KR101177374B1 - Method for estimating position of vehicle using Interacting Multiple Model filter - Google Patents

Abstract

The vehicle position estimation method using the interactive multi-model filter according to the present invention includes the steps of calculating the kinematic model prediction value with the kinematic model extended Kalman filter and calculating the dynamic model prediction value with the kinematic model extended Kalman filter, Calculating a kinematic model filtering value that is a value between the model predicted value and the measured value indicating the current position of the vehicle, and calculating a dynamic model filtering value that is a value between the dynamic model predicted value and the measured value indicating the current position of the vehicle And fusing the kinetic model filtering value and the kinematic model filtering value to calculate a final correction value. By using the vehicle position estimation method using the interactive multi-model filter according to the present invention, the position of the vehicle can be stably estimated even in a place where GPS reception is poor, and even when an error occurs in the calculation of the extended Kalman filter. In addition, the ratio of the weights added to the data calculated from the kinematic model extended Kalman filter and the data calculated from the dynamic model extended Kalman filter can be set more reasonably to estimate the position of the vehicle more accurately.

Description

Method for estimating position of vehicle using Interacting Multiple Model filter}

The present invention relates to a method of estimating the position of a vehicle using an interactive multi-model filter, and more particularly, to a method of estimating the position of a vehicle by fusing GPS position information and sensor information in a vehicle.

Vehicle navigation system is a cutting-edge technology that provides a variety of services, such as route guidance, traffic guidance, surrounding information and additional information provided by receiving the location information from knowing the current position of the vehicle and combined with the geographic information based on it.

The basic navigation system used to identify the position of the vehicle in the vehicle navigation system is the GPS (Global Positioning System), which is used not only for automobiles but also for airplanes and ships. It can be a system.

At this time, as a method of estimating the position of the vehicle using GPS, a method of estimating the position of the vehicle using an interactive multiple model filter having an extended Kalman filter has been recently developed. Is being applied. The Extended Kalman Filter is composed of a Kinematic Model Extended Kalman Filter and a Dynamic Model Extended Kalman Filter, and the interactive multi-model filter is a vehicle calculated from a Kinematic Model Extended Kalman Filter. It is configured to estimate the final position of the vehicle by fusing the coordinate value and angular velocity of the vehicle and the coordinate value and angular velocity of the vehicle calculated from the Dynamic Model Extended Kalman Filter.

In this case, the GPS cannot be used in areas such as tunnels, overpasses, and forests of buildings that cannot receive radio waves, and may not be able to continuously provide location or increase errors depending on the surrounding environment. When the current position of the vehicle is estimated by inputting only the data including the coordinate values of the vehicle into the extended Kalman filter, the actual position of the vehicle cannot be accurately estimated. Therefore, in order to compensate for the above disadvantages, the speed and angular velocity values of the vehicle that can be obtained through the wheel speed sensor and the steering angle sensor mounted on the vehicle as well as the data received from the GPS. By inputting the extended Kalman filter, a method of estimating the current position of the vehicle even when GPS reception is poor is widely used.

On the other hand, it is relatively accurate to estimate the position of the vehicle with the Kinematic Model expansion Kalman filter when the slip does not occur, and when the slip is generated, the dynamic model is extended. Since it is relatively accurate to estimate the position of the vehicle with the Kalman filter, the coordinates of the vehicle calculated from the Kinematic Model extended Kalman filter and the coordinates of the vehicle calculated from the angular velocity and dynamic model Extended Kalman filter In the case of the low speed at which the slip of the vehicle does not occur relatively when the value and the angular velocity are fused, the weight of the vehicle is more weighted to the coordinate value and the angular velocity of the vehicle calculated from the Kinematic Model Extended Kalman Filter. In the case of high speeds that are relatively high, more vehicle coordinates and angular velocities are calculated from Dynamic Model Extended Kalman Filter. Place the weights. However, even at low speeds, the slip of the vehicle may occur, and even at high speeds, the slip of the vehicle may not occur. Therefore, weighting according to the speed of the vehicle may not accurately estimate the position of the vehicle. There is a problem.

The present invention has been proposed to solve the above problems, and by estimating the location of the vehicle by fusing the GPS location information and the sensor information in the vehicle, it is possible to stably estimate the location of the vehicle even in a location where GPS reception is poor. By correcting the vehicle position predicted value calculated by the Kalman filter to the vehicle position measured value, the error value can be reduced even when an error occurs in the calculation of the extended Kalman filter. It is an object of the present invention to provide a method for estimating a vehicle position that can more accurately estimate a vehicle position by more reasonably setting weights added to data calculated from a model extended Kalman filter.

The vehicle position estimation method according to the present invention includes a mixing unit that mixes input data, an extended Kalman filter composed of a kinematic model extended Kalman filter and a dynamic model extended Kalman filter, and a combination unit fused with the input data. A method of estimating a vehicle position using an interactive multi-model filter, the method comprising: receiving, by the mixing unit, a kinematic model initial value and a dynamic model initial value of a vehicle; Calculating, by the mixing unit, a kinematic model mixing value and a dynamic model mixing value by mixing the kinematic model initial value and the dynamic model initial value; The kinematic model extended Kalman filter receiving the kinematic model mixing value to calculate a kinematic model prediction value, and the kinematic model extended Kalman filter receiving the kinematic model mixing value to calculate a dynamic model predicted value ; The kinematic model extended Kalman filter calculates a kinematic model filtering value that is a value between the measured value indicating the current position of the vehicle and the kinematic model predicted value, and the dynamic model extended Kalman filter indicates the current position of the vehicle Calculating a dynamic model filtering value that is a value between the value and the dynamic model prediction value; And combining the dynamic model filtering value and the kinematic model filtering value to calculate a final correction value by the combining unit.
After the calculating of the final correction value, the interactive multi-model filter further comprises the step of determining whether or not to input the position estimation stop signal of the vehicle, if the position estimation stop signal of the vehicle is inputted, the process is terminated, If the position estimation stop signal is not inputted, the method returns to the initial step of receiving the kinematic model initial value and the dynamic model initial value.
When entering the kinematic model initial value and the dynamic model initial value for the first time, the position-related data of the vehicle received from the GPS is set to the kinematic model initial value and the dynamic model initial value, and the vehicle When the step of returning to the input of the kinematic model initial value and the dynamic model initial value through the step of determining whether the position estimation stop signal is inputted, the kinematic model filtering value and the dynamic model filtering calculated in the previous cycle Values are set to the kinematic model initial value and the kinetic model initial value, respectively.
The interactive multi-model filter further includes an update unit 130 that weights the data, and before the calculating of the final correction value 30, the update unit filters the kinematic model filtering value and the measured value. The difference between the measured value and the kinematic model filtering value and the measured value among the kinematic model filtering value and the measured value among the kinematic model filtering value is greater than that of the measured value. The method may further include setting a large weight to a small filtering value, and the calculating of the final correction value may include: a weight set by the update unit to the dynamic model filtering value and the kinematic model filtering value. And then fuse to calculate the final correction value.
In the setting of the weighting unit, the weight is set at a ratio inversely proportional to the difference between the kinematic model filtering value and the measurement value and the difference ratio between the dynamic model filtering value and the measurement value.
In the calculating of the kinematic model mixing value 12 and the dynamic model mixing value, the mixing unit mixes the kinematic model initial value with the kinematic model initial value by a predetermined ratio to convert the kinematic model mixing value. Calculate and proportionally mix the kinematic model initial value with the kinematic model initial value to calculate the kinematic model mixing value.
In the calculating of the kinematic model mixing value and the dynamic model mixing value, the kinematic model initial value is added to the kinematic model initial value and the dynamic model initial value. The mixing ratio varies according to the user's selection.
The measured value includes the current angular velocity of the vehicle and the current coordinates of the vehicle, and the measured value is determined by the angular velocity of the vehicle measured by the angular velocity measuring sensor mounted on the vehicle and the current coordinates of the vehicle received from the GPS. Is calculated.
The angular velocity measuring sensor is applied as a yaw rate sensor.
The vehicle position estimation method according to the present invention includes a mixing unit that mixes input data, an extended Kalman filter composed of a kinematic model extended Kalman filter and a dynamic model extended Kalman filter, and a combination unit fused with the input data. In the method for estimating the vehicle position by using an interactive multi-model filter, the kinematic model extended Kalman filter calculates the kinematic model prediction value, the dynamic model extended Kalman filter is predicted by the dynamic model Calculating; The kinematic model extended Kalman filter calculates a kinematic model filtering value that is a value between the measured value indicating the current position of the vehicle and the kinematic model predicted value, and the dynamic model extended Kalman filter indicates the current position of the vehicle Calculating a dynamic model filtering value that is a value between the value and the dynamic model prediction value; And combining the dynamic model filtering value and the kinematic model filtering value to calculate a final correction value by the combining unit.
The interactive multi-model filter further includes an update unit for weighting the data, and before the calculating of the final correction value, the update unit differs between the kinematic model filtering value and the measured value and the dynamic model. By comparing the difference between the filtered value and the measured value, a weighted value that is larger than the filtered value having a larger difference between the measured value among the kinematic model filtered value and the dynamic model filtered value is smaller than the measured value. The method may further include setting the final correction value, wherein the combining unit adds weights set by the update unit to the dynamic model filtering value and the kinematic model filtering value, and then fuses the final values. Calculate a correction value.
The setting of the weighting unit by the updating unit is configured to set the weights at a ratio inversely proportional to the difference between the kinematic model filtering value and the measurement value and the difference ratio between the dynamic model filtering value and the measurement value.
The measured value includes the current angular velocity of the vehicle and the current coordinates of the vehicle, and the measured value is determined by the angular velocity of the vehicle measured by the angular velocity measuring sensor mounted on the vehicle and the current coordinates of the vehicle received from the GPS. Is calculated.
The angular velocity measuring sensor is applied as a yaw rate sensor.

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

By using the vehicle location estimation method using the interactive multi-model filter according to the present invention, by estimating the location of the vehicle by fusing the GPS location information and the sensor information in the vehicle, it is possible to stably estimate the location of the vehicle even in a place where GPS reception is poor. By correcting the vehicle position predicted value calculated by the extended Kalman filter to the vehicle position measured value, the error value can be reduced even when an error occurs in the calculation of the extended Kalman filter, and it is calculated from the kinematic model extended Kalman filter. There is an advantage that the position of the vehicle can be estimated more accurately by more rationally setting the weights added to the data and the data calculated from the dynamic model extended Kalman filter.

In addition, the vehicle position estimation method using the interactive multi-model filter according to the present invention can estimate the position of the vehicle stably and accurately, significantly improving the stability of the unmanned autonomous vehicle system and the intelligent transportation system, and The effect of remarkably improving accuracy can be obtained.

1 is a block diagram showing the configuration of an interactive multi-model filter for implementing a vehicle position estimation method according to the present invention.
2 is a flowchart of a vehicle position estimation method using an interactive multi-model filter according to the present invention.

Hereinafter, an embodiment of a vehicle position estimation method using an interactive multi-model filter according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram showing the configuration of an interactive multi-model filter for implementing a vehicle position estimation method according to the present invention, Figure 2 is a flow chart of a vehicle position estimation method using an interactive multi-model filter according to the present invention. .

When the vehicle moves, the low speed and low angular speed rarely causes slip and the high speed and high angular speed often cause frequent slip, so that the position of the moving vehicle can be estimated more accurately. The Interacting Multiple Model Filter (IMM filter), which estimates the position of the vehicle according to the model, is mainly used. The interactive multi-model filter includes a kinematic model extended Kalman filter for analyzing a model in which slip occurs rarely and a dynamic model extension for analyzing a model in which slip occurs relatively much. A Dynamic Model Extended Kalman Filter is provided and configured to estimate the position of the vehicle according to the driving environment of the vehicle.

At this time, the current vehicle position estimation system is mostly configured to estimate the position of the vehicle using only the signal received from the GPS, the estimated vehicle when missing satellite signal, signal error, large error of position information, etc. There is a problem that the position of and the actual position of the vehicle is greatly spaced apart.

The vehicle position estimation method using the interactive multi-model filter 100 according to the present invention has been devised to solve the above problems, and is similar to the kinematic model extended Kalman filter 122. By calculating the position-related data of the vehicle received from the GPS using the Interacting Multiple Model Filter (IMM filter) equipped with the Dynamic Model Extended Kalman Filter (124). The current position of the vehicle is estimated, but the kinematic model extended Kalman filter 122 and the kinematic model extended Kalman filter 124 are not estimated to estimate the position of the vehicle. And the position data of the vehicle calculated by the dynamic model extended Kalman filter 124 as the current vehicle position measurement value measured through various sensors attached to the vehicle and GPS. It is characterized in that it is configured to estimate the position of the vehicle more accurately.

When performing a vehicle position estimation method using the interactive multi-model filter 100 according to the present invention, first, position-related data of a vehicle to be analyzed by a kinematic model on the premise that slippage of the vehicle does not occur, The kinematic model initial value 11 and position-related data of the vehicle to be analyzed as the dynamic model, that is, the dynamic model initial value 21 are received (S10). At this time, the vehicle may move to a kinematic model in which the slip hardly occurs, but the vehicle may move to a dynamic model in which a large amount of slip is generated. On the contrary, the vehicle moves to a dynamic model in which a large amount of slip is generated. The vehicle may move with a kinematic model that does not. Therefore, the kinematic model initial value 11 is not directly substituted into the kinematic model extended Kalman filter 122 and the kinematic model initial value 21 is not directly substituted into the kinematic model extended Kalman filter 124, The kinematic model mixing value 12 is calculated by mixing the kinematic model initial value 21 with the kinematic model initial value 11 by a predetermined ratio, and the kinematic model initial value 11 is obtained by the kinematic model initial value 21. ), After calculating a dynamic model mixing value 22 by mixing a predetermined ratio (S20), the kinematic model mixing value 12 and the dynamic model mixing value 22 are kinematic model extended Kalman filter 122, respectively. And kinematic model expansion Kalman filter 124, respectively. At this time, the ratio of mixing the kinematic model initial value 21 to the kinematic model initial value 11 and the ratio of mixing the kinematic model initial value 11 to the dynamic model initial value 21 are the vehicle movement model. The probability of transforming the kinematic model from the kinematic model and the vehicle movement model is determined according to the probability of converting the kinematic model from the kinematic model. In this way, the mixing unit 110 for mixing the initial values of the two models is applied to the conventional interactive multi-model filter 100, and a detailed description thereof will be omitted.

When the kinematic model mixing value 12 and the dynamic model mixing value 22 are transferred to the kinematic model extended Kalman filter 122 and the dynamic model extended Kalman filter 124, respectively, the kinematic model extended Kalman filter The kinematic model prediction value 13 and the dynamic model prediction value 23 are calculated through the reference numeral 122 and the dynamic model extended Kalman filter 124 (S30). Meanwhile, in the process of calculating the kinematic model prediction value 13 and the dynamic model prediction value 23, the vehicle information obtained from the wheel speed sensor 200 and the steering angle sensor 300 may be fused with movement information. The method of fusing the vehicle's movement information obtained from the sensor attached to the vehicle when calculating the vehicle position prediction value using the filter 120 is also applied to the conventional interactive multi-model filter 100. Description is omitted.

At this time, the vehicle position predicted value calculated from the extended Kalman filter 120 may appear as a linear graph according to time variation. However, when the driving conditions of the vehicle change rapidly, the vehicle position calculated from the extended Kalman filter 120 may vary. A large difference may occur between the predicted value and the actual vehicle position value. On the other hand, the current position-related data of the vehicle, i.e., the measured value calculated from the vehicle angular velocity measured by the yaw rate sensor 500 of the vehicle and the vehicle coordinates received from the GPS 400, is unique to the GPS 400. Although it is not possible to show the linear graph according to the time change due to the measurement error, there is a characteristic that a large difference with the actual position value of the vehicle does not occur even if the driving conditions of the vehicle change drastically. Therefore, the method for estimating a vehicle position according to the present invention is characterized by being configured to reduce the disadvantages of the predicted value and the disadvantages of the measured value by fusing the predicted value and the measured value.

That is, in the vehicle position estimation method according to the present invention, after the kinematic model prediction value 13 and the dynamic model prediction value 23 are calculated, the vehicle angular velocity and the GPS 400 measured by the yaw rate sensor 500 of the vehicle are calculated. A measurement value indicating a current position of the vehicle is calculated using the vehicle coordinates received from the vehicle, and the dynamic model filtering is a value between the dynamic model prediction value 23 and the measurement value. A value 24 is calculated and a kinematic model filtering value 14, which is a value between the kinematic model prediction value 13 and the measured value, is passed (S40). In this case, the sensor for measuring the angular velocity of the vehicle is not limited to the above-mentioned yaw rate sensor 500, and any sensor may be replaced as long as the sensor is mounted on the vehicle and can measure the angular velocity of the vehicle.

In this way, when the predicted value and the measured value are fused, even though the predicted value is calculated to be largely different from the actual vehicle position value, the filtered value is adjusted to be closer to the actual vehicle position value. The advantage is that you can. In this case, the median value between the predicted value and the measured value may be calculated as a filtering value, or the filtering value may be calculated by giving a greater weight to any one of the predicted value and the measured value. For reference, not only the coordinates of the vehicle but also the angular velocity data of the vehicle may be included from the filtered value in which the measured values are fused, so that the position of the vehicle may be more accurately predicted.

When the kinematic model filtering value 14 and the kinetic model filtering value 24 are calculated as described above, the kinematic model filtering value 14 and the kinetic model filtering value 24 are fused to the final correction value 30. In step S60, the combination unit 140 that fuses the value calculated from the kinematic model extended Kalman filter 122 and the value calculated from the dynamic model extended Kalman filter 124 is conventional. The same applies to the interactive multi-model filter of 100, which will not be described in detail. At this time, the vehicle position estimation method according to the present invention, the kinematic model filtering value 14 calculated from the kinematic model extended Kalman filter 122 and the dynamic model filtering value calculated from the dynamic model extended Kalman filter 124 Rather than calculating the average value of 24 as the final correction value 30, any filtering value of the kinematic model filtering value 14 and the dynamic model filtering value 24 is determined by the measurement value of the vehicle (a yaw rate sensor ( And determining whether the vehicle is close to the vehicle angular velocity measured by the vehicle 500 and the current position-related data of the vehicle calculated from the vehicle coordinates received from the GPS 400, and weighting the filtering value close to the measured value to determine the final correction value (30). May be configured to

That is, the vehicle position estimation method according to the present invention, before the step (S60) of calculating the final correction value 30, the difference between the kinematic model filtering value 14 and the measured value and the dynamic model filtering By comparing the difference between the value 24 and the measured value, the difference between the measured value among the kinematic model filtered value 14 and the dynamic model filtered value 24 is greater than that of the measured value. The method may further include setting a large weight to a filtering value having a small difference (S50). As described above, when the step S50 of setting weights to the kinematic model filtering value 14 and the dynamic model filtering value 24 is added, the calculating of the final correction value 30 is performed at step S60. It may be configured to calculate the final correction value 30 in consideration of the weights in the mechanical model filtering value 24 and the kinematic model filtering value 14. For example, when the distance between the vehicle position coordinate of the kinematic model filtered value 14 and the vehicle position coordinate of the measured value is farther than the distance between the vehicle position coordinate of the dynamic model filtered value 24 and the measured vehicle position coordinate. The final correction value 30 will be calculated to be close to the dynamic model filtering value 24.

Further, the setting of the weight may include weighting at a rate inversely proportional to the difference between the kinematic model filtering value 14 and the measurement value and the difference between the dynamic model filtering value 24 and the measurement value. It can be configured to set. For example, the vehicle position coordinate of the kinematic model filtering value 14 is (20, 10), the vehicle position coordinate of the kinematic model filtering value 24 is (10, 20), and the kinematic model filtering value 14 The kinematic model filtering value 14 when the distance between the vehicle position coordinate and the vehicle position coordinate of the measured value is three times the distance between the vehicle position coordinate of the dynamic model filtering value and the vehicle position coordinate of the measured value Is given a weight of 0.25 and the kinetic model filtering value 24 is given a weight of 0.75, where the X coordinate value is (20 × 0.25) + (10 × 0.75) = 12.5 and the Y coordinate value is (10 × 0.25 ) + (20 x 0.75) = 17.5, and the vehicle position coordinates of the final correction value 30 are (12.5, 17.5). In FIG. 1, the updater 130 separately adds weights to the dynamic model filtering value 24 and the kinematic model filtering value 14, but the updater 130 may be combined with the combination unit 140. Combined into one, the combination unit 140 is configured to perform both the process of setting weights to the dynamic model filtering value 24 and the kinematic model filtering value 14 and the process of calculating the final correction value 30. Can be.

Meanwhile, in the vehicle position estimation method according to the present invention, the final correction value 30 as described above should be continuously calculated for each unit time, and after the step of calculating the final correction value 30, The method may further include determining whether an input signal has been inputted. If the vehicle position estimation stop signal is input, all processes are terminated. If the vehicle position estimation stop signal is not input, the kinematic model initial value 11 and the dynamic model are determined. It should be configured to return to the step of receiving the initial value 21.

In this case, when it is returned to the step of receiving the kinematic model initial value 11 and the dynamic model initial value 21 through the step of determining whether the position estimation stop signal of the vehicle is input, it is calculated in the previous cycle. The kinematic model filtering value 14 and the kinetic model filtering value are set to the kinematic model initial value 11 and the dynamic model initial value 21, respectively. In this way, the values calculated by the kinematic model extended Kalman filter 122 and the dynamic model extended Kalman filter 124 in the previous cycle are inputted as the kinematic model initial value 11 and the dynamic model initial value 21. The calculation process is similarly used in the conventional interactive multi-model filter 100, a detailed description thereof will be omitted.

On the other hand, when entering the step of receiving the kinematic model initial value 11 and the dynamic model initial value 21 again after one cycle of each of the above-mentioned steps, as calculated above, The kinematic model filtering value 14 and the dynamic model filtering value may be set to the kinematic model initial value 11 and the dynamic model initial value 21, respectively, but the kinematic model initial value 11 and When entering the step of receiving the dynamic model initial value 21 for the first time, since the calculated kinematic model filtering value 14 and the dynamic model filtering value do not exist, the position-related data of the vehicle received from the GPS is It should be set to the kinematic model initial value 11 and the kinetic model initial value 21.

In addition, in the vehicle position estimation method using the interactive multi-model filter 100 according to the present invention, the step of calculating the kinematic model mixing value 12 and the dynamic model mixing value 22 may be omitted (S20). . That is, in the vehicle position estimation method using the interactive multi-model filter 100 according to the present invention, the vehicle position predicted value when the vehicle moves under the condition that there is no slip, that is, the kinematic model predicted value 13 is obtained by the kinematic model. Computing with the extended Kalman filter 122, and calculating the vehicle position prediction value, ie, the dynamic model predicted value 23, when the vehicle is moved under the slip condition with the dynamic model extended Kalman filter 124 (S30). And the vehicle angular velocity measured by the yaw rate sensor 500 of the vehicle and the vehicle coordinates received from the GPS to calculate the current position-related data of the vehicle, that is, the measured value, and calculate the kinematic model prediction value 13 and Calculating a kinematic model filtering value 14 that is a value between the measured values and calculating a dynamic model filtering value 24 that is a value between the dynamic model predicted value 23 and the measured value (S40). Wow It may be configured to include a step (S60) of fusing the dynamic model filtering value 24 and the kinematic model filtering value 14 to calculate the final correction value 30.

Of course, a step S50 of setting weights to the kinematic model filtering value 14 and the kinetic model filtering value 24 may be added, wherein the weights are kinematic model filtering value 14 as mentioned above. And the difference between the measured value and the difference between the kinetic model filtering value 24 and the measured value.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the scope of the present invention is not limited to the disclosed exemplary embodiments. It will also be appreciated that many modifications and variations will be apparent to those skilled in the art without departing from the scope of the present invention.

100: interactive multi-model filter 110: mixing unit
120: Extended Kalman Filter
122: kinematic model extended Kalman filter
124: Dynamic Model Extended Kalman Filter
130: update unit 140: combination unit
200: wheel speed sensor 300: steering angle sensor
400: GPS 500: yaw rate sensor

Claims (14)

A mixing unit 110 for mixing the input data, an extension kalman filter 120 composed of a kinematic model extended Kalman filter 122 and a dynamic model extended Kalman filter 124, and a combination that fuses the input data. In the method for estimating the vehicle position using the interactive multi-model filter 100 including the unit 140,
Receiving, by the mixing unit 110, a kinematic model initial value 11 and a dynamic model initial value 21 of the vehicle (S10);
Calculating the kinematic model mixing value 12 and the dynamic model mixing value 22 by mixing the kinematic model initial value 11 and the dynamic model initial value 21 by the mixing unit 110. (S20);
The kinematic model extended Kalman filter 122 receives the kinematic model mixing value 12 to calculate a kinematic model prediction value 13, and the kinematic model extended Kalman filter 124 mixes the kinematic model. Receiving a value 22 and calculating a dynamic model prediction value 23 (S30);
The kinematic model extended Kalman filter 122 calculates the kinematic model filtered value 14, which is a value between the measured value indicating the current position of the vehicle and the kinematic model predicted value 13, and the kinematic model extended Kalman. The filter 124 calculating a dynamic model filtering value 24 which is a value between the measured value indicating the current position of the vehicle and the dynamic model prediction value 23 (S40);
Calculating, by the combining unit 140, a final correction value 30 by fusing the dynamic model filtering value 24 and the kinematic model filtering value 14;
Vehicle location estimation method comprising a.
The method of claim 1,
After the calculating of the final correction value (30) (S60), the interactive multi-model filter 100 further comprises a step (S70) of determining whether or not to input the vehicle position estimation stop signal, the position estimation of the vehicle If the stop signal is inputted, all processes are terminated, and if the vehicle position estimation stop signal is not inputted, the process returns to the step of receiving the kinematic model initial value 11 and the dynamic model initial value 21. Vehicle position estimation method.
The method of claim 2,
When entering the step S10 of receiving the kinematic model initial value 11 and the dynamic model initial value 21 for the first time, the position-related data of the vehicle received from the GPS is the kinematic model initial value 11. And dynamic model initial value (21),
When returning to the step S10 of receiving the kinematic model initial value 11 and the dynamic model initial value 21 through the step S70 of determining whether to input the position estimation stop signal of the vehicle, the previous cycle And the kinematic model filtering value (14) and the kinetic model filtering value calculated at are set to the kinematic model initial value (11) and the kinetic model initial value (21), respectively.
The method of claim 1,
The interactive multi-model filter 100 further includes an updater 130 that weights the data.
Before calculating the final correction value 30 (S60), the update unit 130 and the difference between the kinematic model filtering value 14 and the measured value and the dynamic model filtering value 24 and The difference between the measured values and the difference between the measured values is greater than that of the measured values among the kinematic model filtered value 14 and the dynamic model filtered value 24. The method further includes setting a large weight in step S50,
Computing the final correction value (30) (S60), the combination unit 140 is the dynamic model filtering value 24 and the kinematic model filtering value 14 to the update unit 130. And a fusion after adding the weights set to calculate the final correction value (30).
The method of claim 4, wherein
The step of setting the weight by the updater 130 (S50) may be based on the difference between the kinematic model filtering value 14 and the measured value and the difference between the kinematic model filtering value 24 and the measured value. And set weights at inversely proportional rates.
The method according to any one of claims 1 to 5,
Computing the kinematic model mixing value 12 and the dynamic model mixing value 22 (S20),
The mixing unit 110 calculates the kinematic model mixing value 12 by mixing the kinematic model initial value 21 with the kinematic model initial value 11 by a predetermined ratio, and calculates the kinematic model initial value. And (21) calculating the dynamic model mixing value (22) by mixing the initial kinematic model initial value (11) with a predetermined ratio.
The method of claim 6,
The ratio of mixing the kinematic model initial value 21 with the kinematic model initial value 11 in the step S20 of calculating the kinematic model mixing value 12 and the dynamic model mixing value 22. And the ratio of mixing the kinematic model initial value (11) to the dynamic model initial value (21) is variable according to the user's selection.
The method according to any one of claims 1 to 5,
The measured value includes the current angular velocity of the vehicle and the current coordinates of the vehicle,
The measured value is calculated based on the angular velocity of the vehicle measured by the angular velocity measuring sensor pre-mounted on the vehicle, and the current coordinates of the vehicle received from the GPS (400).
9. The method of claim 8,
The angular velocity measuring sensor is a yaw rate sensor (500).
A mixing unit 110 for mixing the input data, an extension kalman filter 120 composed of a kinematic model extended Kalman filter 122 and a dynamic model extended Kalman filter 124, and a combination that fuses the input data. In the method for estimating the vehicle position using the interactive multi-model filter 100 including the unit 140,
The kinematic model extended Kalman filter 122 calculates the kinematic model prediction value 13 of the vehicle, and the kinematic model extended Kalman filter 124 calculates the dynamic model predicted value 23 of the vehicle (S30). );
The kinematic model extended Kalman filter 122 calculates the kinematic model filtered value 14, which is a value between the measured value indicating the current position of the vehicle and the kinematic model predicted value 13, and the kinematic model extended Kalman. The filter 124 calculating a dynamic model filtering value 24 which is a value between the measured value indicating the current position of the vehicle and the dynamic model prediction value 23 (S40);
Calculating, by the combining unit 140, a final correction value 30 by fusing the dynamic model filtering value 24 and the kinematic model filtering value 14;
Vehicle location estimation method comprising a.
The method of claim 10,
The interactive multi-model filter 100 further includes an updater 130 that weights the data.
Before calculating the final correction value 30 (S60), the update unit 130 and the difference between the kinematic model filtering value 14 and the measured value and the dynamic model filtering value 24 and The difference between the measured values and the difference between the measured values is greater than that of the measured values among the kinematic model filtered value 14 and the dynamic model filtered value 24. The method further includes setting a large weight in step S50,
Computing the final correction value (30) (S60), the combination unit 140 is the dynamic model filtering value 24 and the kinematic model filtering value 14 to the update unit 130. And a fusion after adding the weights set to calculate the final correction value (30).
The method of claim 11,
In step S50 of setting the weight by the updater 130, the difference between the kinematic model filtering value 14 and the measurement value and the difference ratio between the dynamic model filtering value 24 and the measurement value And set a weight at a rate inversely proportional to the vehicle position estimation method.
The method according to any one of claims 10 to 12,
The measured value includes the current angular velocity of the vehicle and the current coordinates of the vehicle,
The measured value is calculated based on the angular velocity of the vehicle measured by the angular velocity measuring sensor pre-mounted on the vehicle, and the current coordinates of the vehicle received from the GPS (400).
The method of claim 13,
The angular velocity measuring sensor is a yaw rate sensor (500).

Images