tree traversal units (TTUs), artificial intelligence accelerators (AIAs), deep learning accelerators (DLAs), arithmetic logic units (ALUs), application-specific integrated circuits (ASICs), floating-point units (FPUs), input / output (I / O) elements, peripheral component interconnect (PCI) or peripheral component interconnect express (PCIe) elements, and / or the like.

[0223] The Communication Interface 510 can include one or more receivers, transmitters, and / or transmit-receivers that enable the Computing Device 500 to communicate with other Computing Devices over an electronic communications network, including wired and / or wireless communication. The Communication Interface 510 can include components and functionality to enable communication over any number of different networks, such as wireless networks (e.g., Wi-Fi, Z-Wave, Bluetooth, Bluetooth LE, ZigBee, etc.), wired networks (e.g., communicating over Ethernet or InfiniBand), low-power wide area networks (e.g., LoRaWAN, SigFox, etc.), and / or the Internet.In one or more embodiments, the logic unit(s) 520 and / or the communication interface 510 may include one or more data processing units (DPUs) to transfer data received via a network and / or through the connection system 502 directly to (e.g., a memory of) one or more GPU(s) 508.

[0224] The I / O ports 512 enable the computing device 500 to be logically coupled with other devices, including the I / O components 514, the presentation component(s) 518, and / or other components, some of which may be built into (e.g., integrated with) the computing device 500. Illustrative I / O components 514 include a microphone, mouse, keyboard, joystick, gamepad, game controller, satellite dish, scanner, printer, wireless device, etc. The I / O components 514 can provide a natural user interface (NUI) that processes air gestures, speech, or other physiological inputs generated by a user. In some cases, inputs can be transferred to a suitable network element for further processing.A NUI can implement any combination of speech recognition, stylus recognition, facial recognition, biometric recognition, gesture recognition both on-screen and off-screen, air gestures, head and eye tracking, and touch recognition (as described in more detail in the present disclosure) in conjunction with a display of the computing device 500. The computing device 500 can include depth cameras, such as stereoscopic camera systems, infrared camera systems, RGB camera systems, touchscreen technology, and combinations thereof for gesture detection and recognition. Additionally, the computing device 500 can include accelerometers or gyroscopes (e.g., as part of an inertial measurement unit (IMU)) that enable motion detection.In some examples, the output of the accelerometer or gyroscope can be used by the Computing Device 500 to render immersive augmented reality or virtual reality.

[0225] The power supply 516 can include a hardwired power supply, a battery power supply, or a combination thereof. The power supply 516 can provide power to the computer 500 to enable the components of the computer 500 to operate.

[0226] The presentation component(s) 518 can include a display (e.g., a monitor, touchscreen, television screen, head-up display (HUD), other display types, or a combination thereof), speakers, and / or other presentation components. The presentation component(s) 518 can receive data from other components (e.g., the GPU(s) 508, the CPU(s) 506, etc.) and output the data (e.g., as an image, video, sound, etc.). EXEMPLARY DATA CENTER

[0227] Fig. Figure 6 illustrates an exemplary data center 600 that can be used in at least one embodiment of the present disclosure. The data center 600 can include a data center infrastructure layer 610, a framework layer 620, a software layer 630, and / or an application layer 640.

[0228] As in Fig. As shown in Figure 6, the data center infrastructure layer 610 can include a resource orchestrator 612, clustered compute resources 614, and node compute resources (“node CRs”) 616(1)-616(N), where “N” is any positive integer. In at least one embodiment, the node CRs 616(1)-616(N) can include, but are not limited to, any number of central processing units (CPUs) or other processors (including DPUs, accelerators, field-programmable gate arrays (FPGAs), graphics processing units or graphics processing units (GPUs), etc.), storage devices (e.g., dynamic read-only memory), storage devices (e.g., solid-state or disk drives), network input / output (NW I / O) devices, network switches, virtual machines (VMs), power modules, and / or cooling modules, etc. In some embodiments, one or more node CRs can be selected from the node CRs.s 616(1)-616(N) correspond to a server that has one or more of the computing resources mentioned above. Additionally, in some embodiments, the node CRs 616(1)-616(N) may include one or more virtual components, such as vGPUs, vCPUs and / or the like, and / or one or more of the node CRs 616(1)-616(N) may correspond to a virtual machine (VM).

[0229] In at least one embodiment, the grouped compute resources 614 can include separate groupings of node CRs 616 located in one or more racks (not shown), or many racks located in data centers at different geographic locations (also not shown). Separate groupings of node CRs 616 within the grouped compute resources 614 can include grouped compute, network, storage, or memory resources that can be configured or allocated to support one or more workloads. In at least one embodiment, multiple node CRs 616, which include CPUs, GPUs, DPUs, and / or other processors, can be grouped in one or more racks to provide compute resources to support one or more workloads.The one or more racks can also contain any number of power modules, cooling modules and / or network switches in any combination.

[0230] The resource orchestrator 612 can configure or otherwise control one or more node CRs 616(1)-616(N) and / or grouped compute resources 614. In at least one embodiment, the resource orchestrator 612 can include a software design infrastructure (SDI) management entity for the data center 600. The resource orchestrator 612 can include hardware, software, or a combination thereof.

[0231] In at least one embodiment, as in Fig. As shown in Figure 6, the framework layer 620 can include a job scheduler 632, a configuration manager 634, a resource manager 636, and / or a distributed file system 638. The framework layer 620 can include a framework to support software 632 of software layer 630 and / or one or more application(s) 642 of application layer 640. The software 632 or the application(s) 642 can each include web-based service software or applications, such as those provided by Amazon Web Services, Google Cloud, and Microsoft Azure. The framework layer 620 can, but is not limited to, be some kind of free and open-source software web application framework, such as Apache Spark™ (hereinafter "Spark"), which can use the distributed file system 638 for large-scale data processing (e.g., "Big Data").In at least one embodiment, the job scheduler 632 can include a Spark driver to facilitate the scheduling of workloads supported by different layers of the data center 600. The configuration manager 634 can be capable of configuring different layers, such as the software layer 630 and the framework layer 620, including Spark and the distributed file system 638, to support large-scale data processing. The resource manager 636 can be capable of managing clustered or grouped compute resources that are mapped to or allocated to support the distributed file system 638 and the job scheduler 632. In at least one embodiment, clustered or grouped compute resources can include the grouped compute resource 614 in the data center infrastructure layer 610.The Resource Manager 636 can coordinate with the Resource Orchestrator 612 to manage these mapped or allocated computing resources.

[0232] In at least one embodiment, the software 632, which is included in software layer 630, may include software used by at least sections of the node CRs 616(1)-616(N), the grouped computing resources 614, and / or the distributed file system 638 of framework layer 620. One or more types of software may include, but are not limited to, internet website search software, email virus scanning software, database software, and streaming video content software.

[0233] In at least one embodiment, the application(s) 642 included in the application layer 640 may include one or more types of applications used by at least sections of the node CRs 616(1)-616(N), the grouped compute resources 614, and / or the distributed file system 638 of the framework layer 620. One or more types of applications may, but are not limited to, include any number of a genomic application, a cognitive computation, and a machine learning application, 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.

[0234] In at least one embodiment, any configuration manager 634, resource manager 636, and resource orchestrator 612 can implement any number and type of self-modifying actions based on any amount and type of data acquired in any technically feasible way. Self-modifying actions can relieve a data center operator of data center 600 of potentially making poor configuration decisions and potentially avoiding underutilized and / or poorly performing sections of a data center.

[0235] The Data Center 600 may include tools, services, software, or other resources for training one or more machine learning models or for predicting or deriving information using one or more machine learning models according to one or more embodiments described herein. For example, one or more machine learning models may be trained by calculating weight parameters according to a neural network architecture using software and / or computing resources described in this disclosure relating to the Data Center 600.In at least one embodiment, trained or deployed machine learning models corresponding to one or more neural networks can be used to derive or predict information using resources described in the present disclosure relating to the data center 600, by using weight parameters calculated by one or more training techniques, such as, but not limited to, those described herein.

[0236] In at least one embodiment, the data center can use 600 CPUs, application-specific integrated circuits (ASICs), GPUs, FPGAs, and / or other hardware (or equivalent virtual computing resources) to perform training and / or inference using the resources described above. Furthermore, one or more software and / or hardware resources described herein can be configured as a service to enable users to train or perform information inference, such as image recognition, speech recognition, or other artificial intelligence services. Exemplary network environments

[0237] Network environments suitable for use in implementing embodiments of the 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 run on one or more instances of the computing device(s) 500. Fig. 5 be implemented - e.g., each device may include similar components, features, and / or functionality to the computing device(s) 500. Additionally, if backend devices (e.g., servers, NAS, etc.) are implemented, the backend devices may be included as part of a data center 600, an example of which is given herein in relation to Fig. Section 6 is described in more detail.

[0238] Components of a network environment can communicate with each other over one or more networks, which can be wired, wireless, or both. The network can include multiple networks or a network of 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 a public switched telephone network (PSTN), and / or one or more private networks. If the network includes a wireless telecommunications network, components such as a base station, a communications mast, or even access points (as well as other components) can provide wireless connectivity.

[0239] Compatible network environments can include one or more peer-to-peer network environments—in which case a server may not be included in a network environment—and one or more client-server network environments—in which case one or more servers may be included in a network environment. In peer-to-peer network environments, the functionality described herein with respect to one or more servers can be implemented on any number of client devices.

[0240] In at least one embodiment, a network environment can include one or more cloud-based network environments, a distributed computing environment, a combination thereof, etc. A cloud-based network environment can 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. A framework layer can include a framework to support software of a software layer and / or one or more application(s) of an application layer. The software or application(s) can each include web-based service software or applications. In embodiments, one or more of the client devices can use the web-based service software or applications (e.g.,by accessing the service software and / or applications via one or more application programming interfaces (APIs). The framework layer can, but is not limited to, be a type of free and open-source software web application framework, such as one that can use a distributed file system for large-scale data processing (e.g., "Big Data").

[0241] A cloud-based network environment can provide cloud computing and / or cloud storage, performing any combination (or one or more parts thereof) of the computing and / or data storage functions described herein. Each of these different functions can be distributed across multiple locations of central or core servers (e.g., one or more data centers, which may be distributed across a state, region, country, the globe, etc.). If a connection to a user (e.g., a client device) is relatively close to one or more edge servers, one or more core servers may designate at least one part of the functionality for the edge server(s). A cloud-based network environment can be private (e.g., restricted to a single organization), public (e.g., available to many organizations), and / or a combination thereof (e.g., a hybrid cloud environment).

[0242] The client device(s) may include at least some of the components, features, and functionality described herein with respect to Fig.The 5 described exemplary computing device(s) include 500.As an example, and not as a limitation, a client device may be embodied as a personal computer (PC), laptop computer, mobile device, smartphone, tablet computer, smartwatch, portable computer, personal digital assistant (PDA), MP3 player, virtual reality headset, global positioning system (GPS) or device, video game player, video camera, surveillance device or system, vehicle, boat, flying ship, virtual machine, drone, robot, handheld communication device, hospital device, gaming device or system, entertainment system, vehicle computer system, embedded control system, remote control, device, consumer electronics device, workstation, edge device, any combination of these delimited devices, or any other suitable device.

[0243] The disclosure may be described in the general context of computer code or machine-usable instructions, including computer-executable instructions such as program modules, which are executed by a computer or other machine, such as a personal data assistant or other handheld device. In general, program modules, including routines, programs, objects, components, data structures, etc., refer to code that performs specific tasks or implements certain abstract data types. The disclosure may be implemented in a variety of system configurations, including handheld devices, consumer electronics, general-purpose computers, specialized computing devices, etc. The disclosure may also be implemented in distributed computing environments where tasks are performed by remote processing devices connected by a communications network.

[0244] As used herein, a recitation of "and / or" in reference to two or more elements should be interpreted as meaning only one element or a combination of elements. For example, "element A, element B and / or element C" may include only element A, only element B, only element C, element A and element B, element A and element C, element B and element C, or elements A, B, and C. Additionally, "at least one of element A or element B" may include at least one of element A, at least one of element B, or at least one of element A and at least one of element B. Furthermore, "at least one of element A and element B" may include at least one of element A, at least one of element B, or at least one of element A and at least one of element B.Instead, a first element that is “based on a second element” includes cases where the first element is based on the second element, but may also be based on one or more additional elements.

[0245] The subject matter of this disclosure is described herein with specificity to satisfy legal requirements. However, the description itself is not intended to limit the scope of this disclosure. Rather, the inventors have considered that the claimed subject matter could also be carried out in other ways to include various steps or combinations of steps similar to those described in this document, in conjunction with other present or future technologies. Although the terms "step" and / or "block" may be used herein to denote various elements of methods employed, these terms should not be interpreted as implying a particular sequence between or between different steps disclosed herein, unless, and except, the sequence of individual steps is explicitly described. QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited patent literature

[0000] US 16 / 101.232

[0149] Cited non-patent literature

[0000] Society of Automotive Engineers (SAE) “Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles” (Standard No. J3016-201806, published on June 15, 2018, Standard No. J3016-201609, published on September 30, 2016

[0107]

Claims

[1] Method comprising: Analyzing sensor data corresponding to an operating range of a machine using a variety of analysis techniques, the variety of analysis techniques including: at least two analysis techniques that exhibit computational diversity by performing different types of computational analyses on the sensor data; and at least two analysis techniques that exhibit implementation diversity by being implemented on different types of computing platforms; Determining a detection of an object within the operating range based at least on analyzing; and Controlling one or more operations of the machine based at least on the detection. [2] The method of claim 1, wherein the sensor data comprises stereoscopic image data. [3] The method of claim 1, wherein the plurality of analysis techniques include a plurality of depth techniques that vary in how depth is determined from one or more reference points. [4] The method of claim 3, wherein the plurality of depth techniques includes at least a semi-global matching (SGM), a Bi3D, and an efficient semi-supervised depth (ESS). [5] The method of claim 1, wherein the different types of computing platforms include one or more of: a central processing unit (CPU), a graphics processing unit (GPU), an optical flow accelerator (OFA), a deep learning accelerator (DLA), a programmable vision accelerator (PVA), a video input (VI) controller, an image signal processor (ISP), or a video image composition (VIC) engine. [6] The method of claim 1, wherein determining the detection of the object includes: Determining a plurality of individual detection results corresponding to respective analysis techniques of the plurality of analysis techniques; and Determining an overall detection result based on at least the plurality of individual detection results. [7] The method of claim 6, wherein determining the overall detection result includes: Determining the overall detection result by combining the plurality of individual detection results, wherein one or more individual detection results of the plurality of individual detection results are weighted with respect to the combining based at least on respective confidence levels with respect to the one or more individual detection results. [8] The method of claim 6, wherein determining the overall detection result includes: Determining the overall detection result based on whether a threshold number of the plurality of individual detection results indicates a detection. [9] The method of claim 8, wherein the threshold number is based on at least one target safety level associated with the operating range. [10] System comprising: one or more processors to effectuate operations comprising: Analyzing sensor data corresponding to a detection zone using a plurality of analysis techniques, wherein the plurality of analysis techniques includes computational diversity and implementation diversity; Determining a detection of an object within the detection zones based at least on the analyzing; and Perform one or more operations based at least on the detection. [11] The system of claim 10, wherein the sensor data comprises stereoscopic image data. [12] The system of claim 10, wherein the plurality of analysis techniques include a plurality of depth techniques that vary in how depth is determined from one or more reference points. [13] The system of claim 12, wherein the plurality of depth techniques includes at least one of semi-global matching (SGM), Bi3D, and efficient semi-supervised depth (ESS). [14] The system of claim 10, wherein the plurality of analysis techniques are implemented on a plurality of computing platforms including one or more of: a central processing unit (CPU), a graphics processing unit (GPU), an optical flow accelerator (OFA), a deep learning accelerator (DLA), a programmable vision accelerator (PVA), a video input (VI) controller, an image signal processor (ISP), or a video image composition (VIC) engine. [15] The system of claim 10, wherein determining the detection of the object includes: Determining a plurality of individual detection results corresponding to respective analysis techniques of the plurality of analysis techniques; and Determining an overall detection result based on at least the plurality of individual detection results. [16] The system of claim 15, wherein determining the overall detection result includes: Determining the overall detection result by combining the plurality of individual detection results, wherein one or more individual detection results of the plurality of individual detection results are weighted with respect to the combining based at least on respective confidence levels with respect to the one or more individual detection results. [17] The system of claim 15, wherein determining the overall detection result includes: Determining the overall detection result based on whether a threshold number of the plurality of individual detection results indicates a detection. [18] The system of claim 17, wherein the threshold number is based on at least one target security level associated with the detection zone. [19] The system of claim 10, wherein the system is comprised of at least one of the following: 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 simulation; a system for performing collaborative content creation for 3D assets; a system for performing deep learning operations; a system for displaying at least one of augmented reality content, virtual reality content, or mixed reality content; a system for hosting one or more real-time streaming applications; a system implemented using an edge device; a system implemented using a robot; a system for performing conversational AI operations; a system that implements one or more large language models (LLMs); a system for generating synthetic data; a system that contains one or more virtual machines (VMs); a system that is at least partially implemented in a data center; or a system implemented at least in part using cloud computing resources. [20] System comprising: Processing circuitry for performing one or more operations associated with a machine based on at least one final detection result, wherein the final detection result is determined based on at least two or more individual detection results calculated using two or more different algorithms executing on two or more different hardware components of the machine.

Images