COMPUTER-IMPLEMENTED METHOD FOR CLOUD-BASED SCALABLE POSITIONING SYSTEMS

Abstract

A system and a method comprise identifying multiple vehicles as a first vehicle group, wherein the first vehicle group has a front end and a rear end, each of the front end and rear end of the first vehicle group executing a respective latency cover model, receiving sensor data from each vehicle of the first vehicle group collected by a sensor system of the vehicle, and categorizing the sensor data as ordered or unordered. The system and method further comprise processing the ordered sensor data using the front end executing the respective latency cover mode and the unordered sensor data using the rear end executing the respective latency cover model, and generating a position estimate of each vehicle in the first vehicle group.

Description

INTRODUCTION

[0001] The present disclosure relates generally to cloud-based scalable positioning systems for vehicles and in particular to a suitable computer-implemented method for this purpose.

[0002] To provide an overview of the general state of the art, reference is made here in advance to the publications DE 10 2022 001 720 B3, DE 10 2016 214 601 A1 and US 2023 / 0 112 004 A1.

[0003] Vehicle positioning systems are essential for modern transportation, enabling the real-time tracking and management of multiple vehicles. These systems typically use a combination of global positioning systems (GPS), cellular networks, and other sensor data to determine the precise location of each vehicle. This information is then sent to a central server, where it can be used for various applications such as fleet management, navigation, and traffic monitoring. In fleet management, for example, operators can monitor the location and status of all vehicles, optimizing routes and improving efficiency. Similarly, navigation systems use real-time positioning data to provide accurate directions and traffic updates to drivers. These systems are essential for ensuring the safety and efficiency of transportation networks.

[0004] However, the algorithms used in autonomous driving systems to process positioning data are computationally intensive. These algorithms must continuously analyze massive amounts of data from multiple sensors to make real-time driving decisions. This computational demand can lead to delays in processing data packets, especially when the data is transmitted over networks with varying latency. Additionally, data packets from individual vehicles can arrive out of order, further complicating the processing and integration of this information. These challenges highlight the need for more effective and scalable solutions to meet the growing demands of positioning systems for autonomous vehicles. SUMMARY

[0005] According to the invention, a computer-implemented method for cloud-based scalable positioning systems is presented, characterized by the features of claim 1.

[0006] When the procedure is executed in data processing hardware, it causes the data processing hardware to perform operations that include identifying multiple vehicles as a first vehicle group, where the first vehicle group has a front end and a rear end. Here, each of the front end and rear end of the first vehicle group executes a respective latency concealment model. The operations also include receiving sensor data from each vehicle in the first vehicle group, collected by a sensor system on the vehicle, and categorizing the sensor data as either ordered or unordered.The operations also include processing the ordered sensor data using the front end of the first vehicle group, which executes the respective latency occlusion model, processing the unordered sensor data using the rear end of the first vehicle group, which executes the respective latency occlusion model, and generating a position estimate for each vehicle in the first vehicle group.

[0007] Implementations of the disclosure may include one or more of the following optional features. In certain implementations, categorizing the sensor data as either ordered or unordered involves adding the unordered sensor data to a batch of unordered sensor data. In these implementations, processing the unordered sensor data may involve processing the batch of unordered sensor data at a predetermined time threshold.

[0008] In certain examples, each of the latency cover models is configured to receive sensor data as input and generate a position estimate for each vehicle in the first vehicle group as output. In these examples, each latency cover model can execute a corresponding position prediction model. Here, the sensor data for each vehicle in the first vehicle group can include initial sensor data detected by a sensor system on that vehicle, and / or subsequent sensor data detected by other vehicles in the first vehicle group, and / or baseline sensor data detected by a base station communicating with the first vehicle group.

[0009] In certain implementations, the operations further include receiving a position prediction for each vehicle in the first vehicle group. In these implementations, the operations may further include identifying one or more of the sensor data elements as missing based on the corresponding temporal data of the sensor data and inserting the received position prediction of the vehicle into a factor graph of the first vehicle group based on the corresponding temporal data of the identified missing sensor data. In certain examples, the operations further include identifying that a first vehicle from the first vehicle group has joined a second vehicle group and extracting the sensor data associated with that first vehicle from the front end of the first vehicle group.In these examples, the operations may further include reconstructing, using the extracted sensor data associated with the first vehicle, a factor graph of the second vehicle group from the front end of the first vehicle group such that it contains the first vehicle.

[0010] Furthermore, a system for cloud-based scalable positioning is described, comprising data processing hardware and storage hardware that communicate with the data processing hardware. The storage hardware stores instructions which, when executed by the data processing hardware, cause the data processing hardware to perform operations that include identifying multiple vehicles as a first vehicle group, where the first vehicle group has a front end and a rear end. Here, each of the front end and rear end of the first vehicle group executes a respective latency concealment model. The operations also include receiving sensor data from each vehicle in the first vehicle group, collected by a sensor system on the vehicle, and categorizing the sensor data as either ordered or unordered.The operations also include processing the ordered sensor data using the front end of the first vehicle group, which executes the respective latency occlusion model, processing the unordered sensor data using the rear end of the first vehicle group, which executes the respective latency occlusion model, and generating a position estimate for each vehicle in the first vehicle group.

[0011] This aspect can include one or more of the following optional features. In certain implementations, categorizing sensor data as either ordered or unordered involves adding the unordered sensor data to a batch of unordered sensor data. In these implementations, processing the unordered sensor data may involve processing the batch of unordered sensor data against a predefined time threshold.

[0012] In certain examples, each of the latency cover models is configured to receive sensor data as input and to generate the position estimate of each vehicle in the first vehicle group as output. In these examples, each of the latency cover models can execute a corresponding position prediction model. In certain implementations, the operations further include identifying that a first vehicle from the first vehicle group has joined a second vehicle group and extracting the sensor data associated with the first vehicle from the front end of the first vehicle group. In these implementations, the operations may further include reconstructing, using the extracted sensor data associated with the first vehicle from the front end of the first vehicle group, a factor graph of the second vehicle group such that it includes the first vehicle.

[0013] Furthermore, a latency masking method is described which, when executed in data processing hardware, causes the data processing hardware to perform operations that include receiving sensor data for a vehicle, where the sensor data contains spatial and temporal data, and receiving a position prediction of the vehicle. The operations also include generating a factor graph for the vehicle's position and performing factor graph optimization by mapping the spatial and temporal data of the sensor data to predict the vehicle's position.

[0014] This aspect can include one or more of the following optional features. In certain implementations, the vehicle's sensor data includes initial sensor data detected by a sensor system on the respective vehicle, and / or subsequent sensor data detected by other vehicles in the initial vehicle group, and / or baseline sensor data detected by a base station communicating with the initial vehicle group. In these implementations, generating the vehicle position factor graph can involve identifying one or more of the sensor data elements as missing based on the corresponding temporal data of the sensor data and inserting the received vehicle position prediction into the factor graph based on the corresponding temporal data of the missing sensor data.

[0015] The details of one or more implementations of the disclosure are set forth in the accompanying drawings and the description below. Further aspects, features, and advantages will become clear from the description, the drawings, and the claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described here serve only to illustrate selected configurations; they show: Fig. 1 A schematic view of an example system for cloud-based scalable positioning; Fig. 2 a schematic view of example components of Fig. 1; Fig. 3. A schematic view of example components of Fig. 1; Fig. 4A, Fig. 4B Schematic views of vehicle movement between connected groups of the system of Fig. 1; Fig. 5 a schematic view of example components of Fig. 1; Fig. 6. A flowchart of an exemplary sequence of operations for a cloud-based scalable positioning system procedure; and Fig. 7. A flowchart of an exemplary sequence of operations for a latency concealment procedure.

[0017] Throughout the drawings, corresponding reference symbols denote corresponding parts. DETAILED DESCRIPTION

[0018] With reference to Fig. In certain implementations, System 100 comprises several vehicles 10, 10a-c, forming a group 20 that communicate with each other and with a base station 30 via a network 40. Additionally, System 100 includes a remote System 60 that communicates with the group 20 of vehicles 10 and / or the base station 30 via the network 40. The network 40 may include a local wireless network (WLAN) that enables communication and interoperability among the vehicles 10, the base station 30, and the remote System 60 within the vicinity of the vehicles 10. Thus, the network can include 40 wireless fidelity (Wi-Fi) (e.g., IEEE 802.11), low-rate personal networks (e.g., IEEE 802.15.4), worldwide interoperability for microwave access (WiMAX), 3G, 4G, long-range development (LTE), 5G, a digital subscriber line (DSL), Bluetooth, near field communication (NFC), or other wireless standards or Ethernet (e.g., IEEE 802.3).The vehicles 10 and / or the base station 30 may additionally contain one or more access points (APs) (not shown) configured to enable wireless communication between the vehicles 10, the base station 30 and / or the remote system 60.

[0019] As shown, the vehicles 10 in group 20 and / or the remote system 60 operate a connected group system 200 ( Fig. 2) is configured to map spatial and temporal data, define a graph-based optimization problem, and compute the position of each vehicle 10 in the group 20, while mitigating the effect of network latency and unsynchronized data. In short, and as described in more detail below, conventional positioning systems estimate a vehicle's position by solving computational graphs (e.g., a factor graph, a position graph, etc.) using sensor measurements from a single vehicle 10 in general. In these positioning systems, the computational graphs are used to identify the position of the individual vehicle 10 using the individual sensor data collected by the vehicle 10. Additionally, these positioning systems may receive disordered sensor data, which requires continuous updates to the factor graph that can delay the position estimates.

[0020] In contrast, the system of 200 connected groups, executed by the group of 20 vehicles 10, is configured to receive sensor data 202 from the multiple vehicles 10, identify which of the multiple vehicles 10 are in close proximity to each other, and group the vehicles 10 together to utilize vehicle-to-vehicle (V2V) measurements, which accurately map the one or more vehicles 10 in relation to each other. Here, due to the inherent mobility of vehicles 10, the system of 200 connected groups dynamically identifies groups 20 of vehicles 10 that are in close proximity to each other at any given time, in order to improve the flexibility and scalability of the connected group system 200 while limiting the computational complexity of increasingly connected vehicles 10.For example, the system of 200 connected groups generates position estimates 332 for each vehicle 10 in group 20, which are to be used in downstream applications of vehicle 10 as well as by the other vehicles 10 in group 20. Advantageously, the system of 200 connected groups effectively employs batch and parallel processing by executing a front-end system 210, which provides fast response times for time-critical data 202 for the position estimate 332, and a separate rear-end system 220, which performs periodic background processing of disordered and / or delayed sensor data 202 to ensure the accuracy of the position estimate 332.Here, the front-end system 210 can generate / provide an initial factor graph for the group of 20 vehicles and update the initial factor graph with a background factor graph generated by the rear-end system 220 after it has completed its periodic background processing. As described in more detail below, the front-end system 210 and the rear-end system 220 are each configured to perform a graph recalculation of their respective factor graphs. In this graph recalculation process, each respective occlusion model 300a, 300b can update the perceptions (i.e., the probabilities or estimates) associated with each variable (i.e., location estimate 322) in the factor graph.This graph recalculation / update is performed after changes to the factor nodes or variable nodes of the factor graph, or when new sensor data 202 are available.

[0021] In the example shown, the system of 200 connected groups is implemented in the vehicles 10a-10c. However, the system of 200 connected groups can be implemented in other propulsion systems such as motorcycles, trucks, all-terrain vehicles, agricultural equipment, trains, aircraft, and the like. Each vehicle 10a-10c contains a respective data processing hardware 12a-12c and a respective storage hardware 14a-14c that stores instructions which, when executed in the data processing hardware 12, cause the data processing hardware 12 to perform operations. Each vehicle 10a-10c also contains one or more respective sensors 16a-16c configured to acquire / receive sensor data 202.The one or more sensors 16 can include one or more long-range radar sensors, camera sensors capable of capturing image data, global positioning systems (GPS), speedometers, odometers, accelerometers, wireless distance measurement systems, inertial measurement units (IMUs), etc. The sensor data 202 can include the dynamics of the vehicle 10, such as speed, yaw, and acceleration, as well as wireless measurements such as time-of-flight (TOF), angle of incidence (AoA), etc., and can be transmitted to the system of 200 connected groups at 10 Hertz (Hz) via the network 40 (i.e., 5G wireless transmission).

[0022] The remote system 60 (e.g., a server, a cloud computing environment) also contains data processing hardware 62 and storage hardware 64, which stores instructions that, when executed on the data processing hardware 62, cause the data processing hardware 62 to perform operations. In certain examples, the execution of the connected group system 200 is shared across group 20 of vehicles 10 and the remote system 60. In other examples, the remote system 60 runs the connected group system 200, with the remote system 60 acting as a central host / controller. In additional examples, the connected group system 200 runs on one or more of the vehicles 10 in group 20 (i.e., shared across multiple vehicles 10).

[0023] As in Fig. As shown in Figure 1, the system of 200 connected groups receives as input the respective sensor data 202 for each of the vehicles 10a-10c. The sensor data 202 can contain spatial data 204 and corresponding temporal data 206 (e.g., timestamps). Based on the relative proximity of the vehicles 10a-10c, the system of 200 connected groups can group the vehicles 10a-10c into an initial group 20. Here, the system of 200 connected groups can effectively use several types of sensor data 202 for each vehicle 10 in the group 20.For example, the respective sensor data 202 for each vehicle 10 may include original sensor data 202 detected by the sensor system 16 of the respective vehicle 10, and / or next sensor data 202 detected by the respective sensor systems 16 of the other vehicles 10 in the group 20 of a vehicle 10, and / or base sensor data 202 detected by the base station 30 in communication with the vehicle 10.

[0024] After receiving the sensor data 202, a system 210 of the front end of the connected group system 200 filters the sensor data 202 by categorizing the sensor data 202 as either ordered sensor data 202I or unordered sensor data 202O. Here, the system 210 of the front end can process the ordered sensor data 202I using a respective latency covert model 300a ( Fig. 3) immediately process to generate a fast response location estimate 332 for each vehicle 10 in the group of 20 vehicles 10a-10c. In parallel, the rear-end system 220 receives the unordered sensor data 202O and packs the unordered sensor data 202O into sets 242 to avoid redundant updates to the location estimate 332 for each vehicle 10 in the group 20. After each predetermined time threshold (e.g., every ten (10) time cycles) has elapsed, the rear-end system 220 processes the batch 242 of unordered sensor data 202O using a respective latency cover model 300b ( Fig. 3) to generate a high-throughput location estimate 332 for each vehicle 10 in group 20 of vehicles 10a-10c. In certain implementations, the system of 200 connected groups can replace the fast-response location estimate 332 generated by the front-end system 210 with the location estimate 332 generated by the rear-end system 220.

[0025] With reference to Fig. 2 and Fig. 3. The front-end system 210 can contain a data filter module 230 and the latency cover model 300a, while the rear-end system 220 contains a stack module 240 and the latency cover model 300b. The data filter module 230 is configured to receive the sensor data 202 for each vehicle 10 in group 20 and to categorize the sensor data 202 as either ordered sensor data 202I or unordered sensor data 202O. For example, the data filter module 230 can categorize the sensor data 202 based on their corresponding temporal data 206, which specifies the timestamp of the corresponding sensor data 202. In these cases, the temporal data 206 can indicate whether the corresponding sensor data 202 are still synchronized with the current sensor data 202 being processed by the system 200 of connected groups, and / or a common understanding of a time of the connected group system 200.In certain implementations, the data filter module 230 can identify high-priority sensor data 202 (e.g., safety-critical sensor data 202) as ordered sensor data 202I to ensure that the higher-priority sensor data 202 is processed by the latency cover model 300a, which is executed by the front-end system 210. Conversely, the data filter module 230 can categorize time-consuming sensor data 202 as unordered sensor data 202O and output the unordered sensor data 202O to the back-end system 220, which executes the latency cover model 300b, using background batch processing to maintain the fast response time of the latency cover model 300a of the front-end system 210.In certain cases, the data filter module 230 identifies sensor data 202 as unordered sensor data 202O if an upload time for the sensor data 202 into the system of 200 connected groups exceeds a threshold (e.g. 0.2 seconds).

[0026] The sensor data 202, identified as unordered sensor data 202O, can then be received as input into the stacking module 240, which can maintain / hold incoming unordered sensor data 202O in stacks 242 for processing by the latency concealment model 300b. Here, the stacking module 240 is configured to receive the unordered sensor data 202O from the data filtering module 230 and add the unordered sensor data 202O to a stack 242 of unordered sensor data 202O, and simply trigger / initiate the execution of the latency concealment model 300b to process the stack 242 of unordered sensor data 202O at a predefined time threshold. For example, the latency concealment model 300b can process stacks 242 of unordered sensor data 202O only every N seconds, where N is any number of seconds, such as... B. can contain two (2), five (5), ten (10), etc.

[0027] With reference to Fig. Figure 3 shows the latency cover model 300. As should be recognized, while the system of 200 connected groups executes the latency cover models 300a and 300b in the front-end system 210 and the back-end system 220, respectively, the underlying architecture of each of the latency cover models 300a and 300b is the same. Accordingly, the latency cover model 300 of Fig. 3 both the latency cover model 300a, which is executed by the system 210 of the front end, and the latency cover model 300b, which is executed by the system 220 of the rear end.

[0028] As shown, the latency cover model 300 includes an embedding module 310, a motion prediction model 320, and a factor graph module 330. Additionally, the latency cover model 300 has access to an embedded state data store 340, located in the respective memory hardware 14 of one or more of the vehicles 10 and / or the memory hardware 64 of the remote system 60. The embedding module 310 is configured to receive sensor data 202 (e.g., the ordered sensor data 2021 and / or the unordered sensor data 2020) as input and to generate an embedded state 312 of the corresponding sensor data 202 as output. For example, the generated embedded state 312, which is output at each time step, can be stored in the embedded state data store 340. Here, the sensor data 202 contains the spatial data 204 and the corresponding temporal data 206.In particular, because the latency covert model 300 may receive incomplete sensor data 202 (i.e., one or more elements of the received sensor data 202 are missing and / or delayed), the latency covert model 300 is configured to effectively use the historical embedded states 312 stored in the embedded states data memory 340 to predict a location of the vehicle 10 based on the historical embedded states 312 for any given time, which can be used to generate a factor graph when an embedded state 312 is missing / disregarded for one or more of the elements of the sensor data 202.

[0029] In particular, the motion prediction model 320 of the latency occlusion model 300 is configured to receive as input the previous embedded states 312 of vehicle 10 and / or the previous embedded states 312 of the other vehicles 10 in the group 20, and to generate as output a position prediction 322 of vehicle 10. Here, the motion prediction model 320 uses the historical embedded states 312 of the vehicles 10 in the group 20 to derive the best position estimate of a vehicle 10. The factor graph module 330 then receives as input the position prediction 322 of vehicle 10 and the embedded states 312 of vehicle 10 and generates a factor graph for a location estimate 332 of vehicles 10 in group 20. Here, the factor graph module 330 generates the factor graph using the position prediction 322 and the embedded states 312 of vehicle 10 and / or the other vehicles 10 in group 20.In certain cases, the factor graph module 330 links the embedded state 312 with the position prediction 322 when each node of the factor graph is generated.

[0030] In certain cases, the factor graph module 330 identifies that one or more elements of the sensor data 202 are missing, based on its corresponding temporal data 206. In these cases, where a vehicle 10 has missing and / or delayed sensor data 202, the factor graph module 330 can use the position prediction 322, which corresponds to the temporal data 206 of the missing and / or delayed sensor data 202, to populate the factor graph. Furthermore, it can minimize any error in the factor graph using any embedded states 312 of sensor data 202 from vehicle 10, collected by the other vehicles 10 in the group 20. When the missing and / or delayed sensor data 202 arrive, the factor graph module 330 can update the historical factor graph of the embedded states 312 so that the next position calculation is as accurate as possible.

[0031] When the factor graph is generated using the embedded states 312 and the position prediction 322, the factor graph module 330 performs factor graph optimization by mapping the spatial data 204 and the temporal data 206 of the sensor data 202 to predict the position estimate 332 of each of the vehicles 10 in the group 20. Afterwards, the position estimate 332 and the factor graph can be stored in the embedded states data memory 340 for future predictions by the motion prediction model 320 and / or updates by delayed sensor data 202.

[0032] With renewed reference to Fig. The latency cover model 300a of system 210 at the front end receives as input the ordered sensor data 202I, which has been classified by the data filter module 230, and generates as output the position estimate 332 for each of the vehicles 10 in the group 20. In parallel, during each time threshold period, the stack module 240 of system 220 at the rear end of the stack 242 feeds unordered sensor data 202O into the latency cover model 300b. The latency cover model 300b then generates as output its own position estimate 332 for each of the multiple vehicles 10 in the group 20.The front-end system 210 can receive the position estimate 332 generated by the latency occlusion model 300b of the rear-end system 220, and can update / recalculate the current factor graph and perform a factor graph optimization to generate an updated / recalculated position estimate 332 for each vehicle 10 of the multiple vehicles 10 in the group 20.

[0033] With reference to Fig. 4A, Fig. 4B and Fig. 5. Based on the mobility of vehicles 10, the system of 200 connected groups dynamically adjusts and / or reforms the composition of vehicles 10 in a specific group 20. For example, as in Fig. 4A and Fig. As shown in Figure 4B, environments 400a and 400b contain a first group 20a and a second group 20b. Furthermore, as shown in Fig. As shown in Figure 5, each group 20a, 20b has a corresponding system 210a, 210b of the front end and a corresponding system 220a, 220b of the rear end (220a is not shown). In the vicinity of 400a (i.e. Fig. 4A) The system identifies 200 connected groups as the first group 20a, which contains three (3) vehicles 10, vehicle 10a, vehicle 10b and vehicle 10c, and the second group 20b, which contains three (3) vehicles 10, vehicle 10d, vehicle 10e and vehicle 10f. However, as in Fig. As shown in Figure 4B, the environment 400b has been modified such that vehicle 10a is now in closer proximity to vehicles 10d-10f. Here, the system of 200 connected groups can update the first group 20a to remove vehicle 10a, leaving only two (2) vehicles 10b and 10c. Additionally, the system of 200 connected groups updates the second group 20b to include vehicle 10a, which is now in close proximity to the three (3) vehicles 10d-10f, so that the second group 20b now contains four (4) vehicles 10a and 10d-10f.

[0034] As in Fig. As shown in Figure 5, after the system of 200 connected groups has identified the updated first group 20a and the updated second group 20b, the system of 200 connected groups extracts the sensor data 202a and the embedded states 312a corresponding to the vehicle 10a that has migrated from the first group 20a to the second group 20b from the front-end system A 210a to the rear-end system B 220b. Here, the rear-end system B 220b can perform a graph reconstruction (i.e., using its respective occlusion latency model 300b) to include the vehicle 10a in the factor graph and the resulting position estimate 332 for the vehicles 10 in the second group 20b.Afterwards, the rear end system B 220b can provide as input to the front end system B 210b the resulting position estimate 332 and the factor graph containing the additional vehicle 10a that has joined group 20a, with the front end system B 210b further updating the position estimate 332 for each vehicle 10a, 10d-10f in group 20b.

[0035] Fig. Section 6 contains a flowchart of an exemplary sequence of operations for a Procedure 600 for cloud-based scalable positioning systems. The Procedure 600 can be described with reference to Fig. 1- Fig. 5 will be described. Data processing hardware (e.g., the data processing hardware 12a-12c, 62 of Fig. 1) can execute commands directed to memory hardware (e.g., memory hardware 14a-14c, 64 of Fig. 1) are stored to perform the exemplary sequence of operations for procedure 600.

[0036] In Operation 602, the procedure 600 comprises identifying several vehicles 10, 10a-c as a first group 20 of vehicles 10. The first group 20 of vehicles 10 contains a front-end system 210 and a rear-end system 220, each of the front-end system 210 of the first group 20 of vehicles 10 and the rear-end system 220 of the first group 20 of vehicles 10 executing a respective latency concealment model 300a, 300b. In Operation 604, the procedure 600 comprises receiving from each vehicle 10 of the first group 20 of vehicles 10 respective sensor data 202, collected by a sensor system 16 of the vehicle 10.

[0037] In operation 606, the procedure 600 further includes categorizing the sensor data 202 as either ordered sensor data 202I or unordered sensor data 202O. In operation 608, the procedure 600 includes processing the ordered sensor data 202I using the system 210 of the front end of the first group 20 of vehicles 10, which executes the respective latency occlusion model 300a. In operation 610, the procedure 600 further includes processing the unordered sensor data 202O using the system 220 of the rear end of the first group 20 of vehicles 10, which executes the respective latency model 300b. In operation 612, the procedure 600 further includes generating a position estimate 332 for each vehicle 10 in the first group 20 of vehicles 10.

[0038] Fig. Section 7 contains a flowchart of an exemplary sequence of operations for a latency masking procedure 700. The procedure 700 can be described with reference to Fig. 1- Fig. 5 will be described. A data processing hardware (e.g., the data processing hardware 12a-12c, 62 of Fig. 1) can execute instructions stored in memory hardware (e.g., memory hardware 14a-14c, 64 of Fig. 1) are stored to perform the exemplary sequence of operations for procedure 700.

[0039] In Operation 702, Procedure 700 includes receiving sensor data 202 for a vehicle 10. Here, the sensor data 202 contains spatial data 204 and temporal data 206. In Operation 704, Procedure 700 also includes receiving a position prediction 222 of the vehicle 10. Furthermore, in Operation 706, Procedure 700 includes generating a factor graph for a position estimate 332 of the vehicle 10. In Operation 708, Procedure 700 also includes performing a factor graph optimization by matching the spatial data 204 and the temporal data 206 of the sensor data 202 to predict the position estimate 332 of the vehicle 10.

Claims

[1] Computer-implemented method which, when executed in data processing hardware (12a-12c, 62), causes the data processing hardware (12a-12c, 62) to perform operations which include: Identifying multiple vehicles (10) as a first vehicle group (20a), wherein the first vehicle group (20a) has a front end and a rear end, and each of the front end of the first vehicle group (20a) and the rear end of the first vehicle group (20a) executes a respective latency concealment model (300a, 300b); Receiving sensor data from each vehicle of the first vehicle group (20a) collected by a sensor system (16) of the vehicle (10); Categorizing sensor data as one of ordered and disordered; Processing the ordered sensor data using the front end of the first vehicle group (20a) which executes the respective latency occlusion model (300a, 300b); Processing the disordered sensor data using the rear end of the first vehicle group (20a), which executes the respective latency occlusion model (300a, 300b); and Generating a location estimate for each vehicle (10) in the first vehicle group (20a). [2] Computer-implemented method according to claim 1, wherein categorizing the sensor data as one of ordered and unordered comprises adding the unordered sensor data to a stack of unordered sensor data. [3] Computer-implemented method according to claim 2, wherein the processing of the disordered sensor data comprises processing the stack of disordered sensor data to a predetermined time threshold. [4] Computer-implemented method according to claim 1, wherein each of the latency occlusion models (300a, 300b) is configured to receive the sensor data as input and to generate the position estimate of each vehicle of the first vehicle group (20a) as output. [5] Computer-implemented method according to claim 4, wherein each of the latency occlusion models (300a, 300b) executes a respective position prediction model. [6] Computer-implemented method according to claim 5, wherein the sensor data for each vehicle (10) of the first vehicle group (20a) comprise the following: original sensor data detected by a sensor system (16) of the respective vehicle (10); and / or next sensor data detected by further vehicles (10) in the first vehicle group (20a); and / or Basic sensor data detected by a base station in communication with the first vehicle group (20a). [7] Computer-implemented method according to claim 1, wherein the operations further comprise receiving a position prediction of each vehicle (10) of the first vehicle group (20a). [8] Computer-implemented method according to claim 7, wherein the operations further comprise: Identifying one or more of the sensor data elements as missing based on the corresponding temporal data of the sensor data and Inserting the received position prediction of the vehicle (10) into a factor graph of the first vehicle group (20a) based on the corresponding temporal data of the identified missing sensor data. [9] Computer-implemented method according to claim 1, wherein the operations further comprise: Identify that a first vehicle (10) of the first vehicle group (20a) has joined a second vehicle group (20b); and Extracting the sensor data associated with the first vehicle from the front end of the first vehicle group (20a). [10] Computer-implemented method according to claim 9, wherein the operations further comprise reconstructing, using the extracted sensor data associated with the first vehicle (10), from the front end of the first vehicle group (20a) of a factor graph of the second vehicle group (20b) to accommodate the first vehicle (10).

Images