Method for locating a vehicle

Abstract

The present invention relates to a method for locating a first vehicle comprising an agent having access to a first piece of locating information relating to the position of the first vehicle, wherein at least one auxiliary signal carrying at least one second piece of locating information relating to the position of at least one second vehicle is provided, wherein the first and the at least one second piece of locating information are combined within the agent of the first vehicle, taking into account the spatial arrangement of the vehicles, in order to determine the position of the vehicle.

Description

[0001] R. 410249 - 1 - Description Title Method for locating a The invention provides a method for locating a vehicle and an arrangement for carrying out the method. State of the Art Vehicle localization is a process for identifying the location of a vehicle, in particular a motor vehicle, regardless of whether the vehicle is stationary or moving. Most modern vehicles are equipped with a localization system. In most cases, this localization system is based on GNSS (GNSS: Global Navigation Satellite System), but not necessarily. In the case of a driving application, the vehicle is equipped with several other sensors, such as radar, cameras, ultrasonic sensors, etc., in addition to this localization system. In general, localization systems can vary in their degree of accuracy. Therefore, there are higher-quality localization systems and lower-quality localization systems. As a rule, the higher-quality systems are more expensive.Consequently, the goal is to achieve sufficient accuracy at the lowest possible cost. Disclosure of the invention R. 410249 -. 2 -According to the invention, a method for locating a vehicle according to claim 1 is introduced. Furthermore, an arrangement according to claim 10 is introduced, which is suitable for carrying out the method. The proposed method serves to locate a first vehicle, and therefore the method serves to determine the position of this first vehicle. The first vehicle comprises an agent that has access to first location information relating to the position of the first vehicle. Within the method, at least one auxiliary signal is provided that carries at least one second item of location information relating to the position of at least one second vehicle. The first and the at least one second item of location information are combined within the agent of the first vehicle, taking into account the spatial arrangement of the vehicles, in order to determine the position of the vehicle.If there is only one second vehicle, then there is only one second piece of location information relating to the position of the one second vehicle. If there are two second vehicles, a first second vehicle and a second second vehicle, then there is one first piece of second location information relating to the position of the first second vehicle and one second piece of second location information relating to the position of the second second vehicle, and so on. The pieces of second location information are provided by at least one auxiliary signal. If there are two pieces of second location information, then there are two auxiliary signals. Within the agent of the first vehicle, the first location information is merged with the at least one auxiliary signal carrying location information from at least one other vehicle, typically a neighboring vehicle.During the merging of information, the spatial arrangement of the first vehicle and the one or more other vehicles is taken into account. R. 410249 -. 3 -This invention addresses the case where multiple vehicles share information about their localization parameters to improve localization for some of them. In one embodiment, the first agent of the first vehicle is connected to a lower-quality localization system, and the at least one second vehicle, or the second agent within the at least one second vehicle, has access to a higher-quality localization system. As a result, the position of the first vehicle can be determined with high accuracy without the first vehicle having to have a higher-quality and expensive localization system. An agent is a module, typically a software module, associated with a vehicle and capable of determining the vehicle's position using pieces of localization information.An agent that receives a signal representing localization information from a higher-quality localization system is referred to herein as a hg agent or agent. hg An agent that receives a signal representing location information from a lower-order location system is referred to herein as an lg agent or agent lgThe proposed method implements an information-sharing approach to improve the localization performance of agents by using information from other vehicles. For this purpose, an analytical method for such merging has been developed, and the required algorithms are presented herein. The approach presented enables conventional localization systems, for example, GNSS-based systems, to merge information with another auxiliary signal, for example, radar, to enable such merging of information from neighboring agents. In one embodiment, the purpose is to share information about the location of an agent that is equipped with a higher-quality localization system (agent ^^ ) to another neighboring agent equipped with a lower-order localization system R. 410249 - 4 - (Agent ^^). This allows the agent ^^Improve its self-localization. This method can be applicable to multi-agent (MA) settings. The required payloads to enable such a merging algorithm are: a localization system (GNSS), an auxiliary beacon, radar, or other range detector, and a low-bandwidth communication system to enable the sending and receiving of messages between the agents. Unlike existing approaches, this approach does not require the use of any angular information or sharing of information about angular data for this projection. The advantage of this method is to improve the location of some agents through a low-bandwidth communication system by using available information from their neighboring agents.Some conventional fusion schemes assume that many IMUs (Inertial Measurement Units) are placed on the same rigid body. In this way, they project measurements from one sensor to another, using the rigid body assumption. Other works assume geometric knowledge, in the form of point clouds (PCs) about the environment, acquired, for example, by a lidar. These PCs can be further sensed by the other agent, and can thus derive knowledge about their own position and pose. This form of information sharing requires high bandwidth, as these PCs cannot be easily shared by some conventional low-bandwidth communication systems. The described method does not assume a rigid body and does not require a high-bandwidth communication system.Additionally or alternatively, the location information of the at least one auxiliary signal can be used to verify the first location information, which is typically extracted from a signal of a location system. R. 410249 -. 5 -The arrangement described herein is suitable for carrying out the proposed method. The arrangement can be implemented in hardware and / or software. Furthermore, the arrangement can be integrated into an electronic control unit (ECU) or, in one embodiment, be an ECU. Furthermore, the arrangement can be associated with an agent within a motor vehicle or can be a component, e.g., a software component, of such an agent. Brief Description of the Drawings Figure 1 shows an example of information sharing in a driving application. Figure 2 shows an illustration of the overall merging scheme. Figure 3 shows another illustration scheme. Figure 4 shows a timeframe for a specific case. Figure 5 shows an illustration of the method according to the invention for two hg agents and one lg agent.Figure 6 shows a scenario with two vehicles, illustrating one embodiment of the proposed method. It is understood that the features mentioned above and described below can be used not only in the specified combination, but also in other combinations or alone, without departing from the scope of the invention. The invention is schematically illustrated in the drawings by means of exemplary embodiments, and is described below with reference to R. 410249. 6 -explained in detail with reference to the drawings. It is understood that the description in no way limits the scope of the present invention and that it is merely an illustration of embodiments of the invention. Description of Embodiments Figure 1 shows an example of information sharing in a driving application. The drawing shows three agents: Agent_lg 10 at t = 0, Agent_2_hg 12 at t = 0 and Agent_1_hg 14 at t = 0. Therefore, Figure 1 shows one lg agent 10 and two hg agents 12, 14. The described approach according to an embodiment of the invention implements the projection of the ℎ^ agent onto the ^^ agent(s). This projection is valid when using the radar signal, which provides the geometric connection between these agents, since the agents are not placed on the same rigid body.The position of ^^^^^^^ (^^^) is projected along the line of sight (LOS) onto the ^^^^^ to obtain ^^^ , that is, the position of the ℎ^ agent available to the ^^ agent, by projecting the location of the ℎ^ agent onto the ^^ agent. This method is, of course, applicable to many more hg and lg agents, as can be seen in Figure 1, which illustrates this method for two hg agents 12, 14 and one lg agent 10. In this way, another location measurement for the lg agent is synthesized, available through this projection. Then, using these physical low-order measurements together with the projected high-order measurements, a merging algorithm is implemented, which calculates P. ^^updated. The updated position of the lg agent, ^P^^, is the linear combination of these physical measurements together with the synthetic or projected measurements. This idea of ​​collaborative navigation (CN) is used as a complementary filter to the integrated extended Kalman filter (EKF) already running on all agents, especially on the lg agents. R. 410249 - 7 -An extended Kalman filter (EKF) is the nonlinear version of the Kalman filter that linearizes via an estimation of the current mean and covariance. Broadly speaking, this merging mechanism can be divided into three sub-modules: estimation, projection, and updating of the coefficients. Furthermore, there are three main configurations in which this idea could be used: (i) after the EKF (ii) before the EKF (iii) for high-frequency (HF) radar measurements. The estimation of the coefficients is the same for all three configurations, as it depends on the sensor parameters. The ^ parameter is the coefficient for the covariance, and the ^ parameter is the coefficient for the state. For further details, refer to Figure 4. CN after EKF (Figure 2): Figure 2 shows an illustration of the overall merging scheme, using a CN filter after EKF on the lg agent.The drawing shows first agent 150 using a low-cost system. Within first agent 50, there are observer 52, IMU 54, RADAR system 56, EKF 58, and model 60. To illustrate the projection step, within 70, there is first agent 50 using a low-cost INS, second agent agent 272, and further agents represented by agent n 74. In 70, agents 72 and 74 are high-quality agents, and agent 50 is a low-quality agent. The purpose is to use information from high-quality agents 70, 72 to improve the self-localization parameters of low-quality agent 50. Information from high-quality agents 72, 74 is projected and transmitted to agent 50 for use. ^. ^^ is represented by arrow 53. This symbol represents the position of the agent l, which is provided, for example, by a GNSS device. This observation (measurement) could be entered into the EKF. R. 410249 - 8 - ^ ^ ^ ⃗ is represented by arrow 55. These are the velocity and angle subsections output by the integrated low-order vehicle IMU. =^^^,^, ^^,^, … ^ is represented by arrow 57. This is a vector of the range measurement from agent l to all other agents in the vicinity of the ego vehicle. X is represented by arrow 59. The output of the model in the form of an updated (corrected) status of the vehicle. The status contains the localization parameters. 0.^⃗ is represented by arrow 61. This is an observation (0) specified for the model, with uncertainty (^⃗ ). In this case, the EKF 58 outputs the location of the ^^ agent by default at a certain rate. Furthermore, the CN model takes ^ ^^and the range measurements to the ^^ agents as inputs. In this case, the parameters of the ℎ^ agents are projected to the ^^ agents along the LOS calculated from the locations themselves, as shown in Figure 5. From this projection, ^ as the location of the ^^ agent extracted from the ℎ^ agents. If the CN is then activated after EFK 58, ^ ^^ ^ ^ with ^ ^^, estimated by its own EKF, is superimposed, as seen in Figure 6. The covariance is also updated by using the ^ parameter.CN before EKF (Figure 3): Figure 3 shows an illustration of the overall merging scheme, where a CN filter was used before EKF on the lg agent. The drawing shows first agent 1100 using a low-cost system. Within the first agent 100, there are observer 102, IMU 104, RADAR system 106, EKF 108, and model 110. To illustrate the projection step, within 120, there is the first agent 100 using a low-cost INS, second agent agent 2 122, and further agents represented by agent n 124. Block 120 has the same meaning as block 70 in Figure 2, with agents 122 and 124 as high-order agents, and agent 100 as low-order agent. R. 410249 - 9 - ^ ^^ is represented by arrow 111. As 52 above.is represented by arrow 117. As 55 above. ... ^ is represented by arrow 115. As 57 above. X is represented by arrow 119. As 59 above. 0. ^⃗ is represented by arrow 113. As 61 above. The difference between blocks 119 and 59 is the order of the sub-modules. In 59, the EKF output is the input to the information sharing module. In 119, the information sharing module feeds the EKF. One study showed superior performance for block type 119. In this case, the projection to ^ ^^ ^ ^ to obtain, the same as in the previous case. The main difference is that for the update equation, the superposition is done with respect to the predicted value of the state and the projected value from the CN. The merging coefficients are shown below: ^ ^ ^ = ^ ^^ , ^^ = ^ ^ ^ ^^ ^^ ^ ^ ^^,^ , ^ = 1, … , ^, these are the merging coefficients. The weight specified for each location to output the location prediction as an overlay of all input location predictions. ^ ^ Is the weight given for the measurement, and ^^, ^ = 1, …, is the weight of all other merge location predictions. ^⃗ = ^∑ ^ ^ ^^ ^ ^^ ^ ^^⃗, here the ^⃗ vector is normalized to a unit vector.^^ = 0.2, ^^ = 0.8, the model hyperparameters. How the uncertainty from the location of neighboring agents and the radar measurements are weighted. R. 410249 - 10 - How the estimated location covariance matrix is ​​postprocessed. Previous work ignores this part and therefore achieves inferior results, while this term is considered here. Therefore, it is possible to improve performance because the model's covariance better represents the state. The projection equations for case (i) - CN after EKF for the lg agent are shown below: ^^ = ^^^^2^^^^,^ , ^^^^, the LOS calculation as the angle from the low-order agent to each of the high-order agents near the ego vehicle. The estimated location of the ego vehicle, from the individual projected measurements of all other agents, combined with the LOS. ^ ^^ ^^ (^ + 1) = ∑^ ^^ ^^^ ^^ ∙^^^,^ (^ + 1), The superposition after weighting all projected measurements and by using the projection coefficients ^⃗.Below, the update equations for case (i) - CN after EKF for the lg agent are shown:^^^^(^ + 1) = ^^^^^(^ + 1) + ^ ^^^^ (^ + 1), the updated location from the model taking into account the natural measurement and the measurements generated by the neighboring vehicles. the updated covariance taking into account the new covariance update ^. HF radar measurements (Figure 4): Figure 4 shows time frame 200 for case (iii) - HF radar measurements. Index k stands for the time during which an external (GNSS-like) measurement is available. Index m R. 410249 - 11 -represents the times between these measurements in which radar samples are collected. There is a first column 202 for low-order measurements and a second column 204 for high-order measurements. In this case, the scenario is treated for a case where an EKF is running at a low frequency, and the radar measurements are given at a high frequency. Although the radar measurements are given at a high frequency, the CN is still activated at a low frequency because these projections are noisy. The radar measurements are projected and collected between two activations of the CN filter, to formulate, and then d^^ the projection matrix This matrix contains all projections of all ℎ^ agents for all times between consecutive CN filter activations. These samples are then averaged and overlaid with the last state estimate to calculate the input for the EKF after the CN. Projection equations for the HF radar case are shown below. The projection is performed by each agent, subscript I, and for each time, subscript m is high frequency. is the LOS of each high-frequency radar measurement to the ego vehicle.^ ^^ ^^,^ (^) is the predicted location of the low-order agent, estimated from the projection of the location of the hg agents along the LOS, at high frequency, for each timestamp (m). A typical value for m is the frequency of the radar measurements, for example, 5 or 10 Hz. Update equations for the HF radar case are shown below. Formulation of the measurement matrix X ^^ ^ ^,^, their weighted mean for each Agenten x ^^ and to ^^ Final merge for all agents to X ^^ to R. 410249 - 12 - The overlay is performed as usual to construct the improved measurement for the EKF after this CN filter. to form a single low-frequency measurement. ∙ ^ ^^ ^^,^ (^ + 1) , projection of all agents to produce a single generated measurement at low frequency.^^^^(^ + 1) = ^^^^^(^ + 1) + ^ ^^^^ (^ + 1), updated location of the model taking into account the natural measurement and the measurements generated by the neighboring vehicles. Updated covariance taking into account new covariance update ^. The described method is, in at least one of its embodiments, applicable to: 1. a single ^^ agent and a single ℎ^ agent. 2. a single ^^ agent and multiple ℎ^ agents. 3. multiple ^^ agents and a single ℎ^ agent. 4. multiple ^^ agents and multiple ℎ^ agents. Figure 5 shows an illustration of this method for two hg agents and one lg agent. The drawing shows model 300, which includes relative position 302, position coefficient 304, covariance coefficient 306, block 308 for estimating position, and block 308 for estimating covariance. ^ ^^^ is represented by arrow 301. Localization measurements provided for the model by the GNSS sensors. R. 410249 - 13 - ^ ^^^is represented by arrow 303, range measurements from the range sensors.^⃗ is represented by arrow 305, the uncertainty associated with the location measurements, as mentioned above under 301. ^ ^^ ^ ^ is represented by arrow 311, the location of the low-order agent, extracted by projecting the locations of the high-order agents along the estimated LOS.^ is represented by arrow 309, the weighting factor of the ego location and the estimated locations of the neighboring agents.^ is represented by arrow 307, the post-processing coefficient for the covariance matrix. X is represented by arrow 313, the input location measurement of the ego vehicle.^ is represented by arrow 317, the output location from the information sharing module.^ is represented by arrow 315, the covariance estimate from the self-localization algorithm.^ ^is represented by arrow 319, postprocessing for covariance. This should better represent the estimated status ^^ after information sharing. Finally, a method for synthesizing and merging location information from multiple agents is implemented. An example of such merging for the case of driving is provided. These ideas and algorithms could also be used for other applications, such as aviation, construction, indoor robotics, and much more. The main contributions and differences from previous work: R. 410249 - 14 -(i) Here, no other external sensor is used apart from the range. Known methods involve some angular measurements to enable this fusion. Here, this fusion is performed only by location and range. The angular information is formulated by the LOS. (ii) In this fusion, no information about the quality of the measurements is assumed. Any hand-crafted features of the ^ matrix of the EKF are neither changed nor added. (iii) The possibility to explain the update of the covariance matrix is ​​added. This is not addressed in known methods. This applies to the case where the CN is activated after the EKF; therefore, the EKF covariance must also be updated. Figure 6 shows a scenario according to an embodiment of the described method. The drawing shows the first vehicle 400 with localization system 402, in this case a lower-order localization system.Furthermore, Figure 6 shows second vehicle 420 adjacent to first vehicle 400. This second vehicle 420 also includes localization system 422, in this case a higher quality localization system. The localization system 402 within first vehicle 400 provides signal 404 carrying first localization information to the position of first vehicle 400. This information can be extracted by agent 406 in the first vehicle. Accordingly, the localization system 422 in the second vehicle provides signal 424 carrying second localization information to the position of second vehicle 420. This information can be extracted by agent 426 in first vehicle 400. Now, within first vehicle 400, there is arrangement 410 that receives the first localization information and the second localization information, which is transmitted via communication system 430 with respect to line of sight (LOS) 440 between R. 410249 -. 15 -the first vehicle 400 and the second vehicle 420.

Claims

R. 410249 - 16 -Claims 1. A method for locating a first vehicle (400), comprising an agent (406) having access to first location information relating to the position of the first vehicle (400), wherein at least one auxiliary signal is provided carrying at least one second piece of location information relating to the position of at least one second vehicle (420), the first and the at least one second piece of location information being combined within the agent (406) of the first vehicle (400) while taking into account the spatial arrangement of the vehicles (400, 420) to determine the position of the first vehicle (400).

2. The method according to claim 1, wherein the at least one piece of auxiliary information is provided by projection.

3. The method according to claim 2, wherein a line of sight (LOS) (440) is used during the projection to take into account the spatial arrangement of the vehicles (400, 420). 4.Method according to one of claims 1 to 3, wherein first localization information is provided by a lower-order localization system.

5. Method according to one of claims 1 to 4, wherein the at least one second localization information is provided by a higher-order localization system.

6. Method according to one of claims 1 to 5, wherein the localization system is based on GNSS. R. 410249 - 17 - 7. The method according to any one of claims 1 to 6, wherein the localization system is based on radar and / or lidar.

8. The method according to any one of claims 1 to 7, wherein the method is divided into three steps: estimation, projection, and updating of the coefficients.

9. The method according to any one of claims 1 to 8, wherein an extended Kalman filter (EKF) (58) is used.

10. An arrangement for localizing a vehicle, suitable for carrying out a method according to any one of claims 1 to 9.

Images