Location estimation system

Abstract

To provide a position estimation system for increasing estimation precision of a vehicle position by dead reckoning while estimating a vehicle position by utilizing a magnetic marker.SOLUTION: A position estimation system 1 includes: a magnetic positioning circuit 153 for measuring a position of a vehicle on the basis of a magnetic marker 10; a dead reckoning circuit 151 for estimating a position of a vehicle by dead reckoning based on an inner field of a vehicle; and a filtering circuit 157 for estimating the amount of state including a position of a vehicle on the basis of a position of a vehicle measured by the magnetic positioning circuit 153 and a position of a vehicle estimated by the dead reckoning circuit 151. The amount of state can be expressed by a state equation including a sideslip angle as angular deviation between a yaw angle of a vehicle and a traveling direction of a vehicle in a variable, and the sideslip angle is estimated by an approximate expression of regular circle turning.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] The present invention relates to a position estimation system for estimating the position of a vehicle.

Background Art

[0002] Conventionally, a magnetic positioning system has been proposed that uses magnetic markers arranged at intervals on a road to measure the position of a vehicle (see, for example, Patent Document 1 below). In this system, the vehicle position is measured based on the magnetic markers detected by the vehicle. For example, in this system, the amount of lateral displacement of the vehicle with respect to the magnetic marker is measured, and the vehicle position is measured by shifting the position by the amount of lateral displacement from the laying position of the magnetic marker.

[0003] The system of Patent Document 1 below has a function of measuring the vehicle position by magnetic positioning, and also has a function of estimating the vehicle position by dead reckoning (DR, inertial navigation, dead reckoning, inertial navigation). Dead reckoning is a method of estimating the position of a vehicle's destination based on internal information of the vehicle such as wheel speed and yaw rate. With dead reckoning using internal information such as wheel speed and yaw rate, it is possible to estimate the vehicle position even during the period from detecting any magnetic marker until detecting a new magnetic marker.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] In magnetic positioning systems that utilize magnetic markers, ensuring the accuracy of vehicle position estimation during the period between detecting one magnetic marker and detecting a new one is crucial. To achieve advanced vehicle control such as autonomous driving, it is necessary to improve the accuracy of vehicle position estimation using dead reckoning.

[0006] This invention has been made in view of the aforementioned conventional problems, and aims to provide a position estimation system that estimates vehicle position using magnetic markers while improving the accuracy of vehicle position estimation by dead reckoning. [Means for solving the problem]

[0007] The present invention is a system for estimating the position of a vehicle moving on a floor or road surface on which a magnetic marker is placed. A detection unit for detecting the magnetic marker, A magnetic positioning unit that determines the vehicle's position based on the magnetic marker detected by the detection unit, A DR unit estimates the vehicle's position by dead reckoning based on internal information acquired from within the vehicle, The system includes a state estimation unit that estimates state quantities, including the vehicle's position, based on the vehicle's position measured by the magnetic positioning unit and the vehicle's position estimated by the DR unit. The state quantities estimated by the state estimation unit can be expressed by a state equation that includes the sideslip angle, which is the angular difference between the vehicle's yaw angle and the vehicle's direction of movement, as a variable. The aforementioned sideslip angle is a physical quantity estimated by an approximate formula for steady-state circular turning in a vehicle position estimation system. [Effects of the Invention]

[0008] In the position estimation system of the present invention, state variables including the vehicle's position are estimated based on the vehicle's position observed by the magnetic positioning unit and the vehicle's position estimated by the DR unit. In the position estimation system of the present invention, the sideslip angle is estimated by an approximate formula for steady-state circular turning, and the state variables are represented by a state equation with the sideslip angle as a variable.

[0009] In the position estimation system of the present invention, the sideslip angle is used to estimate the position of the vehicle. By using the sideslip angle, which is the angular difference between the vehicle's yaw angle and the vehicle's direction of movement, the accuracy of position estimation by dead reckoning can be improved. In particular, in the position estimation system of the present invention, the sideslip angle is estimated by an approximate formula for steady-state circular turning. The sideslip angle obtained by the approximate formula for steady-state circular turning can be used to estimate state variables, including the vehicle's position. [Brief explanation of the drawing]

[0010] [Figure 1] Diagram illustrating the position estimation system in Example 1. [Figure 2] Perspective view of the magnetic marker in Example 1. [Figure 3] A block diagram showing the configuration of the vehicle in Example 1. [Figure 4] A block diagram showing the configuration of the position estimation system in Example 1. [Figure 5] A graph showing the change in magnetic measurement values ​​in the forward and backward directions when passing through a magnetic marker in Example 1. [Figure 6] This graph shows an approximate curve of the distribution of magnetic measurement values ​​in the vehicle width direction from each magnetic sensor Cn when the magnetic detection unit is positioned directly above the magnetic marker in Example 1. [Figure 7] A diagram illustrating the processing steps performed by magnetic positioning in Example 1. [Figure 8] Block diagram of the position estimation algorithm in Example 1. [Figure 9] A flowchart showing the position estimation process in Example 1. [Figure 10] A block diagram showing the vehicle configuration in Example 2. [Figure 11] Block diagram of the position estimation algorithm in Example 2. [Figure 12] Block diagram of the position estimation algorithm in Example 3.

Mode for Carrying Out the Invention

[0011] Conventionally, under the assumption that the vehicle speed and steering angle are constant, discussions on steady circular turning that further develop the two-wheel model are known. However, in actual driving, the assumption that the speed and steering angle are constant does not hold. On the other hand, when the vehicle is at a low speed and both the speed change and steering angle change are small, it is possible to regard the speed and steering angle as constant for a certain short period of time (quasi-steady state). In the position estimation system of the present invention, in such a situation where the speed and steering angle can be regarded as constant, the sideslip angle is calculated by an approximate formula for steady circular turning.

[0012] Embodiments of the present invention will be specifically described below using the following examples. (Example 1) This example relates to a position estimation system 1 for accurately estimating the vehicle position (the position of the vehicle). This content will be described using FIGS. 1 to 9.

[0013] The position estimation system 1 of this example is a system configured to improve the estimation accuracy of the vehicle position by dead reckoning using the magnetic marker 10 on the road surface as shown in FIG. 1. The position estimation system 1 employs a position estimation algorithm including state estimation by an Extended Kalman Filter (EKF). In this example, a bus is exemplified as the vehicle 2.

[0014] The vehicle position estimated by the position estimation system 1 is input, for example, to a vehicle ECU 20 for realizing driving support control such as lane-keeping driving or autonomous driving of the vehicle 2 (FIG. 1). The vehicle ECU 20 uses the vehicle position estimated by the position estimation system 1 and executes the driving control of the vehicle 2. The vehicle ECU 20 controls, for example, an engine throttle actuator, a steering actuator, a brake actuator, etc. (not shown) so that the vehicle 2 travels along a preset route.

[0015] In this example, magnetic markers 10 are arranged at 2m intervals along the path 1R on which vehicle 2 travels. Each magnetic marker 10 (Figure 2) consists of, for example, a columnar magnet 10M with a diameter of 30mm and a height of 28mm, and is housed and embedded in a housing hole (not shown) drilled in the road surface. A wireless tag 10T is attached to the upward end face of each magnetic marker 10 (the end face of the magnet 10M). The magnet 10M (magnetic marker 10) is a permanent magnet whose magnetic dipole moment is aligned with the axial direction, and both end faces exhibit different magnetic polarities. In this example, each magnetic marker 10 is embedded so that the north pole faces upward.

[0016] The wireless tag 10T is a sheet-like electronic component in which an IC chip is surface-mounted on a resin film sheet with an antenna pattern printed on it. The wireless tag 10T operates on external power supply and wirelessly outputs pre-stored tag information. The tag information includes the installation position of the magnetic marker 10 (marker position) and the absolute bearing connecting adjacent magnetic markers 10 (marker bearing).

[0017] As shown in Figures 3 and 4, the position estimation system 1 includes an IMU (Inertial Measurement Unit) 17 for measuring yaw rate, magnetic sensor units 11F and R for detecting magnetic markers 10, wheel speed sensors 12L and R attached to the left and right rear wheels 202, steering angle sensors 13 attached to the front wheels 201 which are steering wheels, a tag reader 14 for performing wireless communication with the wireless tag 10T, and a calculation unit 15 for executing a position estimation algorithm.

[0018] Here, the IMU 17, wheel speed sensors 12L and 12R, and steering angle sensor 13 are examples of internal sensors for acquiring internal information of the vehicle 2, such as yaw rate, wheel speed, and steering angle of the steering wheels. The magnetic sensor units 11F and 11R and tag reader 14 are devices that constitute a magnetic positioning unit that measures the vehicle's position and orientation using magnetic markers 10. This magnetic positioning unit is an example of an external sensor that acquires external information of the vehicle 2.

[0019] The steering angle sensor 13 is a sensor that detects the steering angle, which is the angle at which the front wheels 201, which are the steering wheels, are steered.

[0020] The wheel speed sensors 12L and 12R (Figures 3 and 4) are sensors that measure the rotational speed (wheel speed) of the rear wheels 202. The wheel speed sensors 12L and 12R output a sensor signal, which is a measurement signal of the wheel speed. The wheel speed sensors 12L and 12R are attached to the rear wheels 202, which have a fixed steering direction, rather than the front wheels (steering wheels) 201. Wheel speed sensor 12L is attached to the left rear wheel 202, and wheel speed sensor 12R is attached to the right rear wheel 202.

[0021] The magnetic sensor units 11F and 11R are examples of detection units for detecting magnetic markers 10. The magnetic sensor units 11F and 11R share the same specifications except for their mounting position. The magnetic sensor unit 11F is positioned at the front of the vehicle 2 in the longitudinal direction. The magnetic sensor unit 11R is positioned at the rear. The sensor span, which is the distance between the magnetic sensor units 11F and 11R in the longitudinal direction of the vehicle 2, is 2m. This sensor span of 2m coincides with the spacing of the magnetic markers 10 arranged along the path 1R. The magnetic sensor units 11F and 11R can simultaneously detect adjacent magnetic markers 10 in the direction of the road.

[0022] As shown in Figure 4, the magnetic sensor unit 11F·R includes multiple magnetic sensors Cn (where n is an integer from 1 to 15), a detection processing circuit 112 including a CPU (not shown), etc. The magnetic sensor unit 11F·R, which is an example of a detection unit for detecting the magnetic marker 10, has a long, slender rod shape. In the magnetic sensor unit 11F·R, 15 magnetic sensors Cn are arranged in a straight line at equal intervals of 10 cm. Both magnetic sensor units 11F·R are mounted on the vehicle 2 so as to be aligned with the vehicle width direction, with the magnetic sensor C8 positioned in the center in the vehicle width direction.

[0023] The magnetic sensor Cn (Figure 4) is an MI sensor that detects magnetism using the known MI effect (Magneto Impedance Effect). The MI effect is a magnetic effect in which the impedance of a linearly assembled magnetometer (e.g., amorphous wire) changes in accordance with the strength of the magnetic field acting in its longitudinal direction. In the magnetic sensor Cn, two magnetometers are assembled orthogonally to each other, thereby enabling the detection of magnetism acting in two orthogonal axial directions. In this example, magnetic sensor units 11F and 11R are attached to the vehicle 2 so that each magnetic sensor Cn can detect the magnetic components in the longitudinal and width directions of the vehicle 2. The magnetic sensor Cn, being an MI sensor, is highly sensitive and can reliably detect the magnetism acting on the magnetic marker 10.

[0024] Each detection processing circuit 112 (Figure 4) of the magnetic sensor units 11F and R is an arithmetic circuit that performs marker detection processing and other operations to detect the magnetic marker 10. These detection processing circuits 112 are configured using a CPU (not shown) that performs various calculations, memory elements such as ROM (Read Only Memory) and RAM (Random Access Memory) (not shown), etc.

[0025] The detection processing circuit 112 acquires the sensor signals output by each magnetic sensor Cn and executes marker detection processing. In addition to detecting the magnetic marker 10, the lateral deviation relative to the magnetic marker 10 is measured during the marker detection processing. Lateral deviation is the amount of displacement in the vehicle width direction relative to the magnetic marker 10. In this example, the center of the magnetic sensor unit 11F·R where the magnetic sensor C8 is located is set as the reference unit center, and the lateral deviation of the unit center relative to the magnetic marker 10 is measured. The marker detection result output by the detection processing circuit 112 in response to the execution of marker detection processing includes information on the lateral deviation relative to the magnetic marker 10, in addition to the fact that the magnetic marker 10 has been detected.

[0026] The tag reader 14 (Figures 3 and 4) is a wireless device that performs wireless communication with the wireless tag 10T held on the magnetic marker 10. The tag reader 14 is configured to operate the wireless tag 10T by wireless power supply to wirelessly output tag information and to receive that tag information. As described above, the tag information includes at least position information representing the laying position of the magnetic marker 10 (marker position), and the absolute bearing (marker bearing) connecting adjacent magnetic markers 10.

[0027] The tag reader 14 starts wireless power supply in response to the detection of the magnetic marker 10 by the front magnetic sensor unit 11F, thereby activating the wireless tag 10T. Then, once the tag reader 14 receives tag information, it temporarily stops its operation. Through this operation, the tag reader 14 reads the tag information from the wireless tag 10T held by the magnetic marker 10 detected by the front magnetic sensor unit 11F.

[0028] The arithmetic unit 15 (Figures 3 and 4) is a unit that executes a position estimation algorithm for estimating the vehicle's position. The arithmetic unit 15 includes an electronic circuit board (not shown) on which a CPU (central processing unit) for performing various calculations, memory elements such as ROM (read-only memory) and RAM (random access memory) are mounted.

[0029] The arithmetic unit 15 is equipped with external input / output ports. Signals and information acquired by the arithmetic unit 15 via these input / output ports include marker detection results output by the magnetic sensor units 11F and 11R, tag information read by the tag reader 14, sensor signals from the wheel speed sensors 12L and 12R and the steering angle sensor 13, and the yaw rate output by the IMU 17.

[0030] The marker detection results include, as described above, a statement that a magnetic marker 10 has been detected, as well as the lateral deviation relative to the magnetic marker 10. The tag information includes, as described above, information on the marker position and marker orientation. The marker position is the installation position of the magnetic marker 10. The marker orientation is the absolute orientation connecting adjacent magnetic markers 10. Here, the tag information that can be obtained in response to the detection of a magnetic marker 10 is, as described above, the tag information of the wireless tag 10T held in the magnetic marker 10 detected by the front magnetic sensor unit 11F.

[0031] The arithmetic unit 15 realizes the following circuit functions by having the CPU execute a software program read from ROM. (1) Dead reckoning circuit 151: A circuit that is an example of a DR unit that estimates the vehicle position and vehicle orientation by dead reckoning based on internal information of the vehicle 2. (2) Magnetic positioning circuit 153: A circuit that is an example of a magnetic positioning unit that performs magnetic positioning. Magnetic positioning is a process that uses magnetic markers 10 to determine the vehicle's position and direction. (3) Parameter setting circuit 155: A circuit for setting the parameters of the EKF (Extended Kalman Filter). In this example, the center of gravity position, which is one of the vehicle parameters, is set as the parameter. (4) Filtering circuit 157: A circuit that improves the accuracy of vehicle position estimation by updating the observation data of the EKF.

[0032] Next, we will explain (a) dead reckoning, (b) marker detection process, (c) magnetic positioning, and (d) position estimation algorithm in order, and then explain (e) the flow of the position estimation process.

[0033] (a) Dead reckoning Dead reckoning is a process that estimates state quantities such as vehicle position and vehicle orientation based on internal information of vehicle 2, namely vehicle speed V and steering angle δ. The steering angle δ is a physical quantity based on the sensor signal of the steering angle sensor 13, and is the angle at which the front wheels 201, which are the steering wheels, are steered.

[0034] In this example, a state vector containing the vehicle position (x, y), vehicle orientation θ, and the vehicle parameter, the center of gravity position lr, is defined as shown in Equation 1. The coordinate reference point in the position estimation system 1 in this example is the center of gravity position of vehicle 2.

number

[0035] The dead reckoning in this example is a process that estimates a state vector based on internal information such as vehicle speed V, steering angle δ, and yaw rate γ. In this example, the internal information of vehicle 2 is defined by the input vector in equation 2.

number

[0036] In this example, the state equation is defined as shown in Equation 3 to estimate the state vector. The input vector consists of vehicle speed V, steering angle δ, and yaw rate γ. The state equation is an equation for estimating the new state vector at time (k) based on the past state vector at time (k-1) when the input vector is applied.

number

[0037] The vehicle speed V in the state equation of equation 3 can be calculated using equation 4 based on the sensor signals of the internal sensor, the wheel speed sensor 12L·R. Vr is the wheel speed of the right rear wheel 202 based on the sensor signal of wheel speed sensor 12R. Vl is the wheel speed of the left rear wheel 202 based on the sensor signal of wheel speed sensor 12L.

number

[0038] Furthermore, the equation of state (Equation 3) includes the sideslip angle β as a variable, which is the angular difference between the vehicle's yaw angle and the direction of vehicle movement. In this particular example, the sideslip angle β is estimated using Equation 5, which is an approximation of steady-state circular turning.

number

[0039] According to Equation 5, the sideslip angle β can be estimated from the vehicle 2's speed V and steering angle δ. Therefore, an IMU is not essential for calculating the sideslip angle β. Furthermore, in this example, the sideslip angle β is calculated in an open-loop manner, making it possible to integrate it into EKF.

[0040] (b) Marker detection process The marker detection process is performed by the magnetic sensor unit 11. In the description common to magnetic sensor units 11F and 11R, it is referred to as magnetic sensor unit 11. The magnetic sensor unit 11 performs the marker detection process at a frequency of 3 kHz. As described above, the magnetic sensor Cn can measure the magnetic components of the vehicle 2 in the longitudinal direction and the magnetic components in the vehicle width direction. For example, when this magnetic sensor Cn moves along the longitudinal direction of the vehicle 2 and passes directly over the magnetic marker 10, the magnetic measurement value in the longitudinal direction changes over time, as shown in Figure 5, with the sign reversing before and after the magnetic marker 10, and crossing zero at the position directly above the magnetic marker 10.

[0041] While vehicle 2 is in motion, when a zero-crossing Zc1 occurs in which the positive and negative signs of the magnetic measurement values ​​in the longitudinal direction detected by any one of the magnetic sensors Cn are reversed, it can be determined that the magnetic sensor unit 11 is positioned directly above the magnetic marker 10. When a zero-crossing Zc1 occurs in the longitudinal magnetic measurement values ​​in this manner, the detection processing circuit 112 determines that the magnetic marker 10 has been detected.

[0042] However, it is difficult to immediately detect the magnetic marker 10 when the magnetic sensor unit 11 is positioned directly above it. This is because the zero-crossing Zc1 cannot be detected unless changes in magnetic measurement values ​​(magnetic measurement values ​​in the forward and backward directions) are obtained not only during the period before the zero-crossing Zc1, but also during the period after the zero-crossing Zc1. In the marker detection process of this example, the change in magnetic measurement values ​​during the period from when the magnetic sensor unit 11 passes directly above the magnetic marker 10 until it moves, for example, 200 mm away from the magnetic marker 10 is necessary to detect the zero-crossing Zc1. The period required to detect the zero-crossing Zc1 varies depending on the size and magnetic force of the magnetic marker, the mounting height of the magnetic sensor unit, etc. The period required to detect the zero-crossing Zc1 can be set as needed.

[0043] For example, let's consider the movement of a magnetic sensor with the same specifications as magnetic sensor Cn along a straight line in the vehicle width direction that passes directly over the magnetic marker 10. The magnetic measurement value in the vehicle width direction from this magnetic sensor will reverse in sign on both sides of the magnetic marker 10, and will change to cross zero at the position directly above the magnetic marker 10. In the case of a magnetic sensor unit 11 in which 15 magnetic sensors Cn are arranged in the vehicle width direction, the sign of the magnetic measurement value in the vehicle width direction will differ depending on whether the magnetic sensor Cn is located on the left or right side of the position of the magnetic marker 10 (Figure 6).

[0044] Figure 6 illustrates an approximate curve of the distribution of magnetic measurement values ​​in the vehicle width direction for each magnetic sensor Cn. In the distribution shown in this figure, the zero-crossing point Zc2, where the sign of the magnetic measurement value reverses, appears directly above the magnetic marker 10. The position of the magnetic marker 10 in the vehicle width direction can be determined as the midpoint between two adjacent magnetic sensors Cn on either side of the zero-crossing point Zc2.

[0045] The detection processing circuit 112 measures the lateral deviation, which is the amount of displacement of the vehicle 2 in the vehicle width direction relative to the magnetic marker 10. In this example, the unit center (position of magnetic sensor C8), which is the center of the magnetic sensor unit 11, is set as the reference point for the lateral deviation. For example, in Figure 6, the position of the zero cross Zc2 corresponding to the magnetic marker 10 is at a position corresponding to C9.5, which is roughly midway between C9 and C10. As described above, the distance between magnetic sensors C9 and C10 is 10 cm, so the lateral deviation of the unit center relative to the magnetic marker 10 (amount of displacement in the vehicle width direction) is (9.5-8) × 10 cm = 15 cm.

[0046] (c) Magnetic positioning In magnetic positioning, the vehicle position and direction are determined based on the tag information received by the tag reader 14 and the marker detection results by the magnetic sensor units 11F and R. Here, as described above, the tag information includes the marker position (laying position) of the magnetic marker 10 detected by the front magnetic sensor unit 11F, and the marker direction representing the direction connecting adjacent magnetic markers 10. The marker detection result when a magnetic marker 10 is detected includes the lateral deviation of the unit center (center of the magnetic sensor unit 11F) relative to that magnetic marker 10.

[0047] As shown in Figure 7, the magnetic positioning circuit 153 determines the angle of deviation of the vehicle direction dirV relative to the marker direction dirR based on the lateral deviation measured by the magnetic sensor units 11F and R (detection units). Then, by adding this angle of deviation to the absolute direction of the marker direction dirR, the vehicle direction dirV is determined. Here, the marker direction dirR is the absolute direction information included in the tag information.

[0048] Furthermore, the magnetic positioning circuit 153 determines the position of the unit center by shifting its position by the amount of the lateral deviation related to the marker detection result, based on the marker position related to the tag information, i.e., the installation position of the magnetic marker 10 detected by the front magnetic sensor unit 11F. In addition, the vehicle position is determined based on the position of this unit center.

[0049] (d) Position estimation algorithm The position estimation algorithm in this example, as shown in Figure 8, is an algorithm that estimates a state vector (Equation 1) consisting of state quantities such as vehicle position and vehicle orientation by state estimation using an extended Kalman filter (EKF).

[0050] The position estimation algorithm is centered around an EKF (Equipment Kinematics Function) for position estimation. The EKF receives positioning results (vehicle position and vehicle orientation) from a magnetic positioning unit (magnetic positioning circuit 153), which is an example of an external sensor, and sensor signals from an IMU 17, wheel speed sensors 12L and 12R, and steering angle sensor 13, which are examples of internal sensors.

[0051] In addition to the state equation shown in Equation 3 above, the position estimation algorithm defines the observation equation shown in Equation 6.

number

[0052] The state equation (Equation 3) is an equation that represents state variables such as vehicle position, and is used to estimate new states from past states. The observation equation is an equation for estimating observed values ​​related to the state. In the position estimation algorithm in this example, the vehicle position and vehicle orientation determined by the magnetic positioning described above are used as observed values.

[0053] A filtering circuit 157, which is an example of a state estimation unit, performs observation updates based on the positioning results (observed values) measured by the magnetic positioning circuit 153 and the vehicle position etc. (state vector) estimated by the dead reckoning circuit 151, thereby improving the estimation accuracy of the vehicle position etc.

[0054] (e) Position estimation process The flow of the position estimation process using the position estimation algorithm in this example will be explained with reference to the flowchart in Figure 9. The calculation unit 15 acquires sensor signals from the internal sensors at each program execution cycle (S11) and performs vehicle position estimation by dead reckoning (S12). The internal sensors in this example are wheel speed sensors 12L and R, steering angle sensor 13, and IMU 17. The sensor signals acquired in step S11 are the wheel speed measurement signals from the wheel speed sensors 12L and R, the steering angle δ from the steering angle sensor 13, and the yaw rate γ from the IMU 17. In the estimation by dead reckoning, the vehicle orientation θ (direction) is estimated along with the vehicle position (x, y).

[0055] Furthermore, the calculation unit 15 acquires the marker detection result from the magnetic sensor unit 11 (S13) and determines whether or not the magnetic marker 10 has been detected (S14). If the magnetic marker 10 is not detected (S14: NO), the vehicle position (x, y) estimated in step S12 is output as the result of the position estimation process. On the other hand, if the magnetic marker 10 is detected (S14: YES), the calculation unit 15 performs the estimation of state quantities such as the vehicle position by executing steps S22 and S23, as described below (state estimation).

[0056] When the magnetic marker 10 is detected (S14: YES), the calculation unit 15 measures the vehicle position (x, y) and vehicle direction θ based on the marker position and direction (S22, magnetic positioning). The calculation unit 15 then uses the vehicle position (x, y), etc., measured by magnetic positioning as observed values ​​to perform observation updates using EKF and estimate the state of the vehicle position (x, y), etc., due to dead reckoning (S23, observation update). By performing observation updates using EKF, the calculation unit 15 corrects the vehicle position (x, y), etc., estimated by dead reckoning, thereby improving the estimation accuracy. The observation update is a correction using the well-known Kalman gain.

[0057] The position estimation system 1 in this example, configured as described above, is a system that uses magnetic markers 10 to improve estimation accuracy through dead reckoning. In dead reckoning, it is more accurate to consider the sideslip angle of vehicle 2. The center of gravity position, a vehicle parameter used in calculating the sideslip angle, changes due to passengers getting on and off the bus, and therefore needs to be estimated in real time.

[0058] Conventional methods for real-time parameter estimation include, for example, adaptive observers that define parameters as state variables in the EKF and estimate them using the EKF. However, if the sideslip angle is calculated outside the EKF for self-localization using a linear observer, that linear observer cannot be incorporated into the EKF as a model. Therefore, it is impossible to define the centroid position as a state variable in the EKF for self-localization and estimate it using an adaptive observer.

[0059] On the other hand, in the position estimation system 1 of this example, the sideslip angle is estimated from the vehicle speed and steering angle using an approximate formula for steady-state circular turning. Therefore, an IMU or other device for measuring lateral acceleration is not required. Furthermore, since the sideslip angle can be calculated in an open-loop manner in the position estimation system 1 of this example, it can be incorporated into the extended Kalman filter for position estimation. Consequently, in the position estimation system 1 of this example, the center of gravity position, which is a vehicle parameter necessary for calculating the sideslip angle, can be estimated by the extended Kalman filter for position estimation.

[0060] In this example, a bus is used as an example of Vehicle 2. Vehicle 2 could be a regular passenger car, a transport vehicle used within a factory or other facility, or any other vehicle. Furthermore, it could be a vehicle driven by a driver, an autonomous vehicle, or a vehicle operated by self-driving technology.

[0061] In this example, a magnetic positioning unit (magnetic positioning circuit 153) that uses a magnetic marker 10 to determine the vehicle's position is shown as an example of an external sensor. Alternatively, a GNSS (Global Navigation Satellite System) that receives satellite radio waves to determine its own position may be used as an external sensor instead of, or in addition to, the magnetic positioning unit.

[0062] (Example 2) This example is a modified version of the dead reckoning configuration based on the position estimation system 1 of Example 1. The dead reckoning in this example is wheel odometry-based, which estimates the vehicle position and orientation using only the left and right wheel speeds. This will be explained with reference to Figures 10 and 11.

[0063] The position estimation system 1 in this example (Figure 10) is a system based on the configuration of Example 1, but with the IMU and steering angle sensors (reference numerals 17 and 13 in Figure 3) omitted. The position estimation algorithm in this example (Figure 11) is an algorithm that estimates a state vector consisting of state quantities such as vehicle position (x, y) and vehicle orientation θ by state estimation using an extended Kalman filter (EKF), similar to Example 1.

[0064] In the position estimation algorithm of this example, the state vector (Equation 1), the vehicle speed calculation formula (Equation 4), and the observation equation (Equation 6) are the same as in Example 1, while the input vector (Equation 7), the state equation (Equation 3), the yaw rate calculation formula (Equation 8), and the sideslip angle calculation formula (Equation 9) differ from those of Example 1. Furthermore, the position estimation system 1 of this example employs dead reckoning using wheel odometry of the rear wheels 202. Therefore, in this example, the center of the rear axle (referred to as the rear axle center) that supports the rear wheels 202 to which the wheel speed sensors 12L·R are attached is set as the coordinate reference point.

[0065]

number

number

number

[0066] Here, we will explain how to derive the formula for calculating the side-slip angle of the rear axle center in formula 9, based on the side-slip angle of the center of gravity of vehicle 2 (formula 5 above). In this explanation, to distinguish between the side-slip angle of the center of gravity and the side-slip angle of the rear axle center, we will denote the side-slip angle of the center of gravity as β and the side-slip angle of the rear axle center as βr. From the two-wheeled model of steady-state circular turning, the side-slip angle βr of the rear axle center can be expressed as shown in formula 10, based on the side-slip angle β of the center of gravity.

number

[0067] From the approximate formula for steady-state circular turns, the yaw rate γ is given by equation 11.

number

number

[0068] In the case of wheel odometry, the steering angle δ is unknown, so by rearranging equation 12 using equation 11, we can derive equation 13.

number

[0069] In well-known wheel odometry, the vehicle speed V can be calculated from the left and right wheel speeds as shown in Equation 14, and the yaw rate γ can be calculated from the left and right wheel speeds and the tread width lt as shown in Equation 15.

number

number

[0070] By substituting equations 14 and 15 into equation 13, we can derive the formula for calculating the sideslip angle of the rear axle center as shown in equation 16.

number

[0071] According to the position estimation system 1 in this example, the accuracy of vehicle position estimation using wheel odometry (dead reckoning) can be improved by utilizing magnetic markers 10. In the position estimation system 1 in this example, there is no need to measure the yaw rate. Therefore, the vehicle position estimation is not affected by drift errors of the yaw rate sensor caused by temperature, aging, etc. The other components and effects are the same as in Example 1.

[0072] (Example 3) This example demonstrates how to perform parameter estimation using EKF by setting the tread width lt of the rear wheel 202, which is a vehicle parameter, as a parameter based on the position estimation system 1 of Example 2. This will be explained with reference to Figure 12.

[0073] In the position estimation system 1 of this example (Figure 12), the vehicle speed calculation formula (Equation 4), observation equation (Equation 6), input vector (Equation 7), yaw rate calculation formula (Equation 8), and sideslip angle calculation formula (Equation 9) are the same as in Example 2, while the state vector (Equation 17) and state equation (Equation 18) are different from those in Example 2.

number

number

[0074] In wheel odometry, the yaw rate γ is calculated from the wheel speeds Vr and Vl and the tread width lt. The tread width lt changes due to factors such as vehicle roll and suspension stroke caused by the load weight. Since changes in the tread width lt affect the yaw rate, changes in the tread width lt significantly impact the accuracy of dead reckoning.

[0075] Therefore, in the position estimation system 1 of this example, the estimation accuracy of the yaw rate γ can be improved by estimating the tread width lt as a parameter, thereby improving the accuracy of position estimation by dead reckoning. The other components and effects are the same as in Example 2.

[0076] Although specific examples of the present invention have been described in detail as shown in the embodiments above, these examples only disclose an example of the technology covered by the claims. Needless to say, the claims should not be interpreted restrictively based on the configuration or numerical values ​​of the specific examples. The claims encompass technologies obtained by various modifications, changes, or combinations of the above examples using prior art or the knowledge of those skilled in the art. [Explanation of Symbols]

[0077] 1. Position estimation system 10 Magnetic Markers 11F·R Magnetic Sensor Unit (Detection Section) 112 Detection Processing Circuit 12L·R Wheel Speed ​​Sensor 13. Steering angle sensor 14 Tag Leaders 15 arithmetic units 151 Dead Reckoning Circuit (DR Section) 153 Magnetic positioning circuit (magnetic positioning section) 155 Parameter setting circuit 157 Filtering circuit (state estimation unit) 17 IMU 2 vehicles 20 Vehicle ECU 201 Front wheels (steering wheels) 202 Rear wheel

Claims

1. A system for estimating the position of a vehicle moving on a floor or road surface on which magnetic markers are placed, A detection unit for detecting the magnetic marker, A magnetic positioning unit that determines the vehicle's position based on the magnetic marker detected by the detection unit, A DR unit estimates the vehicle's position by dead reckoning based on internal information acquired from within the vehicle, The system includes a state estimation unit that estimates state quantities, including the vehicle's position, based on the vehicle's position measured by the magnetic positioning unit and the vehicle's position estimated by the DR unit. The state quantities estimated by the state estimation unit can be expressed by a state equation that includes the sideslip angle, which is the angular difference between the vehicle's yaw angle and the vehicle's direction of movement, as a variable. The aforementioned sideslip angle is a physical quantity estimated by an approximate formula for steady-state circular turning in a vehicle position estimation system.

2. In claim 1, the approximate formula for steady-state circular turning is an approximate formula that includes the position of the vehicle's center of gravity as a variable, A vehicle position estimation system in which the state variables include, in addition to the vehicle's position, at least the position of the vehicle's center of gravity.

3. In claim 1, the vehicle is equipped with left and right wheels whose steering direction is fixed, and the internal information is the speed of the left and right wheels. The aforementioned approximate formula for steady-state circular turning is a vehicle position estimation system in which the variables include the position of the vehicle's center of gravity, the speeds of the left and right wheels, and the tread width, which is the distance between the left and right wheels in the vehicle width direction.

4. A vehicle position estimation system according to claim 3, wherein the state quantities include, in addition to the position of the vehicle, at least the position of the vehicle's center of gravity and the tread width.

Images