Method for longitudinal control when approaching a traffic cluster

Abstract

A system and method for longitudinal control upon approaching a traffic cluster includes receiving sensor data collected from one or more sensors in communication with a vehicle, processing the sensor data to determine that it indicates an approaching cluster, and verifying the approaching cluster by comparing the approaching cluster to track data and determining that the approaching cluster does not match the track data. The method also includes generating a cluster confidence score for the approaching cluster based on a number of objects included in the approaching cluster and initiating an open-loop longitudinal response to the approaching cluster.The method further includes performing a transition optimization for the open-loop longitudinal response to determine a handover point for the vehicle, and transitioning from the open-loop longitudinal response to a closed-loop longitudinal response at the handover point.

Description

INTRODUCTION

[0001] The present invention generally relates to a method for longitudinal control when approaching a traffic cluster.

[0002] A generic procedure is essentially described in DE 10 2018 209 388 A1.

[0003] Further state of the art can also be found in DE 10 2017 011 299 A1.

[0004] Generally, when cruise control (e.g., longitudinal control) is activated for a moving vehicle, sensors at the front of the vehicle continuously receive sensor data to detect objects on the road that the vehicle may strike. If the objects are easily identifiable (i.e., within a high-fidelity range) and likely to impact the vehicle (i.e., within a path of travel of the vehicle), the cruise control can initiate vehicle deceleration using a closed-loop control system based on the approaching object. The sooner the cruise control detects an approaching object, the more time it has to decelerate the vehicle to avoid shock to the driver or vehicle occupants.In particular, objects located at extreme distances from a moving vehicle are generally difficult to detect in terms of initiating or regulating the vehicle's deceleration. Accordingly, reliably detecting objects beyond conventional sensor ranges can allow the vehicle to decelerate without disturbing the driver. SUMMARY

[0005] According to the invention, a computer-implemented method for longitudinal control when approaching a traffic congestion is presented, which is characterized by the features of claim 1.

[0006] When executed on computing hardware, the method causes computing hardware to perform operations that include receiving sensor data collected from one or more sensors in communication with a vehicle, the sensor data identifying one or more data fragments that are outside a high-fidelity range of the vehicle, and processing the sensor data to determine that the sensor data indicates an approaching cluster. Wherein, the approaching cluster includes two or more data objects identified in the one or more data fragments.The operations also include verifying the approaching cluster by comparing the approaching cluster to track data and determining that the approaching cluster does not match the track data, generating a cluster confidence score for the approaching cluster, the cluster confidence score based on the number of objects identified in the approaching cluster, and initiating an open-loop longitudinal response to the approaching cluster based on the cluster confidence score. The operations further include performing transition optimization for the open-loop longitudinal response to determine a handoff point for the vehicle, and transitioning from the open-loop longitudinal response to a closed-loop longitudinal response at the handoff point.

[0007] Implementations of the invention may include one or more of the following optional features. In some implementations, the one or more sensors communicatively coupled to the vehicle comprise at least one of a long-range radar and a camera sensor. In some examples, the one or more sensors are disposed within the vehicle. In some implementations, the route data comprises at least one of road geometry, lane counts, or established structures. In some examples, the cluster confidence score is further based on the duration for which each of the objects identified in the approaching cluster is detected, the speed of each of the objects identified in the approaching cluster, and the distance between the vehicle and the approaching cluster.In some implementations, initiating an open-loop longitudinal response to the approaching cluster includes determining a response level based on the cluster confidence score of the approaching cluster.

[0008] In some examples, performing the transition optimization on the open-loop longitudinal response to determine the handoff point for the vehicle includes identifying a target speed based on a current speed of the vehicle and transitioning from the open-loop longitudinal response to the closed-loop longitudinal response when a speed of the vehicle equals the target speed. In these examples, the target speed may be further based on a cluster speed of the approaching cluster and a distance between the vehicle and the approaching cluster. In some implementations, the one or more data fragments are identified from more than one of the one or more sensors. In some examples, the operations further include updating a speed of the vehicle when the one or more objects enter the high-fidelity range of the vehicle.

[0009] Furthermore, a system for longitudinal control when approaching a traffic cluster is described, which system comprises data processing hardware and storage hardware in communication with the data processing hardware.

[0010] Instructions are stored in the memory hardware that, when executed by the computing hardware, cause the computing hardware to perform operations that include receiving sensor data collected by one or more sensors in communication with a vehicle, the sensor data identifying one or more data fragments that are outside a high-fidelity range of the vehicle, and processing the sensor data to determine that the sensor data indicates an approaching cluster. Wherein, the approaching cluster comprises two or more data objects identified in the one or more data fragments.The operations also include verifying the approaching cluster by comparing the approaching cluster to track data and determining that the approaching cluster does not match the track data, generating a cluster confidence score for the approaching cluster, the cluster confidence score based on the number of objects identified in the approaching cluster, and initiating an open-loop longitudinal response to the approaching cluster based on the cluster confidence score. The operations further include performing transition optimization for the open-loop longitudinal response to determine a handoff point for the vehicle, and transitioning from the open-loop longitudinal response to a closed-loop longitudinal response at the handoff point.

[0011] This aspect may include one or more of the following optional features. In some implementations, the one or more sensors communicatively coupled to the vehicle include a long-range radar and / or a camera sensor. In some examples, the one or more sensors are located within the vehicle. In some implementations, the route data includes road geometry, number of lanes, and / or confirmed structures. In some examples, the cluster confidence score is further based on the duration for which each of the objects identified in the approaching cluster is detected, the speed of each of the objects identified in the approaching cluster, and the distance between the vehicle and the approaching cluster.In some implementations, initiating an open-loop longitudinal response to the approaching cluster includes determining a response level based on the cluster confidence score of the approaching cluster.

[0012] In some examples, performing the transition optimization on the open-loop longitudinal response to determine the handoff point for the vehicle includes identifying a target speed based on a current speed of the vehicle and transitioning from the open-loop longitudinal response to the closed-loop longitudinal response when a speed of the vehicle equals the target speed. In these examples, the target speed may be further based on a cluster speed of the approaching cluster and a distance between the vehicle and the approaching cluster. In some implementations, the one or more data fragments are identified from more than one of the one or more sensors. In some examples, the operations further include updating a speed of the vehicle when the one or more objects enter the high-fidelity range of the vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The drawings described herein are for illustrative purposes only and are intended to illustrate selected configurations. Fig. Figure 1 is a schematic representation of an example system for longitudinal control when approaching a traffic cluster. Fig. 2 is a schematic representation of example components of the system from Fig. 1. Fig. Figure 3 is a schematic view of a vehicle incorporating the system of Fig. 1 when approaching a traffic cluster. Fig. 4 is a classification flow diagram for the system of Fig. 1. Fig. 5 is a flowchart of an exemplary arrangement of operations for a method of longitudinal control when approaching a traffic cluster.

[0014] Corresponding reference symbols indicate corresponding parts in the drawings. DETAILED DESCRIPTION

[0015] Fig. 1 illustrates an example system 100 including a vehicle 10 and / or a remote system 60 communicatively coupled to the vehicle 10 via a network 40. The vehicle 10 and / or the remote system 60 execute a longitudinal control system 200 that operates when a user activates a cruise control function of the vehicle 10. In the examples shown, the longitudinal control system 200 is implemented in a vehicle 10. However, the longitudinal control system 200 may also be implemented on other computing devices (e.g., computing devices communicatively coupled to the vehicle 10), such as, among others, a smartphone, a tablet, a smart display, a desktop / laptop, a smart watch, a smart appliance, or smart glasses / headset.The vehicle 10 includes computing hardware 12 and memory hardware 14 storing instructions that, when executed on the computing hardware 12, cause the computing hardware 14 to perform operations. The vehicle 10 further includes one or more sensors 16 configured to collect / receive sensor data 202. The one or more sensors 16 may include one or more long-range radar sensors and / or one or more camera sensors for collecting image data.

[0016] The remote system 60 (e.g., server, cloud computing environment) also includes computing hardware 62 and storage hardware 64 storing instructions that, when executed on the computing hardware 62, cause the computing hardware 62 to perform operations. In some examples, the longitudinal control system 200 is executed jointly by the vehicle 10 and the remote system 60. As described below with reference to Fig. 2 and Fig. 3, the longitudinal control system 200 executing on the vehicle 10 and / or the remote system 60 executes a cluster detection module 400, a verification module 210, a cluster evaluator 220, an open-loop module 230, and a transition module 240. The longitudinal control system 200 is configured to receive sensor data 202 acquired from one or more sensors 16 communicatively coupled to the vehicle 10 and generate an open-loop longitudinal response 232 and an optimized transition point 242 for a speed of the vehicle 10 at which to transition to a closed-loop longitudinal response 242 while minimizing disruption to a driver of the vehicle 10.

[0017] With reference to Fig. 1-3, during travel of the vehicle 10, the vehicle 10 executes the longitudinal control system 200, which receives as input the sensor data 202 collected by the one or more sensors 16 communicatively coupled to the vehicle 10. The sensor data 202 may include one or more data fragments 302 that are located outside a high-fidelity region 306 of the vehicle 10. The high-fidelity region 306 may generally refer to an area or distance beyond the vehicle where the longitudinal control system 200 detects objects with high confidence. Within the high-fidelity region 306, closed-loop control (i.e., standard adaptive cruise control) based on the nearest object in the path of the vehicle 10 is generally sufficient to maintain a safe speed and following distance.For example, the high-fidelity range 306 may encompass the space up to 120 meters in front of the vehicle 10. However, in other embodiments, the high-fidelity range 306 may encompass any distance from the vehicle 10 at which the longitudinal control system 200 can reliably detect objects. Because the sensor data 202 comprising the one or more data fragments 302 are located outside (e.g., beyond) the high-fidelity range 306, it may be particularly difficult to confirm that the one or more data fragments 302 are objects 304 in the path of the vehicle 10.To resolve the identity of the one or more data fragments 302, and in response to receiving the sensor data 202, the longitudinal control system 200 generates as output an open-loop longitudinal response 232 that initiates a smooth deceleration of a speed of the vehicle 10, a handover point 242 for the speed of the vehicle 10, and a closed-loop longitudinal response 244 that initiates a secondary deceleration (e.g., a standard cruise control) of the vehicle 10. In particular, the handover point 242 for the vehicle 10 is optimized to minimize disturbances and / or shocks to the driver and passengers in the vehicle 10.

[0018] With reference to Fig. 2 and Fig. 3, the longitudinal control system 200 includes a cluster detection module 400 configured to process the sensor data 202 including the one or more data fragments 302 and, if one or more data fragments 302 are confirmed as a threshold number of detected objects 304, identify the detected objects 304 as a cluster 402 that the vehicle 10 is approaching. In some implementations, the cluster detection module 400 identifies the one or more data fragments 302 as an approaching cluster 402 if the cluster detection module 400 detects two (2) or more objects 304 in the one or more data fragments 302. For example, as in Fig. 3, the sensor data 202 includes three (3) data fragments 302a-302c in the path (e.g., lane) of the vehicle 10. Then, the cluster detection module 400 executes a classification model to determine whether the one or more data fragments 302 are accepted or rejected as a detected object 304.

[0019] With reference to Fig. 4 shows the classification model of the cluster detection module 400. The cluster detection module 400 processes the received sensor data 202, including the one or more data fragments 302, and determines whether the sensor data 202 may contain an object 304, and checks whether the one or more data fragments 302 are located in a lane on the road on which the vehicle 10 is traveling. If the sensor data 202 does not contain any data fragments 302 or the data fragments 302 are not located in a lane on the same road as the vehicle 10, they are rejected. If the data fragments 302 are located in a lane on the same road as the vehicle 10, the classification model determines which of the one or more sensors 16 of the vehicle detected the one or more data fragments 302.In particular, the classification model determines whether the data fragments 302 are i) detected by both radar and camera sensors, ii) only by radar sensors, or iii) only by camera sensors. If the data fragments 302 are detected by both radar sensors and camera sensors, the classification model determines whether the data fragments 302 are moving in the same direction as the vehicle 10, whether they were previously moving but are currently stopped, or whether they have not moved.

[0020] If one of these conditions is met, the data fragments 302 are accepted as a recognized object 304. If none of these conditions are met, the data fragments 302 are rejected.

[0021] If the data fragments 302 are detected only by radar sensors, the classification model determines whether the data fragments 302 are moving in the same direction as the vehicle 10 or whether they were previously moving but are currently stopped. If neither of these conditions is met, the data fragments 302 are rejected. If either of these conditions is met, the classification model determines whether the data fragments 302 are within a threshold distance of track data 204 of the vehicle 10. If the data fragments 302 are within the threshold distance of the track data 204, the data fragments 302 are accepted as a detected object 304. However, if the data fragments 302 are not within the threshold distance of the track data 204, the data fragments 302 are rejected.Similarly, if the data fragments 302 are captured only by camera sensors, the classification model determines whether the data fragments 302 are moving in the same direction as the vehicle 10 or whether they were previously moving but are now stopped. If neither of these conditions is met, the data fragments 302 are rejected. If either of these conditions is met, the classification model determines whether the data fragments 302 are within a threshold distance from the vehicle 10. If the data fragments 302 are within the threshold distance from the vehicle 10, the data fragments 302 are accepted as a detected object 304. However, if the data fragments 302 are not within the threshold distance from the vehicle 10, the data fragments 302 are rejected.

[0022] Referring again to the example in Fig. 3, the sensor data 202 includes ten (10) data fragments 302a-302j on the same road as the vehicle 10. Specifically, three (3) data fragments 302a-302c are located in the path (e.g., the lane) of the vehicle 10, while seven (7) data fragments 302d-302j are located in lanes adjacent to the lane of the vehicle 10 (e.g., to the left and right of it). Here, the cluster detection module 400 executes the classification model and, after processing the data fragments 302a-302j, determines whether the data fragments 302 constitute a sufficient number of detected objects 304 to indicate an approaching cluster 402. In particular, the cluster detection module 400 identifies / detects that the data fragments 302b-302j contain objects 304a-304i. Conversely, the cluster detection module 400 detects (e.g., via the classification model) that the data fragment 302a does not contain an object.Based on the detected objects 304a-304i that meet (and exceed) the threshold (e.g., two objects 304), the cluster detection module 400 determines that the sensor data 202 indicates an approaching cluster 402.

[0023] In particular, the threshold for the number of objects 304 that the cluster detection module 400 detects to determine that the sensor data 202 indicates an approaching cluster 402 may be selectively changed based on the characteristics of the data fragments. For example, if a fragment 302 in the path of the vehicle 10 is confirmed as a detected object 304, the cluster detection module 400 may apply a lower threshold to determine that the sensor data 202 indicates an approaching cluster 402. In other words, if the cluster classification model determines that a data fragment 302 in the same lane as the vehicle 10 is a detected object 304, the cluster detection module 400 may apply a threshold of one (1) detected object 304 and determine that the data fragment 302 in the same lane as the vehicle 10 indicates an approaching cluster 402.Conversely, if the cluster classification model does not determine / identify any data fragments 302 in the same lane as the vehicle 10 (i.e., the data fragments 302 are on the road but not in the same lane as the vehicle 10), the cluster detection module 400 may apply a higher threshold (e.g., six detected objects 304) when determining whether the sensor data 202 indicates an approaching cluster 402.

[0024] With further reference to Fig. 2, after the cluster detection module 400 determines that the sensor data 202 indicates an approaching cluster 402, the verification module 210 qualifies the detected approaching cluster 402 based on known track data 204. Specifically, the verification module 210 receives the approaching cluster 402, including the sensor data 202 and the track data 204, and verifies whether the approaching cluster 402 includes any known objects in the track data 204. The track data 204 may include one or more of the road geometries on which the vehicle 10 is traveling, such as hills or bumps, the number of lanes on the road the vehicle is traveling on, or confirmed / known structures such as bridges or traffic signs. In implementations where the track data 204 matches one of the detected objects 304 within the approaching cluster 402, the verification module 210 may discard the detected object 304.For example, if one of the detected objects 304 in the approaching cluster 402 corresponds to an overpass in the track data 204, the verification module 210 determines that the overpass is not an object with which the vehicle 10 will come into contact and removes the object (i.e., the overpass) from the approaching cluster 402. Conversely, if the track data 204 does not correspond to any of the detected objects 304 in the approaching cluster 402, the verification module 210 verifies the approaching cluster 402.

[0025] Once the approaching cluster 402 has been verified by the verification module 210, the longitudinal control system 200 generates a cluster confidence score 222 for the approaching cluster 402. In particular, a cluster evaluator 220 of the longitudinal control system 200 receives as input the number of objects 304 identified in the approaching cluster 402 and generates as output a cluster confidence score 222 indicating a confidence that the approaching cluster 402 is to be acted upon.In doing so, the cluster evaluator 220 may additionally receive a duration of time each of the detected objects 304 is in the approaching cluster 402, a speed of each of the detected objects 304 in the approaching cluster 402, a distance between the vehicle 10 and the approaching cluster 402, a number of detected objects 304 in the same lane as the vehicle 10, and / or a number of sensors 16 that have reported / received sensor data 202. For example, if a detected object 304 in the approaching cluster 402 includes an object 304 that was detected for a duration exceeding the threshold duration, the cluster evaluator 220 may generate a high cluster confidence score 222 for the approaching cluster 402.

[0026] After the cluster evaluator 220 evaluates the characteristics of the approaching cluster 402 and generates the cluster confidence score 222, the open-loop module 230 may identify an escalation level for the approaching cluster 402 and, based on the escalation level, generate an open-loop longitudinal response 232 to the approaching cluster 402. For example, the open-loop module 230 may include any number (e.g., three (3)) of escalation levels based on the cluster confidence score 222 and the sensor data 202.The first escalation level may trigger / initiate an open-loop longitudinal response 232 that includes a gentle reduction in the speed of the vehicle 10, the second escalation level may trigger / initiate an open-loop longitudinal response 232 that includes a moderate reduction in the speed of the vehicle 10, and the third escalation level may trigger / initiate an open-loop longitudinal response 232 that includes a rapid reduction in the speed of the vehicle 10. For example, the open-loop module 230 may accumulate a timer for each detected object 304 in the lane of the vehicle 10 to determine how long the detected object 304 has been in the lane of the vehicle 10. Each escalation level may include a different threshold duration of the detected object 304.

[0027] The first escalation level may be selected, for example, when there are no detected objects 304 in the same lane / path as the vehicle 10 and the time since detection by the longitudinal control system 200 is greater than a calibration threshold and either i) a detected object 304 is present on the same road as the vehicle 10 and the number of detected objects 304 is greater than a counter threshold (e.g., two), or ii) a navigation system (e.g., route data 204) indicates deceleration and the number of detected objects 304 is greater than a counter threshold (e.g., two), or iii) the number of detected objects 304 is greater than a counter threshold (e.g., two).At the first escalation level, the open-loop module 230 may continue to receive sensor data 202 and transition from the first escalation level to the second escalation level if the duration for which the object 304 is detected exceeds a second-level timer threshold. In this case, the second-level timer threshold may include a 2D calibration table with axes for the number of detected objects 304 in the approaching cluster 402 and the speed of the vehicle 10. At the second escalation level, the open-loop module 230 may continue to receive sensor data 202 and transition from the second escalation level to the third escalation level if the duration for which the object 304 is detected exceeds a third-level timer threshold.In this case, the timer threshold for the third stage may also include a 2D calibration table with axes for the number of detected objects 304 in the approaching cluster 402 and the speed of the vehicle 10. As previously mentioned, the timer thresholds may be different for each escalation stage. At each of the escalation levels (e.g., escalation levels one, two, and three), the open-loop module 230 may continue to receive sensor data 202 and deactivate the open-loop longitudinal response 232 if i) a vehicle is detected in the path of the vehicle 10, ii) no object 304 is detected for the calibration threshold duration, iii) the number of detected objects 304 is less than the counter threshold (e.g., two objects 304), iv) the accumulated timer is greater than the calibration threshold duration, or v) the driver of the vehicle 10 overrides the open-loop longitudinal response 232 with the accelerator pedal or the brake.

[0028] With further reference to Fig. 2, the open-loop module 230 identifies the escalation level and, based on the escalation level, determines the open-loop longitudinal response 232 once the open-loop module 230 receives the sensor data 202 and the cluster confidence score 222. Thereafter, the longitudinal control system 320 triggers the open-loop longitudinal response 232, thereby initiating the reduction in speed of the vehicle 10. While the longitudinal control system 200 executes the open-loop longitudinal response 232, the transition module 240 performs a transition optimization on the open-loop longitudinal response 232 to determine a handoff point 244 for the vehicle 10 that avoids / limits abrupt speed changes between the open-loop longitudinal response 232 and a closed-loop longitudinal response 242.The transition module 240 can identify a current speed of the vehicle 10 and a target speed of the vehicle 10, with the transition point 244 creating a smooth transition between the open-loop longitudinal reaction 232 and the closed-loop longitudinal reaction 242. The transition point 244 can be expressed as follows: Vk∼[Vvehicle,ΔVCluster,dCluster]

[0029] Where V k the transfer point 244, expressed as a transfer speed of the vehicle 10, at which the longitudinal control system ideally switches from open-loop control to closed-loop control, V Fahrzeug the current speed of the vehicle 10, ΔV Cluster the speed difference between the vehicle 10 and the approaching cluster 402 and d Cluster the distance between the vehicle 10 and the approaching cluster 402.

[0030] The transition module 240 determines the transfer point 244 (ie V k ) by first determining a motion profile of the vehicle 10 and the approaching cluster 402. The motion profile of the vehicle 10 can be expressed by the following quintic polynomial [V(t), x(t)]: V(x)=c0+c1x+c2x2+c3x3+c4x4+c5x5 where equation 2 for c0, c1x, c2x 2 , c3x 3 , c4x 4 , c5x 5 taking into account V,V̇,V̈|v i v f is solved. Based on the solved [V, x], the transition module 240 calculates the speed reduction as follows: a=vf2−vi22Δx,limx→D ax>threshold where a is the desired acceleration of the vehicle 10, V i the initial speed of the vehicle 10 and V f denote the final speed of the vehicle 10. If a xis less than a threshold acceleration, the transition module 240 identifies the transfer point 244 depending on a speed reduction measure and the current speed V Fahrzeug of the vehicle 10. The transfer point 244 can have a target speed V k and a target acceleration a k include: {Vk,ak}=K{speed reduction rate, VxVehicle}

[0031] Conversely, the transition module 240 can increase the reaction distance of the transfer point 244 if a x is greater than a threshold acceleration. In other words, the transition module 240 may set the transition point 244 such that a transition from the open-loop longitudinal control 232 to the closed-loop longitudinal control 242 occurs when the vehicle 10 is farther away from the approaching cluster 402.

[0032] After the transition module 240 performs the transition optimization to determine the handover point 244 for the vehicle 10, and when the vehicle 10 reaches the handover point 244 (e.g., the speed and / or the speed reduction), the longitudinal control system 200 transitions control from the open-loop longitudinal response 232 to the closed-loop longitudinal response 242. Because the longitudinal control system 200 continues to receive sensor data 202 as the vehicle 10 moves along its path and approaches the cluster 402, the longitudinal control system may update the speed of the vehicle when the one or more objects 304 in the approaching cluster 402 enter the high-fidelity region 306 of the vehicle 10.In other words, as the vehicle 10 approaches the cluster 402, the speed of the vehicle 10 may be updated if the presence of an object 304 is confirmed or rejected once it is within the high-fidelity range 306 of the vehicle 10.

[0033] Fig.5 includes a flowchart of an exemplary arrangement of operations for a method 500 for longitudinal control when approaching a traffic cluster. In operation 502, the method 500 includes receiving sensor data 202 collected from one or more sensors 16 in communication with a vehicle 10, wherein the sensor data 202 identifies one or more data fragments 302 that are outside a high-fidelity region 306 of the vehicle 10. In operation 504, the method 500 also includes processing the sensor data 202 to determine that the sensor data indicates an approaching cluster 402, wherein the approaching cluster includes two or more data objects 304 identified in the one or more data fragments 302.In operation 506, the method 500 further includes verifying the approaching cluster 402 by comparing the approaching cluster to the track data 204 and determining that the approaching cluster 402 does not match the track data 204.

[0034] At operation 508, the method 500 also includes generating a cluster confidence score 222 based on the number of objects 304 identified in the approaching cluster 402. In some implementations, the cluster confidence score 222 is higher when at least one of the objects 304 identified in the approaching cluster 402 is in the same lane as the vehicle 10. The method 500 further includes, at operation 510, initiating an open-loop longitudinal response 232 to the approaching cluster 402. At operation 512, the method 500 also includes performing a transition optimization for the open-loop longitudinal response 232 to determine a handoff point 242 for the vehicle 10. At operation 514, the method 500 further includes, at the handoff point 242 for the vehicle 10, transitioning from the open-loop longitudinal reaction 232 to a closed-loop longitudinal reaction 244.

Claims

[1] A computer-implemented method that, when executed on data processing hardware, causes the data processing hardware to perform operations comprising: Receiving sensor data collected by one or more sensors in communication with a vehicle, the sensor data identifying one or more data fragments that are outside a high-fidelity range of the vehicle; Processing the sensor data to determine that the sensor data indicates an approaching cluster, wherein the approaching cluster comprises two or more data objects identified in the one or more data fragments; and Generating a cluster confidence score for the approaching cluster, the cluster confidence score based on the number of objects identified in the approaching cluster; characterized by Verifying the approaching cluster by comparing the approaching cluster with track data and determining that the approaching cluster does not match the track data; Initiating an open-loop longitudinal response to the approaching cluster based on the cluster confidence score; Performing a transition optimization on the open-loop longitudinal reaction to determine a handover point for the vehicle; and at the transfer point, transition from the open-loop longitudinal reaction to a closed-loop longitudinal reaction. [2] The method of claim 1, wherein the one or more sensors in communication with the vehicle comprise at least one of a long-range radar and a camera sensor. [3] The method of claim 1, wherein the one or more sensors are located within the vehicle. [4] The method of claim 1, wherein the route data comprises at least one of: road geometry; Number of lanes; or existing buildings. [5] The method of claim 1, wherein the cluster confidence score is further based on: a duration for which each of the objects identified in the approaching cluster is detected; a speed of each of the objects identified in the approaching cluster; and the distance between the vehicle and the approaching cluster. [6] The method of claim 1, wherein initiating an open-loop longitudinal response to the approaching cluster comprises determining a response level based on the cluster confidence score of the approaching cluster. [7] The method of claim 1, wherein performing the transition optimization on the open-loop longitudinal response to determine the handoff point for the vehicle comprises identifying a target speed based on a current speed of the vehicle, and transitioning from the open-loop longitudinal response to the closed-loop longitudinal response when a speed of the vehicle equals the target speed. [8] The method of claim 7, wherein the desired speed is further based on a cluster speed of the approaching cluster and a distance between the vehicle and the approaching cluster. [9] The method of claim 1, wherein the one or more data fragments are identified by more than one of the one or more sensors. [10] The method of claim 1, further comprising updating a speed of the vehicle when the one or more objects enter the high-fidelity range of the vehicle.

Images