Tracking different parts of an articulated vehicle

Abstract

This document describes techniques and systems related to tracking different parts of an articulated vehicle. The vehicle uses a radar system that can distinguish between non-articulated vehicles and articulated vehicles, which by definition have multiple parts that can pivot in different directions in order to turn or follow a curve. The radar system obtains detections that indicate another vehicle driving nearby. When the detections indicate that the other vehicle is articulated, the radar system tracks each identifiable part, rather than tracking all of the parts together. A bounding box is generated for each identifiable part; the radar system monitors the speed of each bounding box separately and simultaneously. The multiple bounding boxes that are drawn enable the radar system to accurately track each connected part of the articulated vehicle, including detecting whether any movement occurs between two connected parts in order to accurately locate the vehicle as it is being driven.

Description

BACKGROUND

[0001] Radar systems can enable vehicles to detect both stationary and non-stationary objects at different ranges. Radar provides improved performance over other sensors used by vehicles for autonomous driving or other driving safety or driving assistance functions; radar-equipped vehicles can detect and track objects in a variety of driving conditions, including weak light, rain, snow, and fog. Nonetheless, in certain conditions, some radar systems can report false detections or fail to detect objects altogether. Inaccurate radar detections can be reported in situations where the radar system has difficulty distinguishing between multiple objects (e.g., two vehicles driving in close proximity) or articulated objects (e.g., a truck towing a trailer or a multi-section bus). As a result, operator intervention and / or compromised driving safety can result. SUMMARY

[0002] This document describes techniques and systems related to tracking different parts of articulated vehicles. In some examples, a radar system for installation on a first vehicle includes at least one processor. The at least one processor is configured to track a second vehicle driving in a field of view of the radar system. The at least one processor is configured to track the second vehicle by generating a first bounding box associated with a first part of the second vehicle and a second bounding box associated with a second part of the second vehicle, and determining whether the second vehicle is an articulated vehicle based on a first velocity vector associated with the first bounding box and a second velocity vector associated with the second bounding box. The at least one processor is further configured to, in response to determining that the second vehicle is an articulated vehicle, perform a driving maneuver by separately and simultaneously tracking the first bounding box and the second bounding box in the field of view.

[0003] This document also describes methods performed by the techniques and components summarized above as well as other configurations of radar systems set forth herein, and apparatuses for performing those methods.

[0004] This summary introduces simplified concepts related to tracking different parts of articulated vehicles and is further described by the detailed description and the accompanying drawings. This summary is not intended to identify essential features of the claimed subject matter nor is it intended for determining the scope of the claimed subject matter. BRIEF DESCRIPTION OF DRAWINGS

[0005] The details of one or more aspects of tracking different parts of articulated vehicles are described in this document with reference to the following figures. Like numbers refer to like features and components throughout the figures:

[0006] Figure 1-1An example environment of a vehicle having a radar system configured to track different parts of an articulated vehicle is shown in accordance with the present disclosure;

[0007] Figure 1-2 Another example environment of a vehicle having a radar system configured to track different parts of an articulated vehicle is shown in accordance with the present disclosure;

[0008] Figure 2 An example vehicle including a radar system configured to track different parts of an articulated vehicle is shown in accordance with the present disclosure;

[0009] Figure 3 An example environment showing some drawbacks of using a conventional radar system that is not configured to track different parts of an articulated vehicle is shown;

[0010] Figures 4-1 to 4-2 An example environment showing further details of using a radar system configured to track different parts of an articulated vehicle is shown in accordance with the present disclosure;

[0011] Figure 5 A process for tracking different parts of an articulated vehicle is shown in accordance with the present disclosure; and

[0012] Figure 6-1 And Figure 6-2 Aspects of accuracy improvement functionality of a radar system configured to track different parts of an articulated vehicle are shown in accordance with the present disclosure. DETAILED DESCRIPTION

[0013] SUMMARY

[0014] In some vehicles, including cars equipped with autonomous driving and advanced safety features, perception systems (e.g., radar systems, lidar systems, camera systems, other distance sensors) are used to accurately locate the position of the vehicle relative to nearby other objects. The perception system reports the relative position and size of nearby vehicles; the output of the perception system can be used as input to another system (e.g., an autonomous driving system, an advanced safety system), improving situational awareness and driving safety, including the safety of passengers of other vehicles.

[0015] Rectangular bounding boxes are often used to convey the relative position and size of another vehicle relative to the vehicle. The size of the bounding box approximates the observed group of detections relative to the position of the vehicle or other detections and / or bounding boxes in the field of view. Using rectangular bounding boxes can still present some drawbacks.

[0016] When multiple vehicles are positioned close to one another (e.g., when the vehicles are tailgating by driving unsafely in the same lane with little spacing between them), a perception system can mistake a large group of detections as a single vehicle (e.g., the single vehicle is longer than one car). Rather than drawing multiple bounding boxes to depict each of the different vehicles, the perception system can inaccurately generate a bounding box that contains only one entire group of detections. Furthermore, inaccurate long vehicle tracking is common. Long vehicles are often articulated, which by definition means that the vehicle includes a combination of two or more rigid parts that are configured to pivot around a common hinge. Tractor-trailers, accordion-style buses, and truck-attached campers are some examples of articulated vehicles. In either case (where the perception system misprocessed the group of detections), the rectangular bounding box can not always accurately represent the relative position and size of one or more objects in the environment.

[0017] A perception system can have difficulty accurately maintaining a bounding box around a group of detections related to a group of vehicles or a group of vehicle parts, especially when each of the members of the group is allowed to move independently, even if only slightly changing its own direction or velocity. These differences in velocity can cause the bounding box to stretch as the group of detections grows or shrink as the group of detections decreases (e.g., when an articulated angle measured at a connecting hinge between articulated parts increases beyond zero degrees). The perception system can overload the vehicle’s onboard computer hardware trying to resolve the group of detections to keep the bounding box at its original size (e.g., as the articulated angle increases and decreases with turns and road curvature).

[0018] Furthermore, inaccurately drawn bounding boxes can be particularly troublesome for autonomous driving or advanced cruise control systems that rely on the perception system to make driving decisions. Due to a poorly drawn bounding box that was drawn in error, a vehicle can incorrectly think that it is safe or unsafe to drive in an adjacent lane, especially when driving around a curve. In the real world, a tractor-trailer can be safely parked in its lane of travel, but to a perception system or other system that relies on its output, the tractor-trailer can appear to be an out-of-control vehicle or a vehicle that requires an unsafe spacing distance, which can result in a manual override that requires the operator to take back control.

[0019] This document describes techniques and systems related to tracking different parts of an articulated vehicle. In particular, the techniques and systems provide ways to estimate the hinge point and articulation angle of an articulated vehicle given perception data obtained from a perception system. For ease of description, the described techniques are primarily focused in the context of radar-based tracking, including radar tracking for automotive applications. However, the techniques can be applicable to other types of tracking, including other types of tracking in automotive applications, as well as radar and other types of tracking in other non-automotive contexts. Moreover, for ease of description, the articulated vehicle has two separate parts unless otherwise noted, although the techniques are generally applicable to tracking all parts of an articulated vehicle separately and simultaneously, including those articulated vehicles with more than two parts.

[0020] As one example, a vehicle uses a radar system that can distinguish between non-articulated vehicles and articulated vehicles, which by definition have multiple parts that can pivot in different directions to help turn or follow a curve on the road. The radar system obtains radar detections that indicate another vehicle driving nearby. When the radar detections indicate that the other vehicle can be an articulated vehicle, the radar system tracks each identifiable part separately, rather than tracking all parts together. A bounding box is generated for each identifiable part; the radar system monitors the speed of each bounding box separately and simultaneously. The multiple bounding boxes drawn enable the radar system to accurately track each connected part of the articulated vehicle, including detecting whether any movement occurs between two connected parts, in order to accurately position the vehicle when the two vehicles share the road. By configuring the perception system to communicate articulated vehicles as two or more distinct bounding boxes in its output, the techniques and systems improve driving safety and situational awareness.

[0021] Applications of the described techniques can benefit vehicle computer systems, including executing a driving software stack that can include an object fusion module configured to perform matching and grouping of outputs of multiple perception systems. Moreover, threat assessment and trajectory planning modules that control the trajectory of the vehicle can benefit from receiving more accurate definitions of the edges of a target. With a more accurate representation of an articulated vehicle than a single rectangular box, accuracy of downstream autonomous driving and advanced safety features that rely on the representation (e.g., radar output) can be improved.

[0022] Example Environment

[0023] Figure 1-1An example environment 100 of a vehicle having a radar system configured to track different parts of an articulated vehicle is shown in accordance with the present disclosure. In accordance with the techniques, devices, and systems of the present disclosure, the environment 100 includes a vehicle 102 equipped with a radar system 104 configured to track different parts of an articulated vehicle. Output from the radar system 104 can enable operation of the vehicle 102. The range, approach angle, or velocity of an object can be determined by the radar system 104 or derived from output in the form of radar data.

[0024] Although shown as a car, the vehicle 102 can represent other types of motorized vehicles (e.g., a motorcycle, a bus, a tractor, a tractor-trailer vehicle, or construction equipment), non-motorized vehicles (e.g., a bicycle), rail vehicles (e.g., a train or tram), watercraft (e.g., a boat or ship), aircraft (e.g., an airplane or helicopter), or a spacecraft (e.g., a satellite). In general, the vehicle 102 represents any moving platform, including a moving mechanical or robotic equipment, that can benefit from having a radar representation of the environment 100.

[0025] Detected in the field of view of the radar system 104 are multiple moving objects 106, 108, and 110 (sometimes also referred to as “targets of interest”). The moving objects 108 and 110 are referred to as non-articulated vehicles 108 and 110. In contrast, the object 106 is referred to as an articulated vehicle 106, which includes at least two distinguishable parts connected by a hinge. In general, the objects 106, 108, and 110 are composed of one or more materials that reflect radar signals, or in other examples, are composed of a suitable reflective medium for enabling detection by some other type of perception sensing system. Depending on the application, the objects 106, 108, and 110 can represent detection of individual targets, one or more clutters of radar detections, one or more clusters of radar detections, and / or one or more clouds of radar detections. Throughout the present disclosure, detections, clutters, clusters, and / or clouds of radar detections are represented by small circles (dots), where each small dot represents one or more examples of radar detections.

[0026] The radar system 104 is configured to be installed as part of the vehicle 102. In the depicted environment 100, the radar system 104 is installed near or integrated within a front portion of the vehicle 102 to detect objects and avoid collisions. The radar system 104 can be a mechanically replaceable component, part, or system of the vehicle 102 that can need to be replaced or repaired due to failure over the lifetime of the vehicle 102. The radar system 104 can include an interface to at least one automotive system. The radar system 104 can output, via the interface, signals based on electromagnetic energy received by the radar system 104. The output signals from the radar system 104 represent radar data and can take a variety of forms.

[0027] At least one automotive system of the vehicle 102 relies on radar data output from the radar system 104. Examples of such automotive systems include a driver assist system, an autonomous driving system, or a semi-autonomous driving system. Another example of a system that can rely on radar data provided by the radar system 104 can include a fusion tracker that combines sensor data from various perception sensors, including the radar system 104, to generate a multi-sensor representation of the environment 100. A benefit of operating the fusion tracker using multiple bounding box representations of the articulated vehicle instead of a single bounding box is that the fusion tracker can operate more efficiently and with higher accuracy. The fusion tracker can quickly combine radar data with other high resolution sensor data that is registered with the radar data. With the radar tracking individual portions of the articulated vehicle, the fusion tracker can convey relative changes in motion between the individual portions of the articulated vehicle in its fusion output, providing a more accurate sensor fusion output than in the case of fusion tracking with a conventional radar system that is not configured to track articulated vehicles according to the described techniques.

[0028] The automotive systems of the vehicle 102 can use radar data provided by the radar system 104 to perform functions, also referred to as vehicle operations. In the environment 100, the radar system 104 can detect and track multiple moving objects 106, 108, and 110 by transmitting and receiving one or more radar signals through an antenna system. For example, a driver assistance system can provide blind spot monitoring and generate an alert indicating a potential collision with the object 106 detected by the radar system 104. To do so, the radar system 104 can transmit electromagnetic signals between 100 and 400 gigahertz (GHz), between 4 and 100 GHz, or between approximately 70 and 80 GHz. The radar system 104 includes a transmitter (not shown) and at least one antenna element to transmit the electromagnetic signals. The radar system 104 includes a receiver (not shown) and at least one antenna element (which can be the same or different from the transmitting element) to receive reflected versions of the electromagnetic signals. The transmitter and receiver can be incorporated together on the same integrated circuit (e.g., a transceiver integrated circuit or package), or separately on different integrated circuits or chips.

[0029] The radar system 104 can track the objects 106, 108, and 110 as they appear to travel in the field of view. For example, as the radar system 104 detects more and more radar detections of a higher count, the radar system 104 can detect and track one or more clusters of radar detections, such as in the rear, middle, and / or front of the articulated vehicle 106. Without determining whether the articulated vehicle 106 is articulated, the radar system 104 can at least determine whether the various clusters of radar detections leaving the object 106 represent the same vehicle or object.

[0030] The radar system 104 can create a bounding box 112 for the entire object 106. To further delineate the bounding box 112, the radar system 104 can determine whether these various clusters of radar detections are stationary or non-stationary; whether the clusters of radar detections associated with the object 106 are moving at approximately the same or different speeds (rates and directions); and whether the distance (spacing) between each cluster of the object 106 is changing or nearly constant. Similarly, the radar system 104 can create a bounding box 114 for the object 108 and a bounding box 116 for the object 110. As a follow-up to the example of the driver-assistance system receiving radar data from the radar system 104, the radar data can indicate the dimensions of the bounding boxes 112, 114, and 116 to the driver-assistance system, which the driver-assistance system can use to determine the locations of the objects 106, 108, and 110, e.g., for determining when it is safe or unsafe to change lanes. Based on the radar data output from the radar system 104, an autonomous driving system can move the vehicle 102 to a particular location on the road while avoiding collisions with the objects 106, 108, and 110.

[0031] Hinge determination

[0032] Figure 1-2 Another example environment 100-1 is shown in which a vehicle according to the present disclosure has a radar system configured to track different portions of a hinged vehicle. The environment 100-1 is an example of the environment 100 in which the vehicle 102 is passing or driving near the hinged vehicle 106 while both vehicles 102 and 106 are driving around a turn.

[0033] The radar data output from the radar system 104 can enable an autonomous driving system of the vehicle 102 to determine whether to perform an emergency brake, whether to perform a lane change, whether to adjust a speed, or whether to take any driving action when approaching the hinged vehicle 106. The autonomous driving system can base these determinations on the size, position, and motion of the various portions of the hinged vehicle 106 relative to a safety margin that the vehicle 102 uses to drive through traffic.

[0034] Other radar systems can track the hinged vehicle 106 with only a single bounding box 112 (as shown by both Figure 1-1 and Figure 1-2 When driving in a straight line, a single bounding box of a hinged vehicle can represent the size and position of the vehicle with slight accuracy. However, when the road curves and lets the hinged vehicle turn with it, the single bounding box approximates with multiple errors compared to the true size and position of the vehicle. As shown by the example environment 100-2, the single bounding box 112 can not represent the size and position of the hinged vehicle 106 with sufficient accuracy for the autonomous driving system of the vehicle 102 to safely drive through traffic. Figure 1-2As shown, the bounding box 112 inaccurately reports the position of the articulated vehicle 106 such that the articulated vehicle 106 appears to be crossing the travel lane of the vehicle 102. If the vehicle 102 is equipped with a radar system that is not configured according to the techniques of the present disclosure, it can be erroneously determined that the articulated vehicle 106 is turning into or out of the travel lane. Thus, radar data output from these other radar systems can sometimes be inaccurate, especially near turns. Thus, always using a single bounding box to represent and track an articulated vehicle can impact safe driving, especially when driving on curvy or windy roads.

[0035] Unlike these other types of radar systems, the radar system 104 is configured to determine whether a tracked object is articulated. Rather than outputting radar data with a single bounding box that roughly approximates the articulated as a static, non-articulated shape, the radar system 104 can track individual portions of an articulated vehicle and output radar data with bounding boxes sized and positioned to correspond to the individual portions being tracked. For example, the radar system 104 can generate a bounding box 112-1 at a front portion of the articulated vehicle 106 and further generate a bounding box 112-2 near a rear portion of the articulated vehicle 106. As will become clear below, the radar system 104 can track an estimated hinge point between the front and rear portions of the articulated vehicle 106 and maintain a similarity of the actual motion of the articulated vehicle 106 as the bounding boxes 112-1 and 112-2 are repositioned and rotated around the estimated hinge point.

[0036] By using at least two bounding boxes 112-1 and 112-2 to separately track multiple portions of the articulated vehicle 106 rather than generating a single bounding box 112, the radar system 104 enables the vehicle 102 to safely drive past or near the articulated vehicle 106 with smooth driving maneuvers that are free of hesitation or jerking. The radar system 104 does not distort the detection in the rough manner of using only the bounding box 112. As such, when the radar system 104 detects the articulated vehicle 106, the radar data output to the driver assist system provides a highly accurate size and position representation of the different articulated portions of the articulated vehicle 106 as they appear in real life. This enables the vehicle 102 to drive in an autonomous or semi-autonomous mode in a smooth and predictable manner that is similar to the driving style of a confident driver operating the vehicle 102 under manual control in a similar scenario.

[0037] To implement tracking of articulated portions, the radar system can initially use the bounding box 112 to represent the object 106. Since most articulated vehicles are longer than standard passenger vehicles, the radar system 104 can apply a vehicle length-based criterion to the bounding box 112 before expending computational resources to determine whether the represented object 106 is articulated. This initial filtering of smaller vehicles prevents computational resources from having to determine whether every object is an articulated vehicle, which improves the computational efficiency of the radar system 104.

[0038] The radar system 104 can utilize a threshold length (e.g., greater than 3.0 meters), which can be used as an indicator to classify the object 106 as likely articulated or likely not articulated when compared to the dimensions of the bounding box 112. In other words, whether an object is articulated can depend on whether the object is greater than a length threshold (e.g., greater than a regular passenger car length). The radar system 104 can compare the length of the bounding box 112 to the length threshold to determine whether the object 106 has the potential to be articulated, even if all portions of the radar data appear to be fixed in a non-articulated manner (e.g., when driving on a straight road), as shown in FIG. 1. This length threshold can be applied by the radar system 104 prior to determining whether the object is actually articulated or as a condition to determining whether the object is actually articulated. In this way, if the estimated bounding box length is greater than the threshold length, the object is classified as a likely articulated vehicle that is suitable for further processing to determine whether articulation exists. With a bounding box that does not exceed the threshold length, the radar system 104 can classify the object as a short vehicle that is not a likely articulated vehicle, and thus a single bounding box can be used to track the short vehicle. Figure 1-1

[0039] ​Whether a vehicle is articulated or non-articulated is generally not of concern for shorter vehicles, which can easily fit within the safety margins of a travel lane. This initial length criterion, which can be applied by the radar system 104, derives some of its benefit from the relationship that exists between vehicle length and turning radius; long vehicles tend to have wide turning radii. Such wide turning radii present challenges for driving, particularly when other vehicles are traveling in adjacent lanes, or when driving in narrow streets where parked cars and other static moving objects share the road, which can require reliance on some form of articulation. Because articulation joints allow for pivoting of the two sections, articulated vehicles, such as the articulated vehicle 106, can make tighter turns (without encroaching on adjacent lanes or shoulders) than non-articulated vehicles with comparable length, such as the non-articulated vehicle 108. For vehicles of standard length, such as the object 110, use of a joint and articulation configuration is not necessary, but is more frequently applied to vehicles longer than the standard length. Thus, the radar system 104, in response to determining that the object 110 is not of sufficient length, can refrain from determining whether the object 110 is articulated or non-articulated, and instead track the object 110 as a non-articulated vehicle by default. That is, the final determination as to whether a vehicle is articulated cannot be determined based on length alone.

[0040] Another possible indicator of an articulated vehicle is the behavior of the vehicle during a turn maneuver or when driving around a curve. The radar system 104 can initially treat both the articulated vehicle 106 and the non-articulated vehicle 108 as possible articulated vehicles until the radar system 104 can capture sufficient information about the size and location of any intermediate section, which typically occurs during a turn. That is, ultimately, when a road curves or a possible articulated vehicle turns, each individual section that makes up the articulated vehicle 106 can be observed as appearing to move at different speeds as the vehicle drives further into the curve. Articulation enables the articulated vehicle 106 to follow a curve more closely than a non-articulated vehicle, such as the non-articulated vehicle 108. In practice, this means that the front section of the articulated vehicle 106 will have a radar detection group with a speed 120-1 that is different from the speed 120-2 of a radar detection group captured at the rear section of the articulated vehicle 106. A detection group at the front of the non-articulated vehicle 108 will appear in the radar data with a speed that is somewhat uniform with a detection group at the rear of the non-articulated vehicle 108.

[0041] Example Device

[0042] Figure 2An example vehicle 102-1 including a radar system 104-1 configured to track different portions of an articulated vehicle is shown in accordance with the present disclosure. Radar system 104 is an example of radar system 104-1. Vehicle 102-1 is an example of vehicle 102.

[0043] Radar system 104-1 can be part of an object detection and tracking system 202. In addition to radar system 104-1, object detection and tracking system 202 can include a lidar system 204, an imaging system 206, and / or other systems that can be used to detect and track objects. However, radar system 104-1 can operate as a standalone system without needing to communicate with or use data from lidar system 204 and / or imaging system 206. Additionally, object detection and tracking system 202 can perform the techniques and methods described herein by using radar data from radar system 104-1 alone.

[0044] Vehicle 102-1 also includes vehicle-based systems 210, such as a driver assistance system 212 and / or an autonomous driving system 214. Vehicle-based systems 210 use radar data from radar system 104-1 to perform functions. For example, driver assistance system 212 tracks articulated vehicles (e.g., object 106) and / or non-articulated vehicles (e.g., objects 108 and 110), monitors their proximity to vehicle 102-1, and generates alerts indicating potential collisions or unsafe distances to vehicles 102-1 driving alongside. In this case, radar data from radar system 104-1 (e.g., targets of interest, radar detections of clutter(s), radar detections of cluster(s), and / or radar detections of cloud(s)) indicate whether vehicle 102-1 can safely drive alongside other vehicles in the field of view.

[0045] As another example, on a windy road, driver assistance system 212 suppresses false alerts in response to radar data indicating that an articulated vehicle (e.g., object 106) driving in an adjacent lane is turning into the lane of travel of vehicle 102-1. In this way, driver assistance system 212 can avoid falsely warning the driver of vehicle 102-1 that an articulated vehicle is driving unsafely close to or colliding with vehicle 102-1. By suppressing these false alerts, driver assistance system 212 avoids confusing or unnecessarily worrying the driver of vehicle 102-1.

[0046] The autonomous driving system 214 can move the vehicle 102-1 to a particular location while avoiding collisions or unsafe proximity to vehicles driving alongside the vehicle 102-1. Radar data provided by the radar system 104-1 can provide information about the location and motion of other objects to enable the autonomous driving system 214 to perform emergency braking, perform a lane change, or adjust the speed of the vehicle 102-1. Additionally, the autonomous driving system 214 of the vehicle 102-1 can determine whether a vehicle driving alongside is a hinged vehicle. When driving alongside a hinged vehicle, the autonomous driving system 214 of the vehicle 102-1 performs driving maneuvers by separately, and simultaneously, tracking different portions of the hinged vehicle, as described further below.

[0047] The radar system 104-1 includes a communication interface 220 to transmit radar data to the vehicle-based system 210 or another component of the vehicle 102-1 over a communication bus of the vehicle 102-1. Generally, the radar data provided by the communication interface 220 is in a format available to the object detection and tracking system 202. In some implementations, the communication interface 220 can provide information to the radar system 104-1, such as the speed of the vehicle 102-1 or whether a turn signal is on or off. The radar system 104-1 can use this information to properly configure itself. For example, the radar system 104-1 can determine whether a selected object (e.g., 106) is stationary by comparing the Doppler of the selected object to the speed of the vehicle 102-1. Alternatively, the radar system 104-1 can dynamically adjust the field of view or intra-lane azimuth based on whether a right or left turn signal is on.

[0048] The radar system 104-1 also includes at least one antenna array 222 and at least one transceiver 224 to transmit and receive radar signals. The antenna array 222 includes at least one transmit antenna element and a plurality of receive antenna elements separated in azimuth and elevation directions. In some cases, the antenna array 222 also includes a plurality of transmit antenna elements to enable a multiple-input multiple-output (MIMO) radar capable of transmitting multiple different waveforms at a given time (e.g., each transmit antenna element transmits a different waveform). The antenna elements can be circularly polarized, horizontally polarized, vertically polarized, or a combination thereof.

[0049] Using the antenna array 222, the radar system 104 can form steered or non-steered and wide or narrow beams. Steering and shaping can be achieved through analog beamforming or digital beamforming. One or more transmit antenna elements can have, for example, a non-steered omnidirectional radiation pattern, or can produce a wide, steerable beam to illuminate a large spatial volume. To achieve target angular accuracy and angular resolution, receive antenna elements can be used to generate hundreds of narrow, steered beams with digital beamforming. In this way, the radar system 104-1 can effectively monitor the external environment and detect one or more portions of the articulated vehicle, such as the first portion of the articulated vehicle and the second portion of the articulated vehicle.

[0050] The transceiver 224 includes circuitry and logic to transmit and receive radar signals via the antenna array 222. Components of the transceiver 224 can include amplifiers, mixers, switches, analog-to-digital converters, or filters to condition the radar signals. The transceiver 224 also includes logic to perform in-phase / quadrature (I / Q) operations, such as modulation or demodulation. Various modulations can be used, including linear frequency modulation, triangular frequency modulation, stepped frequency modulation, or phase modulation. The transceiver 224 can be configured to support continuous wave or pulsed radar operation. The transceiver 224 can be used to generate a frequency spectrum (e.g., a range of frequencies) for the radar signals that can span between one gigahertz to four hundred gigahertz (GHz), for example, between 4 GHz to 100 GHz, or between approximately 70 GHz to 80 GHz. The bandwidth can be on the order of hundreds of megahertz, or on the order of gigahertz.

[0051] The radar system 104-1 also includes one or more processors 226. The processor 226 can be implemented using any type of processor, such as a central processing unit (CPU), a microprocessor, a multi-core processor, etc. Although the processor 226 is shown as part of the radar system 104-1, the processor 226 can be part of the object detection and tracking system 202 and can support the lidar system 204 and the imaging system 206 in addition to the radar system 104-1.

[0052] The object detection and tracking system 202, including the radar system 104-1, also includes one or more computer-readable media (CRM) 230 (e.g., computer-readable storage media), and the CRM 230 does not include propagating signals. The CRM 230 can include various data storage media such as volatile memory (e.g., dynamic random access memory, DRAM), non-volatile memory (e.g., flash memory), optical media, magnetic media, etc. The CRM 230 can include instructions (e.g., code, algorithms) that can be executed using the processor 226. The instructions stored in the CRM 230 (not shown) partially interpret, manipulate, and / or use sensor data 232 that can also be stored in the CRM 230. The sensor data 232 includes radar data of the radar system 104-1 (e.g., radar-detected clutters, radar-detected clusters, and / or radar-detected clouds). The sensor data 232 can also include lidar data of the lidar system 204 and imaging data (e.g., video, still frames) of the imaging system 206. In one aspect, the instructions stored in the CRM 230 include a vehicle tracker 234 and an articulated vehicle tracker 236.

[0053] The vehicle tracker 234 can share some similarities with existing vehicle trackers and can detect and track stationary objects and / or non-stationary objects. The vehicle tracker 234 can determine whether various clusters of radar detections are reflected from one object or from multiple objects. The vehicle tracker 234 can generate a bounding box for each detected vehicle near the vehicle 102-1. The vehicle tracker 234 can determine and track a location, a center of mass, and a velocity vector for each bounding box. However, unlike other existing vehicle trackers, the vehicle tracker 234 can make different classifications and handling of long vehicles near the vehicle 102-1. Specifically, once the vehicle tracker 234 determines that a vehicle reaches or exceeds a threshold length (e.g., greater than 3.0 meters), the vehicle tracker 234 triggers the articulated vehicle tracker 236 to make a determination as to whether the target is an articulated vehicle.

[0054] The articulated vehicle tracker 236 helps determine whether a vehicle is articulated or non-articulated. If the articulated vehicle tracker 236 determines that a vehicle is articulated, the articulated vehicle tracker 236 determines the location of the articulation point, where the articulation point couples or connects a first portion (first segment) and a second portion (second segment) of the articulated vehicle. Using the articulated vehicle tracker 236, the radar system 104-1 of the vehicle 102-1 can track the first portion and the second portion of the articulated vehicle separately and simultaneously. The articulated vehicle tracker 236 can generate a first bounding box associated with the first portion (e.g., the front end portion) of the possible articulated vehicle and a second bounding box associated with the second portion (e.g., the rear end portion) of the possible articulated vehicle. The articulated vehicle tracker 236 enables the radar system 104 to track the first bounding box and the second bounding box of the suspected articulated vehicle separately and simultaneously.

[0055] The radar system 104-1 can use the vehicle tracker 234 and the articulated vehicle tracker 236 simultaneously. The articulated vehicle tracker 236 can also use the lidar system 204 to estimate the closest edge of the articulated vehicle from the vehicle 102-1. Before describing the articulation determination in detail, next, Figure 3 Some other existing vehicle trackers are described that can use only a single bounding box instead of using multiple boxes to track articulated vehicles, as is done with the radar system 104-1.

[0056] Figure 3 An example environment 300 is shown that displays some of the shortcomings of using a conventional radar system that is not configured to track different portions of an articulated vehicle. The environment 300 includes a portion of a road that is turning to the right. A tractor-trailer vehicle 306, which is an articulated vehicle, drives alongside a vehicle 302 that is equipped with a legacy radar system that cannot tell whether a target is articulated or not. Unlike the vehicle 102-1, the vehicle 302 uses the radar system 304 without the help of the articulated vehicle tracker 236. The existing radar system 304 is unable to determine that the tractor-trailer vehicle 306 is an articulated vehicle. Instead, the existing radar system 304 can detect and track the tractor-trailer vehicle 306 as non-articulated. This can be a reasonable approach if the vehicle 302 and the tractor-trailer vehicle 306 are always driving on straight road lanes, but this approach fails when the vehicles are driving on a curved or winding portion of a road, as shown. Figure 3 The approach fails when the vehicles are driving on a curved or winding portion of a road, as shown.

[0057] In one respect, radar system 304 can detect one or more clusters of radar detections anywhere at the rear, front, and middle of the tractor-trailer vehicle 306. Radar system 304 can then determine the relationship between all radar detections and a single vehicle (in...) Figure 3 In this context, the radar system 304 is associated with the tractor-trailer vehicle 306. The radar system 304 can then create a single bounding box 312 for the entire tractor-trailer vehicle 306. (As shown...) Figure 3 As shown, without using the articulated vehicle tracker 236, the tracked velocity vector 320-1 is inconsistent with the velocity vector 320-2 of the tractor-trailer vehicle 306. More importantly, the bounding box 312 does not accurately represent the position of the tractor-trailer vehicle 306 relative to the vehicle 302 on the road. Instead, when both vehicles 320 and 306 are capable of turning right, the bounding box 312 appears to intrude into the vicinity of the vehicle 302 within an unsafe separation distance. Therefore, by using the radar system 304, the driving system of the vehicle 302 (e.g., an autonomous driving system) may overcorrect the movement of the vehicle 302 and cause the vehicle 302 to unsafely drive into another lane, drive off the road, accelerate, brake, or perform any other unnecessary driving maneuvers, which may reduce driving safety and passenger comfort.

[0058] Hinge point and hinge angle

[0059] Figures 4-1 to 4-2 Example environments 400-1 and 400-2 are shown, illustrating further details of a radar system configured for tracking different parts of an articulated vehicle according to the use of this disclosure. (See Figures 1 and 2.) Figure 2 Environments 400-1 and 400-2 are described in the background. Each of environments 400-1 and 400-2 includes vehicle 102-2, which is an example of vehicles 102 and 102-1. Driving in the lane adjacent to vehicle 102-2 is a semi-trailer 106-2, which is an example of objects 106-1 and 106.

[0060] First pay attention Figure 4-1 In response to radar system 104-2 identifying semi-trailer 106-2 as a long vehicle, vehicle tracker 234 invokes articulated vehicle tracker 236 for further processing of radar data generated by radar system 104-2 to determine whether the object tracked by vehicle tracker 234 is articulated.

[0061] Regardless of how the vehicle tracker 234 handles the semi-trailer 106-2, the articulated vehicle tracker 236 begins tracking the suspected articulated vehicle by locating a hinge point between the two portions of the suspected articulated vehicle. For example, the articulated vehicle tracker 236 can tell that the large detection group behind the semi-trailer is moving in unison with another detection group in front of the semi-trailer. The hinge point 422-1 can be determined at the intersection between the velocity vector 420-1 of the rear portion and the velocity vector 420-2 of the front portion. The articulated vehicle tracker 436 can generate a first bounding box 412-1 around the rear portion and a second bounding box 412-2 around the front portion. The articulated vehicle tracker 236 determines that the hinge point 422-1 is between the first bounding box 412-1 and the second bounding box 412-2 such that they do not overlap. In this case, the articulation angle 424-1 between the first bounding box 412-1 and the second bounding box 412-2 is approximately zero degrees. The determination of the hinge point 422-1 can be difficult to resolve when the articulation angle 424-1 is close to zero.

[0062] Switching to Figure 4-2 , the articulated vehicle tracker 236 can wait until the vehicle 102-2 and the semi-trailer 106-2 turn or drive on a curved road before establishing the hinge point 422-2. In this case, the articulation angle 424-2 between the first bounding box 412-1 and the second bounding box 412-2 is greater than zero degrees. The articulated vehicle tracker 236 can extrapolate the velocity vector 420-1 and the velocity vector 420-2 to find the intersection that makes the hinge point 422-2 estimate.

[0063] In extrapolating to the hinge point 422-2, the articulated vehicle tracker 236 can periodically update its calculations to improve its tracking of the semi-trailer 106-2. For example, as the turn becomes more and more sharp, the articulated vehicle tracker 236 can identify a greater and greater difference in the velocity vectors 420-1 and 420-2. This increase in the articulation angle 424-2 provides increased accuracy of the hinge point 422-2. Depending on the capabilities of the radar system 104-2, the hinge point 422-2 can become valid for subsequent use in controlling the vehicle 102-1 in response to the degree of certainty in the calculations being made. The articulated vehicle tracker 236 can output a degree of certainty or confidence associated with its radar data calculations, including the hinge point 422-2. The degree of certainty can be lower when the articulation angle 424-2 is close to zero degrees, and the confidence in the hinge point 422-2 increases when the articulation angle 424-2 deviates from zero degrees.

[0064] Based on the estimated hinge point 422-2 and the articulation angle 424-2, the articulated vehicle tracker 236 further determines the position and orientation of one or more side edges 426 of the tracked articulated vehicle based on an estimated width of the object 106-2 being tracked. The width 428 of the object 106-2 being tracked corresponds to the estimated width of either of the bounding boxes 412-1 and 412-2. The articulated vehicle tracker 236 uses the width 428 of the object 106-2, the hinge point 422-2, and the articulation angle 424-2 to estimate the position of the side edges 426 of the vehicle. In some cases, a remote processing service can assist in analyzing radar data collected during this period to help resolve the position of the side edges 426. Low-pass filtering can be used to reduce noise in these estimates. Long-term understanding of the hinge point 422-2 and the articulation angle 424-2 determined over time can be used as feedback information to help reduce noise levels and improve the accuracy of the bounding boxes 412-1 and 412-2, which are now associated with the motion of the individual articulated sections. The articulated vehicle tracker 236 can output the position of the closest edge 426 of the semi-trailer 106-2 to the vehicle 102 as an indication of a safety buffer.

[0065] In addition to determining the width 428 of the second bounding box 412-2, the articulated vehicle tracker 236 can also determine the width 428 of the first bounding box 412-1, the total width 428 of the semi-trailer 106-2 being estimated from the width 428 of the first bounding box 412-1 and the width 428 of the second bounding box 412-2 (e.g., by averaging, by taking the larger value). To estimate the closest edge 426 of the semi-trailer 106-2, the articulated vehicle tracker 236 uses the hinge point 422-2, the articulation angle 424-2, and the width 428 to estimate the portion of the edge 426 between the two articulated sections. By providing an accurate representation of the height of the semi-trailer 106-2 as it turns, the articulated vehicle tracker 236 enables the vehicle 102-2 to safely drive near the semi-trailer 106-2 without generating any false pre-collision warnings that can occur if the semi-trailer 106-2 is tracked using only one bounding box instead of two bounding boxes 412-1 and 412-2.

[0066] Example processes

[0067] Figure 5A process 500 for tracking different portions of an articulated vehicle according to the present disclosure is shown. The process 500 is shown as a set of operations (or actions) performed, but is not necessarily limited to the order or combinations shown herein. Further, any one or more of the operations can be repeated, combined, or re-ordered to provide other methods. In the following discussion, reference is made to the entities detailed in other figures, which are referenced by example only. The process 500 is not limited to being performed by one entity or multiple entities.

[0068] At 502, a vehicle driving in a field of view of a radar system is tracked using a single bounding box. For example, the radar system 104 can track the object 106 in the field of view using the bounding box 112.

[0069] At 504, it is determined whether the vehicle is a long vehicle. For example, the radar system 104 compares the length of the bounding box 112 to a length threshold. If the length is less than the threshold, the “No” branch from 504 is taken and the object 106 is tracked using only the bounding box 112. However, in response to determining that the object 106 is a long vehicle based on the length exceeding the length threshold, the “Yes” branch from 504 is taken. This point in the process 500 is consistent with the vehicle tracker 234 invoking the articulated vehicle tracker 236, which can represent two parallel tracking schemes until the radar system 104 is confident that the object 106 can be tracked using only one bounding box or whether multiple bounding boxes should be used in the articulated case.

[0070] At 506, a bounding box for a front portion of the vehicle and another bounding box for a rear portion of the vehicle are determined. For example, the radar system 104 can generate the bounding boxes 112-1 and 112-2.

[0071] At 508, a velocity vector determined for each of the two bounding boxes generated for the front portion and the rear portion is determined. For example, the bounding boxes 112-1 and 112-2 are characterized by the velocity vector 120-1 and the velocity vector 120-1.

[0072] At 510, it is determined whether the vehicle is articulated. For example, velocity vectors 120-1 and 120-2 are monitored. An articulation point can be determined by estimating the intersection between the two velocity vectors 120-1 and 120-2, particularly when the object 106 is traveling around a curve and the front and rear portions, if articulated, are allowed to move in different directions. If the articulation angle at the articulation point remains near zero degrees, even during a turning maneuver, then the “No” branch from 510 is taken and it is determined that the vehicle is non-articulated. If the articulation angle at the articulation point increases above zero degrees, particularly during a turning maneuver, then the “Yes” branch from 510 is taken and it is determined that the vehicle is articulated.

[0073] At 512, the vehicle is tracked using two bounding boxes generated for the front and rear portions instead of a single bounding box. For example, two bounding boxes 112-1 and 112-2 can replace the bounding box 112. In other examples, the bounding box 112 can be generated in addition to the two bounding boxes 112-1 and 112-2, with the bounding box 112 designated as a less accurate solution for the object 106.

[0074] At 514, a driving maneuver is performed based on the two bounding boxes generated for the front and rear portions. For example, the radar system 104 outputs an indication of the two bounding boxes 112-1 and 112-2 to an autonomous driving system. With the estimated closest edges of the object 106 determined from the dimensions of the bounding boxes 112-1 and 112-2, the vehicle 102 can safely drive near or otherwise close to the object 106 without receiving false alarms or false reports of potential collision situations.

[0075] As another example, consider a radar system 104 tracking an object 108 using a bounding box 114. In this example, the object 108 is a school bus, and thus is non-articulated.

[0076] For example, at 504, the radar system 104 compares the length of the bounding box 114 to a length threshold. If the length is less than the threshold, then the “No” branch from 504 is taken and the object 108 is tracked using only the bounding box 114. However, in response to determining that the object 108 is a long vehicle based on the length exceeding the length threshold, the “Yes” branch from 504 is taken.

[0077] At 506, the radar system 104 produces a bounding box for a front portion of the object 108 and determines another bounding box for a rear portion of the object 108.

[0078] At 508, the velocity vectors determined for each of the two bounding boxes generated for the front portion and the rear portion of the object 108 are evaluated to determine whether the object 108 is articulated at 510. For example, the articulation point can be determined by estimating the intersection between the two velocity vectors, particularly when the object 108 is traveling around a curve and the front portion and the rear portion, if articulated, are allowed to move in different directions. If the articulation angle at the articulation point remains near zero degrees, even during a turning maneuver, the “No” branch from 510 is taken and it is determined that the object 108 is non-articulated. Non-articulated vehicles are tracked using only a single bounding box 114. The edges of the bounding box 114 can be used to perform driving maneuvers to keep the vehicle 102 outside the path of the object 118, especially when in a neighboring lane during a turn.

[0079] Accuracy improvement

[0080] Figure 6-1 And Figure 6-2 Aspects of accuracy improvement functionality of a radar system configured to track different portions of an articulated vehicle according to the present disclosure are shown. In each of the various examples described above, there is an implicit assumption that the velocity vector of a steerable portion of an articulated vehicle (e.g., the tractor portion of a tractor-trailer vehicle) is parallel to the longitudinal axis of that portion, and thus, can be used as a direct indication of the angular orientation of the velocity vector of that steerable portion. In practice, if the radar detects in association with a front portion of a steerable portion (e.g., the tractor portion of a tractor-trailer vehicle), the angular orientation of the velocity vector of the steerable portion can be reported as more closely related to the direction in which the front wheel is pointed.

[0081] For example, Figure 6-1 An environment 600 is included in which the radar system 104 reports a velocity vector 604 of a steerable portion of a tractor-trailer vehicle. Figure 6-1 The direction 602 (angular orientation) of the longitudinal axis of the steerable portion of the tractor-trailer vehicle is also represented in the center. The steerable portion is connected to the trailer portion at an articulation point 622; both portions of the tractor-trailer are intended to pivot at the articulation point 622 to affect an articulation angle 624 calculated between them. The result of assuming that the velocity vector 604 is parallel to the direction 602 of the longitudinal axis is that the edge position can occur in an incorrect location, and the articulation point 622 can be incorrect. An accuracy improvement functionality can be invoked by an articulated vehicle tracker such as the articulated vehicle tracker 236, which reduces this inaccuracy.

[0082] To prevent or at least reduce this inaccuracy, the position information, velocity information, and curvature information associated with the front and rear bounding boxes can be used in conjunction with a tractor-trailer vehicle dynamics model to estimate the hitch point 622. The orientation 602 can be computed as a parallel line to the connecting line between the front position of the front bounding box and the hitch point 622.

[0083] For example, Figure 6-2 An example tractor-trailer dynamics model 630 is depicted. The hitch point 622 is located at the intersection of the longitudinal axis of the trailer, a line containing each of the trailer portion and the tractor portion, one center of rotation (COR) of the tractor, and one COR of the trailer. The longitudinal axis of the trailer portion is a line containing the velocity vector of the trailer at the rear position (which can be assumed to be located at the lateral center of the trailer). Each COR can be readily computed from this position, velocity, and curvature information.

[0084] For example, the radar system 104 updates the bounding boxes 112-1 and 112-2 based on the accuracy improvement function. By running a trajectory filter on the bounding boxes 112-1 and 112-2, the radar system 104 outputs the dimensions including the CORs from the position, velocity, and curvature information reported by the trajectory filter. The radar system 104 can compute the line intersection of the respective COR lines to obtain the position of the hitch point 622. The position of the hitch point 622 is assumed to not change over time. Thus, any historical information relating to this position obtained under good conditions (e.g., where the CORs are on opposite ends of the tractor-trailer vehicle) can be used at the current time, even under poor conditions. Subsequently, the articulation angle 624 is computed along with the pointing angle of the tractor portion.

[0085] To make this accuracy improvement idea workable, a number of practical issues need to be overcome. The tracking filter performed by the radar system 104 can have a time lag, especially when reporting curvature estimates. This can result in a significant error level in the CORs computed using the slow tracking filter. Because these CORs are used to compute the line intersecting the longitudinal axis of the trailer, there can be a significant error in the computed hitch point 622. If the two CORs are close to each other, the error in the computed intersection point will be particularly sensitive to the error in the COR positions. In the extreme but possibly common case where the entire tractor-trailer vehicle is in a steady-state turn, the two CORs will be at the same location, thereby making it theoretically difficult to identify the position of the hitch point 622. That is, the orientation of the trailer can not be affected by the computed position of the hitch point 622; only its length can be affected.

[0086] The radar system 104 uses the accuracy improvement functionality to determine an updated velocity of the bounding box 112-1 and an updated velocity of the bounding box 112-2 and determine whether the object 106 is a articulated vehicle. In response to determining whether the object 106 is an articulated vehicle based on the updated velocities, the radar system 104 can output updated radar data that causes the vehicle 102 to perform a driving maneuver by tracking the bounding boxes 112-1 and 112-2 separately or simultaneously.

[0087] Examples

[0088] Some further examples of tracking an articulated vehicle near a vehicle are described below.

[0089] Example 1. A method comprising: tracking, by a first vehicle, a second vehicle driving in a field of view of a radar system of the first vehicle, the tracking of the second vehicle including: generating, using the radar system of the first vehicle, a first bounding box associated with a first portion of the second vehicle and a second bounding box associated with a second portion of the second vehicle; and determining, based on a first velocity vector associated with the first bounding box and a second velocity vector associated with the second bounding box, whether the second vehicle is an articulated vehicle; and in response to determining that the second vehicle is an articulated vehicle, performing, by the first vehicle, a driving maneuver by tracking the first bounding box and the second bounding box separately and simultaneously in the field of view.

[0090] Example 2. The method of example 1, further comprising: in response to the second vehicle reaching or exceeding a threshold length, initiating the determination of whether the second vehicle is an articulated vehicle; in response to the second vehicle being an articulated vehicle, positioning a hinge point on or between the first portion and the second portion of the articulated vehicle; and in response to positioning the hinge point, setting the first bounding box and the second bounding box to not overlap.

[0091] Example 3. The method of example 2, further comprising: in response to the articulated vehicle driving on a curved road while turning or positioning the hinge point; and determining an articulation angle between the first bounding box and the second bounding box, wherein the articulation angle is greater than zero degrees.

[0092] Example 4. The method of example 3, further comprising: determining a width of the first bounding box; determining a width of the second bounding box; estimating a closest edge of the articulated vehicle to the first vehicle by using: the hinge point; the articulation angle; the first width; and the second width; and performing, by the first vehicle, the driving maneuver by avoiding driving unsafely close to the articulated vehicle or colliding with the articulated vehicle.

[0093] Example 5. The method of Example 4, further comprising: further performing, by the first vehicle, the driving maneuver by avoiding edges of the articulated vehicle.

[0094] Example 6. The method of Example 1, further comprising: tracking, by the first vehicle, a third vehicle driving in a field of view of the radar system, the tracking of the third vehicle including: generating, using the radar system, a third bounding box associated with a first portion of the third vehicle and a fourth bounding box associated with a second portion of the third vehicle; determining, based on a third velocity vector associated with the third bounding box and a fourth velocity vector associated with the fourth bounding box, that the third vehicle is a non-articulated vehicle; and replacing the third bounding box and the fourth bounding box with a fifth bounding box associated with the first portion and the second portion of the third vehicle; and in response to determining that the third vehicle is a non-articulated vehicle, performing, by the first vehicle, another driving maneuver by tracking the fifth bounding box associated with the third vehicle in the field of view.

[0095] Example 7. The method of Example 1, further comprising: updating the first bounding box and the second bounding box based on an accuracy improvement function; determining, based on an updated first velocity vector associated with the first bounding box and an updated second velocity vector associated with the updated second bounding box, whether the second vehicle is an articulated vehicle; and in response to determining, based on the updated first velocity vector and the updated second velocity vector, that the second vehicle is an articulated vehicle, tracking at least one of the first bounding box or the second bounding box separately or simultaneously in the field of view to perform the driving maneuver.

[0096] Example 8. A system comprising at least one processor of a first vehicle, the at least one processor configured to: track a second vehicle driving in a field of view of a radar system of the first vehicle, the at least one processor configured to track the second vehicle by: generating a first bounding box associated with a first portion of the second vehicle and a second bounding box associated with a second portion of the second vehicle; and determining, based on a first velocity vector associated with the first bounding box and a second velocity vector associated with the second bounding box, whether the second vehicle is an articulated vehicle; and in response to determining that the second vehicle is an articulated vehicle, performing a driving maneuver by tracking the first bounding box and the second bounding box separately and simultaneously in the field of view.

[0097] Example 9. The system of example 8, wherein the at least one processor is further configured to track the second vehicle by: responsive to the second vehicle reaching or exceeding a threshold length, initiating a determination of whether the second vehicle is an articulated vehicle; responsive to the second vehicle being an articulated vehicle, positioning a hinge point on or between a first portion and a second portion of the articulated vehicle; and responsive to positioning the hinge point, setting the first bounding box and the second bounding box to be non-overlapping.

[0098] Example 10. The system of example 9, wherein the at least one processor is further configured to track the second vehicle by: responsive to the articulated vehicle turning or driving on a curved road, positioning the hinge point; and determining an articulation angle between the first bounding box and the second bounding box, wherein the articulation angle is greater than zero degrees.

[0099] Example 11. The system of example 10, wherein the at least one processor is further configured to: further track the second vehicle by: determining a first length and a first width associated with the first bounding box; determining a second length and a second width associated with the second bounding box; and estimating a closest edge of the articulated vehicle to the first vehicle by using: the hinge point; the articulation angle; the first length and the first width; and the second length and the second width; and perform the driving maneuver by avoiding driving unsafely close to the articulated vehicle or colliding with the articulated vehicle.

[0100] Example 12. The system of example 11, wherein the at least one processor is further configured to: further perform the driving maneuver by avoiding edges of the second vehicle.

[0101] Example 13. The system of example 8, wherein the field of view comprises: a 360 degree field of view; one or more overlapping or non-overlapping 180 degree fields of view; one or more overlapping or non-overlapping 120 degree fields of view; or one or more overlapping or non-overlapping 90 degree fields of view.

[0102] Example 14. The system of example 8, wherein the at least one processor is further configured to: update the first bounding box and the second bounding box based on an accuracy improvement function; determine whether the second vehicle is an articulated vehicle based on an updated first speed associated with the first bounding box and an updated second speed associated with the updated second bounding box; and responsive to determining whether the second vehicle is an articulated vehicle based on the updated first speed and the updated second speed, further perform the driving maneuver by tracking at least one of the first bounding box or the second bounding box in the field of view individually or simultaneously based on whether the second vehicle is an articulated vehicle.

[0103] Example 15. A computer-readable storage medium comprising instructions that, when executed, cause at least one processor of a first vehicle to track a second vehicle driving in a field of view of a radar system of the first vehicle by at least: generating a first bounding box associated with a first portion of the second vehicle and a second bounding box associated with a second portion of the second vehicle; and determining, based on a first velocity vector associated with the first bounding box and a second velocity vector associated with the second bounding box, whether the second vehicle is an articulated vehicle; and in response to determining that the second vehicle is an articulated vehicle, performing a driving maneuver by separately and simultaneously tracking the first bounding box and the second bounding box in the field of view.

[0104] Example 16. The computer-readable storage medium of example 15, wherein the instructions, when executed, further cause the at least one processor to track the second vehicle by at least: in response to the second vehicle reaching or exceeding a threshold length, initiating a determination of whether the second vehicle is an articulated vehicle; in response to the second vehicle being an articulated vehicle, positioning a hinge point on or between the first portion and the second portion of the articulated vehicle; and in response to positioning the hinge point, setting the first bounding box and the second bounding box to be non-overlapping.

[0105] Example 17. The computer-readable storage medium of example 16, wherein the instructions, when executed, further cause the at least one processor to track the second vehicle by at least: in response to the articulated vehicle turning or driving on a curved road, positioning the hinge point; and determining an articulation angle between the first bounding box and the second bounding box, wherein the articulation angle is greater than zero degrees.

[0106] Example 18. The computer-readable storage medium of example 17, wherein the instructions, when executed, further cause the at least one processor to track the second vehicle by at least: determining a first length and a first width of the first bounding box; determining a second length and a second width of the second bounding box; and estimating a closest edge of the articulated vehicle to the first vehicle by using: the hinge point; the articulation angle; the first length and the first width; and the second length and the second width; and further cause the at least one processor to perform the driving maneuver by avoiding driving unsafely close to the articulated vehicle or colliding with the articulated vehicle.

[0107] Example 19. The computer-readable storage medium of example 18, wherein the instructions, when executed, further cause the at least one processor to perform the driving maneuver by at least: performing the driving maneuver by avoiding the edge of the second vehicle.

[0108] Example 20. The computer-readable storage medium of Example 15, wherein the instructions, when executed, further cause the at least one processor to track the second vehicle by at least: updating the first bounding box and the second bounding box based on an accuracy improvement function; determining whether the second vehicle is a hinged vehicle based on an updated first speed associated with the first bounding box and an updated second speed associated with the updated second bounding box; and in response to determining whether the second vehicle is a hinged vehicle based on the updated first speed and the updated second speed, performing a driving maneuver by tracking at least one of the first bounding box and the second bounding box individually or simultaneously in the field of view further based on whether the second vehicle is a hinged vehicle.

[0109] CONCLUSION

[0110] While various embodiments of the present disclosure have been described above and illustrated in the accompanying drawings, it will be understood that the present disclosure is not limited to the aforementioned embodiments, but can be implemented in various ways to practice within the scope of the claims hereinafter. It will be obvious to those skilled in the art from the foregoing description that various changes can be made without departing from the scope of the present disclosure defined by the appended claims. In addition to radar systems, problems associated with double-static conditions can arise in other systems (e.g., image systems, lidar systems, ultrasonic systems) that identify and process tracks from various sensors. Thus, while described as one way to improve radar detection of static objects, the techniques of the foregoing description can be applied to other problems to effectively detect double-static conditions and take appropriate action.

[0111] The use of “or” and grammatically related terms herein is intended to mean no limitation, exclusive or otherwise, unless the context clearly indicates otherwise. As used herein, a phrase referring to “at least one of’ a list of items refers to any combination of those items, including single members. As an example, “at least one of a, b, or c” is intended to cover a, b, c, a-b, a-c, b-c, and a-b-c, as well as any combination with multiples of the same element (e.g., a-a, a-a-a, a-a-b, a-a-c, a-b-b, a-c-c, b-b, b-b-b, b-b-c, c-c, and c-c-c or any other ordering of a, b, and c).

Claims

1. A method for a vehicle, the method comprising: tracking, by a first vehicle, a second vehicle driving in a field of view of a radar system of the first vehicle, the tracking of the second vehicle comprising: generating, using the radar system of the first vehicle, a first bounding box associated with a first portion of the second vehicle and a second bounding box associated with a second portion of the second vehicle; and determining, based on a first velocity vector associated with the first bounding box and a second velocity vector associated with the second bounding box, whether the second vehicle is an articulated vehicle; and in response to determining that the second vehicle is the articulated vehicle, performing, by the first vehicle, a driving maneuver by separately and simultaneously tracking the first bounding box and the second bounding box in the field of view.

2. The method of claim 1, wherein, further comprising: initiating the determination of whether the second vehicle is the articulated vehicle in response to the second vehicle reaching or exceeding a threshold length; positioning, in response to the second vehicle being the articulated vehicle, a hinge point on or between the first portion and the second portion of the articulated vehicle; and and setting, in response to positioning the hinge point, the first bounding box and the second bounding box to be non-overlapping.

3. The method of claim 2, wherein, further comprising: positioning the hinge point in response to the articulated vehicle turning or driving on a curved road; and determining an articulation angle between the first bounding box and the second bounding box, wherein the articulation angle is greater than zero degrees. further comprising:

4. The method of claim 3, wherein, determining a first width of the first bounding box; determining a second width of the second bounding box; estimating, by using: the hinge point; the articulation angle; the first width; and the second width; a closest edge of the articulated vehicle to the first vehicle; and performing, by the first vehicle, the driving maneuver by avoiding driving unsafely close to the articulated vehicle or colliding with the articulated vehicle. further comprising:

5. The method of claim 4, wherein, further performing, by the first vehicle, the driving maneuver by avoiding the edge of the articulated vehicle. tracking the second vehicle further comprises:

6. The method of claim 3, wherein, determining a first length and a first width associated with the first bounding box; determining a second length and a second width associated with the second bounding box; and estimating, by using: the hinge point; the articulation angle; the first length and the first width; and the second length and the second width; a closest edge of the articulated vehicle to the first vehicle; and performing the driving maneuver by avoiding driving unsafely close to the articulated vehicle or colliding with the articulated vehicle.

7. The method of claim 6, wherein, further comprising: further performing the driving maneuver by avoiding the edge of the second vehicle.

8. The method of claim 1, wherein, further comprising: tracking, by the first vehicle, a third vehicle driving in the field of view of the radar system, the tracking of the third vehicle comprising: generating, using the radar system, a third bounding box associated with a first portion of the third vehicle and a fourth bounding box associated with a second portion of the third vehicle; determining, based on a third velocity vector associated with the third bounding box and a fourth velocity vector associated with the fourth bounding box, that the third vehicle is a non-articulating vehicle; and replacing the third bounding box and the fourth bounding box with a fifth bounding box associated with the first portion and the second portion of the third vehicle; and in response to determining that the third vehicle is the non-articulating vehicle, performing, by the first vehicle, another driving maneuver by tracking, in the field of view, the fifth bounding box associated with the third vehicle.

9. The method of claim 1, wherein, further comprising: updating the first bounding box and the second bounding box based on an accuracy improvement function; determining, based on an updated first velocity vector associated with the first bounding box and an updated second velocity vector associated with the updated second bounding box, whether the second vehicle is an articulating vehicle; and in response to determining, based on the updated first velocity vector and the updated second velocity vector, that the second vehicle is the articulating vehicle, tracking at least one of the first bounding box or the second bounding box separately and simultaneously in the field of view to perform the driving maneuver. the field of view comprises:

10. The method of claim 1, wherein, a 360-degree field of view; one or more overlapping or non-overlapping 180-degree fields of view; or one or more overlapping or non-overlapping 120-degree fields of view; or one or more overlapping or non-overlapping 90-degree fields of view.

11. A system for a vehicle, the system comprising at least one processor configured to perform the method of any one of claims 1-10. the at least one processor is for the first vehicle.

12. The system of claim 11, wherein, the at least one processor for the first vehicle is for a radar system of the first vehicle.

13. The system of claim 12, wherein, 14. A system for a vehicle, the system comprising means for performing the method of any one of claims 1-10.

15. A computer-readable storage medium comprising instructions that, when executed, cause at least one processor of a first vehicle to perform the method of any one of claims 1-10. ​

Images