WO2021080332A2

WO2021080332A2 - System for predicting vehicle trajectory by using extended kalman filter in vehicle software defined networking and method therefor, and computer-readable recording medium on which program for performing method is recorded

Info

Publication number: WO2021080332A2
Application number: PCT/KR2020/014441
Authority: WO
Inventors: 송왕철; 아바스 무하마드타히르; 지브란 무하마드알리
Original assignee: 제주대학교 산학협력단
Priority date: 2019-10-21
Filing date: 2020-10-21
Publication date: 2021-04-29
Also published as: KR20210047423A; KR102287546B1; WO2021080332A3

Abstract

According to an embodiment of the present invention, a network slice selection function for identifying and selecting a network slice instance for an interaction with a mobility management entity (MME)/access mobility function (AMF) may be provided by implementing a function of selecting a correct network slice instance appropriately matching a request of a user terminal in a 5G architecture (23.501 3GGP technical specification), in particular, finding out a current state of a network via communication with a VNF orchestration platform.

Description

Vehicle trajectory prediction system using an extended Kalman filter in vehicle software-defined networking, its method, and a computer-readable recording medium in which a program that performs the method is recorded

The present invention relates to a technology capable of stabilizing future vehicle networking by calculating the positions of vehicles on a road in a certain area in an edge cloud.

Vehicle group driving consists of a top-level vehicle driven directly by the driver and a number of sub-vehicles following it. In this case, the lower vehicles may be autonomous driving without driver intervention. Vehicle group driving like this can be said to be an intermediate step toward fully autonomous driving, and technology development is underway in advanced countries such as Europe.

The SARTRE (Safe Road Trains for the Environment) project was researched as a part of EC FP7, and developed a system to safely and conveniently perform vehicle platoon or vehicle group driving when driving on general public highways.

The system provides a safe vehicle group driving service while interoperating with other traffic on the highway. In particular, the system provides autonomous driving for lower-level vehicles, except for top-level vehicles driven by individual drivers.

Software Defined Networking (SDN) is a new network structure or new paradigm, and separates the network into a control plane and a transfer plane. One of the technologies implementing this is OpenFlow. The Open Networking Foundation (ONF), which develops OpenFlow-related standards, has been the most influential until recently, but recently, research on NFV (Network Function Virtualization) has been actively conducted by ETSI (European Telecommunications Standards Institute) and IETF (Internet Engineering Task Force). . Recently, network function virtualization technology has brought about a new change in the overall network architecture, which was mainly hardware-oriented. Network Function Virtualization (NFV), that is, NFV separates hardware and software components of a network, and virtualizes the functions of physical network facilities to create a VM (Virtual Machine) server, hardware equipped with a general-purpose processor, and cloud. It is a concept running on a Ding computer. According to this, various network devices, such as routers, load balancers, firewalls, intrusion prevention, and virtual private networks, can be implemented in software on a general server, thereby avoiding the vendor dependence of network configuration. This is because expensive dedicated equipment can be replaced with general-purpose hardware and dedicated software. Furthermore, it has the advantage of being able to quickly respond to changes in traffic as well as reducing equipment operating costs.

Meanwhile, Software Defined Networking (SDN), that is, SDN technology, is characterized by separating the functions of a complex control plane from a data plane. According to this, the complex functions of the control plane are processed by software, and the data plane performs only simple functions indicated by the control plane, such as forwarding, ignoring, and changing network packets. By applying these technologies, it is possible to develop new network functions with software without complicated hardware constraints, and at the same time, various attempts that were not possible in the previous network structure became possible.

Although the NFV and SDN are separate technologies, they can work complementarily. This is because various network functions implemented in software by NFV can be efficiently controlled using SDN.

SDN is based on two basic principles.

First, SDN must perform Software Defined Forwarding. Software-defined forwarding means that the data forwarding function handled by hardware in the switch/router must be controlled through an open interface and software.

Second, SDN aims at Global Management Abstraction. SDN enables more advanced network management through abstraction.

For example, this abstraction may include a response to an event (such as a topology change or a new flow input) based on the state of the entire network, and the ability to control network elements.

As interest in autonomous vehicles increases, interest in vehicle communication naturally also increases. Stability is also of great interest in vehicle networks in which various types of data are to be transmitted, but because of its mobility, connection of vehicles on the road has been studied for a long time on vehicle-to-vehicle communication, but there is a need to further improve the communication stability. .

Because the vehicle's movement is entirely dependent on the vehicle driver's habits and judgments, the vehicle's trajectory prediction is not deterministic.

Furthermore, in many cases, vehicle trajectory prediction has been limited only to models that predict well when the vehicle operates at a constant acceleration at a constant speed. In addition, the turn model prevents unexpected changes in the vehicle position. As a result, V2V (V2V, Vehicle to Vehicle, vehicle-to-vehicle communication) and V2I (V2I, Vehicle to Infrastructure, communication between vehicle and roadside base stations) were hardly considered as a problem causing communication failure.

The present invention was conceived to solve the above-described problem, and in an embodiment of the present invention, an Extended Kalman Filter (EKF; extended Kalman filter) incorporating an Interacting Multiple Model (IMM) is used. It is possible to provide a method of predicting the vehicle position, and through this, it is possible to provide a system that can stabilize the networking of vehicles in the future by calculating the positions of vehicles on roads in a certain area in the edge cloud.

In addition, in the present invention, since the navigation map contains most of the information on the road, information on the number of lanes on the road can be known, it is possible to correlate it with the vehicle speed to present a vehicle turn model.

In order to solve the above-described problem, in an embodiment of the present invention, in a system for tracking the location of a vehicle including a device capable of Software Defined Networking (SDN) under an edge crowd, tracking the location of the vehicle SDN control device installed in a vehicle applying the edge crowd to a specific target area; An edge controller that receives driving information including vehicle speed, position, and acceleration generated in the vehicle from the SDN control device; The edge controller, based on the driving information provided by the driving means, calculates and accumulates a probability density function by an Interacting Multiple Model (IMM) unit to calculate an expected position of the vehicle. It is possible to provide a vehicle trajectory prediction system using an extended Kalman filter in vehicle software defined networking including a filter (Extended Kalman Filter; EKF) module.

According to an embodiment of the present invention, it is possible to provide a method of predicting a vehicle position on a road by using an Extended Kalman Filter (EKF; Extended Kalman Filter) incorporating an Interacting Multiple Model (IMM), Through this, by calculating the positions of vehicles on roads in a certain area in the edge cloud, there is an effect of stabilizing future vehicle networking.

In addition, in the present invention, since the navigation map contains most of the information on the road, information on the number of lanes on the road can be known, so it is possible to correlate it with the vehicle speed and present a turn model.

The least stable part of vehicle communication is due to the situation in which the communication link between vehicles needs to be changed from time to time due to the mobility of the vehicle.If the location of the vehicle is stably predicted, the controller controls the limit of the wireless network to prevent loss of traffic. It will be possible.

1 is a block diagram showing a vehicle trajectory prediction system using an extended Kalman filter in vehicle software defined networking implemented according to an embodiment of the present invention, and FIG. 2 is a configuration centered on an edge controller among the components of FIG. 1 It is a conceptual diagram showing together.

3 is an application conceptual diagram of implementing a vehicle trajectory prediction system using an extended Kalman filter in vehicle software defined networking according to an embodiment of the present invention.

Advantages and features of the present invention, and a method of achieving them will become apparent with reference to embodiments to be described later in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but will be implemented in a variety of different forms.

In the present specification, the present embodiment is provided to complete the disclosure of the present invention, and to fully inform the scope of the invention to those of ordinary skill in the art to which the present invention pertains. And the invention is only defined by the scope of the claims. Accordingly, in some embodiments, well-known components, well-known operations, and well-known techniques have not been described in detail in order to avoid obscuring interpretation of the present invention.

The configuration provided in the embodiment of the present invention proposes a method of predicting a vehicle position on a road using an Extended Kalman Filter (EKF; Extended Kalman Filter) incorporating an Interacting Multiple Model (IMM), The point is to implement a method that can stabilize the networking of vehicles in the future by calculating the positions of vehicles on the road in a certain area in the edge cloud.

1 is a block diagram showing a vehicle trajectory prediction system (hereinafter referred to as'the present invention') using an extended Kalman filter in a vehicle software defined networking implemented according to an embodiment of the present invention, and FIG. 2 is FIG. It is a conceptual diagram showing the configuration centered on the edge controller among the components of.

Referring to FIGS. 1 to 3, the present invention provides a system for tracking the location of a vehicle including a device capable of Software Defined Networking (SDN) under an edge crowd. SDN control device (110, 120, 130) installed in a vehicle that applies the edge crowd to a specific area, and operation information including vehicle speed, position, and acceleration generated in the vehicle is transmitted from the SDN control device. It may be configured to include an edge controller (EC).

In this case, the edge controller (EC) calculates the probability density function by the IMM (Interacting Multiple Model) unit 210 based on the operation information provided from the SDN control device of the vehicle. And an Extended Kalman Filter (EKF) module 200 that accumulates and calculates an expected position of the vehicle.

Therefore, in the embodiment in which the present invention is implemented, the SDN networking technology is implemented in vehicle communication, and communication is possible through this, and because one SDN control device cannot control the entire network, the edge cloud for each region When is applied, an environment in which information transmission and calculation for roads in the area is possible is premised.

Therefore, it is assumed that if the computing resource in the edge cloud, not the computing resource of the vehicle itself, can continuously calculate the position of the vehicle using the Kalman filter, the SDN controller can perform much more stable communication control.

The least stable part of vehicle communication is due to the situation in which the communication link between vehicles must be changed from time to time due to the mobility of the vehicle.If the position of the vehicle is stably predicted by the system according to the present invention, the controller can overcome the limitations of the wireless network. Beyond that, it becomes controllable so that no traffic loss occurs.

In order to implement this function, as shown in FIG. 2, the edge controller (EC) of the present invention includes an Interacting Multiple Model (IMM) unit based on the driving information provided by the driving means. It includes an extended Kalman filter (EKF) module 210 that calculates and accumulates the probability density function by 210 to calculate the predicted position of the vehicle, and provides a predicted model for predicted dynamic information. It can be configured to include a prediction model providing unit 220.

In particular, the prediction model providing unit derives and provides five KF (Kalman filter) models in consideration of all possible situations of vehicle movement.

Each of these models is set to fit a specific set of scenarios in which the vehicle can be found, i.e. a constant jerk model, a constant acceleration model, a constant speed model, a fixed position model, and a vehicle turning model.

The prediction model provided by the prediction model providing unit 220 may be provided as a total of five models according to the following {Equation 1} to {Equation 5}. (Vehicle position (Xv), vehicle speed (Vv), and distance from the nearest intersection in the direction of vehicle movement (Dv))

{Equation 1: Constant location model (CL)}

(Constant position model CL describes a scenario where the vehicle does not move or the speed is '0'.)

{Equation 2: Constant velocity model (CV)}

(Constant speed model (CV) _{represents the situation of the vehicle when the vehicle with the current position Xv t} _{of the vehicle and the distance DvI t} at the nearest intersection in the vehicle movement direction moves to the registration Vvt. ω is the processing noise covariance (process). noise covariance), which is a constant.)

{Equation 3: Constant acceleration model (CA)}

(Constant acceleration model (CA) represents the situation of the vehicle when the vehicle moves with a constant acceleration (a). In this equation, the state of the vehicle is considered to be in a moving state with a flexible position, and Δ t is the previous period. And thus the variable of (t-1) means the statistics of the previous period, ω is the process noise covariance, which is a constant.)

{Equation 4: Constant Jerk model (constant jerk model; CJ)}

(Constant jerk model (CJ) represents a situation when a vehicle moves at a constant acceleration with a change in acceleration for a certain period of time.)

{Equation 5: Constant vehicle turn model (vehicle turning model; VT)}

(Vehicle's Turning Model (VT) tells the vehicle's SDN control unit that the vehicle is continuously rotating, or two parameters: junction/intersection and lane number (rightmost, left or middle). To express that you will go straight on the same road.)

That is, if the vehicle is in the largest lane on the right, and the DvIt value is nearly '0', the vehicle will turn to the right.

Alternatively, if the vehicle is placed in the left lane and the DvIt value is close to '0', the vehicle will drive to the left. However, vehicles in the middle lane follow a straight path.

The Kalman filter of the present invention operates with a feedback approach, first processes data at time T, and receives feedback from the vehicle's SDN controller in terms of GPS measurement including vehicle speed, position and acceleration. State vectors are formed using these parameters, which are decomposed into x and y components, respectively.

In order to define the state vector for the above-described model in the present invention, based on the satellite positioning system (GPS) as shown in Equation 1, the vehicle position (Xv), the vehicle speed (Vv), and from the nearest intersection in the vehicle movement direction are Three parameters of distance (Dv) were considered.

{Equation 1}

Then the three parameters are decomposed into x and y components, respectively. Thus, for every Xv in the equation, it is denoted as Xxv and Xyv, respectively.

In the above equation, every parameter has two components, x and y, i.e. the position of the x and y coordinates, and the normal and tangential acceleration. In addition, the formula for vehicle position correction and prediction using KF is as follows.

[Vehicle position correction prediction formula]

H is the Jacobian of the parametric model, P is the prediction error covariance, R is the noise covariance of the model, and K is the Kalman gain. x represents the prediction of the vehicle's condition at time k-1, k represents the current time, and A represents Jacobian's system model for the current state.

In general, since a vehicle has its own GIS system such as a navigation map, it is possible to reduce the accuracy of location prediction by recognizing such a road on a curved road, but in an embodiment of the present invention, an equation is presented in consideration of only a straight road. . For the Kalman filter, the predicted covariance P (covariance P) is

(In this case, the vehicle position (Xv), the vehicle speed (Vv), and the distance from the nearest intersection in the vehicle movement direction (Dv). The matrix ′P′ represents the estimated error covariance, and then ′R′ (measured noise ) Are used together with'K' (covariance to estimate the value of Kalman gain) and'H' (Jacobia matrix). The matrix'PP is a data set, describing the association in the observation error between all pairs at the vertical level. In our case, the error covariance for all proposed models is estimated by mapping the filter itself.)

After obtaining all possible scenarios of the vehicle, the probability density function (pdf) is calculated and the most probable model is selected to predict the future position of the vehicle. This is done by the IMM algorithm that accumulates the probability of occurrence for each model using the Markov model, and the final transition metric is defined as all probabilities.

Through the five prediction models (Kalman model), the IMM unit predicts the next point through the feedback processor as follows.

[Feedback Processor]

The above process can be provided as a solution driven in the following manner through the system of the present invention.

First, in an edge crowd, in a method of tracking the location of a vehicle equipped with a device capable of Software Defined Networking (SDN), the vehicle location, speed, and speed, to the edge controller of the edge cloud through the vehicle's SDN device, Step 1 of transmitting driving information including acceleration can be performed.

In this case, in the extended Kalman filter (EKF) module 200 in the edge controller, the probability density by the IMM (Interacting Multiple Model; Interacting Multiple Model Program) unit 210 based on the operation information The second step of calculating and accumulating the function to calculate the expected position of the vehicle may be performed.

In particular, the second step is, in the IMM (Interacting Multiple Model; interactive multiple model program) unit 210, calculates an occurrence probability for a predictive model modeled on the movement of the vehicle, and applies the calculated probability to the corresponding model. The step of identifying the dominance of vehicle position prediction through may be performed.

When the IMM unit 210 identifies the dominant model (from the models according to Equations 1 to 5) through the feedback process, the probability for each iteration is continuously recalculated and calculated from the previous iteration. A new probability value is weighted with respect to the generated probability value, and the probability of each model is calculated through the state transition matrix according to Equation 7 below, based on the calculated information. In this process, the IMM unit 210 calculates a probability density function (pdf) using a Markov (Markov) chain model, and then calculates a cumulative probability for each model.

{Equation 7}

(CL: constant position model, CV: constant speed model, CA: constant acceleration model, CJ: constant jerk model, VT: vehicle turning model)

For example, the current vehicle position is estimated using a constant position model (CL), and then the CL model is re-applied to find the vehicle position in the future, or a probability of applying another model is calculated.

To this end, the IMM unit calculates the probability density function (pdf) using the Markov chain model and then calculates the probability. The model with the highest probability is selected for vehicle position prediction.

3 is a conceptual diagram showing the application of the system (FIGS. 1 and 2) according to the present invention.

The operational state of the present invention is summarized as follows with reference to FIG. 3.

In the Soft-Defined Internet (SD-IoV) of vehicles moving to a specific area, vehicles on the road search cellular networks to transmit information dynamically, and RSU (Road Side Unit) and RSU (Road Side Unit) to communicate data between vehicles. Unit).

In this case, the extended Kalman filter module in the present invention is used for vehicle prediction in the edge controller (EC).

In general, the Kalman filter (KF) provides an efficient recursive method for predicting the future state of a vehicle using a series of mathematical equations. The Kalman filter (KF) with a multi-model approach is a single model, but avoids complex models. Scenarios are divided into sub-scenarios according to each KF model because more accurate results are obtained than scenarios with high complexity. Scenarios for such a vehicle are a non-powered vehicle, a constant acceleration vehicle, a vehicle with a constant speed, and a variable acceleration vehicle (five models proposed in the present invention). In the case of the extended Kalman filter module in the present invention, the predicted position of the vehicle is calculated by calculating and accumulating a probability density function by an IMM (Interacting Multiple Model) unit based on driving information through five models. It is done to calculate.

That is, by combining the scenarios such as movement, acceleration, position, variable, and turn of the vehicle, a complete series of equations can be derived to predict the vehicle position in time (T). This prediction process is performed by the edge controller (EC). After fully predicting the future location of the vehicle, flow rules are installed in each RSU (Road Side Unit) to efficiently deliver data packets in the predicted path. When a flow rule is installed in the RSU (Road Side Unit) and adjacent vehicles, the source vehicle does not repeatedly send a route request to the SDN controller as shown in FIG. 3, and thus a new route can be immediately found.

That is, the present invention introduces a new technique to the soft-definition Internet of a vehicle in order to estimate the location of a vehicle, so that a new route can be secured in advance when a vehicle enters a new road section with another neighboring vehicle. The five position prediction models proposed in the present invention can predict the vehicle position not only on a straight path, but also on a random and curved path.

As interest in autonomous vehicles increases, vehicle communication, the basis of which is essential, is essential. Among vehicle communications, V2V, that is, vehicle-to-vehicle communication, is the field with the lowest stability, so it is essential to supplement this.

When performing communication necessary for autonomous driving through the LTE network, a huge amount of data needs to be exchanged, and since this is only considering V2I, which is a direct communication method, communication through the RSU (Road Side Unit) through V2V becomes indispensable. .

Accordingly, the present invention can provide a model for predicting the position of a vehicle, which is essential for stabilizing such vehicle-to-vehicle communication, and thus, if the SDN controller is combined with a direct stabilization model for vehicle communication in the future, it is considered to have very great competitiveness.

The system and method according to the present invention are intended to bring about stabilization of vehicle communication by estimating the location of the vehicle, and can be performed in the edge cloud.

Therefore, it can be used by communication network operators as well as ITS centers that require communication stabilization of the vehicle by using the communication infrastructure of the backbone network, and can be applied to companies that provide vehicle communication services.

The vehicle position prediction implementation solution function performed through each model implemented through the extended Kalman filter module provided by the present invention can be programmed and provided, and in the present invention, a computer-readable recording of such an application program is recorded. The medium can be executed using a computer or a mobile communication terminal (eg, a smartphone or a tablet PC). In this case, such an application program may be installed on a hard disk of a computer, installed on a CD-ROM or a DVD-ROM, or installed on a USB memory and executed. In addition, it can be installed and executed in various playback devices.

In addition, the communication network to which the system of the present invention is applied is a network capable of wired and wireless communication, and a mobile communication network installed and operated by a communication company (including 3G network, 4G network, 5G network, WiBro network, LTE network), Internet network/ It may include a public switched telephone network (PSTN) network.

The configuration and execution operation of the Kalman filter module applied to the present invention and implemented in the edge controller may be represented by functional block configurations and various processing steps. These functional blocks may be implemented with various numbers of hardware or/and software configurations that perform specific functions. For example, the present invention provides integrated circuit configurations such as memory, processing, logic, and look-up tables, which can execute various functions by controlling one or more microprocessors or by other control devices. Can be adopted.

Similar to how the elements of the present invention can be implemented with software programming or software elements, the present invention includes various algorithms implemented with a combination of data structures, processes, routines or other programming constructs, including C, C++. , Java, assembler, or the like may be implemented in a programming or scripting language. Functional aspects can be implemented with an algorithm running on one or more processors. In addition, the present invention may employ conventional techniques for electronic environment setting, signal processing, and/or data processing. Terms such as "mechanism", "element", "means", and "configuration" may be used broadly, and are not limited to mechanical and physical configurations. The term may include a meaning of a series of routines of software in connection with a processor or the like.

As described above, the technical idea of the present invention has been described in detail in the preferred embodiment, but the preferred embodiment is for the purpose of explanation and not limitation. As such, it will be understood by those of ordinary skill in the art that various embodiments are possible through a combination of the embodiments of the present invention within the scope of the technical idea of the present invention.

Claims

In a system for tracking the location of a vehicle equipped with a device capable of Software Defined Networking (SDN) under an edge crowd,

An SDN control device installed in a vehicle applying an edge crowd to a specific area to be tracked of a vehicle location;

An edge controller that receives driving information including vehicle speed, position, and acceleration generated in the vehicle from the SDN control device;

The edge controller,

Based on the driving information provided by the driving means, an extended Kalman filter that calculates and accumulates a probability density function by an Interacting Multiple Model (IMM) unit to calculate the expected position of the vehicle. ; EKF) module;

Containing,

Vehicle trajectory prediction system using extended Kalman filter in vehicle software-defined networking.
The method of claim 1,

The extended Kalman filter module,

Includes; a prediction model providing unit that provides a predicted model for predicted dynamic information of the vehicle,

The prediction model provided by the prediction model providing unit is determined by the following {Equation 1} to {Equation 5},

Vehicle trajectory prediction system using extended Kalman filter in vehicle software-defined networking.

{Equation 1: Constant location model (CL)}

(Constant position model CL describes a scenario where the vehicle does not move or the speed is '0'.)

{Equation 2: Constant velocity model (CV)}

(Constant speed model (CV) represents the situation of the vehicle when the vehicle with the current position Xv t of the vehicle and the distance DvI t at the nearest intersection in the vehicle movement direction moves to the registration Vvt. ω is the processing noise covariance (process). noise covariance), which is a constant.)

{Equation 3: Constant acceleration model (CA)}

(Constant acceleration model (CA) represents the situation of the vehicle when the vehicle moves with a constant acceleration (a). In this equation, the state of the vehicle is considered to be in a moving state with a flexible position, and Δ t is the previous period. And thus the variable of (t-1) means the statistics of the previous period, ω is the process noise covariance, which is a constant.)

{Equation 4: Constant Jerk model (constant jerk model; CJ)}

(Constant jerk model (CJ) represents a situation when a vehicle moves at a constant acceleration with a change in acceleration for a certain period of time.)

{Equation 5: Constant vehicle turn model (Vehicle turning model; VT)}

(Vehicle's turning model (VT) tells the vehicle's SDN control unit that the vehicle is continuously rotating, or two parameters: junction/intersection and lane number (rightmost, left or middle). To express that you will go straight on the same road.)
The method of claim 2,

The IMM part of the extended Kalman filter module,

The following feedback process is repeated using the five models provided by the prediction model providing unit,

Calculating the probability of occurrence for each individual model, and using the calculated probability to identify which model will dominate the vehicle's position prediction,

Vehicle trajectory prediction system using extended Kalman filter in vehicle software defined networking.
The method of claim 3,

The feedback process,

Performing the prediction process through the following {Equation 6-1} and the update process through {Equation 6-2}, and correcting the prediction error,

Vehicle trajectory prediction system using extended Kalman filter in vehicle software defined networking.

{Equation 6-1}

{Equation 6-2}

( H represents the Jacobian matrix of the parametric model, P is the prediction error covariance, R is the measurement noise covariance of the model, K is the Kalman gain. x represents the prediction of the vehicle's condition at time t-1, and k is the current Represents time, and A represents Jacobian's system model for the current state. )
The method of claim 4,

The IMM unit,

In the case of identifying the dominant model through the feedback process,

It is performed by continuously recalculating the probability for each iteration over the entire driving of the vehicle and weighting a new probability value to the probability value calculated in the previous iteration,

Based on the information calculated through this, the probability of each model is calculated through the state transition matrix according to Equation 7 below,

Vehicle trajectory prediction system using extended Kalman filter in vehicle software defined networking.

{Equation 7}

(CL: constant position model, CV: constant speed model, CA: constant acceleration model, CJ: constant jerk model, VT: vehicle turning model)
The method of claim 5,

The IMM unit,

After calculating the probability density function (pdf) using the Markov chain model, we compute the cumulative probability for each model,

Vehicle trajectory prediction system using extended Kalman filter in vehicle software defined networking.
In the method for tracking the location of a vehicle equipped with a device capable of Software Defined Networking (SDN) under an edge crowd by applying the system according to claim 1,

1 step of transmitting driving information including the position, speed, and acceleration of the vehicle to the edge controller of the edge cloud through the SDN device of the vehicle;

In the Extended Kalman Filter (EKF) module in the edge controller, the probability density function is calculated and accumulated by an Interacting Multiple Model (IMM) unit based on the driving information to predict the vehicle. A second step of calculating a location; Including,

The second step,

In the IMM (Interacting Multiple Model; Interacting Multiple Model Program),

A step of calculating the probability of occurrence of a predictive model modeled on the movement of a vehicle, and identifying the dominance of vehicle position prediction through the corresponding model by applying the calculated probability,

Vehicle trajectory prediction method using extended Kalman filter in vehicle software-defined networking.
The method of claim 7,

The prediction model,

Including a constant location model (Constant location model; CL) according to the following {Equation 1},

Vehicle trajectory prediction method using extended Kalman filter in vehicle software-defined networking.

{Equation 1}: Constant location model

(Constant position model CL describes a scenario where the vehicle does not move or the speed is '0'.)
The method of claim 8,

The prediction model,

Including a constant velocity model (Constant velocity model; CV) according to the following {Equation 2},

Vehicle trajectory prediction method using extended Kalman filter in vehicle software-defined networking.

{Equation 2: Constant velocity model (CV)}

(Constant speed model (CV) represents the situation of the vehicle when the vehicle with the current position Xv t of the vehicle and the distance DvI t at the nearest intersection in the vehicle movement direction moves to the registration Vvt. ω is the processing noise covariance (process). noise covariance), which is a constant.)
The method of claim 9,

The prediction model,

Including a constant acceleration model (Constant acceleration model; CA) according to the following {Equation 3},

Vehicle trajectory prediction method using extended Kalman filter in vehicle software-defined networking.

{Equation 3: Constant acceleration model (CA)}

(Constant acceleration model (CA) represents the situation of the vehicle when the vehicle moves with a constant acceleration (a). In this equation, the state of the vehicle is considered to be in a moving state with a flexible position, and Δ t is the previous period. And thus the variable of (t-1) means the statistics of the previous period, ω is the process noise covariance, which is a constant.)
The method of claim 10,

The prediction model,

Including a constant jerk model (Constant Jerk model; CJ) according to the following {Equation 4},

Vehicle trajectory prediction method using extended Kalman filter in vehicle software-defined networking.

{Equation 4: Constant Jerk model (constant jerk model; CJ)}

(Constant jerk model (CJ) represents a situation when a vehicle moves at a constant acceleration with a change in acceleration for a certain period of time.)
The method of claim 11

The prediction model,

Including a vehicle turning model (Constant vehicle turn model; VT) according to the following {Equation 5},

Vehicle trajectory prediction method using extended Kalman filter in vehicle software-defined networking.

{Equation 5: Constant vehicle turn model (Vehicle turning model; VT)}

(Vehicle's Turning Model (VT) tells the vehicle's SDN control unit that the vehicle is continuously rotating, or two parameters: junction/intersection and lane number (rightmost, left or middle). To express that you will go straight on the same road.)
The method of claim 12,

The second step,

Using the model, the following feedback process is repeated, the probability of occurrence for each individual model is calculated, and the calculated probability is used to identify which model will dominate the vehicle position prediction,

It is performed by continuously recalculating the probability for each iteration over the entire driving of the vehicle and weighting a new probability value to the probability value calculated in the previous iteration,

Based on the information calculated through this, the probability of each model is calculated through the state transition matrix according to the following {Equation 7},

Vehicle trajectory prediction method using extended Kalman filter in vehicle software-defined networking.

{Equation 7}

(CL: constant position model, CV: constant speed model, CA: constant acceleration model, CJ: constant jerk model, VT: vehicle turning model)
A computer-readable recording medium in which a program for performing a vehicle trajectory prediction method using an extended Kalman filter in vehicle software defined networking according to claim 13 is recorded.