Hierarchical multi-object tracker for auto-labeling

Abstract

Provided are methods for hierarchical multi-object tracker for auto-labeling. Some methods include generating tracklets based on detections in sensor data. A weighted track graph is generated based on the tracklets, and the weight graph is converted to a bipartite graph. The bipartite graph is solved to determine at least one best path cover corresponding to respective observed objects, and the detections in the sensor data are labeled with a track identifier corresponding to the track representing the best path cover of a respective observed object. Systems and computer program products are also provided.

Description

Hierarchical Multi-Object Tracker for Auto-LabelingBACKGROUND

[0001] Labeled or annotated data is used to train autonomous systems. Labeled data enables a machine learning model to learn how to perform various functions. Manual labeling of data can be costly and time consuming.BRIEF DESCRIPTION OF THE FIGURES

[0002] FIG. 1 is an example environment in which a vehicle including one or more components of an autonomous system can be implemented;

[0003] FIG. 2 is a diagram of one or more systems of a vehicle including an autonomous system;

[0004] FIG. 3 is a diagram of components of one or more devices and / or one or more systems of FIGS. 1 and 2;

[0005] FIG. 4 is a diagram of certain components of an autonomous system;

[0006] FIGS. 5 is a diagrams of an implementation of a hierarchical multi-object tracker for auto-labeling;

[0007] FIG. 6 is a block diagram of a workflow for a hierarchical multi-object tracker for auto-labeling;

[0008] FIG. 7 shows detections in frames of sensor data;

[0009] FIG. 8 shows a weighted graph;

[0010] FIG. 9A shows a weighted graph with nodes partitioned into two disjoint and independent sets;

[0011] FIG. 9B shows a bipartite graph;

[0012] FIG. 10A shows a workflow corresponding to a first stage tracker;

[0013] FIG. 10B shows a workflow corresponding to a second stage tracker;

[0014] FIG. 11 shows a flowchart of a process for a hierarchical multi-object tracker for auto-labeling.DETAILED DESCRIPTION

[0015] In the following description numerous specific details are set forth in order to provide a thorough understanding of the present disclosure for the purposes of explanation. It will be apparent, however, that the embodiments described by the present disclosure can be practiced without these specific details. In some instances, well-known structures and devices are illustrated in block diagram form in order to avoid unnecessarily obscuring aspects of the present disclosure.

[0016] Specific arrangements or orderings of schematic elements, such as those representing systems, devices, modules, instruction blocks, data elements, and / or the like are illustrated in the drawings for ease of description. However, it will be understood by those skilled in the art that the specific ordering or arrangement of the schematic elements in the drawings is not meant to imply that a particular order or sequence of processing, or separation of processes, is required unless explicitly described as such. Further, the inclusion of a schematic element in a drawing is not meant to imply that such element is required in all embodiments or that the features represented by such element may not be included in or combined with other elements in some embodiments unless explicitly described as such.

[0017] Further, where connecting elements such as solid or dashed lines or arrows are used in the drawings to illustrate a connection, relationship, or association between or among two or more other schematic elements, the absence of any such connecting elements is not meant to imply that no connection, relationship, or association can exist. In other words, some connections, relationships, or associations between elements are not illustrated in the drawings so as not to obscure the disclosure. In addition, for ease of illustration, a single connecting element can be used to represent multiple connections, relationships or associations between elements. For example, where a connecting element represents communication of signals, data, or instructions (e.g., “software instructions”), it should be understood by those skilled in the art that such element can represent one or multiple signal paths (e.g., a bus), as may be needed, to affect the communication.

[0018] Although the terms first, second, third, and / or the like are used to describe various elements, these elements should not be limited by these terms. The terms first,second, third, and / or the like are used only to distinguish one element from another. For example, a first contact could be termed a second contact and, similarly, a second contact could be termed a first contact without departing from the scope of the described embodiments. The first contact and the second contact are both contacts, but they are not the same contact.

[0019] The terminology used in the description of the various described embodiments herein is included for the purpose of describing particular embodiments only and is not intended to be limiting. As used in the description of the various described embodiments and the appended claims, the singular forms “a,” “an” and “the” are intended to include the plural forms as well and can be used interchangeably with “one or more” or “at least one,” unless the context clearly indicates otherwise. It will also be understood that the term “and / or” as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items. It will be further understood that the terms “includes,” “including,” “comprises,” and / or “comprising,” when used in this description specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0020] As used herein, the terms “communication” and “communicate” refer to at least one of the reception, receipt, transmission, transfer, provision, and / or the like of information (or information represented by, for example, data, signals, messages, instructions, commands, and / or the like). For one unit (e.g., a device, a system, a component of a device or system, combinations thereof, and / or the like) to be in communication with another unit means that the one unit is able to directly or indirectly receive information from and / or send (e.g., transmit) information to the other unit. This may refer to a direct or indirect connection that is wired and / or wireless in nature. Additionally, two units may be in communication with each other even though the information transmitted may be modified, processed, relayed, and / or routed between the first and second unit. For example, a first unit may be in communication with a second unit even though the first unit passively receives information and does not actively transmit information to the second unit. As another example, a first unit may be in communication with a second unit if at least one intermediary unit (e.g., a third unit locatedbetween the first unit and the second unit) processes information received from the first unit and transmits the processed information to the second unit. In some embodiments, a message may refer to a network packet (e.g., a data packet and / or the like) that includes data.

[0021] As used herein, the term “if’ is, optionally, construed to mean “when”, “upon”, “in response to determining,” “in response to detecting,” and / or the like, depending on the context. Similarly, the phrase “if it is determined” or “if [a stated condition or event] is detected” is, optionally, construed to mean “upon determining,” “in response to determining,” “upon detecting [the stated condition or event],” “in response to detecting [the stated condition or event],” and / or the like, depending on the context. Also, as used herein, the terms “has”, “have”, “having”, or the like are intended to be open-ended terms. Further, the phrase “based on” is intended to mean “based at least partially on” unless explicitly stated otherwise.

[0022] Reference will now be made in detail to embodiments, examples of which are illustrated in the accompanying drawings. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the various described embodiments. However, it will be apparent to one of ordinary skill in the art that the various described embodiments can be practiced without these specific details. In other instances, well-known methods, procedures, components, circuits, and networks have not been described in detail so as not to unnecessarily obscure aspects of the embodiments.

[0023] General Overview

[0024] In some aspects and / or embodiments, systems, methods, and computer program products described herein include and / or implement a hierarchical offline multiobject tracker for auto-labeling. Data captured by at least one sensor associated with a vehicle is used to recognize and make decisions regarding features of the environment. In some embodiments, the sensor data is used for object detection. In object detection, sensor data is analyzed to label the sensor data with confidence scores that indicate the presence of a particular object class instance within an area of the environment corresponding to the captured sensor data. For example, the sensor data is divided intoone or more bounding boxes, and each bounding box is labeled with the likelihood that the bounding box contains a particular class of object.

[0025] A hierarchical multi-object tracker automatically determines at least one object track associated with detected object. In examples, the sensor data is automatically labeled with at least one object track associated with each detected object. For example, detections are input to a first stage tracker to generate tracklets. The tracklets are input to a second stage tracker, where the tracklets are associated with tracks of detected objects. The sensor data is labeled with at least one track that represents the motion of a detected object. In examples, the labeled data is used to train or otherwise evaluate an autonomous system.

[0026] By virtue of the implementation of systems, methods, and computer program products described herein, techniques for a hierarchical offline multi-object tracker that automatically labels sensor data are described. Some of the advantages of these techniques include the automatic labeling of tracks in sensor data. Tracklet stitching using a pairwise-likelihood graph enables discarding less than best tracklets for better tracking performance. As a result, a runtime of the tracker is reduced. Moreover, false positives are reduced by discarding frames of the track prior to labeling sensor data using the track. In this manner, the hierarchical tracker operates on tracks that are input from different types of trackers, such as a Kalman Filter based tracker, a first stage stitcher with short scan windows, a deep learning based tracker, or any combinations thereof.

[0027] Referring now to FIG. 1 , illustrated is example environment 100 in which vehicles that include autonomous systems, as well as vehicles that do not, are operated. As illustrated, environment 100 includes vehicles 102a-102n, objects 104a-104n, routes 106a-106n, area 108, vehicle-to-infrastructure (V2I) device 110, network 112, remote autonomous vehicle (AV) system 114, fleet management system 116, and V2I system 118. Vehicles 102a-102n, vehicle-to-infrastructure (V2I) device 110, network 112, autonomous vehicle (AV) system 114, fleet management system 116, and V2I system 118 interconnect (e.g., establish a connection to communicate and / or the like) via wired connections, wireless connections, or a combination of wired or wireless connections. In some embodiments, objects 104a-104n interconnect with at least one of vehicles 102a- 102n, vehicle-to-infrastructure (V2I) device 110, network 112, autonomous vehicle (AV)system 114, fleet management system 116, and V2I system 118 via wired connections, wireless connections, or a combination of wired or wireless connections.

[0028] Vehicles 102a-102n (referred to individually as vehicle 102 and collectively as vehicles 102) include at least one device configured to transport goods and / or people. In some embodiments, vehicles 102 are configured to be in communication with V2I device 110, remote AV system 114, fleet management system 116, and / or V2I system 118 via network 112. In some embodiments, vehicles 102 include cars, buses, trucks, trains, and / or the like. In some embodiments, vehicles 102 are the same as, or similar to, vehicles 200, described herein (see FIG. 2). In some embodiments, a vehicle 200 of a set of vehicles 200 is associated with an autonomous fleet manager. In some embodiments, vehicles 102 travel along respective routes 106a-106n (referred to individually as route 106 and collectively as routes 106), as described herein. In some embodiments, one or more vehicles 102 include an autonomous system (e.g., an autonomous system that is the same as or similar to autonomous system 202).

[0029] Objects 104a-104n (referred to individually as object 104 and collectively as objects 104) include, for example, at least one vehicle, at least one pedestrian, at least one cyclist, at least one structure (e.g., a building, a sign, a fire hydrant, etc.), and / or the like. Each object 104 is stationary (e.g., located at a fixed location for a period of time) or mobile (e.g., having a velocity and associated with at least one trajectory). In some embodiments, objects 104 are associated with corresponding locations in area 108.

[0030] Routes 106a-106n (referred to individually as route 106 and collectively as routes 106) are each associated with (e.g., prescribe) a sequence of actions (also known as a trajectory) connecting states along which an AV can navigate. Each route 106 starts at an initial state (e.g., a state that corresponds to a first spatiotemporal location, velocity, and / or the like) and ends at a final goal state (e.g., a state that corresponds to a second spatiotemporal location that is different from the first spatiotemporal location) or goal region (e.g. a subspace of acceptable states (e.g., terminal states)). In some embodiments, the first state includes a location at which an individual or individuals are to be picked-up by the AV and the second state or region includes a location or locations at which the individual or individuals picked-up by the AV are to be dropped-off. In some embodiments, routes 106 include a plurality of acceptable state sequences (e.g., aplurality of spatiotemporal location sequences), the plurality of state sequences associated with (e.g., defining) a plurality of trajectories. In an example, routes 106 include only high level actions or imprecise state locations, such as a series of connected roads dictating turning directions at roadway intersections. Additionally, or alternatively, routes 106 may include more precise actions or states such as, for example, specific target lanes or precise locations within the lane areas and targeted speed at those positions. In an example, routes 106 include a plurality of precise state sequences along the at least one high level action sequence with a limited lookahead horizon to reach intermediate goals, where the combination of successive iterations of limited horizon state sequences cumulatively correspond to a plurality of trajectories that collectively form the high level route to terminate at the final goal state or region.

[0031] Area 108 includes a physical area (e.g., a geographic region) within which vehicles 102 can navigate. In an example, area 108 includes at least one state (e.g., a country, a province, an individual state of a plurality of states included in a country, etc.), at least one portion of a state, at least one city, at least one portion of a city, etc. In some embodiments, area 108 includes at least one named thoroughfare (referred to herein as a “road”) such as a highway, an interstate highway, a parkway, a city street, etc. Additionally, or alternatively, in some examples area 108 includes at least one unnamed road such as a driveway, a section of a parking lot, a section of a vacant and / or undeveloped lot, a dirt path, etc. In some embodiments, a road includes at least one lane (e.g., a portion of the road that can be traversed by vehicles 102). In an example, a road includes at least one lane associated with (e.g., identified based on) at least one lane marking.

[0032] Vehicle-to-lnfrastructure (V2I) device 110 (sometimes referred to as a Vehicle- to-lnfrastructure or Vehicle-to-Everything (V2X) device) includes at least one device configured to be in communication with vehicles 102 and / or V2I infrastructure system 118. In some embodiments, V2I device 110 is configured to be in communication with vehicles 102, remote AV system 114, fleet management system 116, and / or V2I system 118 via network 112. In some embodiments, V2I device 110 includes a radio frequency identification (RFID) device, signage, cameras (e.g., two-dimensional (2D) and / or three- dimensional (3D) cameras), lane markers, streetlights, parking meters, etc. In someembodiments, V2I device 110 is configured to communicate directly with vehicles 102. Additionally, or alternatively, in some embodiments V2I device 110 is configured to communicate with vehicles 102, remote AV system 114, and / or fleet management system 116 via V2I system 118. In some embodiments, V2I device 110 is configured to communicate with V2I system 118 via network 112.

[0033] Network 112 includes one or more wired and / or wireless networks. In an example, network 112 includes a cellular network (e.g., a long term evolution (LTE) network, a third generation (3G) network, a fourth generation (4G) network, a fifth generation (5G) network, a code division multiple access (CDMA) network, etc.), a public land mobile network (PLMN), a local area network (LAN), a wide area network (WAN), a metropolitan area network (MAN), a telephone network (e.g., the public switched telephone network (PSTN), a private network, an ad hoc network, an intranet, the Internet, a fiber optic-based network, a cloud computing network, etc., a combination of some or all of these networks, and / or the like.

[0034] Remote AV system 114 includes at least one device configured to be in communication with vehicles 102, V2I device 110, network 112, fleet management system 116, and / or V2I system 118 via network 112. In an example, remote AV system 114 includes a server, a group of servers, and / or other like devices. In some embodiments, remote AV system 114 is co-located with the fleet management system 116. In some embodiments, remote AV system 114 is involved in the installation of some or all of the components of a vehicle, including an autonomous system, an autonomous vehicle compute, software implemented by an autonomous vehicle compute, and / or the like. In some embodiments, remote AV system 114 maintains (e.g., updates and / or replaces) such components and / or software during the lifetime of the vehicle.

[0035] Fleet management system 116 includes at least one device configured to be in communication with vehicles 102, V2I device 110, remote AV system 114, and / or V21 infrastructure system 118. In an example, fleet management system 116 includes a server, a group of servers, and / or other like devices. In some embodiments, fleet management system 116 is associated with a ridesharing company (e.g., an organization that controls operation of multiple vehicles (e.g., vehicles that include autonomous systems and / or vehicles that do not include autonomous systems) and / or the like).

[0036] In some embodiments, V2I system 118 includes at least one device configured to be in communication with vehicles 102, V2I device 110, remote AV system 114, and / or fleet management system 116 via network 112. In some examples, V2I system 118 is configured to be in communication with V2I device 110 via a connection different from network 112. In some embodiments, V2I system 118 includes a server, a group of servers, and / or other like devices. In some embodiments, V2I system 118 is associated with a municipality or a private institution (e.g., a private institution that maintains V2I device 110 and / or the like).

[0037] The number and arrangement of elements illustrated in FIG. 1 are provided as an example. There can be additional elements, fewer elements, different elements, and / or differently arranged elements, than those illustrated in FIG. 1. Additionally, or alternatively, at least one element of environment 100 can perform one or more functions described as being performed by at least one different element of FIG. 1. Additionally, or alternatively, at least one set of elements of environment 100 can perform one or more functions described as being performed by at least one different set of elements of environment 100.

[0038] Referring now to FIG. 2, vehicle 200 (which may be the same as, or similar to vehicles 102 of FIG. 1 ) includes or is associated with autonomous system 202, powertrain control system 204, steering control system 206, and brake system 208. In some embodiments, vehicle 200 is the same as or similar to vehicle 102 (see FIG. 1 ). In some embodiments, autonomous system 202 is configured to confer vehicle 200 autonomous driving capability (e.g., implement at least one driving automation or maneuver-based function, feature, device, and / or the like that enable vehicle 200 to be partially or fully operated without human intervention including, without limitation, fully autonomous vehicles (e.g., vehicles that forego reliance on human intervention such as Level 5 ADS- operated vehicles), highly autonomous vehicles (e.g., vehicles that forego reliance on human intervention in certain situations such as Level 4 ADS-operated vehicles), conditional autonomous vehicles (e.g., vehicles that forego reliance on human intervention in limited situations such as Level 3 ADS-operated vehicles) and / or the like. In one embodiment, autonomous system 202 includes operational or tactical functionality required to operate vehicle 200 in on-road traffic and perform part or all of DynamicDriving Task (DDT) on a sustained basis. In another embodiment, autonomous system 202 includes an Advanced Driver Assistance System (ADAS) that includes driver support features. Autonomous system 202 supports various levels of driving automation, ranging from no driving automation (e.g., Level 0) to full driving automation (e.g., Level 5). For a detailed description of fully autonomous vehicles and highly autonomous vehicles, reference may be made to SAE International's standard J3016: Taxonomy and Definitions for Terms Related to On-Road Motor Vehicle Automated Driving Systems, which is incorporated by reference in its entirety. In some embodiments, vehicle 200 is associated with an autonomous fleet manager and / or a ridesharing company.

[0039] Autonomous system 202 includes a sensor suite that includes one or more devices such as cameras 202a, LiDAR sensors 202b, radar sensors 202c, and microphones 202d. In some embodiments, autonomous system 202 can include more or fewer devices and / or different devices (e.g., ultrasonic sensors, inertial sensors, GPS receivers (discussed below), odometry sensors that generate data associated with an indication of a distance that vehicle 200 has traveled, and / or the like). In some embodiments, autonomous system 202 uses the one or more devices included in autonomous system 202 to generate data associated with environment 100, described herein. The data generated by the one or more devices of autonomous system 202 can be used by one or more systems described herein to observe the environment (e.g., environment 100) in which vehicle 200 is located. In some embodiments, autonomous system 202 includes communication device 202e, autonomous vehicle compute 202f, drive-by-wire (DBW) system 202h, and safety controller 202g.

[0040] Cameras 202a include at least one device configured to be in communication with communication device 202e, autonomous vehicle compute 202f, and / or safety controller 202g via a bus (e.g., a bus that is the same as or similar to bus 302 of FIG. 3). Cameras 202a include at least one camera (e.g., a digital camera using a light sensor such as a Charge-Coupled Device (CCD), a thermal camera, an infrared (IR) camera, an event camera, and / or the like) to capture images including physical objects (e.g., cars, buses, curbs, people, and / or the like). In some embodiments, camera 202a generates camera data as output. In some examples, camera 202a generates camera data that includes image data associated with an image. In this example, the image data mayspecify at least one parameter (e.g., image characteristics such as exposure, brightness, etc., an image timestamp, and / or the like) corresponding to the image. In such an example, the image may be in a format (e.g., RAW, JPEG, PNG, and / or the like). In some embodiments, camera 202a includes a plurality of independent cameras configured on (e.g., positioned on) a vehicle to capture images for the purpose of stereopsis (stereo vision). In some examples, camera 202a includes a plurality of cameras that generate image data and transmit the image data to autonomous vehicle compute 202f and / or a fleet management system (e.g., a fleet management system that is the same as or similar to fleet management system 116 of FIG. 1 ). In such an example, autonomous vehicle compute 202f determines depth to one or more objects in a field of view of at least two cameras of the plurality of cameras based on the image data from the at least two cameras. In some embodiments, cameras 202a is configured to capture images of objects within a distance from cameras 202a (e.g., up to 100 meters, up to a kilometer, and / or the like). Accordingly, cameras 202a include features such as sensors and lenses that are optimized for perceiving objects that are at one or more distances from cameras 202a.

[0041] In an embodiment, camera 202a includes at least one camera configured to capture one or more images associated with one or more traffic lights, street signs and / or other physical objects that provide visual navigation information. In some embodiments, camera 202a generates traffic light data associated with one or more images. In some examples, camera 202a generates TLD (Traffic Light Detection) data associated with one or more images that include a format (e.g., RAW, JPEG, PNG, and / or the like). In some embodiments, camera 202a that generates TLD data differs from other systems described herein incorporating cameras in that camera 202a can include one or more cameras with a wide field of view (e.g., a wide-angle lens, a fish-eye lens, a lens having a viewing angle of approximately 120 degrees or more, and / or the like) to generate images about as many physical objects as possible.

[0042] Light Detection and Ranging (LiDAR) sensors 202b include at least one device configured to be in communication with communication device 202e, autonomous vehicle compute 202f, and / or safety controller 202g via a bus (e.g., a bus that is the same as or similar to bus 302 of FIG. 3). LiDAR sensors 202b include a system configured to transmitlight from a light emitter (e g., a laser transmitter). Light emitted by LiDAR sensors 202b include light (e.g., infrared light and / or the like) that is outside of the visible spectrum. In some embodiments, during operation, light emitted by LiDAR sensors 202b encounters a physical object (e.g., a vehicle) and is reflected back to LiDAR sensors 202b. In some embodiments, the light emitted by LiDAR sensors 202b does not penetrate the physical objects that the light encounters. LiDAR sensors 202b also include at least one light detector which detects the light that was emitted from the light emitter after the light encounters a physical object. In some embodiments, at least one data processing system associated with LiDAR sensors 202b generates an image (e.g., a point cloud, a combined point cloud, and / or the like) representing the objects included in a field of view of LiDAR sensors 202b. In some examples, the at least one data processing system associated with LiDAR sensor 202b generates an image that represents the boundaries of a physical object, the surfaces (e.g., the topology of the surfaces) of the physical object, and / or the like. In such an example, the image is used to determine the boundaries of physical objects in the field of view of LiDAR sensors 202b.

[0043] Radio Detection and Ranging (radar) sensors 202c include at least one device configured to be in communication with communication device 202e, autonomous vehicle compute 202f, and / or safety controller 202g via a bus (e.g., a bus that is the same as or similar to bus 302 of FIG. 3). Radar sensors 202c include a system configured to transmit radio waves (either pulsed or continuously). The radio waves transmitted by radar sensors 202c include radio waves that are within a predetermined spectrum In some embodiments, during operation, radio waves transmitted by radar sensors 202c encounter a physical object and are reflected back to radar sensors 202c. In some embodiments, the radio waves transmitted by radar sensors 202c are not reflected by some objects. In some embodiments, at least one data processing system associated with radar sensors 202c generates signals representing the objects included in a field of view of radar sensors 202c. For example, the at least one data processing system associated with radar sensor 202c generates an image that represents the boundaries of a physical object, the surfaces (e.g., the topology of the surfaces) of the physical object, and / or the like. In some examples, the image is used to determine the boundaries of physical objects in the field of view of radar sensors 202c.

[0044] Microphones 202d includes at least one device configured to be in communication with communication device 202e, autonomous vehicle compute 202f, and / or safety controller 202g via a bus (e.g., a bus that is the same as or similar to bus 302 of FIG. 3). Microphones 202d include one or more microphones (e.g., array microphones, external microphones, and / or the like) that capture audio signals and generate data associated with (e.g., representing) the audio signals. In some examples, microphones 202d include transducer devices and / or like devices. In some embodiments, one or more systems described herein can receive the data generated by microphones 202d and determine a position of an object relative to vehicle 200 (e.g., a distance and / or the like) based on the audio signals associated with the data.

[0045] Communication device 202e includes at least one device configured to be in communication with cameras 202a, LiDAR sensors 202b, radar sensors 202c, microphones 202d, autonomous vehicle compute 202f, safety controller 202g, and / or DBW (Drive-By-Wire) system 202h. For example, communication device 202e may include a device that is the same as or similar to communication interface 314 of FIG. 3. In some embodiments, communication device 202e includes a vehicle-to-vehicle (V2V) communication device (e.g., a device that enables wireless communication of data between vehicles).

[0046] Autonomous vehicle compute 202f include at least one device configured to be in communication with cameras 202a, LiDAR sensors 202b, radar sensors 202c, microphones 202d, communication device 202e, safety controller 202g, and / or DBW system 202h. In some examples, autonomous vehicle compute 202f includes a device such as a client device, a mobile device (e.g., a cellular telephone, a tablet, and / or the like), a server (e.g., a computing device including one or more central processing units, graphical processing units, and / or the like), and / or the like. In some embodiments, autonomous vehicle compute 202f is the same as or similar to autonomous vehicle compute 400, described herein. Additionally, or alternatively, in some embodiments autonomous vehicle compute 202f is configured to be in communication with an autonomous vehicle system (e.g., an autonomous vehicle system that is the same as or similar to remote AV system 114 of FIG. 1 ), a fleet management system (e.g., a fleet management system that is the same as or similar to fleet management system 116 ofFIG. 1 ), a V2I device (e.g., a V2I device that is the same as or similar to V2I device 110 of FIG. 1 ), and / or a V2I system (e.g., a V2I system that is the same as or similar to V2I system 118 of FIG. 1 ).

[0047] Safety controller 202g includes at least one device configured to be in communication with cameras 202a, LiDAR sensors 202b, radar sensors 202c, microphones 202d, communication device 202e, autonomous vehicle computer 202f, and / or DBW system 202h. In some examples, safety controller 202g includes one or more controllers (electrical controllers, electromechanical controllers, and / or the like) that are configured to generate and / or transmit control signals to operate one or more devices of vehicle 200 (e.g., powertrain control system 204, steering control system 206, brake system 208, and / or the like). In some embodiments, safety controller 202g is configured to generate control signals that take precedence over (e.g., overrides) control signals generated and / or transmitted by autonomous vehicle compute 202f.

[0048] DBW system 202h includes at least one device configured to be in communication with communication device 202e and / or autonomous vehicle compute 202f. In some examples, DBW system 202h includes one or more controllers (e.g., electrical controllers, electromechanical controllers, and / or the like) that are configured to generate and / or transmit control signals to operate one or more devices of vehicle 200 (e.g., powertrain control system 204, steering control system 206, brake system 208, and / or the like). Additionally, or alternatively, the one or more controllers of DBW system 202h are configured to generate and / or transmit control signals to operate at least one different device (e.g., a turn signal, headlights, door locks, windshield wipers, and / or the like) of vehicle 200.

[0049] Powertrain control system 204 includes at least one device configured to be in communication with DBW system 202h. In some examples, powertrain control system 204 includes at least one controller, actuator, and / or the like. In some embodiments, powertrain control system 204 receives control signals from DBW system 202h and powertrain control system 204 causes vehicle 200 to make longitudinal vehicle motion, such as start moving forward, stop moving forward, start moving backward, stop moving backward, accelerate in a direction, decelerate in a direction or to make lateral vehicle motion such as performing a left turn, performing a right turn, and / or the like. In anexample, powertrain control system 204 causes the energy (e.g., fuel, electricity, and / or the like) provided to a motor of the vehicle to increase, remain the same, or decrease, thereby causing at least one wheel of vehicle 200 to rotate or not rotate.

[0050] Steering control system 206 includes at least one device configured to rotate one or more wheels of vehicle 200. In some examples, steering control system 206 includes at least one controller, actuator, and / or the like. In some embodiments, steering control system 206 causes the front two wheels and / or the rear two wheels of vehicle 200 to rotate to the left or right to cause vehicle 200 to turn to the left or right. In other words, steering control system 206 causes activities necessary for the regulation of the y-axis component of vehicle motion.

[0051] Brake system 208 includes at least one device configured to actuate one or more brakes to cause vehicle 200 to reduce speed and / or remain stationary. In some examples, brake system 208 includes at least one controller and / or actuator that is configured to cause one or more calipers associated with one or more wheels of vehicle 200 to close on a corresponding rotor of vehicle 200. Additionally, or alternatively, in some examples brake system 208 includes an automatic emergency braking (AEB) system, a regenerative braking system, and / or the like.

[0052] In some embodiments, vehicle 200 includes at least one platform sensor (not explicitly illustrated) that measures or infers properties of a state or a condition of vehicle 200. In some examples, vehicle 200 includes platform sensors such as a global positioning system (GPS) receiver, an inertial measurement unit (IMU), a wheel speed sensor, a wheel brake pressure sensor, a wheel torque sensor, an engine torque sensor, a steering angle sensor, and / or the like. Although brake system 208 is illustrated to be located in the near side of vehicle 200 in FIG. 2, brake system 208 may be located anywhere in vehicle 200.

[0053] Referring now to FIG. 3, illustrated is a schematic diagram of a device 300. As illustrated, device 300 includes processor 304, memory 306, storage component 308, input interface 310, output interface 312, communication interface 314, and bus 302. In some embodiments, device 300 corresponds to at least one device of vehicles 102 (e.g., at least one device of a system of vehicles 102), and / or one or more devices of network 112 (e.g., one or more devices of a system of network 112). In some embodiments, oneor more devices of vehicles 102 (e.g., one or more devices of a system of vehicles 102), and / or one or more devices of network 112 (e.g., one or more devices of a system of network 112) include at least one device 300 and / or at least one component of device 300. As shown in FIG. 3, device 300 includes bus 302, processor 304, memory 306, storage component 308, input interface 310, output interface 312, and communication interface 314.

[0054] Bus 302 includes a component that permits communication among the components of device 300. In some cases, processor 304 includes a processor (e.g., a central processing unit (CPU), a graphics processing unit (GPU), an accelerated processing unit (APU), and / or the like), a microphone, a digital signal processor (DSP), and / or any processing component (e.g., a field-programmable gate array (FPGA), an application specific integrated circuit (ASIC), and / or the like) that can be programmed to perform at least one function. Memory 306 includes random access memory (RAM), readonly memory (ROM), and / or another type of dynamic and / or static storage device (e.g., flash memory, magnetic memory, optical memory, and / or the like) that stores data and / or instructions for use by processor 304.

[0055] Storage component 308 stores data and / or software related to the operation and use of device 300. In some examples, storage component 308 includes a hard disk (e.g., a magnetic disk, an optical disk, a magneto-optic disk, a solid state disk, and / or the like), a compact disc (CD), a digital versatile disc (DVD), a floppy disk, a cartridge, a magnetic tape, a CD-ROM, RAM, PROM, EPROM, FLASH-EPROM, NV-RAM, and / or another type of computer readable medium, along with a corresponding drive.

[0056] Input interface 310 includes a component that permits device 300 to receive information, such as via user input (e.g., a touchscreen display, a keyboard, a keypad, a mouse, a button, a switch, a microphone, a camera, and / or the like). Additionally or alternatively, in some embodiments input interface 310 includes a sensor that senses information (e.g., a global positioning system (GPS) receiver, an accelerometer, a gyroscope, an actuator, and / or the like). Output interface 312 includes a component that provides output information from device 300 (e.g., a display, a speaker, one or more lightemitting diodes (LEDs), and / or the like).

[0057] In some embodiments, communication interface 314 includes a transceiver-like component (e.g., a transceiver, a separate receiver and transmitter, and / or the like) that permits device 300 to communicate with other devices via a wired connection, a wireless connection, or a combination of wired and wireless connections. In some examples, communication interface 314 permits device 300 to receive information from another device and / or provide information to another device. In some examples, communication interface 314 includes an Ethernet interface, an optical interface, a coaxial interface, an infrared interface, a radio frequency (RF) interface, a universal serial bus (USB) interface, a Wi-Fi® interface, a cellular network interface, and / or the like.

[0058] In some embodiments, device 300 performs one or more processes described herein. Device 300 performs these processes based on processor 304 executing software instructions stored by a computer-readable medium, such as memory 305 and / or storage component 308. A computer-readable medium (e.g., a non-transitory computer readable medium) is defined herein as a non-transitory memory device. A non-transitory memory device includes memory space located inside a single physical storage device or memory space spread across multiple physical storage devices.

[0059] In some embodiments, software instructions are read into memory 306 and / or storage component 308 from another computer-readable medium or from another device via communication interface 314. When executed, software instructions stored in memory 306 and / or storage component 308 cause processor 304 to perform one or more processes described herein. Additionally or alternatively, hardwired circuitry is used in place of or in combination with software instructions to perform one or more processes described herein. Thus, embodiments described herein are not limited to any specific combination of hardware circuitry and software unless explicitly stated otherwise.

[0060] Memory 306 and / or storage component 308 includes data storage or at least one data structure (e.g., a database and / or the like). Device 300 is capable of receiving information from, storing information in, communicating information to, or searching information stored in the data storage or the at least one data structure in memory 306 or storage component 308. In some examples, the information includes network data, input data, output data, or any combination thereof.

[0061] In some embodiments, device 300 is configured to execute software instructions that are either stored in memory 306 and / or in the memory of another device (e.g., another device that is the same as or similar to device 300). As used herein, the term “module” refers to at least one instruction stored in memory 306 and / or in the memory of another device that, when executed by processor 304 and / or by a processor of another device (e.g., another device that is the same as or similar to device 300) cause device 300 (e.g., at least one component of device 300) to perform one or more processes described herein. In some embodiments, a module is implemented in software, firmware, hardware, and / or the like.

[0062] The number and arrangement of components illustrated in FIG. 3 are provided as an example. In some embodiments, device 300 can include additional components, fewer components, different components, or differently arranged components than those illustrated in FIG. 3. Additionally or alternatively, a set of components (e.g., one or more components) of device 300 can perform one or more functions described as being performed by another component or another set of components of device 300.

[0063] Referring now to FIG. 4, illustrated is an example block diagram of an autonomous vehicle compute 400 (sometimes referred to as an “AV stack”). As illustrated, autonomous vehicle compute 400 includes perception system 402 (sometimes referred to as a perception module), planning system 404 (sometimes referred to as a planning module), localization system 406 (sometimes referred to as a localization module), control system 408 (sometimes referred to as a control module), and database 410. In some embodiments, perception system 402, planning system 404, localization system 406, control system 408, and database 410 are included and / or implemented in an autonomous navigation system of a vehicle (e.g., autonomous vehicle compute 202f of vehicle 200). Additionally, or alternatively, in some embodiments perception system 402, planning system 404, localization system 406, control system 408, and database 410 are included in one or more standalone systems (e.g., one or more systems that are the same as or similar to autonomous vehicle compute 400 and / or the like). In some examples, perception system 402, planning system 404, localization system 406, control system 408, and database 410 are included in one or more standalone systems that are located in a vehicle and / or at least one remote system as described herein. In someembodiments, any and / or all of the systems included in autonomous vehicle compute 400 are implemented in software (e.g., in software instructions stored in memory), computer hardware (e.g., by microprocessors, microcontrollers, application-specific integrated circuits (ASICs), Field Programmable Gate Arrays (FPGAs), and / or the like), or combinations of computer software and computer hardware. It will also be understood that, in some embodiments, autonomous vehicle compute 400 is configured to be in communication with a remote system (e.g., an autonomous vehicle system that is the same as or similar to remote AV system 114, a fleet management system 116 that is the same as or similar to fleet management system 116, a V2I system that is the same as or similar to V2I system 118, and / or the like).

[0064] In some embodiments, perception system 402 receives data associated with at least one physical object (e.g., data that is used by perception system 402 to detect the at least one physical object) in an environment and classifies the at least one physical object. In some examples, perception system 402 receives image data captured by at least one camera (e.g., cameras 202a), the image associated with (e.g., representing) one or more physical objects within a field of view of the at least one camera. In such an example, perception system 402 classifies at least one physical object based on one or more groupings of physical objects (e.g., bicycles, vehicles, traffic signs, pedestrians, and / or the like). In some embodiments, perception system 402 transmits data associated with the classification of the physical objects to planning system 404 based on perception system 402 classifying the physical objects.

[0065] In some embodiments, planning system 404 receives data associated with a destination and generates data associated with at least one route (e.g., routes 106) along which a vehicle (e.g., vehicles 102) can travel along toward a destination. In some embodiments, planning system 404 periodically or continuously receives data from perception system 402 (e.g., data associated with the classification of physical objects, described above) and planning system 404 updates the at least one trajectory or generates at least one different trajectory based on the data generated by perception system 402. In other words, planning system 404 may perform tactical function-related tasks that are required to operate vehicle 102 in on-road traffic. Tactical efforts involve maneuvering the vehicle in traffic during a trip, including but not limited to decidingwhether and when to overtake another vehicle, change lanes, or selecting an appropriate speed, acceleration, deacceleration, etc. In some embodiments, planning system 404 receives data associated with an updated position of a vehicle (e.g., vehicles 102) from localization system 406 and planning system 404 updates the at least one trajectory or generates at least one different trajectory based on the data generated by localization system 406.

[0066] In some embodiments, localization system 406 receives data associated with (e.g., representing) a location of a vehicle (e.g., vehicles 102) in an area. In some examples, localization system 406 receives LiDAR data associated with at least one point cloud generated by at least one LiDAR sensor (e.g., LiDAR sensors 202b). In certain examples, localization system 406 receives data associated with at least one point cloud from multiple LiDAR sensors and localization system 406 generates a combined point cloud based on each of the point clouds. In these examples, localization system 406 compares the at least one point cloud or the combined point cloud to two-dimensional (2D) and / or a three-dimensional (3D) map of the area stored in database 410. Localization system 406 then determines the position of the vehicle in the area based on localization system 406 comparing the at least one point cloud or the combined point cloud to the map. In some embodiments, the map includes a combined point cloud of the area generated prior to navigation of the vehicle. In some embodiments, maps include, without limitation, high-precision maps of the roadway geometric properties, maps describing road network connectivity properties, maps describing roadway physical properties (such as traffic speed, traffic volume, the number of vehicular and cyclist traffic lanes, lane width, lane traffic directions, or lane marker types and locations, or combinations thereof), and maps describing the spatial locations of road features such as crosswalks, traffic signs or other travel signals of various types. In some embodiments, the map is generated in real-time based on the data received by the perception system.

[0067] In another example, localization system 406 receives Global Navigation Satellite System (GNSS) data generated by a global positioning system (GPS) receiver. In some examples, localization system 406 receives GNSS data associated with the location of the vehicle in the area and localization system 406 determines a latitude and longitude of the vehicle in the area. In such an example, localization system 406 determines theposition of the vehicle in the area based on the latitude and longitude of the vehicle. In some embodiments, localization system 406 generates data associated with the position of the vehicle. In some examples, localization system 406 generates data associated with the position of the vehicle based on localization system 406 determining the position of the vehicle. In such an example, the data associated with the position of the vehicle includes data associated with one or more semantic properties corresponding to the position of the vehicle.

[0068] In some embodiments, control system 408 receives data associated with at least one trajectory from planning system 404 and control system 408 controls operation of the vehicle. In some examples, control system 408 receives data associated with at least one trajectory from planning system 404 and control system 408 controls operation of the vehicle by generating and transmitting control signals to cause a powertrain control system (e.g., DBW system 202h, powertrain control system 204, and / or the like), a steering control system (e.g., steering control system 206), and / or a brake system (e.g., brake system 208) to operate. For example, control system 408 is configured to perform operational functions such as a lateral vehicle motion control or a longitudinal vehicle motion control. The lateral vehicle motion control causes activities necessary for the regulation of the y-axis component of vehicle motion. The longitudinal vehicle motion control causes activities necessary for the regulation of the x-axis component of vehicle motion. In an example, where a trajectory includes a left turn, control system 408 transmits a control signal to cause steering control system 206 to adjust a steering angle of vehicle 200, thereby causing vehicle 200 to turn left. Additionally, or alternatively, control system 408 generates and transmits control signals to cause other devices (e.g., headlights, turn signal, door locks, windshield wipers, and / or the like) of vehicle 200 to change states.

[0069] In some embodiments, perception system 402, planning system 404, localization system 406, and / or control system 408 implement at least one machine learning model (e.g., at least one multilayer perceptron (MLP), at least one convolutional neural network (CNN), at least one recurrent neural network (RNN), at least one autoencoder, at least one transformer, and / or the like). In some examples, perception system 402, planning system 404, localization system 406, and / or control system 408 implement at least onemachine learning model alone or in combination with one or more of the above-noted systems. In some examples, perception system 402, planning system 404, localization system 406, and / or control system 408 implement at least one machine learning model as part of a pipeline (e.g., a pipeline for identifying one or more objects located in an environment and / or the like).

[0070] Database 410 stores data that is transmitted to, received from, and / or updated by perception system 402, planning system 404, localization system 406 and / or control system 408. In some examples, database 410 includes a storage component (e.g., a storage component that is the same as or similar to storage component 308 of FIG. 3) that stores data and / or software related to the operation and uses at least one system of autonomous vehicle compute 400. In some embodiments, database 410 stores data associated with 2D and / or 3D maps of at least one area. In some examples, database 410 stores data associated with 2D and / or 3D maps of a portion of a city, multiple portions of multiple cities, multiple cities, a county, a state, a State (e.g., a country), and / or the like). In such an example, a vehicle (e.g., a vehicle that is the same as or similar to vehicles 102 and / or vehicle 200) can drive along one or more drivable regions (e.g., single-lane roads, multi-lane roads, highways, back roads, off road trails, and / or the like) and cause at least one LiDAR sensor (e.g., a LiDAR sensor that is the same as or similar to LiDAR sensors 202b) to generate data associated with an image representing the objects included in a field of view of the at least one LiDAR sensor.

[0071] In some embodiments, database 410 can be implemented across a plurality of devices. In some examples, database 410 is included in a vehicle (e.g., a vehicle that is the same as or similar to vehicles 102 and / or vehicle 200), an autonomous vehicle system (e.g., an autonomous vehicle system that is the same as or similar to remote AV system 114, a fleet management system (e.g., a fleet management system that is the same as or similar to fleet management system 116 of FIG. 1 , a V2I system (e.g., a V2I system that is the same as or similar to V2I system 118 of FIG. 1 ) and / or the like.

[0072] Referring now to FIG. 5, illustrated are diagrams of implementations 500 of a hierarchical multi-object tracker for auto-labeling. In some embodiments, implementation 500 includes an AV compute 510, hierarchical multi-object tracker 504, and planning system 506. In some embodiments, the AV compute 510 is the same as or similar to theautonomous vehicle compute 202f of FIG. 2. In some embodiments, the planning system 506 is the same as or similar to the planning system 404 of FIG. 4. Additionally, some embodiments, the hierarchical multi-object tracker 504 is the same as or similar to the hierarchical multi-object tracker 612 of FIG. 6, described later.

[0073] The implementation 500 represents online or offline implementations hierarchical multi-object tracker. Sensors (e.g., cameras 202a, LiDAR sensors 202b, radar sensors 202c, and microphones 202d of FIG. 2) capture data from the environment that maps the features in the environment to quantitative measurements. In an online implementation 500, the sensor data is evaluated in real time as data is captured by sensors of the vehicle 502. For example, sensor data is streamed from multiple sensor sources, and the AV compute 510 makes predictive decisions to control the vehicle 502. In an offline implementation 500, sensor data is recorded over a period of time as the vehicle 502 navigates through the environment. The sensor data is stored in a drive log for offline processing and or analysis.

[0074] For ease of description, sensor data captured by particular sensors are described. However, the sensors can include, for example, sensors that capture dynamics associated with a vehicle (e.g., Inertial Measurement Unit (IMU), encoders, inertial sensors, and positioning sensors (GNSS receivers) as well as sensors that capture information associated with the environment surrounding the vehicle (e.g., camera, LiDAR, radar, and ultrasonic sensors). In examples, camera sensors continuously record surrounding environments in real-time. GPS sensors track vehicle locations in real-time. Additionally, in examples LiDAR and / or radar sensor perceive the locations and distances of other objects relative to the vehicle in real-time. In examples, real-time data is recorded and used in an offline implementation.

[0075] In an offline implementation 500, a hierarchical multi-object tracker 504 obtains detections extracted from recorded sensor data. In some embodiments, the detections refer to representations of detected objects in the environment. The detections are generated by at least one object detection network. An object detection network identifies and localizes objects in each frame of sensor data. The hierarchical multi-object tracker 504 determines tracks 514 associated with detections. In examples, a track refers to a representation of an object trajectory across frames of sensor data. Tracking an objectassociates the detected object with a path in an environment across multiple frames over time. The association of detections corresponding to a same object forms a track corresponding to the object over a period of time. In some embodiments, the hierarchical multi-object tracker 504 includes multiple tracking stages that iteratively determine tracklets (e.g., a series of continuous detections) based on the detections. The tracklets are stitched together in a tracking stage to obtain tracks. The tracks are output by the hierarchical multi-object tracker 504. In some embodiments, the tracks are stored in the drive logs using track identifications (IDs), where the captured sensor data is labeled with a track ID as published by the hierarchical multi-object tracker 504. In some embodiments, the stored sensor data is labeled according to the detections output by the object detection network and the track ID as published by the hierarchical multi-object tracker 504.

[0076] In an online implementation 500, the detections are obtained from a perception system (e.g., perception system 402 of FIG. 4) of the vehicle 502 as the vehicle 502 navigates through an environment in real-time. The hierarchical multi-object tracker 504 generates tracks, and the tracks are transmitted to a planning system 506 (e.g., planning system 404 of FIG. 4) in real-time. The planning system 506 generates data associated with at least one route (e.g., routes 106 of FIG. 1 ) along which a vehicle 502 can travel along toward a destination in view of the tracks 516. In some embodiments, the routes generated by the planning system 506 are based on, at least in part, avoiding traffic conflicts with the tracks generated by the hierarchical multi-object tracker 504.

[0077] As described in the examples of FIG. 5, the hierarchical multi-object tracker 504 is used in an offline implementation to label sensor data. Labeling sensor data enables the development and training of various vehicle functionalities. In some embodiments, an auto-labeling pipeline including the hierarchical multi-object tracker 504 generates labels for recorded data, such as drive logs. In examples, the auto-labeling pipeline also includes a constant velocity motion model that evaluates and annotates the recorded sensor data. The auto-labeling pipeline reduces the manual resources consumed when labeling data while maintaining a high level of accuracy and consistency in the generated labels. Additionally, the auto-labeling pipeline enables quick and efficient annotation of large data sets, further enabling rapid development and training of machine learning models. Forexample, the labeled datasets are used in supervised learning algorithms, as training data for neural networks, as ground truth labels for the evaluation of the AV stack, and for other testing and evaluation purposes. In examples, the labeled sensor data is used to train or evaluate components of the AV stack including perception system 402, planning system 404, localization system 406, or control system 408 of an autonomous navigation system of a vehicle as described with respect to FIG. 4.

[0078] FIG. 6 is a block diagram of a workflow 600 for a hierarchical multi-object tracker for auto-labeling. A hierarchical multi-object tracker 612 generates tracks 622 that are used to automatically label sensor data. In the workflow 600 of FIG. 6, the hierarchical multi-object tracker 612 outputs a track identification (ID) for respective detections 604 input to the hierarchical multi-object tracker 612. In examples, the detections 604 are input to a first stage tracker 614 to generate tracklets 616. A second stage stitcher 618 stitches together tracklets associated with a same object to generate tracks corresponding to the same object. In some embodiments, the sensor data is labeled with tracks.

[0079] In the example of FIG. 6, a detector 602 is used to generate detections 604. In examples, the detector 602 is an object detection network that discovers objects in the environment using the sensor data. The object detection network generates detections for each frame of sensor data. In examples, a frame is a set of one or more data values captured at a timestamp. A record of sensor data (e.g., drive log) includes multiple frames of data captured at a series of continuous timestamps. In examples, the detector is a semantic network that detects objects in captured sensor data. The detector outputs detections 604 that include bounding boxes associated with respective objects in the environment and confidence scores quantifying a likelihood of the presence of object class instances (e.g., cars, pedestrians, or bikes) within the bounding boxes. In examples, the bounding boxes are two-dimensional bounding boxes or three-dimensional bounding boxes.

[0080] For example, an image semantic network (ISN) or LiDAR semantic network (LSN) outputs detections 604 including oriented bounding boxes with a probability that the object within a respective bounding box belongs to a particular class. In examples, an ISN takes as input image data (e.g., captured by a camera sensor) and outputs a set of predicted 2D or 3D bounding boxes for potential objects in the environment andcorresponding confidence scores representing the presence of object class instances within the bounding boxes. The 2D or 3D bounding boxes include information about the size (e.g., dimensions), orientation, and location of a respective object. In examples, the ISN predicts the class of each pixel in an image and outputs a confidence score for each pixel in the image. An example confidence score is a probability value that indicates the probability that the class of the pixel was correctly predicted.

[0081] Similarly, an LSN takes as input at least one point cloud (e.g., captured by a LiDAR) and outputs a set of predicted 2D or 3D bounding boxes for potential objects in the environment and confidence scores representing the presence of object class instances within the bounding boxes. In an example, the LSN receives a plurality of data points that represent the environment. For example, each data point of the plurality of data points in the at least one point cloud is a set of 3D spatial coordinates (e.g., x, y, z coordinates). Using one or more point clouds, the LSN estimates oriented 2D or 3D bounding boxes for object class instances such as cars, pedestrians, and cyclists, based on the point clouds. Similar to an ISN, the predicted 2D or 3D bounding boxes output by the LSN include information about the size, orientation and location of the object. The predicted 2D or 3D set of bounding boxes also include confidence scores for the presence of object class instances within the bounding boxes.

[0082] The hierarchical multi-object tracker 612 includes a first stage tracker 614 and a second stage stitcher 618. In examples, the first stage tracker 614 obtains the detections 604. In examples, the first stage tracker 614 is a Kalman filter based tracker, a stitcher with a short scan time, or a deep learning based tracker. For example, a Kalman filter based tracker uses a Kalman filter to predict and update the location and velocity of an object given the object detections in each frame of sensor data. The Kalman filter based tracker predicts the object’s future state (position, velocity, etc.) based on its current state and a mathematical model of its motion. When new measurements (e.g., from sensors) are available, the Kalman filter updates its prediction by combining the predicted state with the new measurements to produce a more accurate estimate using a weighted average, where higher weights are assigned to the estimates with a higher certainty. In this manner, a tracklet associated with the object is generated.

[0083] In another example, a stitcher is used as the first stage tracker, where the stitcher includes a series of filters that determine the likelihood a detected object belonging to a tracklet. The filters include, for example, temporal filters and kinematic filters. The scan time of the stitcher refers to a length of time (and ultimately the number of frames) for which the sensor data is evaluated. The short scan window may be, for example, less than a second or less than half a second. In examples, the length of the scan window is selected to reduce a runtime of the first stage tracker 614. A temporal filter evaluates temporal information associated with tracklets to ensure that the tracklets to be stitched are associated with different timestamps (e.g., the same frame does not exist in each tracklet). Kinematic filters are applied to tracklets that are not associated with a same frame (e.g., the tracklets occur at different points in time). A kinematic filter evaluates tracklets and prevents stitching of tracks that are kinematically infeasible.

[0084] In examples, a deep learning based tracker includes a machine learning model that receives as input an initial set of object detections and determines a unique identification associated with each of the initial detections. The deep learning based tracker then tracks the detected objects as in motion across frames of the sensor data.

[0085] As shown in the example of FIG. 6, the first stage tracker 614 outputs tracklets 616. In some embodiments, the tracklets are constrained by detections associated with a respective object, as a respective tracklet is a series of continuous detections of a respective object in frames of sensor data. A misdetection (e.g., incorrectly detecting a previously detected object as being a newly detected object), false positive detection (e.g., incorrectly detecting an object as being a previously detected object), or missing detections (e.g., failure to detect an object present in a frame of data) terminates the tracklet. In some embodiments, misdetections, false positive detections, or missing detections occur due to abrupt changes in an object shape and motion, such as during partial occlusions of an object. Misdetections, false positive detections, or missing detections also occur due to changes in illumination or lighting in the environment. Tracklets are further described with respect to FIG. 7, and a first stage tracker workflow is described with respect to FIG. 10A.

[0086] The second stage stitcher 618 of the hierarchical multi-object tracker 612 stitches the tracklets 616 from the first stage tracker 614 to generate a track of the object.Evaluation of the tracklets enables a determination that the tracklets are associated with the same object, and thus correspond to the same track, even with the presence of misdetections, false positive detections, or missing detections. Instead of stitching detections generated by detectors to obtain tracks, the present techniques generate tracklets for stitching to reduce a runtime of the hierarchical multi-object tracker 612. Stitching the tracklets 616 output by the first stage tracker 614 consumes less resources when compared to directly stitching detections 604 generated from the detectors to form tracks. In some embodiments, second stage stitcher 618 associates the tracklets to generate tracks by building a graph using the tracklets 616. Graphs are further described with respect to FIGs. 11 -13. Respective tracks associated with each object are found using the graph. In examples, the graph is iteratively updated based on the captured sensor data.

[0087] FIG. 7 shows detections in frames of sensor data. In examples, the detections are objects detected in the sensor data and are generated by at least one object detection network. In the example of FIG. 7, each column 702, 703, 704, 705, 706, and 707 represents a frame of sensor data at a respective timestamp in a series of continuous timestamps. For each timestamp, the boxes at each row corresponds to a detection of an object in the environment. The same shading for each box indicates that the box belongs to a particular tracklet. The digits inside each box represent a tracklet ID. A tracklet is generated based on detections, where a tracklet corresponds to detections at continuous timestamps. Missing detections in a frame results in a termination of a tracklet. A new tracklet is initialized in response to a missing detection.

[0088] As shown in the example of FIG. 7, the first row 712 of detections create a single tracklet with a tracklet ID of zero, with no detection associated with the tracklet ID of zero in the frame 702 of data. The second row 714 shows five detections that are ultimately associated with a same object. In the second row 714, a detection is missing at a frame 705. For the first three detections in the second row 714, a tracklet ID of one is assigned to a detected object. After the missing detection at frame 705, a new tracklet is initiated for the object with a tracklet ID of two. Thus, the detections in the second row 714 correspond to two tracklets, a first tracklet with a tracklet ID of one, and a second tracklet with a tracklet ID of two.

[0089] Similarly, the object detected at the third row 716 corresponds to two tracklets, a first tracklet with a tracklet ID of three, and a second tracklet with a tracklet ID of four. The detections in the fourth row 718 correspond to a single tracklet with a tracklet ID of five. The detections at the fifth row 720 correspond to two tracklets, a first tracklet with a tracklet ID of six, and a second tracklet with a tracklet ID of seven. As shown in the example of FIG. 7, a tracklet corresponds to a series of continuous detections, and includes one or more detections.

[0090] As discussed with respect to FIG. 6, tracklets are generated by the first stage tracker (e.g., first stage tracker 614 of FIG. 6). The first stage tracker can be, for example, a Kalman filter based tracker, a stitcher with short scan window, or a deep learning based tracker. As shown in FIG. 7, each tracklet is a series of one or more continuous detections. In the example of FIG. 7, each tracklet is shown by detections with the same tracklet ID. In some embodiments, a second stage stitcher (e.g., the second stage stitcher 618 of FIG. 6) generates tracks from the tracklets. In the second stage tracker, tracklets are stitched together to form a respective track for each detected object. In some embodiments, a bipartite graph is built based on the tracklets to determine object tracks using graph-based object tracking. The tracklets form nodes of a weighted graph, where each node of the weighted graph corresponds to detections of the respective tracklet and the edges of the graph are weighted according to a first cost metric generated by the first stage tracker. The weighted track graph is converted to a bipartite graph, where each node of the weighted track graph is represented in two disjoint and independent sets (U, V), where every edge connects to one node in the first set of nodes (U) and one node in the second set of nodes (V). T racks are generated from the bipartite graph by determining a best path cover using the bipartite graph. In examples, best path cover in a bipartite graph is a set of vertex-disjoint paths that cover the vertices in the graph and minimizes the number of paths.

[0091] FIG. 8 shows a weighted graph. In the examples of FIG. 8, the weighted track graph 800 includes three frames of detections at timestamps to, t1 , and t2. For each frame, two detections are shown. Accordingly, at timestamp tO, detections z1 and z4 occur; at timestamp t1 , detections z2 and z5 occur; and at timestamp t2, detections z3 and z6 occur. A first stage tracker generates two tracklets: a first tracklet 802 (shown bya dashed line) associating detections z1 , z2, and z6; and a second tracklet 804 (shown by a dashed line) associating detections z4 and z5. Detection z3 represents a third, single detection tracklet 806.

[0092] The respective weights associated with each edge of the weighted graph are shown as wO, w1 , w2, w3, w4, w5, w6, and w7. As shown in the example of FIG. 8, each detection is a node of the weighted graph, as each edge is associated with a weight that is a cost representing a likelihood that a first detection and a second detection are associated with a same object.

[0093] FIG. 9A shows a weighted graph 900A with nodes partitioned into two disjoint and independent sets. Each node corresponds to a detection of a tracklet. As shown in FIG. 9A, the nodes of the weighted graph a represented by a first set of nodes (') associated with incoming edges and a second set of nodes (”) associated with outgoing edges. As shown in FIG. 9A, the weights are updated and each edge is associated with a weight.

[0094] FIG. 9B shows a bipartite graph 900B. Each node of the bipartite graph is associated with a timestamp, tracklet ID, and state distributions. In some embodiments, edge weights of the weighted graph are assumed to Gaussian. In examples, the assumed distribution enables a prediction of the likelihood of respective weights occurring. The transformation from a weighted graph to a bipartite graph can leverage the properties of Gaussian distributions to improve performance or accuracy of tracks derived from the bipartite graph. Based on the partitioned nodes shown in FIG. 9A, the weighted graph of FIG. 8 is transformed to a bipartite graph shown in FIG. 9B. Using the bipartite graph, determining a track for each object is done by finding a best path cover. In some embodiments, a path cover is a set of directed paths such that every node U, V belongs to at least one path. The best path cover is the path with the lowest cost. By selecting the path with the lowest cost, tracklets are stitched together to form tracks, as the cost between two tracklets associated with the same object will be the lowest cost. In some embodiments, finding the best path cover is done by solving a bipartite matching problem.

[0095] FIG. 10A shows a workflow 1000A corresponding to a first stage tracker. In examples, the first stage tracker is the first stage tracker 614 of FIG. 6. The first stage tracker may be, for example, a Kalman filter based tracker. At block 1002, detections areobtained. In some embodiments, the detections are obtained from detectors such as object detection networks.

[0096] At block 1004, tracks and detections are associated.

[0097] Unmatched detections 1006 are input to block 1010. At block 1010, tracks are established.

[0098] At block 1012, unmatched tracks 1008 persist as unassociated tracks. In some embodiments, unassociated tracks are terminated (1018) after a predetermined time period. In some embodiments, the predetermined time period is two seconds.

[0099] At block 1014, a 3D Kalman filter is used to generate tracklets 1016.

[0100] In some embodiments, for each detection a state vector is generated for input to the Kalman filter, where the state vector is as follows:X = (x, y, z, theta, L, W, H, confidence, vx, vy, vz)

[0101] This information, X, is obtained from the detectors (e.g., detector 602 of FIG. 6). The state space and measurement space of the Kalman filter are the same. In examples, x, y, z indicates the position of the detection, theta represents an orientation of the detection, and L, W, and H represents the length, width, and height associated with the detection. The confidence represents a likelihood that the detection is correctly classified as a particular object class instance. The variables vx, vy, and vz represent a velocity of the object in the x, y, and z directions. The state vector represents a state space that the Kalman filter updates over time, combining the information from a constant velocity motional model to track a detected object. The measurement space represents the variables that are directly measured by the sensors. In some embodiments, the state space and measurement space are the same as the state vector is extracted from captured sensor data.

[0102] In examples, the 3D Kalman filter predicts tracklets 1016, and provides measurement updates to associated tracks. The measurement updates are input to block 1004, where tracks and detections are associated. As a first stage tracker, the 3D Kalman filter generates short tracklets. In examples, the short tracklets are continuous detections associated with an object. For each tracklet, if there is no detection associated a timestamp, the tracklet is terminated and output as a tracklet 1016.

[0103] For the associated tracks at block 1004, a measurement update is applied from the output of the Kalman filter. For unassociated tracks, such as unmatched tracks 1008, they are propagated without a measurement update. For example, the unassociated tracks coast, which means that there are no measurements associated with tracks propagated by the constant velocity motion model. Accordingly, the tracklets are allowed to float for a period of two or three seconds until the tracks are terminated for lack of measurements. When measurements are detected for the track, a new track is created (e.g., birthed) for the new measurement period.

[0104] FIG. 10B shows a workflow 1000B of a second stage tracker. In examples, the second stage tracker is the stitcher 618 of FIG. 6. The second stage tracker is graph based, using a bipartite representation as described above. Each tracklet 1016 is stored as a node in the stitcher, containing: timestamps, tracklet IDs, state distributions, assumed to Gaussian: Nj . a2).

[0105] When building (1020) the bipartite graph, the edge weight is calculated from the costs. Edges with a corresponding weight in the weighted graph are transformed or mapped to respective weighted edges of the bipartite graph. In examples, a second cost metric used to weight edges of the bipartite graph is the negative log of a kinematic likelihood, a type likelihood, a likelihood based on intersection over union (loll) / generalized intersection over union (Gloll), and the like. The second cost metric is derived based on, at least in part, the first cost metric (e.g., a timestamp, a heading, or a velocity of the detections) used to weight edges of the corresponding weighted graph. In examples, the kinematic likelihood refers to the likelihood that the motion from one tracklet leads to the next tracklet. The type likelihood refers to the predicted classification matching from one tracklet to the next tracklet. loU / Gloll are defined using the position of the bounding box. For example, if the bounding boxes from one tracklet to the next are positioned close to each other, they are likely from the same object, depending on how much they overlap. To build the graph, newly added nodes are tracklets created from newly output tracklets from first stage tracker. Newly added edges are located between existing nodes and new nodes that satisfy an evaluation to ensure feasible time, position, and velocity changes across the edges.

[0106] Once the bipartite graph is built, the graph is solved (1022). To solve the bipartite graph, a cost matrix is constructed. In the cost matrix, for each edge, the "from" node is in the rows, and the "to" node is in the columns. Minimum cost bipartite matching is used to find the best path cover. In minimum cost bipartite matching, a minimum cost is found for complete matching between the two vertex subsets of a given weighted bipartite graph.

[0107] The bipartite graph is resolved (1024) to eliminate false positives in tracking. To resolve the graph, the stitched tracks obtained according to the best path cover are cropped. In some embodiments, if the head tracklet (e.g., an initial tracklet associated with a respective track) is more than a predetermined length of time (e.g., nscan_time) in the past, then the accumulated state history stored in its circular buffer for the head tracklet is passed to its descendant tracklet. In examples, the descendant tracklet is a next tracklet of a corresponding track, in chronological order. The head tracklet ID is used as an ID for the descendant node. The node corresponding to the head tracklet is deleted, and incoming edges to the descendant node are deleted.

[0108] In some embodiments, resolving the bipartite graph includes deleting the tail of the tracks (e.g., the end of the track) to improve a recall of the tracker output, enabling auto-labeled data with higher quality. In some embodiments, the tracker outputs the tracks when they are confirmed. To reduce false positives, a predetermined number of measurement updates are used to confirm a track. Additionally, in resolving the graph, the first two frames of each track are discarded, which will lower the recall. Therefore, to increase recall, back-tracing of the tracks by using future information is implemented to change the confirmation status of the first two frames for each track from 'tentative' to 'confirmed' to output the confirmed tracks.

[0109] In some embodiments, the tracks are terminated after a predetermined period of time, such as one or two seconds, after no additional measurements are associated with a respective track. The boxes from the track at the timestamps without measurement updates are output due to the propagation of the bounding boxes being performed based on the constant velocity motion model. Using the constant velocity motion model, and bounding boxes are present even if there is no measurement update. However, these boxes most likely have no matched ground truth, and are considered false positives. Postprocessing is used to eliminate bounding these boxes without measurements from the output, thereby reducing the false positives. This is similar to removing the first few frames of a track, since the tracker uses a few frames with measurements to confirm a track. Therefore, to increase recall, cutting the tail for each track is implemented when there are no additional measurement updates to prevent these boxes from being output. After resolving the bipartite graph, the tracks are output. In some embodiments, outputting the tracks includes labeling sensor data with track IDs.

[0110] Referring now to FIG. 11 , illustrated is a flowchart of a process 1100 for hierarchical offline multi-object tracker for auto-labeling. In some embodiments, one or more of the steps described with respect to process 1100 are performed (e.g., completely, partially, and / or the like) by a hierarchical multi-object tracker, such as the hierarchical offline multi-object tracker 612 of FIG. 6. Additionally, or alternatively, in some embodiments one or more steps described with respect to process 600 are performed (e.g., completely, partially, and / or the like) by another device or group of devices separate from or including the hierarchical offline multi-object tracker 612 such as the AV compute 510 of FIG. 5.

[0111] At block 1102, detections are obtained corresponding to observed objects in sensor data captured from the environment. In examples, detections are detected objects that represent an identified potential object of interest in the sensor data. In some embodiments, the detections include the output of an ISN, LSN, or other object detector.

[0112] At block 1104, tracklets are generated based on the detections. In examples, tracklets are generated using a first stage tracker, such as the first stage tracker 614 of FIG. 6. Tracklets are output by a tracker to determine which of the detections corresponds to the object being tracked. In some embodiments, the first stage tracker is a stitcher, machine learning based, or Kalman filter based. In examples, the tracklet includes nodes that represents a detection corresponding to a single object at a timestamp.

[0113] At block 1106, a weighted track graph is generated based on the tracklets. The weights are assigned to edges between nodes of the tracklets according to at least one first cost metric associated with the nodes. In examples, the cost metric corresponds to a timestamp, a heading, or a velocity of the detections.

[0114] At block 1108, the weighted track graph is converted to a bipartite graph, wherein a node of the bipartite graph corresponds to a respective detection of the tracklets, and edges of the bipartite graph are assigned weights according to at least one second cost metric associated with the nodes. In examples, the converting creates the bipartite graph from all tracklets in the weighted graph, including multiple tracklets corresponding to a single object, e.g., objects with two tracklets due to a missing detection or misdetection.

[0115] In examples, the first cost metric (e.g., a timestamp, a heading, or a velocity of the detections) are used, at least in part, to derive at least one second cost metric associated with a respective nodes. For example, the weighted graph is converted into a bipartite graph by splitting each vertex in the weighted graph. The two vertices belong to different sets in the bipartite graph. In some embodiments, for each edge a with weight in the weighted graph, a corresponding edge is created in the bipartite graph with a weight that is a second cost metric, based on at least in part, the first cost metric. Additionally or alternatively, in some embodiments, for each edge a with weight in the weighted graph, a corresponding edge is created in the bipartite graph with the same weight as the weighted graph, preserving the weighted nature of the graph.

[0116] At block 1110, the bipartite graph is solved to determine at least one best path cover. Solving the bipartite graph is used to determine full tracks for each detected object. At block 1112, the solved bipartite graph is resolved to obtain a track representing the at least one best path cover corresponding to respective observed objects, wherein frames of the track that do not satisfy a predetermined threshold are discarded. In examples, the resolving includes back tracing the tracks and removing the tail of the tracks to improve the recall of the tracker output, which will offer auto-labeled data with higher quality.

[0117] At block 1114, the sensor data is updated by labeling the detections with a track identifier corresponding to the track representing the best path cover of respective observed objects.

[0118] Embodiments

[0119] According to some non-limiting embodiments or examples, provided is method including: obtaining detections corresponding to observed objects in sensor data; generating tracklets based on the detections where a tracklet includes nodes representing a respective observed object at a series of continuous timestamps and edges connectingthe nodes at the series of continuous timestamps; generating a weighted track graph based on the tracklets, where weights are assigned to edges between nodes of the tracklets according to at least one first cost metric associated with the nodes; converting the weighted track graph to a bipartite graph, where a node of the bipartite graph corresponds to a respective detection of the tracklets, and edges of the bipartite graph are assigned weights according to at least one second cost metric associated with the nodes; solving the bipartite graph to determine at least one best path cover corresponding to respective observed objects; resolving the solved bipartite graph to obtain a track representing the at least one best path cover corresponding to respective observed objects, where frames of the track that do not satisfy a predetermined threshold are discarded; and labeling the detections with a track identifier corresponding to the track representing the best path cover of the respective observed object.

[0120] According to some non-limiting embodiments or examples, provided is a system, including: at least one processor, and at least one non-transitory storage media storing instructions that, when executed by the at least one processor, cause the at least one processor to: obtain detections corresponding to observed objects in sensor data; generate tracklets based on the detections, where a tracklet includes nodes representing a respective observed object at a series of continuous timestamps and edges connecting the nodes at the series of continuous timestamps; generate a weighted track graph based on the tracklets, where weights are assigned to edges between nodes of the tracklets according to at least one first cost metric associated with the nodes; convert the weighted track graph to a bipartite graph, where a node of the bipartite graph corresponds to a respective detection of the tracklets, and edges of the bipartite graph are assigned weights according to at least one second cost metric associated with the nodes; solve the bipartite graph to determine at least one best path cover corresponding to respective observed objects; resolve the solved bipartite graph to obtain a track representing the at least one best path cover corresponding to respective observed objects, where frames of the track that do not satisfy a predetermined threshold are discarded; and label the detections with a track identifier corresponding to the track representing the best path cover of the respective observed object.

[0121] According to some non-limiting embodiments or examples, provided is at least one non-transitory storage media storing instructions that, when executed by at least one processor, cause the at least one processor to: obtain detections corresponding to observed objects in sensor data; generating tracklets based on the detections where a tracklet includes nodes representing a respective observed object at a series of continuous timestamps and edges connecting the nodes at the series of continuous timestamps; generating a weighted track graph based on the tracklets, where weights are assigned to edges between nodes of the tracklets according to at least one first cost metric associated with the nodes; converting the weighted track graph to a bipartite graph, where a node of the bipartite graph corresponds to a respective detection of the tracklets, and edges of the bipartite graph are assigned weights according to at least one second cost metric associated with the nodes; solving the bipartite graph to determine at least one best path cover corresponding to respective observed objects; resolving the solved bipartite graph to obtain a track representing the at least one best path cover corresponding to respective observed objects, where frames of the track that do not satisfy a predetermined threshold are discarded; and labeling the detections with a track identifier corresponding to the track representing the best path cover of the respective observed object.

[0122] Further non-limiting aspects or embodiments are set forth in the following numbered embodiments:

[0123] Embodiment 1 : A method including: obtaining detections corresponding to observed objects in sensor data; generating tracklets based on the detections where a tracklet includes nodes representing a respective observed object at a series of continuous timestamps and edges connecting the nodes at the series of continuous timestamps; generating a weighted track graph based on the tracklets, where weights are assigned to edges between nodes of the tracklets according to at least one first cost metric associated with the nodes; converting the weighted track graph to a bipartite graph, where a node of the bipartite graph corresponds to a respective detection of the tracklets, and edges of the bipartite graph are assigned weights according to at least one second cost metric associated with the nodes; solving the bipartite graph to determine at least one best path cover corresponding to respective observed objects; resolving the solvedbipartite graph to obtain a track representing the at least one best path cover corresponding to respective observed objects, where frames of the track that do not satisfy a predetermined threshold are discarded; and labeling the detections with a track identifier corresponding to the track representing the best path cover of the respective observed object.

[0124] Embodiment 2: The method of any preceding claim, where obtaining detections includes: obtaining detections including three dimensional bounding boxes corresponding to a location, size, and velocity of the object.

[0125] Embodiment 3: The method of any preceding claim, where generating tracklets includes generating tracklets using a three dimensional Kalman filter.

[0126] Embodiment 4: The method of any preceding claim, including assigning nodes of the tracklets a respective timestamp, a respective tracklet identification, and a respective state distribution.

[0127] Embodiment 5: The method of any preceding claim, where the at least one second cost metric includes a kinematic likelihood, a type likelihood, a likelihood based on an intersection over union associated with the tracklets, or any combinations thereof.

[0128] Embodiment 6: The method of any preceding claim, where solving the bipartite graph includes minimum bipartite matching to determine the at least one best path cover corresponding to respective observed objects.

[0129] Embodiment 7: The method of any preceding claim, where resolving the bipartite graph includes removing nodes from the bipartite graph corresponding to the discarded frames, where an identification and state history is passed to a descendant node when a node is removed from the bipartite graph, and incoming edges of the removed node are deleted from the bipartite graph.

[0130] Embodiment 8: A system, including: at least one processor, and at least one non- transitory storage media storing instructions that, when executed by the at least one processor, cause the at least one processor to: obtain detections corresponding to observed objects in sensor data; generate tracklets based on the detections, where a tracklet includes nodes representing a respective observed object at a series of continuous timestamps and edges connecting the nodes at the series of continuous timestamps; generate a weighted track graph based on the tracklets, where weights areassigned to edges between nodes of the tracklets according to at least one first cost metric associated with the nodes; convert the weighted track graph to a bipartite graph, where a node of the bipartite graph corresponds to a respective detection of the tracklets, and edges of the bipartite graph are assigned weights according to at least one second cost metric associated with the nodes; solve the bipartite graph to determine at least one best path cover corresponding to respective observed objects; resolve the solved bipartite graph to obtain a track representing the at least one best path cover corresponding to respective observed objects, where frames of the track that do not satisfy a predetermined threshold are discarded; and label the detections with a track identifier corresponding to the track representing the best path cover of the respective observed object.

[0131] Embodiment 9: The system of any preceding claim, where the detections are three dimensional bounding boxes corresponding to a location, size, and velocity of the object.

[0132] Embodiment 10: The system of any preceding claim, where the tracklets are generated by a three dimensional Kalman filter.

[0133] Embodiment 11 : The system of any preceding claim, where nodes of the tracklets are assigned a respective timestamp, a respective tracklet identification, and a respective state distribution.

[0134] Embodiment 12: The system of any preceding claim, where the at least one second cost metric includes a kinematic likelihood, type likelihood, and likelihood based on an intersection over union associated with the tracklets, or any combinations thereof.

[0135] Embodiment 13: The system of any preceding claim, where solving the bipartite graph includes minimum bipartite matching to determine the at least one best path cover corresponding to respective observed objects.

[0136] Embodiment 14: The system of any preceding claim, where resolving the bipartite graph includes removing nodes from the bipartite graph corresponding to the discarded frames, where an identification and state history is passed to a descendant node when a node is removed from the bipartite graph, and incoming edges of the removed node are deleted from the bipartite graph.

[0137] Embodiment 15: At least one non-transitory storage media storing instructions that, when executed by at least one processor, cause the at least one processor to: obtain detections corresponding to observed objects in sensor data; generating tracklets based on the detections where a tracklet includes nodes representing a respective observed object at a series of continuous timestamps and edges connecting the nodes at the series of continuous timestamps; generating a weighted track graph based on the tracklets, where weights are assigned to edges between nodes of the tracklets according to at least one first cost metric associated with the nodes; converting the weighted track graph to a bipartite graph, where a node of the bipartite graph corresponds to a respective detection of the tracklets, and edges of the bipartite graph are assigned weights according to at least one second cost metric associated with the nodes; solving the bipartite graph to determine at least one best path cover corresponding to respective observed objects; resolving the solved bipartite graph to obtain a track representing the at least one best path cover corresponding to respective observed objects, where frames of the track that do not satisfy a predetermined threshold are discarded; and labeling the detections with a track identifier corresponding to the track representing the best path cover of the respective observed object.

[0138] Embodiment 16: The at least one non-transitory storage media of any preceding claim, where obtaining detections includes: obtaining detections including three dimensional bounding boxes corresponding to a location, size, and velocity of the object.

[0139] Embodiment 17: The at least one non-transitory storage media of any preceding claim, where generating tracklets includes generating tracklets using a three dimensional Kalman filter.

[0140] Embodiment 18: The at least one non-transitory storage media of any preceding claim, including assigning nodes of the tracklets a respective timestamp, a respective tracklet identification, and a respective state distribution.

[0141] Embodiment 19: The at least one non-transitory storage media of any preceding claim, where the at least one second cost metric includes a kinematic likelihood, type likelihood, and likelihood based on an intersection over union associated with the tracklets, or any combinations thereof.

[0142] Embodiment 20: The at least one non-transitory storage media of any preceding claim, where solving the bipartite graph includes minimum bipartite matching to determine the at least one best path cover corresponding to respective observed objects.

[0143] In the foregoing description, aspects and embodiments of the present disclosure have been described with reference to numerous specific details that can vary from implementation to implementation. Accordingly, the description and drawings are to be regarded in an illustrative rather than a restrictive sense. The sole and exclusive indicator of the scope of the invention, and what is intended by the applicants to be the scope of the invention, is the literal and equivalent scope of the set of claims that issue from this application, in the specific form in which such claims issue, including any subsequent correction. Any definitions expressly set forth herein for terms contained in such claims shall govern the meaning of such terms as used in the claims. In addition, when we use the term “further comprising,” in the foregoing description or following claims, what follows this phrase can be an additional step or entity, or a sub-step / sub-entity of a previously- recited step or entity.

Claims

WHAT IS CLAIMED IS:1 . A method comprising: obtaining detections corresponding to observed objects in sensor data; generating tracklets based on the detections wherein a tracklet comprises nodes representing a respective observed object at a series of continuous timestamps and edges connecting the nodes at the series of continuous timestamps; generating a weighted track graph based on the tracklets, wherein weights are assigned to edges between nodes of the tracklets according to at least one first cost metric associated with the nodes; converting the weighted track graph to a bipartite graph, wherein a node of the bipartite graph corresponds to a respective detection of the tracklets, and edges of the bipartite graph are assigned weights according to at least one second cost metric associated with the nodes; solving the bipartite graph to determine at least one best path cover corresponding to respective observed objects; resolving the solved bipartite graph to obtain a track representing the at least one best path cover corresponding to respective observed objects, wherein frames of the track that do not satisfy a predetermined threshold are discarded; and labeling the detections with a track identifier corresponding to the track representing the best path cover of the respective observed object.

2. The method of claim 1 , wherein obtaining detections comprises: obtaining detections comprising three dimensional bounding boxes corresponding to a location, size, and velocity of the object.

3. The method of claim 1 , wherein generating tracklets comprises generating tracklets using a three dimensional Kalman filter.

4. The method of claim 1 , comprising assigning nodes of the tracklets a respective timestamp, a respective tracklet identification, and a respective state distribution.

5. The method of claim 1 , wherein the at least one second cost metric comprises a kinematic likelihood, a type likelihood, a likelihood based on an intersection over union associated with the tracklets, or any combinations thereof.

6. The method of claim 1 , wherein solving the bipartite graph comprises minimum bipartite matching to determine the at least one best path cover corresponding to respective observed objects.

7. The method of claim 1 , wherein resolving the bipartite graph comprises removing nodes from the bipartite graph corresponding to the discarded frames, wherein an identification and state history is passed to a descendant node when a node is removed from the bipartite graph, and incoming edges of the removed node are deleted from the bipartite graph.

8. A system, comprising: at least one processor, and at least one non-transitory storage media storing instructions that, when executed by the at least one processor, cause the at least one processor to: obtain detections corresponding to observed objects in sensor data; generate tracklets based on the detections, wherein a tracklet comprises nodes representing a respective observed object at a series of continuous timestamps and edges connecting the nodes at the series of continuous timestamps; generate a weighted track graph based on the tracklets, wherein weights are assigned to edges between nodes of the tracklets according to at least one first cost metric associated with the nodes; convert the weighted track graph to a bipartite graph, wherein a node of the bipartite graph corresponds to a respective detection of the tracklets, and edges of the bipartite graph are assigned weights according to at least one second cost metric associated with the nodes;solve the bipartite graph to determine at least one best path cover corresponding to respective observed objects; resolve the solved bipartite graph to obtain a track representing the at least one best path cover corresponding to respective observed objects, wherein frames of the track that do not satisfy a predetermined threshold are discarded; and label the detections with a track identifier corresponding to the track representing the best path cover of the respective observed object.

9. The system of claim 8, wherein the detections are three dimensional bounding boxes corresponding to a location, size, and velocity of the object.

10. The system of claim 8, wherein the tracklets are generated by a three dimensional Kalman filter.11 . The system of claim 8, wherein nodes of the tracklets are assigned a respective timestamp, a respective tracklet identification, and a respective state distribution.

12. The system of claim 8, wherein the at least one second cost metric comprises a kinematic likelihood, type likelihood, and likelihood based on an intersection over union associated with the tracklets, or any combinations thereof.

13. The system of claim 8, wherein solving the bipartite graph comprises minimum bipartite matching to determine the at least one best path cover corresponding to respective observed objects.

14. The system of claim 8, wherein resolving the bipartite graph comprises removing nodes from the bipartite graph corresponding to the discarded frames, wherein an identification and state history is passed to a descendant node when a node is removed from the bipartite graph, and incoming edges of the removed node are deleted from the bipartite graph.

15. At least one non-transitory storage media storing instructions that, when executed by at least one processor, cause the at least one processor to: obtain detections corresponding to observed objects in sensor data; generating tracklets based on the detections wherein a tracklet comprises nodes representing a respective observed object at a series of continuous timestamps and edges connecting the nodes at the series of continuous timestamps; generating a weighted track graph based on the tracklets, wherein weights are assigned to edges between nodes of the tracklets according to at least one first cost metric associated with the nodes; converting the weighted track graph to a bipartite graph, wherein a node of the bipartite graph corresponds to a respective detection of the tracklets, and edges of the bipartite graph are assigned weights according to at least one second cost metric associated with the nodes; solving the bipartite graph to determine at least one best path cover corresponding to respective observed objects; resolving the solved bipartite graph to obtain a track representing the at least one best path cover corresponding to respective observed objects, wherein frames of the track that do not satisfy a predetermined threshold are discarded; and labeling the detections with a track identifier corresponding to the track representing the best path cover of the respective observed object.

16. The at least one non-transitory storage media of claim 15, wherein obtaining detections comprises: obtaining detections comprising three dimensional bounding boxes corresponding to a location, size, and velocity of the object.

17. The at least one non-transitory storage media of claim 15, wherein generating tracklets comprises generating tracklets using a three dimensional Kalman filter.

18. The at least one non-transitory storage media of claim 15, comprising assigning nodes of the tracklets a respective timestamp, a respective tracklet identification, and a respective state distribution.

19. The at least one non-transitory storage media of claim 15, wherein the at least one second cost metric comprises a kinematic likelihood, type likelihood, and likelihood based on an intersection over union associated with the tracklets, or any combinations thereof.

20. The at least one non-transitory storage media of claim 15, wherein solving the bipartite graph comprises minimum bipartite matching to determine the at least one best path cover corresponding to respective observed objects.

Images