Method and program for determining the actual location of an observed object.

Abstract

To provide a technology capable of improving the accuracy of reproducing traffic conditions. [Solution] A position determination method for observing a relatively moving object using observation equipment, comprising the steps of acquiring perceived positions observed by individual observation equipment, acquiring reliability for each of the observation equipment, and determining the actual position of the object by weighting and integrating each acquired perceived position by its respective reliability.

Description

[Technical Field]

[0001] This invention relates to a method and program for determining the actual location of an object being observed. [Background technology]

[0002] In recent years, with the spread of autonomous driving technology, technologies have been developed to simulate and generate traffic conditions and other factors. For example, in the technology described in Patent Document 1, a range is given to parameters such as the number of vehicles appearing per unit time, speed, and distance between vehicles in the traffic flow model, and random behavior of the vehicles within the parameter range is simulated to generate a random traffic flow. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2024-008144 [Overview of the project] [Problems that the invention aims to solve]

[0004] The technology described in Patent Document 1 allows for the verification of various functions and performance of autonomous driving in a simulated traffic flow generated through simulation. However, the technology described in Patent Document 1 does not consider simulations that reproduce actual traffic conditions, and there is room for improvement in terms of the accuracy of reproduction of actual traffic conditions.

[0005] The present invention aims to provide a technology that can improve the accuracy of reproducing traffic conditions. [Means for solving the problem]

[0006] One aspect of this disclosure is a method for determining the position of an object that is moving relative to it, by observing it with observation instruments, comprising the steps of: obtaining a perceived position observed by each observation instrument; obtaining a confidence level for each observation instrument; and determining the actual position of the object by weighting each obtained perceived position by each confidence level and integrating them.

[0007] Another aspect of this disclosure is a location-determining program for observing a relatively moving object using observation instruments, which causes a computer to perform the steps of: obtaining a perceived position observed by each observation instrument; obtaining a confidence level for each observation instrument; and determining the actual position of the object by weighting and integrating each obtained perceived position by its respective confidence level. [Effects of the Invention]

[0008] According to the present invention, it is possible to provide a technology that can improve the accuracy of reproducing traffic conditions. [Brief explanation of the drawing]

[0009] [Figure 1] This figure shows an example of the system configuration of the location identification system 1 according to the embodiment of this disclosure. [Figure 2] This figure shows an overview of the location identification process according to this embodiment. [Figure 3] This is a functional block diagram of the location tracking server 100 according to this embodiment. [Figure 4] This figure shows an example of the hardware configuration of the location tracking server 100 according to this embodiment. [Figure 5] This flowchart shows an example of the location identification process according to this embodiment. [Figure 6A] This diagram illustrates the correspondence between the observed object 300 and the observed object according to this embodiment. [Figure 6B] A flowchart illustrating an example of the assignment process. [Figure 7] This diagram illustrates the integration of cognitive results through weighted averaging. [Figure 8] It is a schematic diagram for explaining a method of calculating reliability when introducing the observation device 200. [Figure 9] It is a schematic diagram for explaining a method of calculating reliability when there are three or more observation devices 200. [Figure 10] It is a flowchart for explaining a process of updating reliability. [Figure 11A] It is a schematic diagram for explaining a reliability update process in the present embodiment. [Figure 11B] It is a schematic diagram for explaining a reliability update process in the present embodiment. [Figure 12] It is a diagram showing a simulation result by a position identification process according to an embodiment of the present disclosure. [Figure 13] It is a diagram for explaining simulation conditions.

Embodiments for Carrying Out the Invention

[0010] Embodiments of the present invention will be described with reference to the accompanying drawings. In each figure, components denoted by the same reference numerals are assumed to represent the same or similar components.

[0011] <System Configuration><000009IV>FIG. 1 is a diagram showing an example of a system configuration of a position identification system 1 according to an embodiment of the present disclosure. The position identification system 1 according to the present embodiment includes a position identification server 100 and a plurality of observation devices 200.

[0012] As will be described later, the position identification server 100 is configured to identify the actual position of the observation target based on data regarding the position of the observation target 300 recognized by the plurality of observation devices 200.

[0013] In other words, the location identification server according to this embodiment is configured to execute the location identification method according to this embodiment. The location identification method according to this embodiment is a location identification method for observing an observation target that is moving relatively using observation equipment, and is configured to perform the steps of: acquiring a perceived position observed by each observation equipment; acquiring a confidence level for each of the observation equipment; and identifying the actual position of the observation target by weighting each acquired perceived position by each of the confidence levels and integrating them.

[0014] The multiple observation instruments 200_1, 200_2, ..., 200_n (where n is a natural number) shown in Figure 1 are configured to recognize the position of the object being observed. In the following, observation instruments 200_1, 200_2, ..., 200_n may be collectively referred to as observation instrument 200.

[0015] In this embodiment, the observation device 200 is a detection means having the function (also referred to as "recognition" in this invention) of detecting the position of an object to be observed by optical or physical means and outputting information indicating the position of the object to be observed, and includes sensors such as visible light or infrared cameras and lidar measuring instruments. It may also be a control device that combines hardware and software as detection means, such as a terminal device such as a smartphone. The observation device may be installed standalone, or it may be incorporated into other devices. For example, the observation device may be installed in a communication base station that functions as a relay for communication between a mobile terminal or the like and a core network system. For example, a sensor may be installed in the communication base station and configured to recognize the position of an object to be observed located in the vicinity of the communication base station (for example, within a radius of several tens of meters).

[0016] Furthermore, in this embodiment, the objects 300 observed by the observation device 200 (also referred to as "observation targets") may be, for example, pedestrians (pedestrians 300_P1, 300_P2, ..., 300_Pm) or vehicles (vehicles 300_V1, 300_V2, ..., 300_Vn). In this embodiment, for example, a process may be performed to identify the relative positions of observation targets 300, such as pedestrians or vehicles, with respect to the observation device 200, which is equipped with sensors on utility poles or communication base stations.

[0017] The observation device 200 may be configured, for example, by installing a sensor on a utility pole. Alternatively, the observation device 200 may be configured, for example, by installing a sensor on a vehicle. Or, a mobile terminal carried by the driver of a vehicle or a pedestrian may be used to recognize other vehicles and other pedestrians as observation targets and to recognize the positions of other vehicles and other pedestrians. Thus, in this embodiment, the observation device 200, or the driver of a vehicle or pedestrian carrying the observation device 200, may also become the observation target 300.

[0018] As shown in Figure 1, the location identification system 1 according to this embodiment may further include a database (DB) 600. The database 600 may store, for example, information such as the reliability of the observation instrument 200, which will be described later, and information regarding the location of the observation target 300 recognized by the observation instrument 200. The database 600 may also store, for example, other data used in the location identification process performed by the location identification server 100.

[0019] Furthermore, the location identification system 1 according to this embodiment may also include an autonomous driving simulation server 510 and an autonomous driving management server 520. In this embodiment, for example, the autonomous driving simulation server 510 may perform a driving simulation of an autonomous vehicle using the data of the observed object identified by the location identification server 100. Also in this embodiment, for example, the autonomous driving management server 520 may generate traffic information to be provided to the autonomous vehicle actually in motion using the data of the observed object identified by the location identification server 100.

[0020] In the position identification system 1 according to this embodiment, a system configured to simulate real-world events in a virtual space that virtually mimics the real world, such as a digital twin system, may be configured, and the autonomous driving simulation performed by the autonomous driving simulation server 510 may be performed in the digital twin system. In this case, for example, in the digital twin system, the reproduction of current and past traffic conditions in the area of ​​interest may be updated in real time, and a virtual world synchronized with the real world may be reproduced on the computer constituting the digital twin system. That is, in this embodiment, a test run of an autonomous vehicle may be performed on the digital twin based on the real position of the observation target 300 identified based on the position identification process.

[0021] Figure 2 is an illustrative diagram showing an overview of the location identification process performed by the location identification system 1 according to this embodiment.

[0022] First, the position identification server 100 obtains information regarding the position of the observation target 300, which has been recognized by the observation instrument 200, from the observation instrument 200 (S202). At this time, the position identification server 100 may perform preprocessing on the acquired data, such as noise reduction or offset removal.

[0023] As shown in Figure 2, within the target area, for example, there are observation devices 200_T1 and 200_T2 installed on utility poles and observation device 200_V1 installed on a vehicle. Furthermore, observation device 200_T1 recognizes pedestrians 300_Pt11 and 300_Pt12, and vehicle 300_Vt11; observation device 200_T2 recognizes pedestrians 300_Pt22 and 300_Pt23, and vehicle 300_Vt21; and observation device 200_V1 recognizes pedestrian 300_Pv14 and vehicle 300_Vv11. Furthermore, the vehicle 300_Vv11 recognized by the observation device 200_V1 is the observation device 200_V1 itself, and the observation device 200_V1 transmits its own position to the location identification server 100 as the position of vehicle 300_Vv11. In Figure 2, the direction of movement of pedestrian 300_P is indicated by an arrow.

[0024] Next, the location identification server 100 associates the observed objects (S204). In this embodiment, for example, among the pedestrians 300_P and vehicles 300_V recognized by the observation devices 200_T1, 200_T2, and 200_V1 within the target range, observed objects 300 that can be identified as the same are associated with each other. In this embodiment, the association of objects is performed based on the similarity of the trajectories of two objects moving relatively. For example, as will be explained later, the observed objects 300 may be associated with each other by calculating a score regarding the relationship between the observed objects 300 based on the trajectories of the recognized observed objects 300 using Dynamic Time Warping (DTW).

[0025] In the example shown in Figure 2, as indicated by the dashed rectangle in S204, for example, pedestrian 300_Pt12 is associated with pedestrian 300_Pt22. Also, vehicle 300_Vt11 is associated with 300_Vt21 and 300_Vv11. Furthermore, pedestrian 300_Pt11, pedestrian 300_Pt23, and pedestrian 300_Pv14 are not associated with any other pedestrian 300_P.

[0026] Next, the location identification server 100 integrates the associated observation targets (S206). As shown in Figure 2, for example, multiple pedestrians 300_P and vehicles 300_V that have been associated with each other as described above may each be integrated into one. That is, as shown in Figure 2, the observation data obtained by integration will show, for example, pedestrians 300_P1, 300_P2, 300_P3, and 300_P4, and vehicle 300_V1. "Integration" here refers to the process of making the system recognize multiple observation targets as a single observation target as a group.

[0027] In this embodiment, as will be described later, the results recognized by each observation instrument 200 are weighted and integrated according to the reliability of the observation instrument 200. For example, if the reliability of observation instrument 200_T1 is higher than that of observation instrument 200_T2, when integrating the corresponding pedestrians 300_Pt12 and 300_Pt22, the recognized position of pedestrian 300_Pt12 recognized by observation instrument 200_T1 is weighted more heavily than the recognized position of pedestrian 300_Pt22 recognized by observation instrument 200_T2. Therefore, in this embodiment, the integration is performed in such a way that the influence of the recognition results from the observation instrument 200 with higher reliability is greater, thereby improving the accuracy of reproduction.

[0028] As autonomous driving technology develops, for example, digital twin technology is being used to virtually recreate real-world traffic conditions. For instance, by more accurately determining the location of objects recognized by multiple observation devices such as autonomous vehicles and road infrastructure (utility poles and base stations), it can be used to understand and recreate situations in emergencies. Therefore, this embodiment can also be used in a virtual test drive method for autonomous vehicles, in which a test drive of the autonomous vehicle is performed on a digital twin based on the actual location of the observation target 300 identified based on the location identification process.

[0029] To improve the accuracy of traffic condition reproduction, one might consider increasing the number of observation devices and obtaining the final integrated result based on more perception data. However, since individual noise may be superimposed on the observation results from each observation device, it is not necessarily true that increasing the number of observation devices will improve reproduction accuracy. For example, if mobile devices carried by drivers of ordinary vehicles are used as observation devices, the variability in the performance of each observation device will be large, making it difficult to improve reproduction accuracy. Also, for example, if there are obstacles within the area of ​​interest that may hinder accurate observation, and these obstacles move dynamically, the observation results perceived by each observation device may be affected by the obstacles, potentially leading to more noise in the perceived results.

[0030] In this embodiment, for each observation device, a reliability score is calculated according to the location of the observation device, the location of the recognized observation target, the time, etc., and the calculated reliability score is updated. Then, the observation results are integrated by calculating a weighted average value according to the reliability score. This weighted average makes it possible to obtain integrated results that reduce the influence of less accurate observation devices. Therefore, the accuracy of reproducing traffic conditions can be improved.

[0031] <Functional Block Configuration> The location identification server 100 according to this embodiment will be described with reference to Figure 3. Figure 3 is an example of a functional block diagram of the location identification server 100 according to this embodiment. The location identification server 100 includes an observation data acquisition unit 110, a data preprocessing unit 120, a grouping unit 130, a confidence information acquisition unit 140, a recognition result integration unit 150, an integrated result output unit 160, a confidence calculation unit 170, and a confidence update unit 180.

[0032] The observation data acquisition unit 110 is configured to acquire information regarding the recognized location of observation targets such as pedestrians and vehicles from the observation equipment 200 (for example, observation equipment 200_1 and observation equipment 200_2).

[0033] The data preprocessing unit 120 is configured to perform preprocessing such as noise reduction and offset removal on the observation data acquired by the observation data acquisition unit 110.

[0034] The grouping unit 130 is configured to associate the observed objects 300 with each other based on the observed data that has been preprocessed by the data preprocessing unit 120.

[0035] The reliability information acquisition unit 140 acquires reliability information for each observation instrument 200, which is stored in a database 600 located outside the location identification server 100.

[0036] The cognitive result integration unit 150 integrates cognitive results by weighting the cognitive positions of the observed objects 300, which have been associated with each other by the grouping unit 130, according to the confidence level for each observed object 300 obtained by the confidence level information acquisition unit 140.

[0037] The integrated result output unit 160 outputs the integrated result of the cognitive position of each observed object 300 that has been integrated by the cognitive result integration unit 150.

[0038] The reliability calculation unit 170 is configured to calculate the reliability of the observation instruments 200. For example, if reliability information is not stored in the database 600 for some of the observation instruments 200, the reliability calculation unit 170 may calculate the reliability for them, as described later.

[0039] The reliability update unit 180 may be configured to update the new reliability calculated by the reliability calculation unit 170. The accuracy of the recognition results by the observation instrument 200 may change dynamically depending on the presence of obstacles that hinder observation. Therefore, the reliability of the observation instrument 200 may be configured to be updated according to the passage of a predetermined period. Furthermore, not only the reliability calculated by the reliability calculation unit 170, but also the reliability acquired from the database 600 by the reliability information acquisition unit 140 may be updated by the reliability update unit 180 according to the passage of a predetermined period.

[0040] The cognitive result integration unit 150 may perform integration processing of cognitive results from each observation instrument 200 based on the confidence level calculated by the confidence level calculation unit 170, or it may perform integration processing of cognitive results based on the confidence level updated by the confidence level update unit 180.

[0041] The location identification server 100 may further include a storage unit (not shown) for storing, for example, a program executed by the location identification server 100. The storage unit may store, for example, information on the reliability of the observation instrument 200 described above.

[0042] <Hardware Configuration> Figure 4 shows an example of the hardware configuration of the location tracking server 100 according to this embodiment.

[0043] The location-determining server 100 includes a processor 11 such as a CPU (Central Processing Unit) and a GPU (Graphical Processing Unit), a storage device 12 such as memory, an HDD (Hard Disk Drive) and / or an SSD (Solid State Drive), a communication interface 13 for wired or wireless communication, an input device 14 for receiving input operations, and an output device 15 for outputting information. The input device 14 is, for example, a keyboard, a touch panel, a mouse and / or a microphone. The output device 15 is, for example, a display, a touch panel and / or a speaker.

[0044] In this embodiment, the location identification server 100 may consist of one or more physical computers or servers as needed, or it may be configured using a virtual server operating on a hypervisor.

[0045] The entire program or a part thereof, which is executed by the location identification server 100 according to this embodiment, for the process of identifying the actual location of the observed object, may be stored and provided on a computer-readable storage medium such as the storage device 12. Alternatively, the entire program or a part thereof may be provided from outside the location identification server 100 via a communication network to which the location identification server 100 is connected. In the location identification server 100, for example, the processor 11 executes the location identification program according to this embodiment, thereby realizing various operations described later with reference to Figure 5, etc.

[0046] For example, the storage unit of the location identification server 100 described above can be implemented using the storage device 12 provided by the location identification server 100. Furthermore, the observation data acquisition unit 110, the data preprocessing unit 120, the grouping unit 130, the confidence information acquisition unit 140, the recognition result integration unit 150, the integrated result output unit 160, the confidence calculation unit 170, and the confidence update unit 180 can be implemented by the processor 11 of the location identification server 100 executing a program stored in the storage device 12. This program can be stored in a storage medium as described above, and the storage medium storing the program may be a non-transitory computer-readable medium. The non-transitory storage medium is not particularly limited, but may be, for example, a USB memory stick or a CD-ROM.

[0047] These physical configurations are illustrative and do not necessarily have to be independent of each other. For example, the location server 100 according to this embodiment may have a Large-Scale Integration (LSI) that integrates the processor 11 and the storage device 12. Also, as mentioned above, the location server 100 may have a GPU as the processor 11, in which case the GPU executes the above program, thereby realizing various operations described later with reference to Figure 5, etc.

[0048] The location tracking server 100 is not limited to the configuration described above. For example, some functions of the location tracking server 100 may be performed by other information processing devices or servers. Furthermore, the location tracking server 100 may be configured using a cloud server.

[0049] Furthermore, for each database (for example, the database 600 in which confidence information is stored), some or all of the information and data stored in these databases may also be stored in a storage unit provided in the location identification server 100.

[0050] <Processing Procedure> Referring to Figure 5, the location identification process performed by the location identification server 100 according to this embodiment will be explained. Figure 5 is a flowchart showing an example of the location identification process performed by the location identification server 100.

[0051] First, in step S502, the observation data acquisition unit 110 acquires observation data, such as the recognition result of the position of the observation target 300, from the observation instruments 200. For example, if there are multiple observation instruments 200 within the area of ​​interest, observation data from all observation instruments 200 within the area of ​​interest may be acquired. Alternatively, observation data may be acquired from some of the multiple observation instruments 200 within the area of ​​interest. For example, if there are a relatively large number of observation instruments 200 within the area of ​​interest, observation data from observation instruments 200 with high reliability may be selectively acquired. Alternatively, if there is variation in the density of observation instruments 200, the observation instruments 200 from which observation data is selectively acquired may be determined in order to reduce spatial bias in the observation data.

[0052] Furthermore, observational data may be acquired for all observation targets 300 recognized by each observation instrument 200, or observational data may be acquired for some of the multiple observation targets 300 recognized by each observation instrument 200. For example, observational data may be selectively acquired for observation targets 300 that are recognized as being located along roads, etc., where there is considered to be a particularly high level of interest regarding information on traffic conditions, even within the area of ​​interest.

[0053] Next, in step S504, the data preprocessing unit 120 may perform preprocessing on the observation data acquired in step S502. For example, preprocessing such as noise reduction or offset removal may be performed on the acquired observation data.

[0054] Next, in step S506, the grouping unit 130 associates the observed objects. That is, in this embodiment, one or more observation devices 200 that are observing the recognized position of the same observed object 300 are associated based on the similarity of the trajectories shown by the recognized positions of the observed objects 300 observed over time by different observation devices 200. In this embodiment, for example, the observed objects 300 may be associated with each other based on the trajectories of the recognized observed objects 300 using the Dynamic Time Warping (DTW) method.

[0055] Referring to Figure 6A, the mapping of the observation targets 300 will be explained. As shown in Figure 6A, within the example area of ​​interest, there are utility poles 200_T1 and 200_T2 on which the observation equipment 200 is installed. Furthermore, pedestrians 300_Pt11 and 300_Pt12, which are observation targets 300, are recognized by utility pole 200_T1, and pedestrian 300_Pt21 is recognized by utility pole 200_T2. In this embodiment, the dynamic time stretching method is used to determine which of the observed pedestrians 300_Pt11, 300_Pt12, and 300_Pt21 can be mapped.

[0056] As shown in Figure 6A, the observed object 300_Pt11 moves to positions a1_0, a1_1, and a1_2 at times t0, t1, and t2, respectively. Similarly, the observed objects 300_Pt12 and 300_Pt21 move to positions a1_0, a1_1, and a1_2, and positions b1_0, b1_1, and b1_2, respectively, at times t0, t1, and t2. For example, at time t0, the distance between position a1_0 and position b1_0 is smaller than the distance between position a2_0 and position b1_0.

[0057] Therefore, at time t0 only, the observed object 300_Pt11 and observed object 300_Pt21 corresponding to a1_0 are closer to each other than the observed object 300_Pt12 corresponding to a2_0 and the observed object 300_Pt21 corresponding to b1_0, and can be presumed to be corresponding observed objects. However, as time transitions from t0 to t1, observed objects 300_Pt12 and 300_Pt21 move to the left in Figure 6A, while observed object 300_Pt11 moves to the right in Figure 6A. As time transitions from t1 to t2, observed objects 300_Pt12 and 300_Pt21 move further to the left in Figure 6A, while observed object 300_Pt11 moves further to the right in Figure 6A.

[0058] Based on the above, within the area of ​​interest illustrated in Figure 6A, it is suggested that at time t0, observed objects 300_Pt11 and 300_Pt21 may be associated. However, based on the trajectories transitioning to times t0, t1, and t2, it is inferred that observed object 300_Pt12 and observed object 300_Pt21 are associated.

[0059] In this embodiment, the correspondence between the observed objects 300 is determined by focusing on the difference in the trajectories of each observed object 300 and determining whether the trajectories are similar. For example, using dynamic time stretching (DTW), the correspondence between the observed objects 300 can be determined based on the similarity of their trajectories, regardless of their proximity in each time cross-section, based on the score calculated for any two observed objects 300. The score can be calculated, for example, by maximizing the total score of bipartite matching or by heuristic assignment based on a threshold.

[0060] In the example shown in Figure 6A, using the dynamic time stretching method, for example, the score corresponding to the similarity of the trajectories between observed objects 300_Pt11 and 300_Pt21 is 0.1, and the score corresponding to the similarity of the trajectories between observed objects 300_Pt12 and 300_Pt21 is 0.7. In this case, observed objects 300 with higher scores have a higher similarity of trajectories and a higher probability of being matched.

[0061] In this embodiment, by using the dynamic time stretching method, the accuracy of the correspondence between the observed objects 300 can be improved compared to using indicators that do not have time-series information, such as position or distance from the center of gravity at a given time cross-section. Furthermore, with the dynamic time stretching method, even if there are differences in the data sampling period by each observation instrument 200, the score can be calculated regardless of the variation in the number of samples. Therefore, since the score can be calculated regardless of the sampling period by the observation instrument 200, the accuracy of the correspondence between the observed objects 300 can be improved.

[0062] Furthermore, in this embodiment, the following methods may be used as heuristic assignments based on the thresholds described above. Figure 6B is a flowchart showing an example of heuristic assignment processing.

[0063] First, for a given b (for example, b1 or b2), the a with the highest score (for example, a1 or a2) whose score exceeds a predetermined threshold is tentatively assigned. Similarly, tentative assignments of a are made to other b, and tentative assignments are made to all b (S602).

[0064] Next, for b, which is the subject of the provisional allocation to be finalized, if the provisional allocation does not conflict with provisional allocations for other b, it is finalized; if there is a conflict, the provisional allocation for the one with the higher score is finalized (S604).

[0065] Next, exclude a and b which are included in the confirmed allocation (S606).

[0066] Next, it is determined whether the assignment has been completed for all a and b (S608). If it is determined that the assignment has been completed for all a and b (YES in S608), the process ends.

[0067] As described above, if it is determined that the assignment of all a and b has not been completed (e.g., NO in S608), it is determined whether the score exceeds a predetermined threshold (S610). If it is determined that the score exceeds a predetermined threshold (YES in S610), the a with the highest score is provisionally assigned to the b (e.g., b1 or b2) (S602), and the same process (S604-S610) is repeated thereafter.

[0068] The process is repeated for all a and b until the assignment is complete or there are no more combinations whose scores exceed the threshold. Once the assignment is complete for all a and b (S608) or there are no more combinations whose scores exceed the threshold (S610), the process terminates.

[0069] For example, let's consider a case where the scores corresponding to the similarity of trajectories between multiple observed objects are 0.4, 0.7, 0.2, and 0.8, respectively, for example, between observed objects b1 and a1, between observed objects b1 and a2, between observed objects b2 and a1, and between observed objects b2 and a2.

[0070] For example, let's set the threshold to 0.3. As described above, first, we search for matching candidates for b1, and from a1 and a2, we extract a2 with a higher score with b1 and set it as the provisional assignment (b1,a2). Similarly, for b2, we search for matching candidates, and from a1 and a2, we extract a2 with a higher score with b2 and set it as the provisional assignment (b2,a2) (S602). Next, we compare the provisional assignment (b1,a2) with the provisional assignment (b2,a2), and confirm the provisional assignment (b2,a2) with the higher score (S604). Next, we remove the confirmed assignment b2 (S606), determine if there are any observation target b whose assignment has not been completed (S608), and for observation target b1 whose assignment has not been completed, we determine if the score of the provisional assignment (b1,a1) of 0.4 exceeds the threshold (S610). The provisional assignment (b1, a1) has a score of 0.4, which exceeds the threshold of 0.3. Therefore, the assignment (b1, a1) for b1 is confirmed. The process is now terminated.

[0071] Furthermore, if the threshold is set to 0.5, as described above, first, a provisional assignment (b1, a2) is made for b1, and a provisional assignment (b2, a2) is made for b2 (S602). Next, the provisional assignment (b1, a2) and the provisional assignment (b2, a2) are compared, and the provisional assignment (b2, a2) with the higher score is confirmed (S604). After removing b2 from the confirmed assignment (S606), it is determined whether there are any observation target b whose assignment has not been completed (S608). For the observation target b1 whose assignment has not been completed, it is determined whether the score of the provisional assignment (b1, a1) of 0.4 exceeds the threshold (S610). Since the score of the provisional assignment (b1, a1) of 0.4 does not exceed the threshold of 0.5, there is no assignment for b1, and the process is terminated.

[0072] Returning to Figure 5, in step S508, the recognition results for the associated observation target 300 are integrated. In this embodiment, the results recognized by each observation instrument 200 are weighted and integrated according to the confidence level of the observation instrument 200. In this embodiment, the recognized positions of the observation target 300 acquired by the observation instrument 200 are weighted and averaged based on the confidence levels of each observation instrument 200.

[0073] Referring to Figure 7, we will explain the integration of cognitive results by weighting and averaging according to confidence level. In Figure 7, we show an example of integrating the observation target 300_Pt12 by observation instrument 200_T1 and the observation target 300_Pt21 by observation instrument 200_T2, as described above for Figure 6A. That is, at times t0, t1, and t2, the positions a2_0 and b1_0, positions a2_1 and b1_1, and positions a2_2 and b1_2 are weighted and averaged according to confidence level to obtain positions iab_0, iab_1, and iab_2.

[0074] In this embodiment, the reliability is calculated for each observation device 200 according to the time, the position of the observation device 200, and the position of the recognized observation target 300. The weighted average of the calculated reliability values ​​is then calculated to obtain an integrated recognition result.

[0075] In other words, the integrated result of the cognitive outcomes is calculated using the weighted average of formula 1 below.

[0076]

number

[0077]

number

[0078] As described above, the recognized positions of the observed objects 300 can be integrated by weighted averaging according to their reliability, making it possible to determine the actual positions of the observed objects 300. For example, as shown in Equation 2 above, the weighting coefficient for observation instruments 200 with relatively low reliability is relatively small, and the weighting coefficient for observation instruments 200 with relatively high reliability is relatively large. Therefore, by calculating a weighted averaging according to reliability as shown in Equation 1, and multiplying the position of the observed object 300 by different weighting coefficients according to the level of reliability and accumulating the results, an integrated result that takes reliability into account can be obtained. Thus, the accuracy of determining the actual position of the observed object 300 identified by the position identification server 100 can be improved.

[0079] Thus, the recognition results are integrated (S508 (Figure 5)), and the location identification process performed by the location identification server 100 according to this embodiment, as described with reference to Figure 5, is terminated.

[0080] As described above, the location determination method according to the embodiment of this disclosure is a location determination method for observing a relatively moving object using observation instruments, and comprises the steps of: acquiring a perceived position observed by each observation instrument; acquiring a confidence level for each observation instrument; and determining the actual position of the object by weighting and integrating the acquired perceived positions by their respective confidence levels. The location determination method according to this embodiment can provide, for example, a technology that can improve the accuracy of reproducing traffic conditions.

[0081] The method for calculating reliability in this embodiment will be described below. In this embodiment, for example, in the reliability calculation process for calculating the reliability of an observation instrument 200 for which reliability has not been set, a process may be performed to calculate the distance between the actual position of the observation object 300, which is identified based on the perceived position of each observation object 300, and the perceived position of the observation object 300 identified by the observation instrument 200 for which reliability has not been set, and a process to calculate reliability according to the distance.

[0082] For example, if there are already 200 observation instruments in the area of ​​interest, and the reliability of each existing instrument has been evaluated, the reliability of the new instrument 200 may be calculated when introducing a new instrument 200. Alternatively, for example, if there are no 200 observation instruments in the area of ​​interest, the reliability of the new instrument 200 may be calculated when introducing a new instrument 200.

[0083] For example, if observation equipment 200 already exists within the area of ​​interest and new observation equipment 200 is to be introduced, the reliability of the newly introduced observation equipment 200 may be calculated based on whether or not the accuracy improves by adding the observation data shown by the newly introduced observation equipment 200 to the observation data shown by the original set of observation equipment 200 and evaluating it. For example, if the accuracy of identifying an arbitrary observation target 300, which is identified by each observation data shown by the original observation equipment 200, improves by including the observation data from the newly introduced observation equipment 200, the reliability of the newly introduced observation equipment 200 may be calculated to be relatively high. Conversely, if the accuracy of identification decreases, the reliability of the newly introduced observation equipment 200 may be calculated to be relatively low.

[0084] Thus, the reliability of the newly introduced observation instrument 200 may be calculated based on the contribution of the observation data from the newly introduced observation instrument 200 to the position of the observation target 300 identified by the original set of observation instruments 200. Therefore, since the contribution of the newly introduced observation instrument 200 to the accuracy of the position identified for the observation target 300 is considered to correspond to the reliability of the newly introduced observation instrument 200, the reliability of the newly introduced observation instrument 200 may be calculated, for example, based on the method for calculating the Shapley value exemplified in Equation 3 below.

number

[0085] In this embodiment, for example, if there are no observation instruments 200 within the area of ​​interest, and a new observation instrument 200 is introduced, the reliability of the new observation instrument 200 may be calculated. The method for calculating the reliability in this case will be explained below with reference to Figure 8. Figure 8 is a schematic diagram illustrating the method for calculating the reliability when a new observation instrument 200 is introduced.

[0086] As shown in Figure 8, a new utility pole is installed within the area of ​​interest, and observation of the target object 300_Pn1 is started using observation instrument 200_Tn1. Observation instrument 200_Tn1 observes that the position of the target object 300_Pn1 moves as a0_n-2, a0_n-1, a0_n, a0_n+1, and a0_n+2 at times t=n-2, t=n-1, t=n, t=n+1, and t=n+2, respectively. In this case, the reliability of the observation instrument 200_Tn1 may be calculated based on the distance between the position of the observed object 300_Pn1 observed by the observation instrument 200_Tn1 at the time t=n and the position estimated to be where the observed object 300_Pn1 is located at time t=n, calculated based on the movement trajectory a0_n-2, a0_n-1, a0_n+1, and a0_n+2 of the observed object 300_Pn1 at time t=n-2, t=n-1, t=n+1, and t=n+2 in the observation data.

[0087] By using arbitrary polynomial interpolation for positions a0_n-2, a0_n-1, a0_n+1, and a0_n+2, the path that the observed object 300_Pn1 traveled between times t=n-2, t=n-1, t=n+1, and t=n+2 is approximated by curve cn1. As the arbitrary polynomial interpolation, known interpolation methods such as Lagrangian interpolation may be used.

[0088] Assuming that the example observation target 300_Pn shown in Figure 8 moves along a path approximated by curve cn1, it is estimated that at time t=n, it is located at position a0_n_t on curve cn1. Furthermore, based on the observation data from observation instrument 200_Tn1, observation target 300_Pn1 is located at position a0_n at time t=n. Since the curve cn1 calculated by polynomial interpolation is considered to be a path close to the actual movement trajectory of observation target 300_Pn1, as shown in Figure 8, the distance between the observation result from observation instrument 200_Tn1 and the position of observation target 300_Pn1 on the movement path at the point of interest t=n can be assumed to correspond to the observation error of observation instrument 200_Tn1.

[0089] Thus, the confidence level may be calculated based on the distance between the position at time t=n on the trajectory calculated based on the observed position at times other than the time of interest, and the position at time t=n in the observed data. For example, the confidence level may be calculated by multiplying this distance by a predetermined weighting coefficient. Alternatively, the distance between the position at time on the moving curve calculated similarly for multiple times of interest by polynomial interpolation and the observed position at the time of interest may be calculated, and the confidence level may be calculated by multiplying each distance by a predetermined weighting coefficient.

[0090] Furthermore, the confidence level calculated in this manner may be used to calculate the confidence level of the newly added observation instrument 200. For example, even if there is one observation instrument 200 in the area of ​​interest and a new observation instrument 200 is to be added, the confidence level of the observation instrument 200 calculated by the method described above may be used as the confidence level of the already existing observation instrument 200, and the confidence level of the newly added observation instrument 200 may be calculated by calculating the Shapley value shown in Equation 3 for the observation data of the newly added observation instrument 200.

[0091] Furthermore, if two observation instruments 200_k1 and 200_k2 exist within the area of ​​interest, the confidence level for the set of the two observation instruments 200_k1 and 200_k2 may be calculated using the method described below. For example, for the observation data of observation instrument 200_k1 and the observation data of observation instrument 200_k2, curves ck1 and ck2 showing the movement trajectory of the same observation target 300 may be calculated by polynomial interpolation, the amount of deviation from the calculated curves ck1 and ck2 may be calculated for each of the observation instruments 200_k1 and 200_k2, and the average value of the calculated deviation amounts may be calculated as the confidence level for the set of the two observation instruments 200_k1 and 200_k2.

[0092] Furthermore, if there are three or more observation instruments 200 within the area of ​​interest, the confidence level for the set of three or more observation instruments 200 may be calculated using the method described below. As shown in Figure 9, for example, the confidence level for the subset consisting of observation instruments 200_l1, 200_l2, and 200_l3 within the area of ​​interest is calculated. Note that there are other observation instruments 200 within the area of ​​interest.

[0093] As shown in Figure 9, for any observation target 300_l1, at any time t=n, it is determined that observation instruments 200_l1, 200_l2, and 200_l3 are located at positions a2_l1, a2_l2, and a2_l3, respectively. In this embodiment, for example, the centroid g_lp of the position of any observation target 300_l1 (positions a2_l1, a2_l2, and a2_l3) observed by each of the observation instruments 200_l1, 200_l2, and 200_l3 within the area of ​​interest is calculated, and the centroid g_la of the position of the same observation target 300_l1 observed by each of the observation instruments 200 present in the area of ​​interest is calculated. The confidence level of a subset of the observation instruments 200 (observation instruments 200_l1, 200_l2, and 200_l3) may be calculated based on the amount of deviation of the centroid g_lp of the position of the observation target 300_l1 observed by each of the observation instruments 200_l1, 200_l2, and 200_l3 relative to the centroid g_la of the position of the observation target 300_l1 observed by each of the observation instruments 200. The confidence level for a subset of 200 observation instruments consisting of four or more instruments may also be calculated using a similar method.

[0094] The confidence level for a subset of two observation instruments 200, or for a subset of three or four or more observation instruments 200, may be calculated using the method described above. The Shapley value, etc., may then be calculated using the calculated confidence levels, for example, when calculating the confidence level of a newly added observation instrument 200 when two observation instruments 200 already exist in the area of ​​interest, or when calculating the confidence level of a newly added observation instrument 200 when three or more observation instruments 200 already exist.

[0095] The Shapley value exemplified in Equation 3 is generally used to evaluate the contribution of each participant to the organization, and the greater the participant's contribution, the larger the Shapley value. In this embodiment, when evaluating the amount of deviation from the curve calculated by polynomial interpolation using the method described above, the reliability of the observation instrument 200 may be evaluated using the Shapley value calculated by the method exemplified in Equation 3, as described below, so that the reliability decreases as the amount of deviation increases and increases as the amount of deviation decreases.

[0096] For example, as illustrated in Equation 4, evaluation values ​​may be calculated by inverting the minimum and maximum Shapley values ​​of each observation instrument 200.

number

number

number

number

number

[0097] In this embodiment, the evaluation value calculated by formula 4 may be scaled and weighted so that the sum of all 200 observation instruments equals 1, as shown in formula 9 below.

number

[0098] Generally, since traffic volume can vary depending on the time of day, Shapley weights may be updated at predetermined intervals. For example, if a malfunction occurs in the electrical system within a target area that can suddenly affect all or many of the observation instruments 200 within that area, it can become a source of noise in the observation data of all or many of the affected observation instruments 200. For example, by updating the Shapley weights at predetermined intervals, such as after a predetermined period of time has elapsed, it is possible to calculate the Shapley value weights for the target observation instruments 200 while taking into account the effects of noise and other factors on the observation instruments 200 for which the reliability has already been calculated.

[0099] In other words, in this embodiment, a process is performed to update the reliability over time, and the process of updating the reliability may include: calculating a weighting coefficient for the target observation instrument 200 based on the Shapley value at a predetermined observation time; obtaining the reliability of the target observation instrument 200 at the time of the previous update; calculating the amount of reliability update based on the difference between the weighting coefficient calculated at the predetermined observation time and the reliability at the time of the previous update; and updating the reliability by adding the amount of update to the reliability at the time of the previous update.

[0100] The reliability update process in this embodiment will be described below with reference to Figures 10, 11A, and 11B. Figure 10 is a flowchart illustrating the reliability update process. Figures 11A and 11B are schematic diagrams illustrating the reliability update process in this embodiment.

[0101] Figures 11A and 11B schematically show the area of ​​interest, and are schematic diagrams of the area of ​​interest at time t=t0 and time t=t1, respectively. As shown in Figures 11A and 11B, for example, the area of ​​interest is assumed to be a 5x5 mesh, and the confidence level is updated. As shown in Figure 11A, at t=t0, the observation instrument 200_v1, which is a vehicle, is at position (2,1), and the observation target 300_p1 is at position (1,2). Also, as shown in Figure 11B, at t=t1, the observation instrument 200_v1 has moved to position (2,2), and the observation target 300_p1 has moved to position (1,3).

[0102] Figures 11A and 11B show the affected areas due to the reliability update, with the affected areas by observation instrument 200_v1 and observation target 300_p1 indicated by dashed and dotted rectangles, respectively. As shown in Figures 11A and 11B, in this embodiment, both the affected areas by observation instrument 200_v1 and observation target 300_p1 are within a 1-mesh neighborhood and a 3x3 mesh area. Furthermore, in the following explanation, Figures 11A and 11B will be used as an example when there are four observation instruments 200 within the area of ​​interest.

[0103] As shown in Figure 10, first, the Shapley weight at the current observation point is calculated (S1002). Here, for example, the Shapley weight at the observation point at t=t0 shown in Figure 11A is calculated. In this embodiment, the observation point is represented as (t,(ia,ja),(io,jo)). t, i, and j represent time, the vertical axis coordinate in Figure 11A, and the horizontal axis coordinate in Figure 11A, respectively. In Figure 11A, the observation instrument 200_v1 is located at coordinate (ia,ja)=(2,1), and the observation target 300_p1 is located at coordinate (io,jo)=(1,2). Therefore, the observation point for which the Shapley weight is calculated here is represented as (t,(ia,ja),(io,jo))=(t0,(2,1),(1,2)). In the example shown in Figure 11A, the time t=t0=0, and the Shapley weight at this observation point (t0,(2,1),(1,2)) is 0.5.

[0104] Next, observation points that meet the influence range update conditions are extracted (S1004). Here, the combination of the position of the observation instrument 200_v1 and the influence range of the recognized position of the observation target 300_p1 at the next time (t=t1) is extracted as an observation point. For example, 9 x 9 = 81 points such as (t,(ia,ja),(io,jo))=(t1,(1,0),(0,1)), (t1,(1,1),(0,1)), ..., (t1,(3,2),(2,3)) may be extracted as observation points.

[0105] Next, for each of the extracted observation points, we determine the observation point of interest (S1006). Here, for example, (t,(ia,ja),(io,jo))=(t1,(2,2),(1,3)) is determined as the observation point of interest.

[0106] Next, the database 600 is referenced to determine whether or not there is confidence in past data (S1008). For example, the database 600 is referenced to determine whether or not there is confidence in the previous time period.

[0107] If it is determined that past confidence levels are stored in database 600 (YES in S1008), the past confidence levels are retrieved (S1010). Then, the difference between the retrieved past confidence levels and the Shapley weights at the observation points calculated in S1002 is calculated (S1014).

[0108] On the other hand, if it is determined that past confidence levels are not stored in database 600 (NO in S1008), then, for example, the number of 1 / 200 observation instruments is calculated (S1012). As shown above, in the example shown in Figure 11A, there are 4 observation instruments 200, and the number of 1 / 200 observation instruments is calculated as 0.25. Subsequently, the difference between the calculated value of the number of 1 / 200 observation instruments and the Shapley weight at the observation point calculated in S1002 is calculated (S1014).

[0109] In FIG. 11A, an example where t = t0 is shown. The reliability before the current time has not been calculated, and the reliability for the past is not stored in the database 600. Therefore, since it is NO in S1008, 1 / the number of observation devices 200 = 0.25 is calculated in S1012 (S1012), and the difference between 0.25 and the sharp ray weight = 0.5 calculated in S1002 is calculated as 0.25 (S1014).

[0110] Next, a coefficient is calculated from the difference between the current observation point and the observation point of interest (S1016). For example, a coefficient is calculated from the difference between the current observation point (t0, (2, 1), (1, 2)) and the observation point of interest (t0, (2, 2), (1, 3)). At this time, for example, a coefficient may be determined such that it decreases as the total distance between the difference in the positions of the observation devices 200 (here, for example, the difference between (2, 1) and (2, 2)) and the difference in the recognized positions of the observation target 300 (here, for example, the difference between (1, 2) and (1, 3)) increases. The rule for determining the coefficient may be arbitrarily selected. For example, the coefficient = 0.9 is calculated.

[0111] The coefficient may be calculated, for example, based on the following formula 10.

[0112] [Number] Here, c is the coefficient, d agent indicates the difference in the position of the observation device 200, and d object indicates the difference in the recognized position of the observation target 300. Also, α is a coefficient for adjusting the coefficient related to the whole, and may be determined, for example, according to how much the increase and decrease between each observation point of interest are reflected. Also, d agent / β indicates an adjustment term for the influence of the difference in the position of the observation device 200, and d object / β indicates an adjustment term for the influence of the recognized position of the observation target 300.

[0113] Next, the confidence update amount is calculated (S1018). The confidence update amount may be calculated, for example, using the difference value calculated in S1014 and the coefficient value calculated in S1016, based on the formula: confidence update amount = coefficient x difference. Here, the confidence update amount is calculated as 0.9 x 0.25 = 0.225.

[0114] Next, the confidence level is updated based on the confidence update amount (S1020). The updated confidence level may be calculated based on the confidence update amount calculated in S1018 and the past confidence level obtained in S1010 (or 1 / number of observation instruments 200 if past confidence levels are not stored), using the formula: Confidence level (confidence level of the observation point of interest) = Past confidence level (or 1 / number of observation instruments 200) + Confidence update amount. Here, the confidence level of the observation point of interest (t1,(2,2),(1,3)) is calculated to be 0.475. The calculated confidence level of the observation point of interest may be stored in database 600, for example.

[0115] Next, it is determined whether the confidence level has been updated for all observation points (S1022). For example, it is determined whether the confidence level has been updated for all 9x9=81 observation points, such as (t,(ia,ja),(io,jo))=(t1,(1,0),(0,1)), (t1,(1,1),(0,1)), ..., (t1,(3,2),(2,3)). If there are any observation points whose confidence levels have not been updated, the process returns to S1006 and the confidence level update process described above is executed.

[0116] Thus, the confidence level update process at time t=t0 is completed, and then the confidence level update process at time t=t1 is executed.

[0117] As shown in Figure 11B, at time t=t1, the observation instrument 200_v1 has moved to position (ia,ja)=(2,2), and the observation target 300_p1 has moved to position (io,jo)=(1,3).

[0118] The Shapley weight is calculated to be 0.5 at the current observation point (t1,(2,2),(1,3)) (S1002).

[0119] Next, 9 x 9 = 81 points are extracted as observation points that meet the conditions for updating the range of influence, for example, (t,(ia,ja),(io,jo))=(t2,(1,0),(0,1)), (t2,(1,1),(0,1)), ..., (t2,(3,2),(2,3)) (S1004).

[0120] Next, for each of the extracted observation points, a point of interest is determined (S1006). For example, (t,(ia,ja),(io,jo))=(t2,(2,2),(1,3)) is determined as the point of interest.

[0121] Next, the database 600 is referenced to determine whether or not there is confidence in past data (S1008). At time t=t1, the confidence level for time t=t0 is obtained as 0.475 (S1010). The difference between the obtained confidence level for past data and the Shapley weight at the observation point calculated in S1002 is calculated as 0.5 - 0.475 = 0.025 (S1014).

[0122] Next, a coefficient is calculated from the difference between the current observation point and the observation point of interest (S1016), which in this case is calculated to be, for example, 0.9.

[0123] Next, the confidence update amount is calculated based on the formula: confidence update amount = coefficient x difference, resulting in confidence update amount = 0.9 x 0.025 = 0.023.

[0124] Next, the confidence level is updated based on the confidence update amount (S1020), and the updated confidence level = confidence update amount + past confidence level is calculated as 0.475 + 0.023 = 0.498. 0.498 is stored as the confidence level at the observation point of interest (t2,(2,2),(1,3)).

[0125] Next, it is determined whether the confidence level has been updated for all observation points (S1022). If it has been updated for all observation points, the confidence level update process at time t=t1 is completed, and then the confidence level update process at time t=t2 may be executed.

[0126] For example, if noise affecting many observation instruments 200 is introduced at a certain time, the confidence score calculated based on Shapley weights at that time may deviate significantly from the actual confidence score due to the influence of the noise. Also, for example, if multiple observation instruments 200 or multiple observation targets 300 are relatively close to each other at a certain time, the confidence scores calculated at that time will be similar, and in this case as well, even if there is actually a difference in confidence scores between the observation instruments 200, the difference in confidence scores may not be easily apparent. In this embodiment, for example, the confidence score can be updated by the method described above with reference to Figure 10, and since the confidence score is calculated based on observation data from the observation instruments 200 during a predetermined update period, a more reliable confidence score can be obtained compared to when the confidence score is calculated based only on observation data at a certain time.

[0127] Figure 12 is a diagram illustrating the simulation results obtained by the location determination process according to the embodiment of this disclosure. In the simulation shown in Figure 12, the comparative example simulation is the case where the observation data from the observation instrument 200 is integrated using a simple average (dashed line). In the example simulation, the observation data from the observation instrument 200 is the case where it is integrated using a confidence-weighted average (solid line).

[0128] Figure 13 illustrates the simulation conditions. As shown in Figure 13, in the simulation, position determination processing was performed based on observation data obtained by recognizing the position of the observation target 300_s using observation equipment 200 (three observation equipment 200_vs1, 200_vs2, and 200_vs3 shown in Figure 13), which is a vehicle moving at a constant velocity on the circumference Cs. The root mean square error (RMSE) of the error in the recognized position of the observation target 300 was evaluated according to the number of observation equipment 200. As shown in Figure 12, simulations were performed for cases with 2, 3, 4, 5, and 6 observation equipment 200. In the simulation, noise was periodically introduced into the recognition of each observation equipment 200. Furthermore, as shown in Figure 13, in the simulation, when an obstruction 700s was placed within the area of ​​interest, and the obstruction was on the line between the observation instrument 200 and the observation target 300 (in Figure 13, observation instrument 200_vs2), a noise with a gain of 5 was applied.

[0129] As shown in Figure 12, regardless of the number of observation instruments 200 used, the RMSE was reduced in the embodiment (solid line) compared to the comparative example (dashed line). Therefore, it was confirmed that the accuracy of the recognized position of the observation target 300 can be improved by integrating the recognition results of the observation data by taking a confidence-weighted average in the position identification process of this embodiment. It was also confirmed that the accuracy of the recognized position can be further improved by increasing the number of observation instruments 200.

[0130] Figure 12 shows the results obtained from the simulation, but in experiments conducted by the inventors in the field, it was similarly confirmed that the accuracy of the recognized location can be improved by the location identification process according to this embodiment.

[0131] <Summary> According to the embodiments described above, the method for determining the position of an object that is moving relative to it, using observation instruments, is configured to perform the steps of: obtaining the perceived position observed by each observation instrument; obtaining the confidence level for each observation instrument; and determining the actual position of the object by weighting and integrating the obtained perceived positions by their respective confidence levels.

[0132] This makes it possible to provide a technology that can improve the accuracy of traffic condition reproduction. Furthermore, the technology according to this embodiment can contribute to achieving Sustainable Development Goal (SDG) 9, "Build resilient infrastructure, promote inclusive and sustainable industrialization and foster innovation." In addition, the technology according to this embodiment can contribute to achieving Sustainable Development Goal (SDG) 11, "Make cities and human settlements inclusive, safe, resilient and sustainable."

[0133] The embodiments described above are for the purpose of facilitating understanding of the present invention and are not intended to limit its interpretation. The flowcharts, sequences, elements, and their arrangement, materials, conditions, shapes, and sizes described in the embodiments are not limited to those exemplified and can be modified as appropriate. Furthermore, the calculation methods described in the embodiments described above are not limited to those exemplified. It is also possible to partially substitute or combine the configurations shown in different embodiments. [Explanation of Symbols]

[0134] 1...Location identification system, 100...Location identification server, 110...Observation data acquisition unit, 120...Data preprocessing unit, 130...Grouping unit, 140...Confidence information acquisition unit, 150...Recognition result integration unit, 160...Integration result output unit, 170...Confidence calculation unit, 180...Confidence update unit, 200...Observation equipment, 300...Observation target, 300...Observation equipment, 510...Autonomous driving simulation server, 520...Autonomous driving management server, 600...Database

Claims

1. A method for determining the position of an object that is moving relative to it, by observing it with an observation instrument, The steps include obtaining the perceived position observed by individual observation instruments, The steps include obtaining the reliability of each of the aforementioned observation instruments, The steps include identifying the actual location of the observed object by weighting and integrating the recognized locations of the same observed object from among the acquired recognized locations according to the respective confidence levels, A method for determining location, comprising the following features.

2. A step of associating one or more of the observation devices that are observing the same object with respect to the recognition position, based on the similarity of the trajectories shown by the recognition position observed over time by different observation devices. A method for determining location according to claim 1, comprising:

3. The step of identifying the actual location includes the step of weighting and averaging the acquired perceived locations based on each of the confidence levels. The method for determining location according to claim 1.

4. A step of calculating the reliability of the observation instrument for which the reliability has not been set, A step of calculating the distance between the actual position of the observed object, which is determined based on each of the aforementioned perceived positions, and the perceived position of the observed object, which is determined by the observation instrument for which the confidence level has not been set. A method for determining location according to claim 3, further comprising the step of calculating the reliability according to the distance, the step of calculating the reliability according to the distance.

5. The step of calculating the confidence level is: The step includes calculating the confidence level for each of the aforementioned observation instruments based on the Shapley value obtained when the aforementioned recognition position is obtained by an observation instrument for which the confidence level has not been set. The method for determining location according to claim 3.

6. A step of updating the confidence level over time, A step of calculating a weighting coefficient based on the Shapley value at a predetermined observation time for the observation instrument in question, A step of obtaining the reliability of the observation instrument in question at the time of the last update, A step of calculating the amount of the reliability update based on the difference between the weighting coefficient calculated at the predetermined observation time and the reliability at the time of the previous update, The step of updating the reliability by adding the update amount to the reliability at the time of the previous update includes, The step of updating the aforementioned reliability is further included, The method for determining location according to claim 5.

7. A method for virtual test driving an autonomous vehicle, comprising performing a test drive of the autonomous vehicle on a digital twin based on the actual position of the observed object identified according to the method of any one of claims 1 to 6.

8. A position determination program for observing a relatively moving object using observation equipment, The steps include obtaining the perceived position observed by individual observation instruments, A step of obtaining the reliability for each of the aforementioned observation instruments, The steps include identifying the actual location of the observed object by weighting and integrating the recognized locations of the same observed object from among the acquired recognized locations according to the respective confidence levels, A location-finding program that causes a computer to execute a command.

Images