Vehicle control device and vehicle position estimation method

Abstract

The present invention provides a vehicle control device and a vehicle position estimation method that are capable of appropriately learning corrections to a vehicle position estimated based on GNSS in response to environmental changes. To this end, the vehicle position estimation unit of the vehicle control device includes: an absolute position estimation unit that estimates a first vehicle position based on absolute position information acquired from GNSS; a relative position estimation unit that estimates a second vehicle position based on relative position information acquired from outside the vehicle; a driving state determination unit that determines changes in the driving state of the vehicle based on vehicle information or satellite information; a difference calculation unit that calculates a difference after synchronizing the time of the first and second vehicle positions; a learning unit that accumulates the difference as time series data for each driving state and learns a correction to the first vehicle position for each driving state based on the accumulated time series data; and a position correction unit that corrects the first vehicle position based on the correction calculated by the learning unit.

Description

Technical Field

[0001] The present invention relates to a vehicle control device and a vehicle position estimation method for estimating the vehicle position in a map by comparing the surrounding environment structure of the vehicle with structural information recorded in map information. Background Art

[0002] In the past, as a technology for correcting the current position of a vehicle detected by a positioning device, a technology has been proposed that combines the use of satellite positioning to estimate the vehicle's position, and the use of external sensors such as cameras installed in the vehicle to identify landmarks and compare them with the landmark positions recorded in advance in map information. This technology further improves the accuracy of the estimated vehicle position.

[0003] For example, Patent Document 1 states in paragraph 0038 that "the first vehicle position detection unit 23 detects the vehicle's position on the map, i.e., the first vehicle position, based on the measurement results of the positioning unit 2 and the map information in the map database 5." Paragraph 0039 states that "the first vehicle position detection unit 23 corrects the vehicle's position by comparing edge points of a white line extracted from an image captured by a camera with the position information of the white line included in the map information." In other words, Patent Document 1 discloses an automated driving system that improves the estimated accuracy of the vehicle's position by correcting the vehicle's position, as determined by satellite positioning, based on the camera's detection results.

[0004] In addition, claim 1 of patent document 1 states that “an automatic driving system for performing automatic driving control of a vehicle, characterized in that it includes: a positioning unit for measuring the position of the vehicle; a map database for storing map information; a first vehicle position detection unit for detecting the position of the vehicle on the map, i.e., a first vehicle position, based on the measurement result of the positioning unit and the map information of the map database; a driving scene recognition unit for recognizing the driving scene of the vehicle based on the first vehicle position detected by the first vehicle position detection unit and the map information of the map database; a second vehicle position detection unit for recognizing the driving scene of the vehicle based on an image captured by a camera installed in the vehicle or an image captured by a radar sensor installed in the vehicle”. The invention further comprises a method for detecting a second vehicle position on a map by using a detection result of a sensor, a measurement result of a positioning unit, and map information in a map database, based on a position detection process pre-associated with the driving scenario; a determination unit for determining whether a difference between the first vehicle position and the second vehicle position is below a threshold; and an automatic driving control unit for executing the automatic driving of the vehicle based on the second vehicle position if the difference between the first vehicle position and the second vehicle position is below the threshold, and executing the automatic driving control of the vehicle based on the first vehicle position if the difference between the first vehicle position and the second vehicle position is not below the threshold. That is, Patent Document 1 discloses an automatic driving system that compares the first vehicle position with a second vehicle position obtained through position detection processing associated with the driving scenario and switches which vehicle position to use based on the size of the difference between the first vehicle position and the second vehicle position, thereby improving the estimation accuracy of the vehicle position.

[0005] Prior art literature

[0006] Patent Literature

[0007] Patent Document 1: Japanese Patent Application Laid-Open No. 2017-138282 Summary of the Invention

[0008] Problems to be solved by the invention

[0009] The solution disclosed in Patent Document 1 is to select and switch the method of detecting the second vehicle position according to the driving environment such as the type of white line of the lane, the inclination of the driving road surface, and the inside and outside of the tunnel, thereby selecting the position calculation method according to the driving scene to achieve improved accuracy.

[0010] However, the position detection process for determining the second vehicle position only switches the position detection process method according to the environment. For example, the detection position of the vehicle can only be switched according to a fixed environment such as a white line drawn on the road surface or a tunnel.

[0011] On the other hand, in the GNSS (Global Navigation Satellite System) used in the autonomous driving system, the detected vehicle position is output with deviations due to the obstructions such as buildings around the receiver and the configuration of positioning satellites at that date and time. However, since the vehicle position with position deviation is directly used, there is a problem of deterioration in accuracy.

[0012] In view of such problems, an object of the present invention is to provide a vehicle control device and a vehicle position estimation method capable of appropriately learning a correction amount of a vehicle position estimated based on GNSS in response to environmental changes.

[0013] Technical solutions to problems

[0014] In order to solve the above-mentioned problems, the vehicle control device of the present invention has a vehicle position estimating unit for estimating the position of the vehicle, and the vehicle position estimating unit includes: an absolute position estimating unit, which estimates the first vehicle position based on the absolute position information obtained from the GNSS; a relative position estimating unit, which estimates the second vehicle position based on the relative position information obtained from outside the vehicle; a driving state judgment unit, which judges the change in the driving state of the vehicle based on vehicle information or satellite information; a difference calculation unit, which calculates the difference after synchronizing the time of the first vehicle position and the second vehicle position; a learning unit, which accumulates the difference as time series data for each of the driving states, and learns the correction amount of the first vehicle position for each of the driving states based on the accumulated time series data; and a position correction unit, which corrects the first vehicle position based on the correction amount calculated by the learning unit.

[0015] In addition, the vehicle position estimation method of the present invention includes: a step of estimating the first vehicle position based on absolute position information obtained from GNSS; a step of estimating the second vehicle position based on relative position information obtained from outside the vehicle; a step of judging the change in the driving state of the vehicle based on vehicle information or satellite information; a step of calculating the difference after synchronizing the time of the first vehicle position and the second vehicle position; a step of accumulating the difference as time series data for each of the driving states, and learning the correction amount of the first vehicle position for each of the driving states based on the accumulated time series data; and a step of correcting the first vehicle position based on the correction amount obtained through learning.

[0016] Effects of the Invention

[0017] GNSS positioning is affected by obstruction, diffraction, and reflection caused by surrounding structures, errors due to changes in the vehicle's motion, and the combination of satellites used for positioning, resulting in errors relative to the vehicle's exact position. These errors do not significantly fluctuate during periods of constant driving conditions, such as when the surrounding structure environment is stable, vehicle motion changes are minimal, and the combination of satellites used for positioning remains unchanged.

[0018] Therefore, if the vehicle's accurate position can be determined using external sensors while the driving environment remains constant, the deviation from the vehicle's positioning coordinates can be determined, and this deviation can be used to calculate positioning coordinates with improved accuracy. Furthermore, if the driving environment changes, the deviation, or correction amount, can be recalculated to obtain the optimal correction for each driving environment.

[0019] Therefore, according to the vehicle control device and the vehicle position estimation method of the present invention, by learning the appropriate correction amount (offset) of the positioning position according to the driving environment and driving state, the influence of the driving state change is eliminated. Even if the driving environment and driving state change, the high-precision estimation of the vehicle position in the map can be obtained at a high frequency. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a structural diagram of an electronic control unit including the vehicle position estimation unit of the first embodiment.

[0021] Figure 2 This is a block diagram of the vehicle position estimation unit of the first embodiment.

[0022] Figure 3 This is a flowchart showing the processing contents of the relative position estimation unit.

[0023] Figure 4 Schematic diagram of an example of a vehicle turning.

[0024] Figure 5 Schematic diagram showing the steering angle and vehicle heading angle of an example vehicle turning.

[0025] Figure 6 This table determines the driving environment based on vehicle speed.

[0026] Figure 7 This is a schematic diagram showing the timing of estimating the vehicle's position.

[0027] Figure 8 This is a flowchart showing the processing contents of the position correction unit.

[0028] Figure 9 This is a block diagram of the vehicle position estimation unit of the second embodiment.

[0029] Figure 10 This is a flowchart showing the processing contents of the relative position estimation unit of the second embodiment.

[0030] Figure 11 This diagram illustrates how a crosswalk is recognized using an external recognition sensor.

[0031] Figure 12 This diagram is an explanatory diagram for calculating the relative relationship with a crosswalk on a sloped road.

[0032] Figure 13 This diagram is an explanatory diagram for calculating the relative relationship with a crosswalk on a sloped road. DETAILED DESCRIPTION

[0033] Hereinafter, embodiments of the vehicle control device according to the present invention will be described with reference to the drawings.

[0034] Example 1

[0035] Figure 1 This is a block diagram illustrating a vehicle control device 100 including a vehicle position estimation unit 10 according to a first embodiment of the present invention. As shown here, the vehicle control device 100 includes the vehicle position estimation unit 10, an autonomous navigation integration unit 20, a map comparison unit 30, and an automatic driving control unit 40. Furthermore, the vehicle control device 100 is connected to a vehicle information receiving unit 1a, an absolute position sensor 1b, a relative position sensor 1c, an actuator 1d, an HMI (Human Machine Interface) 1e, a storage device storing map data M, and a route generation unit 50.

[0036] The vehicle position estimation unit 10 estimates the vehicle position based on the vehicle information from the vehicle information receiving unit 1 a , the absolute position information from the absolute position acquisition sensor 1 b , and the relative position information from the relative position acquisition sensor 1 c , and sends the position estimation result to the autonomous navigation integration unit 20 .

[0037] The autonomous navigation integration unit 20 calculates the vehicle position based on the vehicle information from the vehicle information receiving unit 1 a and the position estimation result of the vehicle position estimation unit 10 , and outputs the calculated position to the map comparison unit 30 .

[0038] The map comparison unit 30 estimates the vehicle's position on the map based on the position estimation result of the autonomous navigation integration unit 20 and the map data M. The result of the map comparison unit 30 is then sent to the automatic driving control unit 40 .

[0039] The automatic driving control unit 40 controls the actuator 1 d based on the result of estimating the vehicle position in the map and the driving path generated by the path generation unit 50 , thereby realizing automatic driving of the vehicle V.

[0040] The actuator 1d is, for example, various actuators such as a steering system, a driving system, and a braking system for driving the vehicle V. The HMI 1e is, for example, a steering wheel, an accelerator pedal, a brake pedal operated by the driver, and sensors that detect their operation amounts.

[0041] In addition, the vehicle control device 100 is specifically an electronic control unit (ECU) having hardware such as a CPU, a main storage device such as a semiconductor memory, an auxiliary storage device, and a communication device. Various functions such as the vehicle position estimation unit 10 are realized by executing a program loaded into the main storage device by the computing device. However, the details of each part will be described below while appropriately omitting the well-known technologies in the computer field.

[0042] <Own vehicle position estimation unit 10>

[0043] Figure 2 1 is a block diagram illustrating details of the vehicle position estimation unit 10. As shown therein, the vehicle position estimation unit 10 includes an absolute position estimation unit 11, a relative position estimation unit 12, a driving state determination unit 13, a difference calculation unit 14, a learning unit 15, and a position correction unit 16.

[0044] <Absolute Position Estimation Unit 11>

[0045] The absolute position estimation unit 11 estimates the vehicle's position based on signals (absolute position information and satellite positioning status information) received from an absolute position acquisition sensor, such as a GNSS receiver or a receiver capable of receiving positioning radio waves from pseudolites, etc., using one or more sensors for identifying the vehicle's absolute position. Furthermore, the vehicle's position and attitude are estimated using the absolute position acquisition sensors at a high frequency. For example, the GNSS positioning interval is typically around 0.1 to 1 second, which is considered to be a sufficiently high update rate for the route guidance function of the autonomous driving system.

[0046] GNSS positioning receives positioning radio waves transmitted from positioning satellites and measures the distance between the positioning satellites and the receiver antenna based on their arrival time. Because satellite configuration information is superimposed on the positioning radio waves, the positions of the positioning satellites are known, and by obtaining multiple distances to these satellites, the receiver's position can be calculated. However, if there are obstructions between the positioning satellites and the receiver antenna, or if there are nearby structures, the distance between the positioning satellites and the receiver antenna may become inaccurate due to obstruction, diffraction, and reflection. Furthermore, the GNSS receiver can not only determine the vehicle's position, but also its heading and speed, which can be measured by observing the frequency fluctuations of positioning radio waves from each satellite using the Doppler effect.

[0047] In addition, since pseudolites are the same system as GNSS, positioning satellites can be replaced with pseudolites signal transmitters, and they have similar characteristics to GNSS.

[0048] <Relative Position Estimation Unit 12>

[0049] The relative position estimation unit 12 uses the C2X device and / or the tag reader to accurately estimate the vehicle's position. The C2X device and the tag reader obtain the vehicle's relative position relative to the tags installed in the vehicle's driving environment and the position information output from the transmitter, and are therefore also referred to as the relative position acquisition sensor 1c. Furthermore, the relative position estimation unit 12 calculates and outputs the vehicle's absolute position based on the tag or transmitter position information and the vehicle's relative position relative to this information.

[0050] Because the distance between the marker or transmitter observed by the relative position acquisition sensor 1c and the vehicle is relatively short, and thus position errors due to diffraction or reflection are less likely to occur, the position estimation result of the relative position estimation unit 12 can be expected to be more accurate than the position estimation result of the absolute position estimation unit 11. On the other hand, the relative position estimation unit 12 estimates the vehicle's position and posture only in locations where markers are present or where C2X communication is established. Therefore, it can be considered discrete in terms of time and space.

[0051] The C2X device and tag reader described above are explained. Examples of C2X devices include beacon receivers that identify beacon transmitters located in the environment, road-to-vehicle communication devices that communicate with access points such as wireless LAN (Local Area Network) and Bluetooth, and vehicle-to-vehicle communication devices for receiving information from other vehicles. The tag reader uses sensors such as cameras or magnetic sensors to detect the type, position, and posture of characteristic markings, such as road markings and signs, or magnetic tags located in the environment, and transmits this information to the relative position estimation unit 12.

[0052] <Driving State Determination Unit 13>

[0053] Because the GNSS deviation tendency changes due to changes in (1) vehicle information such as turning angle and vehicle speed, and (2) satellite information such as satellite configuration and the number of visible satellites, the driving state judgment unit 13 has either or both the function of judging the persistence of the same driving environment and the function of judging the timing of driving state changes. Regarding vehicle information and reception environment, each parameter value is accumulated as time series information, and it is judged that the driving state has changed when a specified parameter value changes, when the amount of change in each parameter within a certain period of time exceeds a specified threshold, or when the cumulative amount of change in the parameter exceeds a certain value. In addition, the driving state judgment unit 13 sends the output result to the learning unit 15.

[0054] The above-mentioned vehicle information will be explained. This information includes vehicle speed information obtained from a rotational speed sensor located on the axle, a wheel speed pulse sensor located near each wheel that outputs a pulse each time the wheel rotates a certain angle, and a wheel speed pulse count sensor that outputs the number of wheel speed pulses per unit time; steering angle information obtained from a steering angle sensor that outputs the angle of the steering wheel and a steering wheel angle sensor that outputs the amount of steering wheel rotation; and gear position information indicating forward / reverse movement and shifting status.

[0055] The satellite information mentioned above is explained below. Satellite information refers to various parameters related to GNSS positioning, such as the positioning fix status, accuracy indicators such as position accuracy and velocity accuracy, accuracy degradation indicators such as pDOP and vDOP, the number of satellites used for positioning, the received signal strength of each positioning satellite, and the multipath flag, obtained from the GNSS receiver.

[0056] <Difference Calculation Unit 14>

[0057] The difference calculation unit 14 calculates the difference between the vehicle positions output by the absolute position estimation unit 11 and the relative position estimation unit 12. As described above, the absolute position estimation unit 11 outputs the estimated position frequently, but this may result in increased variation depending on the driving conditions. Furthermore, the relative position estimation unit 12 outputs the estimated position less frequently than the absolute position estimation unit 11, but with higher accuracy. Therefore, when calculating the difference between these estimated positions, the estimated positions may not be output at the same time, necessitating a time synchronization function. After synchronizing the times of the outputs from the absolute position estimation unit 11 and the relative position estimation unit 12, the difference is calculated. This difference can be the Euclidean distance obtained by adding the squares of the differences in latitude, longitude, and altitude for each estimated position, or it can be a distance divided into the forward and backward components, the left and right components, and the altitude components of the direction of travel at that time.

[0058] The distance calculated in this way is output to the learning unit 15 .

[0059] <Study Section 15>

[0060] Based on the difference calculated by the difference calculation unit 14, the learning unit 15 calculates a correction amount for correcting the estimated position of the absolute position estimation unit 11. For example, after the driving state determination unit 13 outputs the timing of a driving state change, the difference between the absolute and relative position estimation outputs from the difference calculation unit 14 at the time of initial observation of the marker and the output of the relative position estimation unit 12 may be used directly as the correction amount. Furthermore, this correction amount may be retained and continuously output to the position correction unit 16 until the driving state determination unit 13 next outputs the timing of a driving state change.

[0061] <Position Correction Unit 16>

[0062] The position correction unit 16 receives the outputs of the absolute position estimation unit 11, the driving state determination unit 13, and the learning unit 15, and corrects and outputs the estimated position of the absolute position estimation unit 11. For example, if the driving state determination unit 13 determines that the same driving state is continuing, it directly uses the correction amount from the learning unit 15 and outputs a corrected position by adding the correction amount to the output position of the absolute position estimation unit 11. On the other hand, if the driving state determination unit 13 determines that the driving state is not continuing, it uses a value that reduces the correction amount from the learning unit 15 by a predetermined percentage and outputs a corrected position by the output position of the absolute position estimation unit 11. The position correction unit 16 also outputs an accuracy indicator for the correction result.

[0063] Next, the processing contents of each unit of this embodiment will be described using flowcharts as needed.

[0064] <Processing Contents of Absolute Position Estimation Unit 11>

[0065] First, Japanese Patent No. 6482720 is known as a technical document related to vehicle position detection used in the absolute position estimation unit 11. This publication discloses a positioning device (positioner device) and a positioning method for performing lane-level precision positioning using a GNSS receiver or the like.

[0066] In the field of measurement, there is a high-precision positioning method (carrier phase positioning) that measures the carrier phase of positioning signals from GNSS receivers. This carrier phase positioning method requires a dual-frequency GNSS receiver with a high-precision clock, which is expensive. On the other hand, the existing coded positioning method used in vehicle position estimation devices can be performed using inexpensive single-frequency GNSS receivers.

[0067] However, the accuracy of the clock of such a GNSS receiver is low (approximately 1 microsecond in the case of a vehicle-mounted GNSS receiver). In order to obtain higher positioning accuracy, the bias error of the GNSS receiver clock (receiver clock bias error) needs to be corrected with high precision.

[0068] To solve this problem, Japanese Patent No. 6482720 discloses a positioning device capable of accurately correcting a positioning error caused by a receiver clock bias error of a GNSS receiver.

[0069] The absolute position estimation unit 11 of this embodiment estimates the absolute position by using such conventional technology.

[0070] The absolute position estimation unit 11 outputs not only the positioning position but also satellite information such as which satellite's positioning radio waves were used to calculate the positioning position (positioning satellite) and the radio wave intensity from each satellite.

[0071] <Processing Contents of the Relative Position Estimation Unit 12>

[0072] Next, use Figure 3 The relative position estimating unit 12 is described with reference to the flowchart of FIG. Figure 3 The flowchart shown is continuously executed during the vehicle position estimation operation of the vehicle position estimation unit 10 .

[0073] In step S1, the relative position estimation unit 12 determines whether a marker, beacon, or other reference is present when determining the relative position of the vehicle. The relative position estimation unit 12 can estimate the absolute position of the vehicle if a marker, beacon, or other reference is present within the detection range of a marker reading device or C2X device, a specific example of the relative position acquisition sensor 1c. Therefore, the relative position estimation unit 12 first determines the presence of a marker, beacon, or other reference based on the output of these devices. Specifically, if the reference is an image marker (a two-dimensional code typified by a QR code (registered trademark)), the marker reading device detects a finder pattern and determines the presence of a marker if its contrast and size are above a certain level; otherwise, the C2X device determines the absence of a marker. Alternatively, if the reference is a beacon, the C2X device determines whether it can detect a signal of a specific frequency with a predetermined intensity or above. If so, the C2X device determines the presence of a beacon, etc.; otherwise, the C2X device determines the absence of a beacon, etc.

[0074] In step S2, the relative position estimator 12 calculates the relative distance and relative direction from a reference marker or beacon. This is because the detectable range of the C2X device or marker reader is limited to a certain range, so the distance from the marker or beacon is calculated to improve the accuracy of position recognition. For example, if the reference is a QR code, the distance and relative direction can be calculated based on the apparent size of the time-series pattern between positioning marks. Alternatively, if the reference is a beacon emitting a signal at a specific frequency, the moment of passing directly under the beacon can be detected by detecting changes in the Doppler frequency. Since the beacon is typically located at a different height from the plane in which the C2X device moves during vehicle travel, the distance can be calculated separately for both horizontal and vertical components by taking into account vehicle speed.

[0075] Alternatively, when there are multiple beacons, the frequency change of each beacon can be observed separately, and the relative relationship with each beacon can be calculated based on the principle of triangulation.

[0076] In step S3, the relative position estimation unit 12 reads the absolute position information of the marker or beacon embedded in the signal received from the marker or beacon. For example, in the case of a QR code, information regarding latitude, longitude, and azimuth, as well as index information for searching a database (not shown) can be pre-embedded in the data portion, and this information can be read. In the case of a beacon, information regarding latitude, longitude, and azimuth, as well as index information for searching a database (not shown) can be superimposed on the transmission frequency of the beacon, and the code can be decoded.

[0077] In step S4, the relative position estimation unit 12 calculates the absolute position of the vehicle, calculates the driving direction, and calculates a reliability index for the calculated absolute position and driving direction. The absolute position and driving direction are calculated by combining the relative relationship with the marker or beacon obtained in step S2 and the absolute position and setting direction of the marker or beacon obtained in step S3, so that the time is aligned. For a QR code, the reliability index is defined as a function where the reliability is high if there is no difference between the data portion and the error correction coding portion based on Reed-Solomon coding, etc., and the reliability decreases as the number of bits required for information recovery through error correction processing increases. For beacons, the reliability index is also defined as a function where the greater the difference between the data portion and the checksum, the lower the reliability. Alternatively, when using a camera for external recognition, threshold ranges can be set for contrast in the lighting environment, frequency components in the vehicle's motion, and edge straightness in the object's condition, based on lighting conditions such as dusk or backlighting, vehicle motion such as image jitter caused by high speeds and sharp turns, and object conditions such as contamination and wear. A function can be defined such that the more conditions that fail to meet the threshold ranges, the lower the reliability. The output of this function can be used as a reliability indicator.

[0078] In step S5, the calculation results of the relative position estimation unit 12 are output to the backend processing. The output content includes one or more of the vehicle's absolute position, driving direction, reliability index, and the time information of these observations. For example, if the vehicle position estimation unit 10 is installed in the vehicle cabin and connected to other devices via CAN (Car Area Network), the output processing repackages the output content into CAN output packets and transmits them.

[0079] <Processing Contents of the Driving State Determination Unit 13>

[0080] Next, the driving state determination unit 13 will be described. Based on changes in (1) vehicle information such as turning angle and speed, and (2) satellite information such as satellite configuration and the number of visible satellites, the driving state determination unit 13 calculates changes in driving direction and average speed, and determines whether the same driving state continues or whether the driving state has changed. The following specifically describes how the driving state determination unit 13 determines changes in driving state.

[0081] <How to determine driving state changes caused by left or right turns>

[0082] The driving state determination unit 13 determines the driving state change based on the left or right turn detected from the vehicle information. Specifically, it determines whether the vehicle is turning left or right at an intersection, or entering or merging into a road different from the road it was traveling on. Figure 4 and Figure 5 The driving behavior of the vehicle is estimated based on the observation values ​​from the rotation angle sensor and the steering angle sensor or the steering wheel angle sensor of the above-mentioned axle portion.

[0083] Figure 4 This is a schematic diagram of vehicle V turning approximately 90 degrees to the right. As can be seen from the figure, vehicle V is moving straight from time t = t0 to t2, is turning from time t = t3 to t5, and is moving straight again after turning from time t = t6 to t7.

[0084] Figure 5The steering angle and turning angle of vehicle V at this time are shown. Using a vehicle model such as the Ackermann model, a typical four-wheeled vehicle V is modeled into two wheels: a driving wheel and a steering wheel. The turning radius is calculated based on the steering angle of the steering wheel, and the travel distance is calculated based on the rotation of the driving wheel. This allows the turning angle to be calculated. As at time t3, if a turning angle exceeding a threshold value (thr1) is calculated within a certain period of time, it is considered that the vehicle has made a left or right turn at an intersection, entering or merging onto a different road from the previously traveled road. This indicates that the driving environment (t2) before the steering angle began to change is no longer consistent. If the steering angle returns to neutral (t6) and remains neutral for a certain period of time, the moment the steering angle returns to neutral (t6) is considered to indicate the end of the driving state change. The term "neutral" in this context refers to a state below a certain threshold value (thr1).

[0085] When turning left or right at an intersection, the positional relationship with the obstructions on the left and right of the vehicle changes. For example, when driving in a north-south urban area surrounded by densely populated high-rise buildings, there are no buildings in the north-south direction, i.e., in the front and rear directions of the vehicle, and there are buildings in the east-west direction, i.e., on the left and right sides of the vehicle. At this time, the positioning radio waves directly received from the positioning satellites in the east-west direction are blocked by the buildings. Even if the positioning radio waves can be received, there is a high possibility that they will be reflected on the high-rise buildings and cause multipath. When multipath occurs, the transmission distance of the positioning radio waves is extended compared to the case where the positioning radio waves are directly received. Therefore, it can be considered that when the positioning radio waves that have caused multipath are used to locate the position of the vehicle, the deviation in the direction of the positioning satellite increases. Under the above-mentioned conditions, the positioning radio waves from the positioning satellites in the north-south direction from the perspective of the vehicle are less likely to cause multipath, while the positioning radio waves from the positioning satellites in the east-west direction are more likely to cause multipath, so it is easy to result in deviations in the east-west direction.

[0086] Next, when the vehicle's driving direction changes due to a left or right turn at an intersection, positioning radio waves from positioning satellites located in the east-west direction, as seen from the vehicle, are less likely to experience multipath. However, positioning radio waves from positioning satellites located in the north-south direction are more likely to experience multipath, resulting in a tendency for the vehicle to exhibit a north-south deviation. By detecting left or right turns at intersections, determining when the same driving environment is no longer continuation, or determining when the driving state change has ended, these changes in state can be addressed.

[0087] <How to determine driving state changes caused by vehicle speed changes>

[0088] The driving state determination unit 13 determines a change in the driving state based on the vehicle speed detected from the vehicle information. Specifically, the vehicle speed is used to determine whether the driving environment has changed from the road on which the vehicle was traveling previously, as described below.

[0089] Generally, the speed limit is set to a low speed of about 30 km / h in a narrow environment such as a residential area, and the speed limit is set to about 60 km / h on a main road with multiple lanes on one side. Therefore, by observing the time series change of the average speed during driving, it is possible to judge the change of the driving environment. Specifically, for each driving environment, such as Figure 6 As shown, a speed range and its duration are defined. If the vehicle continues traveling in the speed range for a period of time or longer, the vehicle is judged to have transitioned to that driving state. Furthermore, if the speed range frequently changes and does not meet any conditions, the driving state can be considered as an unknown state.

[0090] Generally speaking, in confined environments such as residential areas, roads are narrow and sidewalks are also narrow, so buildings are often located close to the roadway. In such environments, surrounding positioning satellites are easily obscured, and the combination of satellites used for positioning is prone to frequent switching, resulting in unstable positioning results. On the other hand, on arterial roads, the lanes are wide and there are sidewalks, so buildings are often located at a distance from the roadway. In such environments, it is easier to directly observe positioning satellites than in the narrow environments mentioned above, and positioning results tend to be stable. If the positioning results change from an unstable state to a stable state, or vice versa, it can be determined that the same driving environment is no longer continuing or that the driving state change has ended, thereby enabling the switching of such changes.

[0091] <Method for determining driving status changes due to changes in satellite combination>

[0092] The driving state determination unit 13 determines changes in driving state based on changes in the combination of satellites used, as detected from satellite information. In satellite positioning, when pseudoranges can be calculated for three or four or more satellites, the absolute position of the observation point can be calculated using the intersection of a sphere centered on the position of each satellite and radii of the pseudoranges. However, the calculated pseudoranges are typically affected by various errors, such as satellite orbit error, ionospheric delay error, and tropospheric delay error, so the intersection of the spheres does not always occur at a single point. Therefore, pseudoranges are often calculated for more satellites, and points close to the intersection are calculated using methods such as the least squares method. Consequently, if the number of satellites from which pseudoranges can be calculated deviates toward a specific orientation, or if the satellite configuration used for positioning is biased, the absolute position of the observation point will also be biased. Therefore, when the positioning satellite configuration changes, it is necessary to determine whether the same driving state is continuing or whether the driving state has changed.

[0093] Specifically, if the used satellite information output by the absolute position estimation unit 11 changes by a certain number or more (e.g., three or more satellites) between a certain time ago (e.g., two seconds ago) and the current time, the output driving status has changed. If the used satellites temporarily change between a certain time and then return to the original state, the output driving status remains unchanged.

[0094] For example, at time t -2 At time t, the five satellites G1 to G5 are -1 If the five satellites G1 to G4 and G6 are positioning satellites at time t0, and the six satellites G1 to G4, G6 and G7 are positioning satellites at time t0, it is determined that the driving state has changed. -2 At time t, the five satellites G1 to G5 are -1 If three satellites G1 to G3 at time t0 and five satellites G1 to G5 at time t0 are positioning satellites, it is determined that the driving state has not changed.

[0095] <Method for determining driving state changes caused by changes in positioning accuracy>

[0096] The driving state determination unit 13 determines changes in driving state based on changes in the accuracy of positioning results detected from satellite information. In satellite positioning, when the absolute position of the observation point is calculated from the pseudoranges of multiple positioning satellites as described above, it is possible to calculate the geometric degradation of precision (GDOP) and the degradation of precision (PDOP) with respect to spatial coordinates based on the observation vector (DOP). This is a numerical value calculated based on the configuration of positioning satellites in the sky. In addition, there are also indices based on the value of the loss function when calculating the absolute position using the least squares method described above, and estimated accuracy indices of absolute position calculated based on the positioning state, such as whether it is 3D positioning or 2D positioning. In addition, the GNSS receiver manufacturer's own accuracy index (accuracy value) can also be output, but this is not specifically used in this embodiment. By determining whether these estimated accuracy indices are stable or significantly changing, it is determined whether the same driving state continues or whether the driving state has changed.

[0097] Specifically, if the pDOP changes by a threshold value (e.g., 1.0 or greater) within a certain period of time (e.g., 2 seconds), it can be determined that the driving state has changed. The thresholds used for these accuracy indicators can be fixed values ​​determined through experimentation, or thresholds appropriate for the region can be received using the communication unit 18 described later. The same applies to other estimated accuracy indicators.

[0098] <Processing Contents of Difference Calculation Unit 14>

[0099] Next, the difference calculation unit 14 will be described. The difference calculation unit 14 calculates the difference between the vehicle positions estimated by the absolute position estimation unit 11 and the relative position estimation unit 12. However, since position estimation is not necessarily performed at the same time, the difference between the vehicle positions estimated by the two position estimation units is calculated after time synchronization as described below.

[0100] In the case of a GNSS receiver, the absolute position estimation unit 11 obtains absolute positions at regular intervals of approximately 0.1 to 1 second. Meanwhile, the relative position estimation unit 12 obtains position information of markers, beacons, and their relative positions at irregular intervals of approximately several seconds to several minutes. Therefore, the difference calculation unit 14 performs a difference calculation aligned with the output timing of the relative position estimation unit 12 or the time information of the time when the marker, beacon, or the like was observed, as included in the output of the relative position estimation unit 12.

[0101] For example, the difference calculation unit 14 uses linear interpolation to transform the output of the absolute position estimation unit 11 into a value that aligns with the output of the relative position estimation unit 12, and then calculates the difference. Here, the vehicle position estimated by the absolute position estimation unit 11 at time t is defined as P(t), and the vehicle position estimated by the relative position estimation unit 12 at time s is defined as Q(s). Figure 7 The diagram schematically shows the relationship between the estimation timings of P(t) and Q(s). The absolute position estimation unit 11 estimates P(t n ) is the time t n The time when the relative position estimation unit 12 estimates Q(s) is then s. n Wherein, n is an integer, which increases by 1 every time the absolute position estimation unit 11 performs position estimation. n Time t before n Defined as time s nb , set time s n The time after t n+1 Defined as time s na Among them, b and a are subscripts representing before and after respectively.

[0102] In this case, regarding s1 and s3, b 、s a As described below.

[0103] [Mathematical formula 1]

[0104] s 0b =t0,s 0a =t1...(Formula 1)

[0105] [Mathematical formula 2]

[0106] s 3b =t3,s 3a =t4...(Formula 2)

[0107] Here, let:

[0108] [Number 3]

[0109] Δs 0b =s0-s 0b , Δs 0a =s 0a -s0...(Equation 3)

[0110] [Number 4]

[0111] Δs 3b =s3-s 3b , Δs 3a =s 3a -s3...(Formula 4)

[0112] At this time, linear interpolation can be used to calculate the values ​​P(s0) and P(s3) that are aligned with the output timing of the relative position estimation unit 12 from the output of the absolute position estimation unit 11 as follows.

[0113] [Number 5]

[0114]

[0115] [Number 6]

[0116]

[0117] After calculating P(s0) and P(s3) based on the output of the absolute position estimation unit 11, the difference calculation unit 14 calculates the difference from Q(s0) and Q(s3) at the same time. For example, the difference calculation can use the Euclidean distance obtained by summing the squares of the differences in latitude, longitude, and altitude of each estimated position, and its Euler angle as the difference vector. The difference calculation unit 14 outputs this difference vector to the learning unit 15.

[0118] <Modification of Processing Contents of Difference Calculation Unit 14>

[0119] The above example uses linear interpolation to convert the output of the absolute position estimation unit 11 to a value that aligns with the output of the relative position estimation unit 12. However, the difference calculation unit 14 may also use earlier and later points to perform quadratic interpolation, bicubic interpolation, or the like. This can improve the accuracy of the interpolation approximation and the accuracy of the calculated differences.

[0120] Furthermore, in the case where the output frequency of relative position estimation unit 12 is higher than the output frequency of absolute position estimation unit 11, or when the interval is several times longer and relatively close, difference calculation unit 14 may convert the output of relative position estimation unit 12 to a value aligned with the output timing of absolute position estimation unit 11. This allows for regular difference calculations aligned with the output timing of absolute position estimation unit 11, and can further accelerate the timing of updating the correction amount in learning unit 15 and position correction unit 16.

[0121] Furthermore, the difference calculation unit 14 may calculate a vector, based on the current travel direction, as the difference vector calculated after synchronizing the vehicle's position, divided into a longitudinal component, a lateral component, and a height component. This allows the position correction unit 16 to perform fine position corrections separately for the longitudinal and lateral directions of the vehicle's travel direction.

[0122] <Processing Contents of Learning Unit 15>

[0123] Next, the learning unit 15 will be described. The learning unit 15 accumulates the differences calculated by the difference calculation unit 14 as time-series data and, based on this accumulated time-series data, calculates corrections for the absolute position estimation unit 11. For example, the simplest method involves performing learning at the time the difference calculation unit 14 calculates the differences, directly using the differences as corrections. This method can be used when the relative position estimation unit 12 is highly reliable. It assumes that the recognition error of the markers and transmitters that the relative position estimation unit 12 can recognize is sufficiently small, and the corrections are generated to align with the vehicle position obtained based on the recognition results. By starting the calculations at the time the difference calculation unit 14 outputs the calculation results—that is, at the time the relative position estimation unit 12 observes the position information of the markers and beacons—the corrections for the absolute position estimation unit 11 can be determined quickly after the position information is detected. This allows corrections to be initiated quickly after the latest relative position estimation unit 12 results are calculated.

[0124] Furthermore, the learning unit 15 may also reset the learning state when the change in driving state is complete, and calculate a correction value based on the statistics of the position information obtained multiple times from markers or beacons. Specifically, upon receiving the judgment result of the moment when the driving state change has ended from the driving state judgment unit 13, the learning state or correction value up to that moment is reset. Furthermore, after receiving the output of the difference calculation unit 14 obtained thereafter multiple times, the correction value can be determined as the value that minimizes the loss function based on the least squares method when the variance is below a threshold. Alternatively, a maximum likelihood estimation method can be used. This can eliminate the instability of the external recognition results and can be expected to improve the reliability of the estimated position.

[0125] Furthermore, the learning unit 15 may also use the reliability index of the relative position estimation unit 12. As mentioned above, recognition performance is reduced by the use of error correction information, lighting conditions, vehicle behavior, and object status. Therefore, information from markers or beacons with reliability indexes below a threshold can be excluded from the learning data or calculated with minimal emphasis. Specifically, in the case of the aforementioned least squares method, by multiplying these data points by a coefficient less than 1 when calculating the loss function and adding them, the influence of data points with low reliability indexes can be reduced, and the reliability of the estimated position can be expected to be improved.

[0126] <Processing Contents of Position Correction Unit 16>

[0127] Next, use Figure 8 The flowchart of the position correction unit 16 is described. Figure 8The flowchart shown is executed each time the vehicle position is output by the absolute position estimation unit 11. The position correction unit 16 receives the outputs of the absolute position estimation unit 11, the driving state determination unit 13, and the learning unit 15, and outputs the corrected estimated position and error index.

[0128] In step S11, the position correction unit 16 determines whether the same driving environment is continuing based on the output of the driving state determination unit 13. If the driving state determination unit 13 determines that the same driving environment is no longer continuing and that the driving state is changing (branching on the "No" side), since correction using the correction amount from the learning unit 15 is inappropriate, the output value of the absolute position estimation unit 11 is output without using the correction amount (step S14). On the other hand, if it is determined in step S11 that the same driving state is continuing (branching on the "Yes" side), the process proceeds to step S12.

[0129] In step S12, the position correction unit 16 determines whether learning by the learning unit 15 has been completed. If learning by the learning unit 15 has not yet been completed (the "No" side of the branch), since correction using the correction amount by the learning unit 15 is inappropriate, the output value of the absolute position estimation unit 11 is output as is without using the correction amount (step S14). On the other hand, if learning by the learning unit 15 has been completed or the correction amount is determined using only the primary output result of the relative position estimation unit 12 (the "Yes" side of the branch), the process proceeds to step S13.

[0130] In step S13 , the position correction unit 16 determines whether the correction amount in the learning unit 15 has changed by a threshold or more compared to the previous correction amount.

[0131] If the correction amount output from the learning unit 15 has changed by a value greater than the threshold value compared to the correction amount at the previous moment (branch is one side), the process proceeds to step S15. In step S15, the position correction unit 16 adds a correction amount obtained by interpolating the previous correction amount and the current correction amount to the absolute position estimation amount in order to mitigate the change in the correction amount, and outputs the correction amount. Specifically, assuming that the correction amounts output from the learning unit 15 until the last time are Δx(tN), Δx(t-N+1), ..., Δx(t-1), and assuming that the correction amount output from the learning unit 15 this time is Δx(t), Δc(N)Δx(tN)+Δc(N-1)Δx(t-N+1)+...+Δc(1)Δx(t-1)+Δc(0)Δx(t) can be used as the current correction amount. Here, ΣΔc(t)=1, t=t0~t N .

[0132] On the other hand, if in step S13 there is no change in the correction amount from the previous time by a value greater than the threshold (the "No" side of the branch), the process proceeds to step S16. In step S16, the position correction unit 16 deems that a stable correction amount has been obtained from the learning unit 15, adds the correction amount from the learning unit 15 to the result from the absolute position estimation unit 11, and outputs the result.

[0133] Furthermore, in steps S14, S15, and S16, the position correction unit 16 calculates an accuracy index. For example, the accuracy index may be the radius of the error circle, and in step S14, the accuracy index output by the absolute position estimation unit 11 may be directly output. For example, in step S15, assuming that the accuracy indexes output by the absolute position estimation unit 11 up to the last time are R(tN), R(t-N+1), ..., R(t-1), and assuming that the accuracy index output by the absolute position estimation unit 11 this time is R(t), Δc(N)ΔR(tN)+Δc(N-1)ΔR(t-N+1)+...+Δc(1)ΔR(t-1)+Δc(0)ΔR(t) may be used as the accuracy index for this time. Furthermore, in step S16, the reliability index of the relative position estimation unit 12 may be multiplied by an appropriate multiplier to convert it into an accuracy index.

[0134] While the output from the position correction unit 16 is delayed by various processing times, it is synchronized with the output timing of the absolute position estimation unit 11. The absolute position estimation unit 11 acquires positioning information from GNSS and other sources at a regular rate of approximately 2 Hz, for example. Therefore, it is expected that the position correction unit 16 will also output information at a regular interval. Autonomous driving systems require constant knowledge of the vehicle's current location, and the configuration of this embodiment can meet these requirements.

[0135] As described above, according to the vehicle control device of this embodiment, by learning the appropriate correction amount (offset) of the positioning position corresponding to the driving environment and driving state, the influence of the change in the driving state is eliminated. Even if the driving environment and driving state change, the high-precision estimation of the vehicle position in the map can be obtained at a high frequency.

[0136] Example 2

[0137] Next, the vehicle position estimating unit 10 according to the second embodiment of the present invention will be described. Note that redundant descriptions of points common to the first embodiment will be omitted.

[0138] Figure 9This is a block diagram illustrating the vehicle position estimation unit 10 of this embodiment. As shown here, the vehicle position estimation unit 10 includes an absolute position estimation unit 11, a relative position estimation unit 12, a driving state determination unit 13, a difference calculation unit 14, a learning unit 15, and a position correction unit 16, as well as a learning control unit 17 and a communication unit 18. In addition, the driving state determination unit 13 of this embodiment also receives input of map data M including driving route information. T The relative position estimation unit 12 of this embodiment also inputs the map data M including the landmark information. L In addition, the map data M of this embodiment T and map data M L yes Figure 1 A part of the map data M. The following describes the details of each part, focusing on the features of this embodiment.

[0139] <Absolute Position Estimation Unit 11>

[0140] The absolute position estimation unit 11 is the same as that in the first embodiment, and therefore its description is omitted.

[0141] <Relative Position Estimation Unit 12>

[0142] The relative position estimation unit 12 uses an external recognition sensor, a C2X device, or a tag reading device to accurately estimate the vehicle's position. The external recognition sensor, C2X device, and tag reading device obtain a relative position based on the absolute position of landmarks, tags, and transmitters in the vehicle's driving environment, and are therefore also referred to as a relative position acquisition sensor 1c. The relative position estimation unit 12 then calculates and outputs the vehicle's absolute position based on the vehicle's relative position based on the absolute positions of the landmarks, tags, and transmitters. The location information of the landmarks, tags, and transmitters used at this time can be retrieved and used in map data M containing the landmark information. L The information stored in the may also use the position information output from the marker or transmitter as described in Example 1.

[0143] Since the landmarks observed by the relative position acquisition sensor 1c are relatively close to the vehicle, position errors due to diffraction and reflection are unlikely to occur. Therefore, the position estimation results from the relative position estimation unit 12 can be expected to be more accurate than those estimated by the absolute position estimation unit 11. Meanwhile, the relative position estimation unit 12 estimates the vehicle's position and posture only in locations where landmarks are present or where C2X communication is established. Therefore, this estimation can be considered discrete in terms of time and space.

[0144] The above-mentioned external recognition sensor is explained below. The external recognition sensor refers to, for example, a single-lens camera or a stereo camera, a laser radar (Lidar), a millimeter wave sensor, an ultrasonic sensor, or other measuring equipment.

[0145] Single-lens cameras or stereo cameras are used to capture images of the exterior of the vehicle V. The cameras are installed on the inside of the front window and rear window of the vehicle V. The cameras can also be installed on the left and right sides of the vehicle V, or on top of the license plate outside the vehicle, for example. The cameras capture images of the front and rear of the vehicle V and transmit them to the relative position estimation unit 12. The stereo cameras, in turn, transmit depth information derived from binocular parallax to the relative position estimation unit 12.

[0146] LiDAR, millimeter wave, and ultrasonic sensors are radar sensors that use light, radio waves, sound waves, and the like to detect obstacles around the vehicle V. The radar sensor transmits light, radio waves, sound waves, and the like toward the area surrounding the vehicle V and receives light, radio waves, sound waves, and the like reflected from obstacles. This radar sensor detects the presence of obstacles and information such as their type, speed, and distance, and transmits this information to the relative position estimation unit 12. Obstacles include structures such as guardrails and sign posts, buildings, and moving obstacles such as pedestrians, bicycles, and other vehicles.

[0147] For map data M including landmark information L The map data M L This includes the types and configurations of objects on the ground that can be detected by external recognition sensors. For example, for cameras, these include straight or dashed lines such as white or yellow lines that divide lanes, diagonal lines indicating restricted access, speed limit markings and vehicle classification markings that indicate traffic rules and restrictions, stop lines, crosswalks, and three-dimensional objects that come into view while driving, such as guardrails, traffic lights, signs, road signs, billboards, and the exteriors of stores and buildings. Furthermore, for LiDAR, millimeter-wave, and ultrasonic sensors, these include guardrails, reflective road signs, manhole covers, and sign posts.

[0148] Next, for the map data M including the driving route information T The map data M TThis includes road information, structure information, geographic topography information, and route information. Road information refers to road location information (node ​​information) and connection information (link information) for each location, as well as the number of lanes, road width, speed limit information, the presence and width of shoulders, the presence and width of sidewalks, and other information. Structural information includes the width and height of buildings such as high-rise buildings, residential buildings, and pedestrian bridges, their relative position to the road, the presence and height of soundproof walls or windbreaks installed on the sides of the road, the internal and external conditions of tunnels and rockfall protection roofs, and three-dimensional road information around bridges and elevated roads. Geographic topography information refers to land use conditions at various locations, such as residential areas, commercial areas, farmland, and forests, as well as topographic information including road and land undulation information. Route information can be a driving route generated by the route generation unit 50, or, in the case of mobile vehicles such as buses traveling on a predetermined route, a fixed driving route generated in advance.

[0149] <Driving State Determination Unit 13>

[0150] Because the GNSS deviation tendency changes due to changes in (1) vehicle information such as turning angle and vehicle speed, (2) satellite information such as satellite configuration and the number of visible satellites, and (3) map information including three-dimensional environments such as terrain and surrounding buildings, the driving state judgment unit 13 has either or both the function of judging whether the same driving environment continues or the function of judging the timing of driving state changes. Regarding vehicle information and reception environment, each parameter value is accumulated as time series information, and it is judged that the driving state has changed when a specified parameter value changes, when the amount of change in each parameter within a certain period of time exceeds a specified threshold, or when the cumulative amount of change in the parameter exceeds a certain value. In addition, the driving state judgment unit 13 sends the output result to the learning control unit 17.

[0151] <Learning Control Unit 17>

[0152] The learning control unit 17 controls the initialization of the learning state of the learning unit 15, the updating of the learning content, the learning frequency, and the like. Specifically, it accumulates the results of the driving state determination unit 13 and, based on the time series information thereof, outputs a learning instruction to the learning unit 15. For example, performance can be improved by changing the operation of the learning unit 15 depending on the situation such as long driving on a suburban highway or driving on narrow roads in an urban area with repeated left and right turns.

[0153] When driving on a suburban highway for a long time, the driving state continues in the same environment, and it is expected that the relative position estimation unit 12 can perform high-precision position estimation multiple times, so a learning instruction of the majority of the multiple results obtained within a certain period of time using the difference calculation unit 14 can be output to the learning unit 15.

[0154] On the other hand, when driving on a narrow road in the city with repeated left and right turns, the driving state switches in a short time, so the learning unit 15 can output a learning instruction directly outputting the difference between the absolute position estimation unit and the relative position estimation unit output by the difference calculation unit 14 at the time when the mark is first observed and the relative position estimation unit 12 outputs the change in driving state after the driving state judgment unit 13 outputs the change in driving state such as left and right turns.

[0155] <Difference Calculation Unit 14>

[0156] The difference calculation unit 14 is the same as that described in the first embodiment, and therefore its description is omitted.

[0157] <Study Section 15>

[0158] The learning unit 15 calculates a correction amount for correcting the estimated position of the absolute position estimating unit 11 based on the difference calculated by the difference calculating unit 14 and the learning instruction output by the learning control unit 17. For example, if the learning control unit 17 outputs a learning instruction to output a mode using multiple results obtained within a certain period of time, the differences within the certain period of time are accumulated and the mode is output. Alternatively, a reliability indicator of the correction amount output by the learning unit 15 may be output based on the time since the difference information was obtained from the difference calculating unit 14, the estimated error amount of the absolute position estimating unit 11, and the estimated error amount of the relative position estimating unit 12.

[0159] <Position Correction Unit 16, Communication Unit 18>

[0160] The position correction unit 16 receives the outputs of the absolute position estimation unit 11, the driving state determination unit 13, and the learning unit 15, corrects and outputs the estimated position of the absolute position estimation unit 11, and bidirectionally inputs and outputs information with the communication unit 18. For example, when the driving state determination unit 13 determines that the same driving state is continuing, it directly uses the correction amount from the learning unit 15 and outputs a corrected position by adding the correction amount to the output position of the absolute position estimation unit 11. On the other hand, when the driving state determination unit 13 determines that the driving state is not continuing, it uses a value that reduces the correction amount from the learning unit 15 by a predetermined percentage and outputs a corrected position to the output position of the absolute position estimation unit 11. In addition to the output position and the learning amount, the communication unit 18 also transmits information such as the date and time, and the vehicle type of the vehicle. Furthermore, when correction information is received from outside the vehicle via the communication unit 18, the reliability index of the correction amount obtained from the learning unit 15 can be used to determine whether to use the correction information from the communication unit 18 or the correction information from the learning unit 15.

[0161] The input and output between the communication unit 18 and the communication unit 18 will be described. The communication unit transmits information such as the output position and learning amount from the position correction unit 16, the date and time, the vehicle type information, the reliability index from the learning unit 15, and the vehicle position from the absolute position estimation unit 11 to a server (not shown).

[0162] The processing in the server (not shown) is not directly related to the present invention and will not be described in detail here. However, for example, the server accumulates output positions and learned values, date and time, vehicle type information, learned reliability indicators, and vehicle position transmitted from multiple vehicles to which the present invention is applied. Furthermore, correction values ​​are accumulated for each type of positioning satellite configuration that can be estimated based on the date and time, positioning sensor antenna height that can be estimated based on vehicle type information, and driving location. This allows estimation of the deviation imposed on the absolute position sensor at a certain longitude and latitude at a certain date and time when a specific vehicle is traveling in an environment with a specific satellite configuration. It is assumed that these estimated values ​​are transmitted to the communication unit 18 of the vehicle traveling at that longitude and latitude at that date and time.

[0163] In this way, the communication unit 18 can obtain correction information from a server (not shown) and transmit the correction information to the position correction unit 16 .

[0164] Next, the processing contents of each unit of this embodiment will be described using flowcharts as needed.

[0165] <Processing Contents of Absolute Position Estimation Unit 11>

[0166] The absolute position estimation unit 11 outputs not only the positioning position but also satellite information such as which satellite's positioning radio waves were used to calculate the positioning position (positioning satellite), the radio wave strength from each satellite, etc. The details of the processing are the same as those described in Example 1 and are omitted.

[0167] <Processing Contents of the Relative Position Estimation Unit 12>

[0168] Regarding the processing contents of the relative position estimation unit 12, use Figure 10 The flowchart is described below. Figure 10 The flowchart shown is continuously executed during the vehicle position estimation operation of the vehicle position estimation unit 10 .

[0169] In step S1a, the relative position estimating unit 12 determines whether a landmark, marker, beacon, or the like serves as a reference for determining the relative position of the vehicle. Since the relative position estimating unit 12 can estimate the absolute position of the vehicle when a landmark, marker, beacon, or the like is within the detection range of an external recognition sensor, a marker reader, or a C2X device (specific examples of the relative position acquisition sensor 1c), the relative position estimating unit 12 first determines whether a landmark, marker, beacon, or the like is present based on the output of these devices.

[0170] Regarding the case where the reference is a mark or a beacon, since it has been described in the first embodiment, it is omitted here. L The method of identifying landmarks with an external recognition sensor is described in detail. For example, a use case using a camera, a specific example of an external recognition sensor, to identify crosswalk markings is described. Crosswalk markings are road markings with a periodic pattern of three or more 0.6m-wide white lines spaced 0.6m apart, arranged across the roadway, primarily at intersections.

[0171] Map Data M L The relative position estimator 12 stores information about the length of the white line of the intersection's crosswalk markings, the relative relationship between the end points and the angle between the white line, the number of white lines, the presence of side lines, and other information. Based on the vehicle's position calculated by the absolute position estimator 11, information about landmarks surrounding the vehicle is input into the relative position estimator 12. Specifically, the absolute position estimator 11 receives the vehicle's position, heading, and speed, which are not yet corrected and therefore contain errors. Therefore, a motion vector is calculated, with the vehicle's heading as the vector direction and the vehicle's speed multiplied by time T as the vector length. This vector is then added to the current vehicle position to calculate the vehicle's position after time T has elapsed. A margin area is created between the vehicle's position and the vehicle's position after time T has elapsed, and a certain range is further considered for this interval. The landmark information included in this area is output in ascending order of distance from the vehicle. This allows the landmarks on the vehicle's path to be output to the relative position estimator 12 in the order in which they appear.

[0172] From map data M L When receiving the landmark information of the pedestrian crossing markings on the vehicle's forward route, the pedestrian crossing markings are detected by the camera in step S1a. Figure 11The schematic diagram illustrates the pedestrian crossing marking detection. D1 is the image before processing input from the camera. In step S1a, longitudinal edge detection is performed on the input image to detect the pedestrian crossing marking to obtain D2. Thereafter, the line segment elements are extracted from D2 by performing conventional markings, etc., to obtain the endpoint coordinates and length. After grouping the line segment elements with similar lengths within a threshold value, the endpoint coordinates with a certain interval are extracted for each group to obtain D3. In addition, the interval can also vary within the threshold value because it is affected by the lens distortion of the camera. As mentioned above, the pedestrian crossing marking has more than 3 periodic white line elements, so it is a group of endpoint coordinates with a certain interval of more than 6 at the upper end and the lower end of the line segment element.

[0173] Furthermore, since crosswalk markings exist on the road, the camera's angle and height relative to the vehicle, as well as its field of view, can be used to estimate the vanishing point or horizon height in the image, allowing D4 to be extracted based on the coil elements on the road surface. If such endpoint coordinates are extracted, crosswalk markings are considered successfully recognized.

[0174] Furthermore, in the case of landmarks such as speed limit signs, which have prescribed shapes and text, a template image corresponding to the sign type is prepared in advance and compared with the camera image to determine its presence. Specifically, considering that the apparent size varies with distance, a method (template matching) is used to sequentially overlay a template image, scaled to multiple sizes, with a portion of the camera input image, calculating correlations. Successful recognition is determined by determining the presence of the detection target where the correlation value exceeds a threshold. Similarly to the crosswalk markings described above, the camera's angle of installation relative to the vehicle, its height, and field of view can be considered to limit the area within the image where the sign is detected.

[0175] Furthermore, a large number of prior arts are disclosed for road markings and structures other than crosswalk markings, and by using these, it is possible to determine whether an object exists.

[0176] Thus, in step S1a, it is determined whether a landmark, marker, beacon, etc. exists. If so (the "yes" side of the branch), the process proceeds to the next step, step S2a. If not (the "no" side of the branch), the process again attempts to detect whether a landmark, marker, beacon, etc. exists.

[0177] In step S2a, the relative position estimation unit 12 calculates the relative distance and relative direction to the landmark, marker, beacon, etc. used as a reference. This is because the detection range of the external recognition sensor, C2X device, and marker reading device is limited to a certain range. Therefore, in order to improve the accuracy of position recognition, the distance to the landmark, marker, beacon, etc. is calculated. The case where the reference is a marker or beacon is omitted because it has been described in Example 1. Here, the case where the reference is a marker or beacon is omitted. L A method of identifying landmarks using an external recognition sensor is described.

[0178] For example, a use case of using a camera to identify a crosswalk marking is described. Assume that in the processing of the above-mentioned step S1a, it is determined that a landmark exists, and the endpoint coordinate groups of the upper and lower ends of the line segment element extracted as the crosswalk marking are obtained. Using the setting angle, setting height, field of view, and lens distortion of the camera relative to the vehicle body, and assuming that the road surface directly below the vehicle is flat at least up to the crosswalk marking, the distance to the crosswalk marking when assuming that there is a crosswalk marking on the road can be calculated. Furthermore, by calculating the angle of arrangement of each endpoint of the endpoint coordinate group, the angle of the vehicle relative to the crosswalk marking can be calculated.

[0179] In addition, if the assumption that the road surface directly below the vehicle is flat at least up to the crosswalk marking does not hold true, for example, if the vehicle is on a slope or a curve with an inclination angle, the vehicle can obtain the information from the map data M. L Using the size information of a landmark, it is also possible to calculate the distance to the landmark. For example, for a pedestrian crossing, use Figure 12 Provide explanation.

[0180] like Figure 12 As an example, the top point group Gr is used. U and the lower point group Gr D When the coordinates of the point group Gr are used, the distance to the pedestrian crossing marking and the ground inclination angle can be calculated. In the following, for convenience, although the point group Gr has multiple points, its center of gravity is used as the representative point. The distance from any point A to any point group Gr refers to the distance from point A to the representative point of the point group Gr. As mentioned above, the distance between the endpoint coordinates is 0.6m in the pedestrian crossing specifications, so using the camera information such as the size and field of view observed on the image, the distance from the camera origin O to the lower end of the point group Gr can be calculated separately. D The distance from the camera origin O to the upper end of the endpoint coordinate group Gr U distance.

[0181] Here, if Figure 13 As an example, let the camera origin O and the projection plane S be fixed, and determine the point P on the projection plane S from the camera origin O.H 、P L , the distance to the point P H '、P L ', the inclination of the straight line passing through these two points can be determined, so the point group Gr from the camera origin O to the bottom can be used to calculate the inclination of the straight line passing through these two points. D The distance from the camera origin O to the upper end of the endpoint coordinate group Gr U The inclination of the road can be estimated from the distance.

[0182] In step S3a, the relative position estimation unit 12 determines the absolute position information of the landmark, marker, or beacon. The marker and beacon have been described in the first embodiment and are omitted here. L A method of identifying landmarks using an external recognition sensor is described.

[0183] For example, if the landmark used as a reference is a crosswalk, Figure 11 As shown in D1, for the four corner endpoints P1 to P4, in the map data M L The latitude, longitude, altitude and other location information, the road marking setting direction, white line length, number of white lines, whether there is a side line, etc. are stored in the map data M. The number of white lines, whether there is a side line, and the setting direction of the pedestrian crossing marking identified in step S2a are compared with the map data M. L If the information stored in the is consistent, the longitude and latitude of the four corner endpoints of the crosswalk are read and used as the result.

[0184] In step S4, the relative position estimation unit 12 calculates the absolute position of the vehicle, calculates the driving direction, and calculates a reliability index for the calculated absolute position or driving direction. The absolute position and driving direction are calculated by combining the relative relationship with landmarks, markers, beacons, etc. obtained in step S2a and the absolute position and installation direction of the landmarks, markers, beacons, etc. obtained in step S3a, so that they are aligned at all times. The reliability index is omitted because it has been described in Example 1. The reliability index for the camera used for landmark recognition and the camera used as the external recognition sensor can be defined using the same principles.

[0185] In step S5, the calculation results of the relative position estimation unit 12 are output to the backend processing. The output content includes one or more of the vehicle's absolute position, driving direction, reliability index, and the time information of these observations. For example, if the vehicle position estimation unit 10 is installed in the vehicle and connected to other devices via CAN, the output processing repackages the output content into packets for CAN output and transmits it.

[0186] <When the external recognition sensor is LiDAR or millimeter wave>

[0187] The above description uses an example where the external recognition sensor is a camera, but other external recognition sensors can be used in the same manner. For example, when the external recognition sensor uses a laser radar or millimeter wave, the detection of whether a landmark is recognized in step S1a can be considered as a change in the reflection intensity in the scanning direction, that is, a spatial change in distance information and material information, and such change patterns are recorded in advance in the map data M. L In the above example, it is defined as whether there is a similarity above a threshold compared with the change pattern of the record.

[0188] In addition, the relative relationship with the vehicle in step S2a can define the distance information that can be measured by laser radar or millimeter waves as the distance, and the offset change of the reflection intensity change in the scanning direction as the angle.

[0189] In step S3a, the map data M L Reading landmarks is the same as for the camera.

[0190] The calculation method for absolute position in step S4 is the same as for cameras, and the reliability index can be defined as the reflection intensity of laser radar or millimeter waves. In the case of laser radar or millimeter waves, laser radar or millimeter waves are emitted toward the outside of the vehicle and the reflected waves are measured. Therefore, if the sensitivity is reduced due to the adhesion of raindrops or mud, the measured reflection intensity will decrease. In addition, the measurement data obtained in this case is distorted by the uneven adhesion of raindrops or mud, which can reduce reliability.

[0191] As described above, according to the vehicle control device of this embodiment, the vehicle position can be estimated with higher accuracy by utilizing landmark information or information acquired from the outside via the communication unit.

[0192] Example 3

[0193] Next, the vehicle position estimating unit 10 according to the third embodiment of the present invention will be described. Note that redundant descriptions of points common to the above-described embodiments will be omitted.

[0194] In this embodiment, a method of considering more detailed recognition characteristics when recognizing a landmark using an external recognition sensor in the relative position estimation unit 12 and reflecting the characteristics in the correction method will be described.

[0195] The third embodiment of the present invention is similar to the second embodiment in terms of device structure. The vehicle position estimation unit 10 includes an absolute position estimation unit 11, a relative position estimation unit 12, a driving state determination unit 13, a difference calculation unit 14, a learning unit 15, and a position correction unit 16, and includes map data M of driving route information. T , map data M including landmark information L, a learning control unit 17, and a communication unit 18. The internal processing of the relative position estimation unit 12, the difference calculation unit 14, and the position correction unit 16 is different.

[0196] Hereinafter, details of each part will be described while appropriately omitting the parts that overlap with those in Example 2.

[0197] The recognition characteristics of the relative position estimation unit 12 depend on the shape of the recognition target. The relative position estimation unit 12 can measure with high accuracy in a specific direction, but the accuracy is degraded or the position cannot be recognized at all in other directions.

[0198] Take the above-mentioned pedestrian crossing marking as an example for explanation. Usually, when a motor vehicle is traveling, it passes through the pedestrian crossing marking orthogonally. At this time, the endpoints of the white lines included in the point group Gr are periodic, and there is no feature to distinguish between the white lines, so it is difficult to distinguish when multiple white lines are arranged. That is, it is difficult to distinguish when there are multiple white lines arranged in the left and right directions of the motor vehicle, and there is a possibility of errors when identifying the positions in the left and right directions. On the other hand, for the front and rear directions of the motor vehicle, that is, the direction crossing into the pedestrian crossing marking (the direction orthogonal to the arrangement of the point group Gr), there is no such periodicity, and there is a feature that can uniquely identify the position.

[0199] For example, a vehicle typically travels parallel to the white lane lines. Using the boundary between the white lane lines and the asphalt road surface for positioning makes it easier to identify the vehicle's left-right position, but it's difficult to identify the vehicle's front-back position because the white lane lines are continuous.

[0200] Because there is such a feature that depends on the shape of the identified object and is easy to correct the position in a specific direction, it is possible to pre-define it based on rules, such as only the front-to-back direction (vertical direction) for pedestrian crossings, only the left-to-right direction (horizontal direction) for white lines, and both the front-to-back direction and the left-to-right direction for maximum speed limit markings.

[0201] Alternatively, in the map data M L In the landmark information, whether position correction is permitted in each direction is stored for each landmark. This allows for differentiating between situations such as narrow road crosswalks, where the entire area is easily captured by the external recognition sensor and left-right errors are less likely to occur, and situations such as multi-lane roads with many pedestrians and vehicles, where only partial obstruction can be detected by the external recognition sensor and left-right errors are more likely to occur. This makes it easier to obtain position information appropriate to the environment.

[0202] The relative position estimating unit 12 outputs the position of each landmark with respect to a direction component with respect to the vehicle's traveling direction, which is defined in advance for each landmark.

[0203] The difference calculation unit 14 calculates the difference between the absolute position estimation unit 11 and the relative position estimation unit 12, using the directional component output from the relative position estimation unit 12. Specifically, consider a situation where the absolute position estimation unit 11 indicates that the vehicle is traveling north, and the relative position estimation unit 12 outputs position information regarding a north-south component of a pedestrian crossing marking located east-west on a map. In this case, the difference between the absolute position estimation unit 11 and the relative position estimation unit 12 is calculated as the vehicle's north-south position difference, and the uncalculated difference is output as the vehicle's east-west position difference.

[0204] The learning unit 15 is the same as that in the second embodiment, and therefore its description is omitted.

[0205] The position correction unit 16 corrects and outputs the output of the absolute position estimation unit 11 based on the difference in each orientation output from the learning unit 15. This allows more detailed recognition characteristics to be considered when recognizing landmarks using the external recognition sensor and reflected in the correction method.

[0206] Description of Reference Numerals

[0207] 100…Vehicle control device, 1a…Vehicle information receiving unit, 1b…Absolute position acquisition sensor, 1c…Relative position acquisition sensor, 1d…Actuator, 1e…HMI, 10…Vehicle position estimation unit, 11…Absolute position estimation unit, 12…Relative position estimation unit, 13…Driving state determination unit, 14…Difference calculation unit, 15…Learning unit, 16…Position correction unit, 17…Learning control unit, 18…Communication unit, 20…Autonomous navigation integration unit, 30…Map comparison unit, 40…Automatic driving control unit, 50…Path generation unit, M, M T 、M L …map data.

Claims

1. A vehicle control device having a vehicle position estimating unit for estimating the vehicle position, characterized in that: The vehicle position estimation unit includes: an absolute position estimation unit for estimating the first vehicle position based on absolute position information acquired from the GNSS; a relative position estimating unit for estimating a second vehicle position based on relative position information acquired from outside the vehicle; a driving state determination unit for determining a change in the driving state of the vehicle based on vehicle information or satellite information; a difference calculation unit that calculates a difference after synchronizing the times of the first vehicle position and the second vehicle position; a learning unit that accumulates the difference as time-series data for each of the traveling states and learns a correction amount for the first vehicle position for each of the traveling states based on the accumulated time-series data; and A position correction unit corrects the first vehicle position based on the correction amount calculated by the learning unit and the driving state determined by the driving state determination unit.

2. The vehicle control device according to claim 1, wherein: The second vehicle position is a relative position based on the absolute position of the tag read by the tag reading device or a relative position based on the absolute position of the transmitter or access point device received by the C2X device.

3. The vehicle control device according to claim 1, wherein: Landmark information in which the absolute positions of landmarks around the vehicle are registered can be input into the relative position estimating unit. The second vehicle position is a relative position based on the absolute position of the landmark recognized by the external recognition sensor.

4. The vehicle control device according to claim 3, wherein: The landmark information stores position correction permission information for each direction for each landmark.

5. The vehicle control device according to claim 1, wherein: The driving state determination unit determines that the driving state of the vehicle has changed when detecting that the vehicle has turned left or right or when detecting that the vehicle speed has changed based on the vehicle information.

6. The vehicle control device according to claim 1, wherein: The driving state determination unit determines that the driving state of the vehicle has changed when detecting a change in the positioning satellite used to estimate the first vehicle position or a change in the accuracy of the positioning result based on the satellite information.

7. The vehicle control device according to claim 1, wherein: The difference calculation unit calculates the difference in the traveling direction of the vehicle by distinguishing between the front-back direction and the left-right direction. The position correction unit corrects the first vehicle position by distinguishing between a front-rear direction and a left-right direction with respect to the traveling direction of the host vehicle.

8. The vehicle control device according to claim 1, wherein: The vehicle further includes a learning control unit that controls initialization of a learning state of the learning unit, updating of learning content, or a learning frequency based on a determination result of the driving state determination unit.

9. The vehicle control device according to claim 1, wherein: It also has a communication unit for communicating with the outside of the vehicle. The position correction unit corrects the first vehicle position based on correction information received from outside the vehicle via the communication unit.

10. The vehicle control device according to any one of claims 1 to 9, wherein: Also includes: an autonomous navigation integration unit for calculating a vehicle position based on the vehicle information and the vehicle position estimated by the vehicle position estimation unit; a map comparison unit that estimates the vehicle position on the map based on the vehicle position calculated by the autonomous navigation integration unit and the map data; and The automatic driving control unit controls the steering system, driving system, and braking system of the vehicle based on the vehicle position and driving path in the map calculated by the map comparison unit.

11. A method for estimating the position of a vehicle, characterized in that: include: The step of calculating the first vehicle position based on the absolute position information obtained from the GNSS; The step of estimating the position of the second vehicle based on the relative position information obtained from outside the vehicle; a step of determining a change in the driving state of the vehicle based on vehicle information or satellite information; a step of calculating a difference after synchronizing the time of the first vehicle position and the second vehicle position; a step of accumulating the difference as time series data for each of the driving states, and learning a correction amount of the first vehicle position for each of the driving states based on the accumulated time series data; and A step of correcting the first vehicle position based on a correction amount obtained through learning and the determined traveling state.

Images