KR20190130210A - Apparatus for generating continuous mobility information, method therefor, and unmanned ground vehicle - Google Patents

Abstract

According to one embodiment of the present invention, an apparatus for generating continuous spatial mobility information generates mobility data which represents whether the vehicle may drive on the absolute coordinates by utilizing a method and environmental awareness information obtained after the unmanned vehicle drives on an unpaved/paved road and generates continuous spatial mobility information which represents whether the vehicle may drive on continuous space by utilizing a machine learning technique based on the mobility data so as to display a drivable area with continuity on a wide road plan, thereby enabling an unamend vehicle, which performs a wide road plan, to more safely perform autonomous driving.

Description

Apparatus, method and unmanned vehicle for generating continuous space mobility information {APPARATUS FOR GENERATING CONTINUOUS MOBILITY INFORMATION, METHOD THEREFOR, AND UNMANNED GROUND VEHICLE}

TECHNICAL FIELD The present invention relates to autonomous driving of an unmanned vehicle, and more particularly, to an apparatus, a method, and an unmanned vehicle for generating continuous space mobility information.

In general, in order for an unmanned vehicle to autonomously drive, a wide-area route plan that uses mobility information of several kilometers of operation area to follow the operator's target point before the unmanned vehicle starts autonomous driving must be preceded.

In addition, for more stable autonomous driving, information processing technology that efficiently manages and maintains the similarity between the mobility information and the actual environment indicating whether the actual unmanned vehicle can be driven is essential.

In connection with this, in the related art, a method of constructing a database based on geographic information and sensor information that can be used for autonomous driving of an unmanned vehicle has been proposed to use in an autonomous driving function. In addition, in order to solve problems such as data inconsistency, data size, and storage structure complexity in using terrain information with different data characteristics with heterogeneous classification system, a mobile DB is built by fusing terrain information effectively. A method has been proposed for use in planning broad-area pathways.

However, in the pre-built field environment, the unmanned vehicle is not only movable information that can be driven, but also navigation information of a relatively well-established road environment is mounted on the unmanned vehicle and thus there is a problem of inconsistency with the navigation information utilized for autonomous driving. Uncertainty of wide-area route points that are used in conjunction with environmental awareness information has been reflected in the local route planning.

In addition, the existing mobility information is serviced in the form of a unit grid map, which has caused difficulties in increasing data throughput in the case of high resolution depending on the resolution (10cm, 25cm, 1m, etc.) of the movable DB.

(Patent literature)

Republic of Korea Patent No. 10-1703144 (Registration date January 31, 2017)

Accordingly, in an embodiment of the present invention, the unmanned vehicle generates mobility data indicating whether the vehicle can be driven in absolute coordinates using navigation and environmental recognition information obtained after driving on an unpaved / paved road, and machine learning based on the mobility data. By generating continuous spatial mobility information indicating whether the vehicle can be driven in continuous space using the technique, the unmanned vehicle performing the wide-area planning can be more stably operated by autonomous driving so that the driving area is displayed with continuity. An apparatus, method, and an unmanned vehicle for generating continuous space mobility information are provided.

An apparatus for generating continuous spatial mobility information for an unmanned vehicle according to an embodiment of the present invention, comprising: a navigation information generation unit for generating navigation information by obtaining absolute coordinates of a route on which an unmanned vehicle travels; An environment recognition information generation unit for generating environment recognition information for the surrounding area of the mobile terminal; and an operability indicating whether the vehicle is capable of traveling in absolute coordinates corresponding to the path and the surrounding area based on the navigation information and the environment recognition information ( a mobility data generation unit for generating mobility data and machine learning based on the mobility data to calculate a driving possibility probability for continuous space between absolute coordinates in which the mobility data is generated; On the basis of the probable probability of travel, whether or not the vehicle can be driven for each absolute coordinate corresponding to the continuous space Indicating to a continuous space comprising a movable continuous space information generator for generating a mobility information.

The navigation information generator may be configured to obtain absolute coordinates of the route based on GPS information on a current position of the unmanned vehicle.

The environment recognition information generation unit may generate the environment recognition information using an environment recognition sensor mounted in the unmanned vehicle.

The environment sensor may be at least one of a LIDAR, a camera, and a RADAR.

The apparatus may further include a movable database (DB) for storing the movable data and the continuous space movable information.

)로 구성된 로지스틱 회귀 분류기를 이용하여 상기 주행 가능 확률을 산출하는 것을 특징으로 한다.The continuous spatial mobility information generation unit may include a plurality of sparse kernels and weights assigned to each sparse kernel for an area including the path and the peripheral area.

The driving probability is calculated using a logistic regression classifier consisting of a).

The continuous spatial mobility information generation unit may update the weight through the machine learning using stochastic gradient descent (SGD).

In addition, a method of generating continuous spatial mobility information for an unmanned vehicle according to an embodiment of the present invention, the method comprising: generating navigation information by obtaining absolute coordinates of a route on which an unmanned vehicle travels; Generating environmental awareness information for generating environmental awareness information, and generating mobility data indicating whether the vehicle can travel on absolute coordinates corresponding to the route and the surrounding area based on the navigation information and the environmental awareness information. And performing machine learning on the basis of the mobility data to calculate a driving possibility probability for a continuous space between absolute coordinates in which the mobility data is generated, and based on the driving probability. Continuous spatial mobility information indicating whether or not driving is possible for each absolute coordinate corresponding to And a step of sex.

The generating of the navigation information may include obtaining absolute coordinates of the route based on the GPS information of the current position of the unmanned vehicle.

The generating of the environmental recognition information may include generating the environmental recognition information using an environmental recognition sensor mounted on the unmanned vehicle.

The method may further include storing the mobility data and the continuous space mobility information.

The calculating of the driving probability may include: driving the driving apparatus using a logistic regression classifier comprising a plurality of sparse kernels and weights assigned to the sparse kernels, for the region including the route and the peripheral region. Calculating a probability.

In addition, the weight is characterized by being updated through the machine learning using a stochastic gradient descent (SGD).

In addition, as an unmanned vehicle according to an embodiment of the present invention, the unmanned vehicle includes a device for generating continuous space mobility information, the device for generating the continuous space mobility information, for the path traveled by the unmanned vehicle A navigation information generation unit for acquiring absolute coordinates to generate navigation information, an environment recognition information generation unit for generating environment recognition information for a peripheral area of the path, and the path based on the navigation information and the environment recognition information And a mobility data generator for generating mobility data indicating whether the vehicle is capable of traveling in absolute coordinates corresponding to the peripheral area, and performing machine learning based on the mobility data to generate the mobility information. Compute the run possibility probability for the continuous space, and based on the run possibility probability Representing the driving availability of the corresponding absolute coordinates, each unit comprises a continuous space mobility information generator for generating a continuous space mobility information.

According to an embodiment of the present invention, the unmanned vehicle generates mobility data indicating whether the vehicle can be driven on absolute coordinates using navigation and environmental recognition information obtained after driving on an unpaved / paved road, and machine learning based on the mobility data. By generating continuous spatial mobility information indicating whether the vehicle can be driven in continuous space using the technique, the unmanned vehicle performing the wide-area planning can be more stably operated by autonomous driving so that the driving area is displayed with continuity. Make sure

1 is a detailed block diagram of an apparatus for generating continuous space mobility information according to an embodiment of the present invention.
2 is an exemplary view of driving data extracted through navigation information according to an embodiment of the present invention.
3 is an exemplary view of driving data extracted through environment recognition information according to an embodiment of the present invention.
4 is an exemplary diagram illustrating continuous space mobility information learned through navigation information-based travelable data according to an embodiment of the present invention.
5 is an exemplary diagram of continuous space mobility information learned through driving data based on environmental awareness information according to an embodiment of the present invention.
6 is an operation control flowchart of the apparatus for generating continuous spatial mobility information according to an embodiment of the present invention.

Hereinafter, with reference to the accompanying drawings will be described in detail the operating principle of the present invention. In the following description of the present invention, when it is determined that a detailed description of a known function or configuration may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. Terms to be described later are terms defined in consideration of functions in the present invention, and may be changed according to intentions or customs of users or operators. Therefore, the definition should be made based on the contents throughout the specification.

1 is a detailed block diagram of an apparatus for generating continuous space mobility information according to an embodiment of the present invention.

Hereinafter, an operation of each component of the apparatus 100 for generating continuous space mobility information according to an embodiment of the present invention will be described in detail with reference to FIG. 1.

First, when the unmanned vehicle autonomously travels according to the wide-area route plan, the navigation information generation unit 102 obtains absolute coordinates of the route traveled by the unmanned vehicle and generates the navigation information of the unmanned vehicle.

In this case, the navigation information may refer to global positioning system (GPS) based location information on a route on which the unmanned vehicle travels. That is, the navigation information generation unit 102 detects the GPS system to obtain GPS information about the current position of the unmanned vehicle and the route traveled by the unmanned vehicle, and based on the GPS information, absolute coordinates of the route on which the unmanned vehicle travels. Create

The environment recognition information generation unit 104 is connected to environmental sensors such as cameras, lidars, and radars mounted on an unmanned vehicle, and includes a surrounding of a preset range on a path on which the unmanned vehicle travels. Generates environmental awareness information about the area.

At this time, among the above environmental sensors, the rider scans the terrain of a predetermined area in front of the driverless vehicle to detect terrain characteristics and obstacles. The size of the region may be set to about 40m × 40m, but is not limited thereto.

Accordingly, the environment recognition information generation unit 104 generates environment awareness information about the path around the unmanned vehicle on the basis of the terrain characteristics and obstacles in front of the driverless vehicle provided from the lidar. Such environment recognition information may mean, for example, information about an area in which the vehicle can travel among the surrounding areas of the path on which the unmanned vehicle travels.

The mobility data generation unit 106 may be configured in a path and a peripheral area in which the unmanned vehicle travels based on the navigation information generated by the navigation information generation unit 102 and the environmental recognition information generated by the environment recognition information generation unit 104. Recognizes the corresponding absolute coordinates, and generates label information indicating whether the unmanned vehicle for each absolute coordinate can travel. In addition, the mobility data generation unit 106 stores label information for each absolute coordinate in the mobility DB 108.

)정보와 해당 위치에 대한 주행 가능 또는 불가능 여부를 표시한 레이블 정보(

)를 생성한다. 예컨대, 특정 위치가 주행 가능하다면 레이블 정보

이 할당될 수 있고, 특정 위치가 주행 불가능하다면 레이블 정보

이 할당될 수 있다.Hereinafter, the operation of the movable data generating unit 106 will be described in more detail. The movable data generating unit 106 is a two-dimensional space with respect to the route of the unmanned vehicle and the surrounding area of the route based on the environmental recognition information and navigation information. Position on

) And label information indicating whether or not you can travel for that location.

) For example, label information if a specific position is available for driving

Label information can be assigned, if a specific location is impossible to drive

May be assigned.

)정보와 해당 위치에 대한 주행 가능 또는 불가능 여부를 표시한 레이블 정보(

)를 매칭시킨 가동성 데이터(

)를 가동성 DB(100)에 저장한다. Subsequently, the movable data generation unit 106 is positioned in the two-dimensional space (

) And label information indicating whether or not you can travel for that location.

Mobility data matching)

) Is stored in the movable DB (100).

)가 정의될 수 있으며, 환경 인식 정보를 통해 무인 차량이 주행한 경로의 주변 영역에 대해 도 3에서와 같은 주행 가능 데이터(

)와 주행 불가능 데이터(

)가 정의될 수 있다. Accordingly, the driving possible data as shown in FIG.

) Can be defined, and the driving data (as shown in FIG.

) And non-driving data (

) Can be defined.

In addition, the continuous space mobility information generation unit 110 performs machine learning based on the label information for each absolute coordinate stored in the mobility DB 108 to define the probability of running in the continuous space between the absolute coordinates. do.

In addition, the continuous space mobility information generation unit 110 generates label information indicating whether the vehicle can travel by absolute coordinates corresponding to the continuous space based on the probability of driving defined above, thereby generating mobility information of the continuous space. Mobility information for the generated continuous space is also stored in the mobility DB 108.

)를 기초로 아래의 [수학식 1]과 같이 스파스(Sparse) 커널(Kernel)

과 학습이 필요한 가중치(

)로 구성된 로지스틱 회귀 분류기(Logistic regression classifier)를 통해 기계 학습을 수행하여 무인 차량이 주행한 경로와 경로 주변의 주행 가능한 영역에 대응되는 절대 좌표들 사이의 연속 공간에 대한 주행 가능 확률을 정의한다. 이때, 연속 공간 가동성 정보 생성부(110)는 연속 공간상에 가동성 정보를 정의하기 위해 Hilbert maps 기법을 활용할 수 있으나, 이에 한정되는 것은 아니다.Hereinafter, the operation of the continuous space mobility information generating unit 110 will be described in more detail. The continuous space mobility information generating unit 110 is capable of driving / impossible data on two-dimensional space obtained from the movable DB 108.

) Sparse Kernel as shown in Equation 1 below.

And weights that require learning (

Machine learning is performed by using a logistic regression classifier consisting of) to define the probable probability of continuous space between absolute coordinates corresponding to the route driven by the unmanned vehicle and the range around the route. In this case, the continuous space mobility information generator 110 may use the Hilbert maps technique to define the mobility information on the continuous space, but is not limited thereto.

)에 대해서도 정의할 수 있다. In addition, the continuous space mobility information generating unit 110 is impossible to travel (

) Can also be defined.

는 아래의 [수학식 2]와 같이 표현된다.Sparse kernel

Is expressed by Equation 2 below.

)를 학습하기 위한 목적함수

를 아래의 [수학식 3]에서와 같이 Negative log-likelihood로 정의할 수 있다.In addition, the continuous space mobility information generating unit 110 is a weight (

Objective function for learning

Can be defined as a negative log-likelihood as shown in [Equation 3] below.

)를 점진적으로 갱신시킬 수 있다.Here, the continuous space mobility information generating unit 110 uses Stochastic Gradient Descent (SGD) as shown in Equation 4 below to obtain a weight (

) Can be updated gradually.

)는 가동성 DB(108)로 저장되어 동일 또는 신규 지역 데이터의 사전 정보로 활용될 수 있다. The weights learned as above (

) May be stored in the mobility DB 108 and used as dictionary information of the same or new area data.

FIG. 4 illustrates continuous spatial mobility information machine-learned through the runnable data of the navigation information, and FIG. 5 illustrates continuous spatial mobility information machine-learned through the runable / disabled data of the environmental recognition information.

As shown in FIG. 4 and FIG. 5, even if the resolution of the map of the travelable route is increased, the area in which the driverless vehicle can run can be displayed with continuity even when the continuous space mobility information is used. There is no inconsistency between navigation information.

6 illustrates an operation control flow of the apparatus for generating continuous spatial mobility information according to an embodiment of the present invention. Hereinafter, an embodiment of the present invention will be described in detail with reference to FIGS. 1 to 6.

First, when the unmanned vehicle autonomously travels according to the wide-area path plan, the continuous space mobility information generating apparatus 100 obtains absolute coordinates of a path on which the unmanned vehicle travels and generates the navigation information of the unmanned vehicle (S600). In this case, such navigation information may refer to GPS-based location information of a route on which the unmanned vehicle travels.

Subsequently, the continuous space mobility information generating apparatus 100 is connected to environmental sensors such as a camera, a lidar, and a radar mounted on the unmanned vehicle, and is configured on a path in which the unmanned vehicle travels. Environment-aware information about the surrounding area of the range is generated (S602).

At this time, the continuous space mobility information generating device 100 generates environmental awareness information about the path around the unmanned vehicle on the basis of information on the terrain and obstacles in front of the unmanned vehicle provided from the lidar, etc. . Such environment recognition information may mean, for example, information about an area in which the vehicle can travel among the surrounding areas of the path on which the unmanned vehicle travels.

Subsequently, the apparatus 100 for generating continuous space mobility information recognizes absolute coordinates corresponding to a path and a surrounding area in which the unmanned vehicle travels based on the navigation information and environment recognition information, and indicates whether the unmanned vehicle can travel by absolute coordinates. Create label information. In addition, the mobility data generated by matching the label information for each absolute coordinate is generated and stored in the mobility DB (S604).

Thereafter, the apparatus 100 for generating continuous space mobility information performs machine learning on the basis of the label information for each absolute coordinate stored in the mobility DB to calculate a probability of traveling in the continuous space between the absolute coordinates (S606).

Subsequently, the apparatus 100 for generating continuous space mobility information generates mobility information on the continuous space by generating label information indicating whether the vehicle is capable of traveling by absolute coordinates corresponding to the continuous space based on the calculated probability of driving, as described above. It is stored in the mobility DB (S608).

At this time, when the mobility information of the continuous space is used, even if the resolution of the map of the travelable route is increased, as shown in FIG. There is no inconsistency between the loaded navigation information.

As described above, according to one embodiment of the present invention, the unmanned vehicle generates mobility data indicating whether the vehicle can be driven on absolute coordinates by using navigation and environmental recognition information obtained after driving on an unpaved / paved road, and the mobility data. Based on the machine learning technique, the continuous space mobility information indicating whether the vehicle can be driven in the continuous space is generated. Allow for autonomous driving.

Combinations of the steps of each flowchart attached to the present invention may be performed by computer program instructions. These computer program instructions may be mounted on a processor of a general purpose computer, special purpose computer, or other programmable data processing equipment such that the instructions performed through the processor of the computer or other programmable data processing equipment are described in each step of the flowchart. It will create a means to perform them. These computer program instructions may be stored in a computer usable or computer readable memory that can be directed to a computer or other programmable data processing equipment to implement functionality in a particular manner, and thus the computer usable or computer readable memory. It is also possible for the instructions stored in the to produce an article of manufacture containing instruction means for performing the functions described in each step of the flowchart. Computer program instructions may also be mounted on a computer or other programmable data processing equipment, such that a series of operating steps may be performed on the computer or other programmable data processing equipment to create a computer-implemented process to create a computer or other programmable data. Instructions for performing the processing equipment may also provide steps for executing the functions described in each step of the flowchart.

In addition, each step may represent a module, segment or portion of code that includes one or more executable instructions for executing a specified logical function (s). It should also be noted that in some alternative embodiments, the functions noted in the steps may occur out of order. For example, the two steps shown in succession may in fact be performed substantially simultaneously or the steps may sometimes be performed in the reverse order, depending on the function in question.

Meanwhile, in the above description of the present invention, specific embodiments have been described, but various modifications may be made without departing from the scope of the present invention. Therefore, the scope of the invention should be determined by the claims rather than by the described embodiments.

102: navigation information generation unit 104: environment recognition information generation unit
106: mobility data generation unit 108: mobility DB
110: continuous space mobility information generation unit

Claims (14)

An apparatus for generating continuous space mobility information for an unmanned vehicle,
A navigation information generation unit for generating navigation information by acquiring absolute coordinates of a route driven by an unmanned vehicle;
An environmental awareness information generation unit for generating environmental awareness information on a peripheral area of the path;
A mobility data generator for generating mobility data indicating whether the vehicle is capable of traveling with respect to absolute coordinates corresponding to the route and the surrounding area, based on the navigation information and the environment recognition information;
Machine learning is performed based on the mobility data to calculate the probability of the running of the continuous space between the absolute coordinates in which the mobility data is generated, and correspond to the continuous space based on the driving probability. A continuous space mobility information generating unit for generating continuous space mobility information indicating whether the driving for each of the absolute coordinates
Continuous space mobility information generating device. The method of claim 1,
The navigation information generation unit,
Obtaining absolute coordinates for the path based on the GPS information about the current position of the unmanned vehicle
Continuous space mobility information generating device. The method of claim 1,
The environment recognition information generation unit,
Generating the environment recognition information using an environment recognition sensor mounted in the unmanned vehicle;
Continuous space mobility information generating device. The method of claim 3, wherein
The environmental sensor is at least one of a LIDAR, a camera or a RADAR.
Continuous space mobility information generating device. The method of claim 1,
A mobility database (DB) for storing the mobility data and the continuous space mobility information;
Continuous space mobility information generating device. The method of claim 1
The continuous space mobility information generation unit,
For a region including the path and the peripheral region, a plurality of sparse kernels and weights assigned to each sparse kernel (

Calculating the probability of driving using a logistic regression classifier
Continuous space mobility information generating device. The method of claim 6
The continuous space mobility information generation unit,
Updating the weight through the machine learning using stochastic gradient descent (SGD)
Continuous space mobility information generating device. A method of generating continuous spatial mobility information for an unmanned vehicle,
Generating navigation information by obtaining absolute coordinates of a route driven by an unmanned vehicle;
Generating environmental awareness information for generating environmental awareness information on a peripheral area of the path;
Generating mobility data indicating whether the vehicle is capable of traveling with respect to absolute coordinates corresponding to the route and the surrounding area, based on the navigation information and the environment recognition information;
Performing machine learning based on the mobility data to calculate a driving probability for continuous space between absolute coordinates in which the mobility data is generated;
Generating continuous space mobility information indicating whether driving is possible for each of the absolute coordinates corresponding to the continuous space based on the driving possibility probability;
How to generate continuous space mobility information. The method of claim 8,
Generating the navigation information,
Obtaining absolute coordinates for the route based on GPS information on a current position of the unmanned vehicle;
How to generate continuous space mobility information. The method of claim 8,
Generating the environmental awareness information,
Generating the environment recognition information by using an environment recognition sensor mounted on the unmanned vehicle.
How to generate continuous space mobility information. The method of claim 8,
Storing the mobility data and the continuous space mobility information;
How to generate continuous space mobility information. The method of claim 8,
Computing the driving probability is,
Calculating a driving probability using a logistic regression classifier composed of a plurality of sparse kernels and weights assigned to each sparse kernel, for an area including the path and the surrounding area;
A method for generating learning based continuous spatial mobility information. The method of claim 12,
The weight is updated through the machine learning using stochastic gradient descent (SGD).
A method for generating learning based continuous spatial mobility information. As an unmanned vehicle,
The unmanned vehicle includes a device for generating continuous space mobility information,
The apparatus for generating the continuous space mobility information,
A navigation information generation unit for generating navigation information by obtaining absolute coordinates of a route on which the unmanned vehicle travels;
An environmental awareness information generation unit for generating environmental awareness information on a peripheral area of the path;
A mobility data generation unit for generating mobility data indicating whether the vehicle can travel with respect to the absolute coordinates corresponding to the route and the surrounding area based on the navigation information and the environment recognition information;
The machine learning is performed based on the mobility data to calculate the probability of the running of the continuous space between the absolute coordinates from which the mobility information is generated, and the absolute coordinates corresponding to the continuous space of each of the absolute spaces are based on the driving probability. And a continuous space mobility information generator configured to generate continuous space mobility information indicating whether the vehicle can be driven.
Driverless vehicles.

Images