Patent Application
20110010131
This application claims priority from Japanese Patent Application No. 2009-163860 filed on Jul. 10, 2009 including the specification, drawings and abstract, the disclosure of which is incorporated herein by reference in its entirety.
Aspects of the present invention relate to a reliability evaluation device, a reliability evaluation method, and a reliability evaluation program which evaluate the reliability of map information.
Known related art (Japanese Patent Application Publication No. JP-A-2006-218950, for example) determines whether an erroneous control has been performed by calculating a road curvature radius based on vehicle behavior and comparing the calculated radius with a curvature radius specified in map information. If an erroneous control was performed, the corner control that referenced the map information is subsequently not performed. The related art also calculates the reliability of map information (Japanese Patent Application Publication No. JP-A-2007-225498, for example).
However, the related art is not capable of evaluating reliability based on a distribution property of a measurement value when a position of a feature or the like specified in map information is measured.
In other words, a curvature, a start position, and the like of a curve segment may be measured based on output signals from various types of sensors, such as an acceleration sensor and a position sensor (GPS or the like) installed in a vehicle. If the various sensors perform these measurements a plurality of times, the measurement values develop a distribution pattern having a wide variety of distribution properties. However, the reliability of measurement values having different distribution properties cannot be evaluated using the same evaluation method. For example, a different index should be used for determining whether a measurement value is reliable when a specific value from a distribution of the measurement value is assumed as the correct value, and when the correct value cannot be assumed but the measurement value is distributed within an allowable margin of error. As described above, the same evaluation method cannot be used to evaluate the reliability of measurement values having different distribution properties. Even comparing a value identified by vehicle behavior and a value specified in map information according to the related art cannot accurately identify the reliability of various types of information included in the map information.
Aspects of the present invention were devised in light of the above problems in order to evaluate reliability depending on a distribution property of a measurement value when measuring a position of a feature or the like specified in map information.
In order to achieve the above, according to an aspect of the present invention, a measurement value for measuring an evaluation object is acquired a plurality of times using a sensor installed in a vehicle, and a reliability evaluation method is selected depending on a dispersion of the measurement value. The reliability of the evaluation object value is then evaluated using the selected reliability evaluation method. Specifically, a plurality of reliability evaluation methods that correspond to the measurement value dispersion is prepared in advance, and the reliability evaluation method is selected depending on the measurement value dispersion. In these reliability evaluation methods, the reliability of the evaluation object value, which specifies a position of a feature or the like specified in the map information, is evaluated based on the measurement value of a sensor installed in the vehicle. Since the measurement value dispersion expresses a spreading property of the distribution for an average value in the measurement value distribution, the measurement value dispersion is an index that indicates whether a specific measurement value can be assumed from the distribution. The reliability of an evaluation object value is thus evaluated using a reliability evaluation method that is selected depending on the measurement value dispersion. Therefore, when the position of a feature or the like specified in map information is measured, the reliability can be evaluated based on the distribution property of the measurement value.
The evaluation object value acquisition unit may extract any evaluation object value from a plurality of evaluation object values included in the map information as the evaluation object. The evaluation object value is a value that can be measured by a sensor installed in the vehicle, and is also information that is stored in a storage medium in advance as part of the map information. Thus, the evaluation object value is not limited provided that it is a value pertaining to a feature or the like that relates to the map information. The evaluation object value may identify a position or curvature of a feature or the like, or a size or amount such as a road width. The evaluation object value may also express a specific meaning such as a flag that indicates the existence of a feature.
The measurement value acquisition unit is not limited provided that it can acquire the measurement value a plurality of times. The measurement value is not limited provided that it is a value for measuring the evaluation object value. In other words, the measurement value and the evaluation object value are values that can be directly compared. If an output value of the sensor employs the same system of units as the evaluation object value, the sensor output value may be used as the measurement value that measured the evaluation object, and a comparison made of the output value and the evaluation object value. If the sensor output value employs a different system of units, one of the output value and the evaluation object value is converted to match the other. The sensor is not limited provided that it is installed in the vehicle and can output a measurement value for measuring an evaluation object that relates to the map information, or a value that may be converted into the measurement value. Thus, in addition to sensors that measure physical quantities such as position, speed, acceleration and the like, the sensor may also be embodied as a camera or the like for measuring the position, speed, acceleration and the like of the vehicle or a feature on the road through image analysis.
The reliability evaluation unit is not limited provided that it selects the reliability evaluation method depending on the measurement value dispersion, and evaluates the reliability of the evaluation object value based on the measurement value. The measurement value dispersion may be directly or indirectly specified. For example, a configuration may be used that pre-associates dispersion magnitudes with a sensor type, a property of the measurement value, a method of computational processing for the measurement value, or the like, and considers the measurement value dispersion to be associated with the sensor type, the property of the measurement value, the method of computational processing for the measurement value, or the like. Obviously, the dispersion may be directly specified by acquiring the measurement value a plurality of times and actually computing the dispersion.
The selection of the reliability evaluation method depending on the dispersion is not limited provided that a configuration is used such that the reliability evaluation method is selected depending on a characteristic of the evaluation object that corresponds to the measurement value dispersion. Specifically, the evaluation object has a characteristic such that the smaller the measurement value dispersion, the more stably the measurement value can be measured. If the measurement value can be stably measured, the accuracy of the evaluation object itself can be evaluated based on the measurement value. For example, if the measurement value dispersion is small, a configuration may be used that selects the reliability evaluation method for evaluating reliability depending on the accuracy of the evaluation object. Meanwhile, an evaluation object having a larger measurement value dispersion is accompanied by a wider measurement value distribution range. Thus, if the measurement value dispersion is large, instead of depending on the measurement value, a configuration may be used that selects the reliability evaluation method for evaluating reliability depending on the measurement probability of measuring a measurement value within a set of measurement values, e.g. within a specific range.
The reliability of the evaluation object value is not limited provided that it is defined such that a higher reliability means a smaller negative effect on the vehicle when the evaluation object value is used for vehicle control. For example, a configuration may be used that considers the evaluation object value that includes a margin of error as the correct value and evaluates reliability depending on the magnitude of the effect on driving when vehicle control is performed. The reliability of the evaluation object value is not limited provided that it is evaluated based on the measurement value. A configuration that specifies a value computed from the measurement value, a value obtained by converting the measurement value, or a plurality of measurement values, and analyzes a distribution of the specified value, and a configuration that compares the specified value with the evaluation object value may be used.
As a specific example of the configuration of the reliability evaluation method, the reliability evaluation method may be defined such that the reliability is evaluated by referencing the measurement value, whose difference with the evaluation object value is within a predetermined threshold, and mutually different predetermined thresholds defined for a plurality of reliability evaluation methods. According to this configuration, by selecting the reliability evaluation method depending on the measurement value dispersion, the predetermined threshold is selected depending on the measurement value dispersion to evaluate reliability.
Specifically, for a small dispersion, the measurement values are distributed in a relatively narrow range when the measurement values are obtained a plurality of times, whereas for a large dispersion, the measurement values are distributed in a relatively wide range when the measurement values are obtained a plurality of times. Therefore, a narrow range may be used as the analysis object in the case of the former, and a wide range may be used as the analysis object in the case of the latter. Therefore, by selecting the reliability evaluation method such that a different threshold is used depending on the measurement value dispersion, the predetermined threshold is selected with the measurement values of a suitable range depending on the measurement value dispersion used as the analysis object.
If a configuration is used that selects the reliability evaluation method for evaluating the reliability by referencing the measurement value in a narrower range for a smaller measurement value dispersion, it is possible to evaluate reliability by referencing the measurement value of a relatively narrow range when the reliability of the evaluation object value can be evaluated using the measurement value of a narrow range having a small dispersion. Meanwhile, when the distribution of the measurement value in a specific range expresses a property of the measurement object, it is possible to evaluate reliability by referencing the measurement value of a relatively wide range. Therefore, the reliability of the evaluation object value is evaluated using the reliability evaluation method depending on the measurement value dispersion.
A configuration may be used that evaluates reliability based on a probability of measuring the measurement value. In other words, the probability of repeatedly measuring a specific measurement value indicates reproducibility and accuracy when measuring the specific measurement value. Hence, focusing on the measurement value with the highest probability of being measured, it is possible to assume that the measurement value with the highest probability of being measured is a value that better measures the evaluation object value. In the case of such an assumption, the measurement value is near the evaluation object value, and a higher measurement probability of the measurement value means that the evaluation object value is closer to the precise value of the evaluation object. Thus, the reliability of the evaluation object value can be evaluated using a range that includes the measurement value with the highest probability of being measured and the measurement probability of the measurement value. Hence, if a first range is defined as a range for determining whether the measurement is near the evaluation object value and the highest measurement probability is used as a first probability, the reliability evaluation method can be defined as evaluating the reliability of the evaluation object value based on the measurement value acquired a plurality of times.
Meanwhile, a probability of measuring the measurement value within a specific range indicates a probability that the measurement value is included within the specific range when the measurement value is acquired a plurality of times, and is an index that indicates whether the measurement value within the specific range is stably measured. Therefore, if an error range allowed for the evaluation object value is used as the specific range, the reliability of the evaluation object value can be evaluated using the probability of measuring the measurement value within the specific range. Hence, if the error range allowed for the evaluation object value is defined as a second range, the reliability evaluation method can be defined as evaluating the reliability of the evaluation object value based on the measurement value acquired a plurality of times.
By defining the reliability evaluation methods as described above, the reliability method can be selected depending on the measurement value dispersion. In other words, the measurement value dispersion is a distribution property when the measurement value is acquired a plurality of times, and among distribution properties characterizing the distribution of the measurement value, the measurement value dispersion is also a distribution property that is closely related to the reliability of the measurement value. Thus, by selecting the reliability evaluation method corresponding to the dispersion, reliability can be evaluated based on the distribution property of the measurement value when a position of a feature or the like specified in map information is measured. For an evaluation object for which the measurement value dispersion is small, a configuration may be used that selects the reliability evaluation method for evaluating reliability based on the first probability, which is a probability of measuring the measurement value with the highest probability of being measured, if the measurement value with the highest probability of being measured exists within the first range from the evaluation object value. For an evaluation object for which the measurement value dispersion is large, a configuration may be used that selects the reliability evaluation method for evaluating reliability based on the second probability, which is a probability of measuring the measurement value within the second range that includes the evaluation object value.
Further, the reliability evaluation method may be defined as evaluating reliability based on a difference between the measurement value with the highest probability of being measured and the evaluation object value, for example. In such case, the reliability of the evaluation object value can be evaluated on the assumption that the measurement value with the highest probability of being measured is the actual measurement value of the evaluation object. In other words, if the measurement value is a value that actually measures the evaluation object using a sensor of the vehicle, the evaluation object value included in the map information and the measurement value can be directly compared to evaluate the reliability of the evaluation object value. Note that, because the difference between the measurement value with the highest probability of being measured and the evaluation object value corresponds to reliability, various configurations are possible, including a configuration that compares the difference and a threshold and evaluates the reliability as low when the difference exceeds the threshold, and a configuration that evaluates reliability by associating the magnitude of difference and reliability.
The probability of measuring the measurement value in a range where the difference between the measurement value and the evaluation object value is equal to or greater than a predetermined value serves as an index that evaluates the proportion of measurement values separate from the evaluation object value by at least a predetermined value, i.e., measurement values that lower the reliability of the evaluation object value measured. Hence, the reliability evaluation method may be defined as evaluating reliability based on the probability of measuring the measurement value in a range where the difference between the measurement value and the evaluation object value is equal to or greater than a predetermined value. In such case, the reliability of the evaluation object value can be evaluated based on the measurement probability of measurement values that lower the reliability of the evaluation object value. For example, a configuration may be used that defines the range where the difference between the evaluation object value and the measurement value is equal to or greater than a predetermined value as a range where the margin of error allowed for the evaluation object value is exceeded. Thus, the reliability of the evaluation target value is considered to be low if the probability of measuring the measurement value in a range where the difference between the measurement value and the evaluation object value is equal to or greater than a predetermined value is at least a predetermined probability.
In addition, the measurement value dispersion may be defined based on various indices. For example, a configuration may be used in which the measurement value dispersion when a sensor that measures an operation of a driver of the vehicle is not included as a sensor that acquires the measurement value is considered to be smaller than the measurement value dispersion when a sensor that measures an operation of the driver of the vehicle is included. In other words, when a sensor that measures an operation of the driver of the vehicle is included as a sensor for acquiring the measurement value, conditions and the like during driving are reflected on the resulting measurement value acquired and the measurement value dispersion is generally large. Therefore, if the magnitude of dispersion is identified based on whether a sensor that measures an operation of the driver of the vehicle is included as a sensor for acquiring the measurement value, the dispersion can be determined without actually specifying the dispersion.
If the reliability of the evaluation object value is low, the dispersion is small, and the measurement probability of the measurement value with the highest probability of being measured is equal to or greater than a predetermined value, a value that specifies the evaluation object may be identified based on the measurement value with the highest probability of being measured, and the evaluation object value in the map information may be updated using the identified value. In other words, if the measurement value dispersion is small, the reliability of the value of the measurement value with the highest probability of being measured can be evaluated as high. Further, if the measurement probability of the measurement value with the highest probability of being measured is equal to or greater than a predetermined value, the measurement value with the highest probability of being measured is highly likely to be closer to the actual value of the evaluation object than the existing evaluation object value. Hence, if the reliability of the existing evaluation object value is evaluated as low, the reliability of the measurement value with the highest probability of being measured is considered to be high and used to update the existing evaluation object value. With this configuration, the reliability of the map information can be increased.
In addition, the reliability evaluation method for evaluating reliability is not limited provided that the reliability evaluation method is selected depending on the distribution property of the measurement value. An index other than dispersion may be used as the index for selecting the reliability evaluation method depending on the distribution property of the measurement value. For example, because the distribution property of the measurement value differs for different evaluation objects, a configuration may be used that selects the reliability evaluation method depending on the evaluation object. According to this configuration, the reliability of the evaluation object value can be evaluated using a reliability evaluation method suited to the evaluation object. Note that various configurations may be used for selecting the reliability evaluation method depending on the evaluation object. For example, a configuration may be used that pre-associates the evaluation object and the reliability evaluation method and selects the reliability evaluation method depending on such correspondence. Another configuration that may be used classifies the evaluation object in advance, pre-associates the classification and the reliability evaluation method, and selects the reliability evaluation method depending on such correspondence. Because the distribution property of the measurement value differs with different measurement methods, a configuration may be used that selects the reliability evaluation method depending on the measurement method of the measurement value. According to this configuration, the reliability of the evaluation object value can be evaluated using a reliability evaluation method suited to the measurement method of the measurement value. Note that the measurement method is a method for measuring the evaluation object value, and is specified depending on the combination of sensor type, measurement value measured by the sensor, and conversion method if the measurement value is converted.
Further, the technique for selecting the reliability evaluation method depending on the measurement value dispersion and evaluating the reliability of the evaluation object value can also be applied in the forms of a program and a method. Moreover, the reliability evaluation device, the reliability evaluation program, and the reliability evaluation method as described above include various type of embodiments. An aspect of the present invention may be implemented in a stand-alone device, and it may be implemented through parts used in common with respective components provided in the vehicle. For example, it is possible to provide a navigation device that is equipped with the reliability evaluation device described above, and to provide the reliability evaluation method and the reliability evaluation program as well. The various aspects of the present invention can also be modified as desired, such as by providing portions in the form of software and portions in the form of hardware, for example. The present invention may also be practiced in the form of a storage medium for a program that controls the reliability evaluation device. The software storage medium may be a magnetic storage medium or a magneto optical storage medium. Furthermore, any storage medium that is developed henceforth can also be considered to be exactly the same.
Hereinafter, embodiments of the present invention will be described in the order shown below.
The map information 30a is used for identifying a vehicle position and providing vehicle route guidance. The map information 30a includes node data that indicate nodes that are set on the roads that the vehicle travels, shape interpolation data that indicates the shapes of the roads between the nodes, link data that indicate connections between the nodes; data that indicate the roads and the features around them, and the like. In the present embodiment, information that specifies curve segments formed in roads and information that specifies the start position and curvature of such curve segments is included in the map information 30a.
The vehicle in the present embodiment (the vehicle installed with the navigation device 10) includes a GPS receiving portion 41, a vehicle speed sensor 42, a gyroscopic sensor 43, a steering sensor 44, an acceleration sensor 45, a gear shift portion 46, and a braking portion 47. The functions of the reliability evaluation program 21 and the functions of the vehicle control program 22 are implemented by coordinated operation among the various portions and the control unit 20.
The GPS receiving portion 41 receives radio signals from GPS satellites and outputs, through an interface that is not shown in the drawing, information for computing a current position of the vehicle. The control unit 20 acquires the current position of the vehicle by receiving such information. The vehicle speed sensor 42 outputs a signal that corresponds to the rotational speed of a wheel of the vehicle. The control unit 20 acquires the speed of the vehicle by receiving the signal through an interface that is not shown in the drawing. The gyroscopic sensor 43 outputs a signal that corresponds to an angular velocity of the vehicle. The control unit 20 acquires the direction in which the vehicle is traveling by receiving the signal through an interface that is not shown in the drawing. The vehicle speed sensor 42 and the gyroscopic sensor 43 are used to correct the current position of the vehicle that is specified by the output signal from the GPS receiving portion 41. The current position of the vehicle is also corrected as necessary based on the driving path of the vehicle.
The steering sensor 44 outputs a signal that specifies a steering angle (a rotation angle from a steering wheel base condition (0 degrees)) of the vehicle. The control unit 20 receives the signal through an interface that is not shown in the drawing, and identifies the timing at which the driver performs a steering operation based on the signal that specifies the steering angle. The acceleration sensor 45 outputs a signal that corresponds to an acceleration of the vehicle. The control unit 20 acquires an acceleration having a component parallel to the lateral direction of the vehicle (lateral acceleration) by receiving the signal through an interface that is not shown in the drawing.
The gear shift portion 46 includes a stepped transmission with a torque converter that has a plurality of gear speeds, such as six forward speeds, one reverse speed, and the like. The gear shift portion 46 can transmit the driving force of an engine to the wheel of the vehicle while using the gear ratios that correspond to the various gear speeds to regulate the engine rotational speed. Through an interface that is not shown in the drawing, the control unit 20 outputs a control signal for switching the gear speed, and the gear shift portion 46 can receive the control signal and switch the gear speed. In the present embodiment, as the gear speed increases to a higher gear, such as from the first forward speed towards the sixth forward speed, the gear ratio becomes smaller.
The braking portion 47 includes a device that controls a wheel cylinder pressure, which regulates the degree of deceleration realized by brakes that are mounted in the vehicle wheels. The control unit 20 outputs a control signal to the braking portion 47 to regulate the wheel cylinder pressure. If the control unit 20 outputs a control signal to the braking portion 47 to increase the wheel cylinder pressure, the braking force of the brakes is increased and the vehicle is decelerated.
The control unit 20 executes the reliability evaluation program 21 to evaluate the reliability of information included in the map information 30a. For this purpose, the reliability evaluation program 21 has an evaluation object value acquisition unit 21a, a measurement value acquisition unit 21b, and a reliability evaluation unit 21c.
The evaluation object value acquisition unit 21a is a module that realizes in the control unit 20 a function for acquiring an evaluation object value whose reliability is to be evaluated from the map information 30a. In the present embodiment, the control unit 20 determines whether there is a curve segment within a predetermined range ahead of the current position of the vehicle, which is specified based on the output signals of the GPS receiving portion 41, the vehicle speed sensor 42, and the gyroscopic sensor 43. If it is determined that there is a curve segment within the predetermined range, the control unit 20 uses the processing of the evaluation object value acquisition unit 21a to find as evaluation objects the curvature related to vehicle control in the curve segment and the start position of the curve segment, after which the values of the curvature and the start position of the curve segment are extracted from the map information 30a and set as evaluation object values. Note that various types of formats may be used for the evaluation object values included in the map information 30a. For example, in addition to a configuration that associates the curve segment and the curvature and stores them in the map information 30a, a configuration that extracts at least three shape interpolation points set on the road of the curve segment and calculates the curvature from the positions of these shape interpolation points may also be used.
The measurement value acquisition unit 21b is a module for realizing in the control unit 20 a function for acquiring a measurement value for measuring the evaluation object value a plurality of times using the sensors mounted in the vehicle. In the present embodiment, if the sensor output values employ the same system of units as the evaluation object value, the sensor output values are used as the measurement value; if the sensor output values and the evaluation object value employ different systems of units, one is converted to match the other. Specifically, if it is determined that there is a curve segment within a predetermined range ahead of the current position of the vehicle, the control unit 20 uses the processing of the measurement value acquisition unit 21b to acquire the output signal of the steering sensor 44 and identify the timing at which the driver performs a steering operation. The vehicle position at such timing is identified by the output signals of the GPS receiving portion 41, the vehicle speed sensor 42, and the gyroscopic sensor 43, and is considered to be the start position of the curve segment. Since the vehicle position at the timing of the steering operation uses the same system of units as the evaluation object value that specifies the start position of the curve segment, the vehicle position at the timing of the steering operation is set as a measurement value and stored in the storage medium 30 as measurement value information 30b.
Alternatively, the control unit 20 uses the processing of the measurement value acquisition unit 21b to identify a lateral acceleration Gt of the vehicle by receiving the output signal of the acceleration sensor 45, and to identify a vehicle speed v by receiving the output signal of the vehicle speed sensor 42. Based on the lateral acceleration Gt and the vehicle speed v, a curvature (Gt/v2) is acquired as a measurement value and stored in the storage medium 30 as measurement value information 30b. In other words, the system of units for the curvature is different from that for the lateral acceleration Gt and the vehicle speed v, which are sensor output values. Therefore, the sensor output values are converted to the same system of units as the curvature, which is an evaluation object value, and used as the measurement value. Note that the measurement value information 30b is associated with information that specifies the measurement objects. For example, the position of the curve segment is associated with a value that specifies the start position of the curve segment and a value that specifies the curvature.
In the present embodiment, when the vehicle reaches a position where there is a curve segment ahead within a predetermined range, the control unit 20 acquires a measurement value related to the curve segment. Thus, according to the present embodiment, the measurement value is acquired a plurality of times in accordance with the plurality of times the vehicle arrives at the curve segment, and the measurement value is accumulated as the measurement value information 30b.
The reliability evaluation unit 21c is a module that realizes in the control unit 20 a function for selecting a reliability evaluation method, and evaluating the reliability of an evaluation object value according to the selected reliability evaluation method. In the present embodiment, the measurement value is classified in advance into two types depending on a dispersion property thereof. Two types of reliability evaluation methods are defined in advance in accordance with the classification, and stored in the storage medium 30 as evaluation method information 30c. According to the present embodiment, the evaluation method information 30c includes first evaluation method information and second evaluation method information. The first evaluation method information applies to a small dispersion, while the second evaluation method information applies to a large dispersion. The evaluation method information 30c also includes information that specifies a value for a threshold representing a range of the measurement value to be used in the reliability evaluation, an index for evaluating reliability, and a probability for evaluating reliability. Specifically, the values for the threshold, index, and probability in the first evaluation method information are ±10%, a measurement value with the highest probability of being measured, and a first probability N (0≦N≦1), respectively. The values for the threshold, index, and probability in the second evaluation method information are ±30%, a measurement value included between the thresholds, and a second probability M (0≦M≦1), respectively.
Specifically, the measurement value of the curvature shown in
The reliability evaluation methods specified by the first and second evaluation method information defined as described above are set so as to evaluate the reliability of the evaluation object value depending on this dispersion property. In other words, a measurement value with which a frequency distribution having a small dispersion can be obtained is distributed in a narrow range. Therefore, the measurement value can be stably measured, and the accuracy of the evaluation object value can be evaluated by directly comparing the measurement value and the evaluation object value. Consequently, for a measurement value with a small dispersion, the measurement value with the highest probability of being measured is assumed to be the actual measurement value of the evaluation object. The control unit 20 thus selects the first evaluation method information from the evaluation method information 30c and evaluates reliability based on the measurement value with the highest probability of being measured.
A measurement value with which a frequency distribution having a large dispersion can be obtained is distributed in a wide range. Therefore, it is difficult to evaluate the evaluation object value based on the measurement value. Instead, the reliability of the evaluation object value is evaluated by treating the measurement value as a set. Consequently, for a measurement value with a large dispersion, the measurement probability of the measurement value being measured within a specific range is considered to be the evaluation index of the actual measurement value of the evaluation object. The control unit 20 thus selects the second evaluation method information from the evaluation method information 30c and evaluates reliability based on the measurement probability of the measurement value being measured within a specific range.
More specifically, in an evaluation using the reliability evaluation method corresponding to a measurement value with a small dispersion, the control unit 20 establishes a range in which a difference between the measurement value and the evaluation object value is within a threshold (a range within ±10% from the evaluation object value) as a first range. If the measurement value with the highest probability of being measured is within the first range from the evaluation object value, and the measurement probability of the measurement value with the highest probability of being measured is equal to or greater than the first probability N, the control unit 20 determines that the reliability of the evaluation object value is high. On the other hand, if the measurement probability of the measurement value with the highest probability of being measured is less than the first probability N, the control unit 20 determines that the reliability of the evaluation object value is low. Note that if there is no measurement value with the highest probability of being measured within the first range, the control unit 20 determines that the reliability of the evaluation object value is low. In other words, if the difference between the measurement value with the highest probability of being measured and the evaluation object value is excessively large, the reliability is evaluated as low.
In an evaluation using the reliability evaluation method corresponding to a measurement value with a large dispersion, the control unit 20 establishes a range in which a difference between the measurement value and the evaluation object value is within a threshold (a range within ±30% from the evaluation object value) as a second range. If the probability of the measurement value within the second range being measured is equal to or greater than the second probability M, the control unit 20 determines that the reliability of the evaluation object value is high. If the probability of the measurement value in a range within ±30% from the evaluation object value being measured is less than the second probability M, the control unit 20 determines that the reliability of the evaluation object value is low. Note that if the measurement probability of the measurement value not within the second range that includes the evaluation object value is higher than the measurement probability of the measurement value within the second range that includes the evaluation object value, the control unit 20 determines that the reliability of the evaluation object value is low. In other words, if there is a high proportion of excessively large differences between the measurement value and the evaluation object value, the reliability is evaluated as low.
Note that because the first range and the second range which specify analysis objects in the reliability evaluation methods are different, different thresholds may also be selected depending on the dispersion. In the present embodiment, the first range is formed so as to be narrower than the second range. Specifically, for a small dispersion, the measurement value is distributed in a relatively narrow range when the measurement value is acquired a plurality of times, whereas for a large dispersion, the measurement value is distributed in a relatively wide range when the measurement value is acquired a plurality of times. Therefore, a narrow range may be used as the analysis object in the case of the former, and a wide range may be used as the analysis object in the case of the latter. According to this configuration, the reliability evaluation method can be selected with the measurement value of a suitable range depending on the measurement value dispersion used as the analysis object.
The first and second probabilities may be set to any value provided that the reliability is high if the value to be compared to each probability exceeds the first probability or the second probability, and the reliability is low if the probabilities are not exceeded. Therefore, the first and second probabilities may be set depending on the precision of required information in the map information 30a. Note that reliability is information that specifies the degree of accuracy of an evaluation object value, and is defined such that a higher reliability means a smaller negative effect on the vehicle when the evaluation object value is used in a vehicle control. According to the present embodiment, an evaluation object value with a relatively small allowable margin of error is set such that the first and second ranges are narrow and the first and second probabilities are large compared to an evaluation object value with a relatively large allowable margin of error. Thus, it is possible to consider an evaluation object value that includes a margin of error as the correct value and evaluate reliability depending on the magnitude of the effect on driving when the vehicle control is performed.
Further, in the present embodiment, the measurement value distribution is evaluated in terms of the frequency distribution. Therefore, the measurement value is analyzed per predetermined interval and the number of measurement values contained in each interval is the frequency. Thus, the measurement probability of the measurement values and the measurement value with the highest probability of being measured are evaluated for each interval. For example, the measurement probability of the measurement values contained in a certain interval is the frequency of that particular interval divided by the overall frequency, and the measurement value with the highest probability of being measured is the median value of the interval that includes the measurement value with the highest probability of being measured in the frequency distribution.
The control unit 20 uses the processing of the reliability evaluation unit 21c to select the reliability evaluation method in the above manner and evaluate the reliability of the evaluation object value based on the measurement value in accordance with the selected reliability evaluation method. Once the reliability of the evaluation object value is identified, the control unit 20 performs a vehicle control suited to the reliability. Specifically, the vehicle control program 22 is a module that realizes in the control unit 20 a function for controlling the vehicle based on the evaluation object value, and the control content for the control is selected depending on the reliability of the evaluation object value.
In the present embodiment, the control unit 20 uses the processing of the vehicle control program 22 to decelerate the vehicle within a segment that comes after the start position of the curve segment but before the start position of a fixed curvature segment. In such case, the vehicle control is set such that an evaluation object value with higher reliability is accompanied by a higher degree of automatic control, and an evaluation object value with higher reliability is also accompanied by a larger control amount. More specifically, the control unit 20 varies the content of the vehicle control based on the reliability of the start position and the curvature of the curve segment.
For example, if the reliability of the start position of the curve segment is high, the control unit 20 starts decelerating the vehicle when the vehicle reaches the start position of the curve segment. However, if the reliability of the start position of the curve segment is low, the control unit 20 starts decelerating the vehicle when the driver starts to steer according to the output signal of the steering sensor 44. In other words, if the reliability of the start position of the curve segment is high, the deceleration timing is automatically determined based on the map information 30a, but if the reliability of the start position of the curve segment is low, the deceleration timing is determined based on the driver's operation.
If the reliability of the curvature is high, the control unit 20 outputs a control signal to the braking portion 47 to generate a braking force for decelerating the current vehicle speed to a target vehicle speed over the course of traveling from the start position of the curve segment to the start position of the fixed curvature segment. However, if the reliability of the curvature is low, the control unit 20 outputs a control signal to the gear shift portion 46 to control the gear ratio so as to become a preferred gear ratio for decelerating the current vehicle speed to a target vehicle speed over the course of traveling from the start position of the curve segment to the start position of the fixed curvature segment. In other words, if the reliability of the curvature is high, the control unit 20 actually applies a deceleration to the vehicle for slowing the current vehicle speed to the target vehicle speed, but if the reliability of the curvature is low, the control unit 20 performs a control such that the engine brake is applied to the vehicle through gear shifting.
As described above, according to the present embodiment, the reliability of an evaluation object value is evaluated using a reliability evaluation method that is selected depending on the measurement value dispersion. Therefore, when the position of a feature or the like specified in map information is measured, the reliability can be evaluated based on the distribution property of the measurement value. Since the vehicle is controlled in accordance with the reliability of the evaluation object value, the vehicle control is surely executed in most cases, and if an erroneous control is performed, the effect of the erroneous control may be minimized.
A vehicle control process carried out by the navigation device 10 with the above configuration will be described next based on a flowchart shown in
The control unit 20 determines whether there is a curve segment within a predetermined range ahead of the vehicle based on the output signals of the GPS receiving portion 41, the vehicle speed sensor 42, and the gyroscopic sensor 43. If it is determined that there is a curve segment, the control unit 20 executes a measurement value acquisition process and the vehicle control process shown in
Meanwhile, the vehicle control process shown in
Next, the control unit 20 uses the processing of the reliability evaluation unit 21c to identify the dispersion of the measurement value (S105). In the present embodiment, the control unit 20 determines whether a sensor that measures an operation of the driver of the vehicle is included as a sensor for acquiring the measurement value of the evaluation object. If a sensor that directly measures an operation of the driver is not included, the control unit 20 considers the measurement value dispersion to be small; if a sensor that directly measures an operation of the driver is included, the control unit 20 considers the measurement value dispersion to be large. In the example shown in
Next, the control unit uses the processing of the reliability evaluation unit 21c to select a reliability evaluation method depending on the dispersion (S110), and evaluates the reliability of the evaluation object value (S115). Specifically, the control unit 20 references the second evaluation method information with regard to the start position P of the curve segment, for which the measurement value is considered to have a large dispersion, and defines the second range that includes the evaluation object value based on a predetermined threshold (±30%). The control unit 20 then acquires the measurement value information 30b to obtain a sum S1 of the frequencies of the measurement values included in the second range, a sum S2 of the frequencies of the measurement values not included in the second range, and a sum S0 of the frequencies of all the measurement values. In
Further, the control unit 20 compares the measurement probability of the measurement values not included in the second range (S2/S0) and the measurement probability of the measurement values included in the second range (S1/S0). If the measurement probability of the measurement values not included in the second range is higher, the reliability of the evaluation object value EP is considered to be low. If the measurement probability of the measurement values not included in the second range is lower, the control unit 20 further determines whether the measurement probability of the measurement values included in the second range is equal to or greater than the second probability M. If the measurement probability of the measurement values included in the second range is equal to or greater than the second probability M, the reliability is considered to be high. However, if the measurement probability of the measurement values included in the second range is less than the second probability M, the reliability is considered to be low.
Meanwhile, with regard to the curvature α, for which the measurement value is considered to have a small dispersion, the control unit 20 defines the first range based on a predetermined threshold (±10%). The control unit 20 then acquires the measurement value information 30b to identify the measurement value with the highest measurement probability (the measurement value with the maximum frequency), and obtains a frequency F1 of a measurement value D with the highest measurement probability and a sum F0 of all frequencies. In
If the measurement value D with the highest measurement probability is not included within the first range, the control unit 20 determines that the reliability of the evaluation object value Eα is low. If the measurement value D with the highest measurement probability is included in the first range, the control unit 20 further determines whether the measurement probability (F1/F0) of the measurement value D with the highest measurement probability is equal to or greater than the first probability N. If the measurement probability of the measurement value included in the first range is equal to or greater than the first probability N, the reliability is considered to be high. However, if the measurement probability of the measurement values included in the first range is less than the first probability N, the reliability is considered to be low.
After the control unit 20 identifies the reliability of the start position P and the curvature α of the curve segment as described above, the control unit 20 uses the processing of the vehicle control program 22 to determine whether the reliability of the start position P of the curve segment is high (S120). If it is determined at S120 that the reliability of the start position P of the curve segment is high, the control unit 20 sets the vehicle control so as to identify the start position P of the curve segment based on the evaluation object value (S125). However, if it is determined at S120 that the reliability of the start position P of the curve segment is not high, the control unit 20 sets the vehicle control so as to identify the start position P of the curve segment based on the measurement value of the steering sensor 44 (S130).
The control unit 20 then uses the processing of the vehicle control program 22 to determine whether the reliability of the curvature α is high (S135). If it is determined at S135 that the reliability of the curvature α is high, the control unit 20 sets the vehicle control so as to execute a deceleration control using the braking portion 47 (S140). However, if it is determined at S135 that the reliability of the curvature α is not high, the control unit 20 sets the vehicle control so as to perform auxiliary deceleration using the engine brake (S145). In other words, once the vehicle arrives at the start position P of the curve segment identified in accordance with the setting at S125 or 5130, the control unit 20 executes deceleration using the braking portion 47 or auxiliary deceleration using the engine brake in accordance with the setting at S140 or S145. More specifically, if the vehicle control is set so as to execute deceleration using the braking portion 47, a control signal is output to the braking portion 47 once the vehicle reaches the start position P of the curve segment, and a deceleration control is performed that generates a braking force for decelerating the current speed of the vehicle to a target vehicle speed within a distance spanning from the start position P of the curve segment to the start position of the fixed curvature segment. If the vehicle control is set so as to execute auxiliary deceleration using the engine brake, a control signal is output to the gear shift portion 46 once the vehicle reaches the start position P of the curve segment, and the gear ratio is controlled so as to become a preferred gear ratio for decelerating the current speed of the vehicle to a target vehicle speed within a distance spanning from the start position P of the curve segment to the start position of the fixed curvature segment.
The present embodiment described above is only one example for practicing the present invention, and various other embodiments may also be used including those that select a reliability evaluation method depending on a measurement value dispersion and evaluate the reliability of an evaluation-target value. For example, the processing procedure is not limited to the procedure shown in
A configuration may be used that does not execute the reliability evaluation of the present invention until the measurement value corresponding to each evaluation object value have been measured at least a certain number of times. Besides a configuration of map information 30a in which the dispersion of the measurement value for measuring the evaluation object is identified each time the vehicle control is performed, the map information 30a may also be configured such that the size of the dispersion for each evaluation object value is identified in advance, and information that specifies the size of the identified dispersion is associated with the evaluation object value.
The object that is controlled based on the reliability of the evaluation object value is not limited to the braking portion and the gear shift portion; a throttle control portion or the like may be used to perform an acceleration control of the vehicle. Further, the deceleration control may combine controls for the braking portion, the gear shift portion, and the throttle control portion. Specifically, the deceleration control may be configured such that an evaluation object value with a higher reliability is accompanied by the use of a larger control amount (degree of deceleration and the like) by the combination of controls.
The evaluation object value is a value that can be measured by a sensor installed in the vehicle, and is also information that is stored in the storage medium in advance as part of the map information. Therefore, the evaluation object value may be any value that pertains to a feature or the like that forms the map information, and the evaluation object value is not limited to the start position and the curvature of a curve segment as described above. For example, the evaluation object value may identify the position of a feature such as a stop line, or a size or amount such as the road width. The evaluation object value may also express a specific meaning such as a flag that indicates the existence of a feature.
The measurement value may be found based on various sensors, and a camera or the like for measuring the position, speed, acceleration and the like of the vehicle or a feature on the road through image analysis may be used as a sensor of the present invention.
Besides a configuration that identifies the dispersion based on a property of the measurement value as described above, a configuration may be used that pre-associates dispersion magnitudes with a sensor type, a method of computational processing for the measurement value, or the like, and considers the measurement value dispersion to be associated with the sensor type, the method of computational processing for the measurement value, or the like. Obviously, the dispersion may actually be computed by acquiring the measurement value a plurality of times. Further, instead of actually analyzing the dispersion, the reliability evaluation method may be selected depending on the evaluation object.
Alternatively, a combination of reliability evaluation methods may be selected depending on the dispersion, or a specific reliability evaluation method may be selected. For example, depending on the measurement value dispersion, either of the following two reliability evaluation methods may be selected: a reliability evaluation method that evaluates reliability based on the difference between the measurement value with the highest probability of being measured and the evaluation object value; and a reliability evaluation method that evaluates reliability based on the probability that a measurement value is measured in a range where a difference between the measurement value and the evaluation object value is equal to or greater than a predetermined value. In such a configuration, the former reliability evaluation method may form all or part of a reliability evaluation method that corresponds to a small dispersion, and the latter reliability evaluation method may form all or part of a reliability evaluation method that corresponds to a large dispersion.
If the reliability of the evaluation object value is low, the dispersion is small, and the measurement probability of the measurement value with the highest probability of being measured is equal to or greater than a predetermined value, a value that specifies the evaluation object may be identified based on the measurement value with the highest probability of being measured, and the evaluation object value in the map information may be updated using the identified value. This configuration can be realized, for example, by adding a map information update unit to the reliability evaluation program 21 shown in
| Number | Date | Country | Kind |
|---|---|---|---|
| 2009-163860 | Jul 2009 | JP | national |
