WO2021049231A1 - Occupied grid map management device - Google Patents

Abstract

The present invention addresses the problem of managing an occupied grid map with which efficient operation is possible while suppressing processing load and memory consumption. Provided is an occupied grid map management device that is mounted as an electronic control device 110 and manages the environment around a vehicle on a cell basis, the device comprising: a map generation unit 114 which generates an occupied grid map 201 on the basis of the outside world recognition sensor data sensed according to the state of the outside world and that of the own vehicle; a map management unit 113 which manages each piece of cell information 301 of the generated occupied grid map 201 as one record of a database on a cell-by-cell basis; an application execution unit 111 which executes application software for automatic driving or driving assistance; and a cell selection unit 112 which selects and duplicates a corresponding cell from the cells managed by the map management unit, in response to a cell acquisition request from the application execution unit.

Description

Occupied grid map management device

The present invention relates to a control technique for a vehicle, particularly an electronic control unit (ECU: Electronic Control Unit) for an automobile. Among them, in particular, it relates to a technology for managing an occupied grid map used for automatic driving / driving support (hereinafter, also simply referred to as automatic driving) technology of a vehicle.

When realizing automatic driving of automobiles in urban areas, automobiles often sense objects around their current position, and based on the sensed information, generate routes that do not collide with them and drive. On the other hand, since there are more objects in urban areas than in expressways and suburbs, it is necessary to generate a travel route in consideration of the objects existing on such roads. In order to generate a traveling route in an environment where multiple objects exist, an occupied grid map (grid map) is used, which divides the space into cells and expresses with probability whether surrounding objects exist in the cells. It is often said that.

On the other hand, when managing the ever-changing environment around the vehicle with cells, it is necessary to handle a large number of cells in the electronic control device, and as the number of cells increases, the processing load increases and access delay to each cell increases. appear. For example, when the area of 50 meters in width × 15 meters in height in front of the vehicle is represented by cells of 0.5 m in length × 0.1 m in width, 15,000 cells are handled in the electronic control device.

Therefore, in the field concerned, in a situation where a large number of cells are handled on an occupied grid map, it is an issue to reduce the processing load of the electronic control device in an environment where the processing in the electronic control device has a high load.

In Patent Document 1, when handling an occupied grid map in an electronic control device, only cells existing in an area sandwiched between both ends of a road are targeted for calculation, and only these limited cells are updated to perform a processing load. And propose a method to reduce memory consumption.

Japanese Unexamined Patent Publication No. 2017-157087

When expanding the area represented by the occupied grid map, the occupied grid map will be represented in many cells. Further, in order to improve the accuracy of the existence probability of the object at each point, it is necessary to increase the resolution of the cells, and the number of cells to be handled increases. In such a case, there is a problem that the number of cells handled in the electronic control device becomes enormous, the processing load increases, and the memory consumption increases dramatically.

Even when the cells to be calculated are limited to the portions sandwiched between both ends of the road as in Patent Document 1, it is difficult to reduce the number of cells to be calculated when the distance between both ends of the road is large.
Further, when it is desired to handle an arbitrary area, since the minimum area is limited to the part sandwiched between both ends of the road, it is difficult to reduce the processing load and memory consumption.

Further, when searching for a cell in which an object exists from among a plurality of cells, for example, in Patent Document 1, it is necessary to confirm and determine all cell information in the area sandwiched by the road edge. Therefore, there is little reduction in the increase in processing load and memory consumption.

In order to solve the above problems, in the present invention, the cell information of the occupied grid map is managed as a database in which the cells constituting the occupied grid map are used as the management unit. For example, the cell information of each cell is managed as one record. As a result, the cell information can be managed in a manageable format for each item constituting the cell information. At this time, it is preferable to provide the occupancy probability, the occupancy state, the relative coordinate information, and the absolute coordinate information in each cell as the items of the cell information.

A more specific aspect of the present invention is an occupied grid map management device that manages an occupied grid map around the vehicle, and uses the data around the vehicle detected by the sensor to obtain cell information of the occupied grid map. A map generation unit for generating, a map management unit that manages the generated cell information with cell information of cells constituting the occupied grid map as a management unit, and the map management in response to a cell information acquisition request. An occupied grid map management device including a cell selection unit that selects cell information corresponding to the cell acquisition request from the cell information managed by the unit.

The present invention also includes a method using the occupied grid map management device and a program for executing the method on the occupied grid map management device.

According to the present invention, by managing the cell information of the occupied grid map, it is possible to improve the access speed to the cell information and to flexibly acquire the cell information according to the operation control.

It is a system block diagram (functional block diagram) of the electronic control device 110 in one Example of this invention. It is a figure which shows the occupied grid map 201 used in one Example of this invention. It is a figure which shows the cell information 301 used in one Example of this invention. It is a figure for demonstrating how to hold the cell information 301 in the time axis direction in one Example of this invention. It is a figure which shows the data registration flow in one Example of this invention. It is a figure which shows the flow which acquires the cell information 301 in one Example of this invention. It is a figure which shows the data acquisition sequence at the time of emergency in one Example of this invention. It is a system block diagram (hardware block diagram) of the electronic control device 110 in one Example of this invention.

An embodiment of the present invention will be described below with reference to the drawings.

The electronic control device 110 shown in FIG. 1 is a diagram showing an example of a system configuration diagram (functional block) according to this embodiment.

The electronic control device 110 is an ECU (Electronic Control Unit) mounted on the vehicle. The electronic control device is configured to include a storage area such as a CPU (Central Processing Unit) and RAM (Rando Access Memory). It also includes an electronic control device 110, an external world recognition sensor group 116, and an in-vehicle electronic control device. The electronic control device 110 includes an application execution unit 111, a cell selection unit 112, a map management unit 113, a map generation unit 114, a cell update unit 115, and an outside world recognition sensor group 116.

The application execution unit 111 has, for example, a function of recognizing the surrounding environment of the vehicle, a function of planning the traveling route of the own vehicle based on the recognition result of the vehicle, a function of calculating the control amount of the vehicle, and the like, and the map management unit 113 has a function of recognizing the surrounding environment of the vehicle. Execute various functions based on the stored information. Further, based on the result of integrating the information of the external world recognition sensor group 116 by the application execution unit 111, the cell information 301 of the occupied grid map 201 stored in the map management unit 113 is updated via the cell update unit 115. The cell information 301 of the occupied grid map 201 refers to the cell information 301 corresponding to the occupied grid map 201, and the contents thereof will be described later.

The cell selection unit 112 holds the address information on the memory in which the occupied grid map 201 is stored and the ID information given to each cell in association with each other. When there is a cell acquisition request from the application execution unit 111, the search condition is analyzed, and the corresponding cell information 301 stored in the map management unit 113 is acquired based on the analysis result.

The map management unit 113 includes, for example, a storage device such as an HDD (Hard Disk Drive), a flash memory, and a ROM (Read Only Memory), and a memory such as a RAM (Random Access Memory). The map management unit 113 holds an occupied grid map 201 in which the traveling environment is divided in a grid pattern. Each grid managed by the occupied grid map 201 is also called a cell, and the size (resolution) of each cell can be arbitrarily defined. It goes without saying that a plurality of occupied grid maps 201 can be provided depending on how the application execution unit 111 is used, and the resolution of each occupied grid map 201 can be freely set. The map management unit 113 manages the occupied grid map 201 as a database. Each cell is managed as a record in the database, and one cell corresponds to one record.

In addition, a unique ID is assigned to each cell, and the ID and the address information in the memory are managed by the cell selection unit 112. Each cell of the occupied grid map 201 holds, for example, the existence probability of an object based on the outside world recognition result of the traveling environment recognized by the outside world recognition sensor group 116. The information held by each cell can be freely set by the developer defining it in advance. Further, the map management unit 113 has a mechanism for periodically deleting the cell information 301 that has not been updated for a certain period of time so that the information after a certain period of time is not accumulated.

The map generation unit 114 associates the outside world recognition information sensed by the outside world recognition sensor group 116 connected to the electronic control device 110 by, for example, CAN (Control Area Network) or Ethernet, with the occupied grid map 201 of the map management unit 113. At this time, the cell that stores the outside world recognition information is based on the position information sensed by the outside world recognition information. Further, for example, the deviation of the position information from the sensing time caused by the transmission delay from the outside world recognition sensor group or the processing delay in the electronic control device 110 is corrected by the map generation unit 114 and then transmitted to the map management unit 113.

The cell update unit 115 is used when the application execution unit 111 assigns object information integrated based on a plurality of different sensor information of the external world recognition sensor group 116 to the occupied grid map 201 of the map management unit 113.

The external world recognition sensor group 116 includes, for example, a camera mounted on a vehicle, a laser radar, a sonar, a lidar (Light Detection and Ringing, Laser Imaging Detection and Ringing), and the like. The object recognized by the outside world recognition sensor may be not only a dynamic object such as another vehicle, a pedestrian, or a two-wheeled vehicle, but also a static object such as a utility pole, a sign, a roadside tree, a hole, or a roadside. Further, information in which the type or shape of the object is not specified, such as point cloud information sensed by Lidar, may be used. In this embodiment, the external recognition sensor group 116 is expressed, but one sensor may be used.

The electronic control device 110 shown in FIG. 1 is a so-called one-chip processor, and is realized by the hardware configuration shown in FIG. Hereinafter, FIG. 8 will be described. Each function of the application execution unit 111, the cell selection unit 112, the map generation unit 114, and the cell update unit 115 is realized by a program. That is, it is realized by the application program 111A, the cell selection program 112A, the map generation program 114A, and the cell update program 115A stored in the main storage device 120 shown in FIG. Each of these programs is executed by the control device (CPU core) 121 to realize each function. Further, the map management unit 113 is realized by the memory 113A and stores the cell information 301 and the occupied grid map 201. Further, the electronic control device 110 is connected to the outside world recognition sensor group 116 via the connection terminal 123. Further, the constituent requirements of the electronic control device 110 are connected to each other via the bus 122.

FIG. 2 shows an example of the occupied grid map 201 held by the map management unit 113. The number of cells in the occupied grid map 201 can be freely changed. Further, the resolution of each cell constituting the occupied grid map 201 can be freely changed, and different resolutions can be set even within the same occupied grid map 201. In addition, the ID information given to the cell can be arbitrarily set, and any information other than numbers can be given as long as the information can uniquely identify the cell.

FIG. 3 is an example of cell information 301 of the occupied grid map 201 shown in FIG. 2 managed by the map management unit 113. Each cell is independently managed as a database record by the map management unit 113. This is created based on the external recognition sensor data.

Each cell contains at least information about the positional relationship with the vehicle. When distance information is used as the positional relationship information, the distance information may be, for example, relative distance information from the own vehicle, absolute position information of GPS (Global Position System), or the like. In addition, there are no restrictions on the method of expressing the distance, such as the Euclidean distance or the polar coordinate system.

In the example shown in FIG. 3, the cell information 301 of the occupancy grid map 201 includes coordinate information (X, Y), update time information, occupancy state, occupancy probability, area information, Ray information, and distance from the sensor of each cell. It holds polar coordinates, the rate of change from the occupancy probability of the previous time to the occupancy probability of the current time, and so on. Of these pieces of information, the "occupancy state", "occupancy probability", and "coordinate information" (absolute coordinates and relative coordinates described later) may be used.

The coordinate information (X, Y) indicates the position in the real world in the cell. The coordinate information includes "absolute coordinates" (for example, latitude / longitude) and "relative coordinates" indicating the relationship with the automobile.

The update time information represents the time information when the state of the cell was updated.

The occupied state indicates whether or not the object recognized by the external world recognition sensor group 116 exists in the corresponding cell. Hereinafter, the content of the occupied state is expressed as occupied or unoccupied.

The occupancy probability represents the probability that the object recognized by the external recognition sensor group 116 exists on the cell. The existence probability may be an existence probability indicating the existence of an object at the current time, or a posterior probability based on the occupancy probability of the current cell from the occupancy probability of the corresponding cell before the current time.

Area information is a unique ID given to each group when multiple cells are grouped and treated as one area. Further, it goes without saying that one cell belongs to a plurality of different regions and can hold information on each region.

Ray information is information on cells through which a certain light beam passes. A plurality of cells may be assigned to one Ray, or a plurality of Rays may be assigned to one cell. By using a sensor such as Lidar, it is possible to estimate the distance from the spatial density of each cell to the foremost object based on multiple rays (Rays) emitted from the sensor. Is.

The distance from the sensor indicates the position of the cell in the real world and the distance from the outside world recognition sensor. For this information, the distance between the recognized object and the own vehicle included in the object recognition information may be used.

Polar coordinates indicate the position in the real world in the cell in a polar coordinate system.

The rate of change represents the degree of protrusion of the object using the rate of change of the occupancy probability of the current time from the occupancy probability of the previous time in the cell.

FIG. 4 describes holding the cell information 301 in the time axis direction. The occupied grid map 201 holds not only the information of the current time but also the past information of the preset time steps. Therefore, it is possible to confirm the state and occupancy probability of the cell going back in the time axis direction with respect to the cell.

FIG. 3 illustrates an example of the process of storing the data recognized by the external world recognition sensor group 116 in the map management unit 113. The object information recognized by the external world recognition sensor group 116 is assigned to the corresponding cell of the occupied grid map 201 of FIG. 2 based on the sensed position information, and a unique ID is assigned.

Using FIG. 5, a registration flow showing a series of processes from the acquisition of recognition information from the external world recognition sensor group 116 to the input to the electronic control device 110 will be described.

First, in step 501, the electronic control device 110 receives the outside world recognition data detected by the outside world recognition sensor group 116. The electronic control device 110 may directly receive the outside world recognition data or may receive it via an application program. When passing through an application program, the application program performs conversion for performing the processing of step 502 and subsequent steps described later on the external recognition data. For example, one of the external world recognition sensor group 116 is a camera, and when handling an image, the RAW data received by the application program is developed to generate the image data.
Object recognition information is an example of external recognition data. The object recognition information includes, for example, the type of the sensor used for recognition, the type of the recognized object, the distance between the recognized object and the own vehicle (distance from the sensor), and the relative speed between the recognized object and the own vehicle. Further, the occupied grid map 201 data (see FIG. 3) such as the update time in each cell constituting the occupied grid map 201 is included.

Next, in step 502, the map generation unit 114 of the electronic control device 110 calculates the position on the occupied grid map 201 in which the object included in the object recognition information exists, based on the object recognition information. This is realized by calculating the storage position of the map management unit 113, which is a database. Then, the map generation unit 114 of the electronic control device 110 stores the cell information 301 (FIG. 3) created based on the external world recognition sensor data (object recognition information) in the storage position of the map management unit 113.

Next, in step 503, the electronic control device 110 determines whether to update the data of the cell information 301. The premise of this process will be described first. The electronic control device 110 needs to update the data of the map management unit 113 with the passage of time. This is because the object recognition information corresponding to each cell changes (for example, movement) as the automobile moves or the object moves over time. Therefore, the data is updated according to this change. However, there are cases where it is not necessary to update at least some cells even after a lapse of time because the vehicle is stopped or the object does not move. In this step, it is determined whether or not each cell or the entire occupied grid map 201 needs to be updated. The contents are as follows.

First, the electronic control device 110 specifies the occupied state of each cell. Then, the electronic control device 110 determines whether the occupied state has changed from occupied to unoccupied to occupied within a predetermined period. When it is determined that the change is made in this way, it is determined that the non-occupied period is equal to or greater than the threshold value t. That is, the difference between the update times, which is the time when the cell occupancy state changes, is compared with the threshold value t to make a determination. The difference in update time indicates the difference between the previous update time and the latest update time.

As a result of the determination, if the threshold value is t or more, it is highly likely that the object detected in the cell has changed from the original object to another object, so it is determined that the objects are not the same. On the contrary, if it is less than the threshold value t, it is judged that the objects are the same because there is a high possibility of a factor such as a data error.

Further, the identity of this object may be judged based on the changing state of neighboring cells. That is, if the neighboring cells also change from occupied to unoccupied to occupied, it is determined that the objects are the same. On the contrary, when the change of the occupied state is different, it is judged that the objects are not the same. The adjacent cell may be a cell whose initial occupied state is occupied as in the cell, or a cell located within a predetermined range with the cell.

If it is determined that they are the same (YES) by the above processing, the process proceeds to step 505, and if it is determined that they are not the same (NO), the process proceeds to step 504.

In step 504, the electronic control device 110 deletes the cell information of the past cell stored in the cell, and the cell information of the current time is newly registered as new cell information.

In step 505, the electronic control device 110 registers the cell information of the past cell stored in the cell as it is.

This completes the registration process for cell information 301. Hereinafter, data acquisition for using the registered cell information 301 will be described with reference to FIG. FIG. 6 describes an example of a process of acquiring cell information 301 managed by the map management unit 113 from the application execution unit 111.

For example, a case where the application execution unit 111 requests the cell selection unit 112 to acquire the cell information of the cell in which the object exists will be described.

First, in step 601 the cell selection unit 112 analyzes the request. That is, the content of the acquisition request is specified.

Next, in step 602, the cell selection unit 112 specifies a storage position in which the cell information 301 corresponding to the specified acquisition request is stored. For example, specify the address information on RAM (Random Access Memory).

Next, in step 603, the cell selection unit 112 refers to the specified address information and copies the cell information stored at the address to the RAM area prepared by the application execution unit 111. The application execution unit 111 that has acquired the cell information generates, for example, a route on which the own vehicle travels based on the acquired information. Then, the vehicle control amount information is calculated based on the generated route, and the vehicle control amount information is transmitted to each in-vehicle electronic control device. Here, regarding the acquisition of the cell information 301 in FIG. 6, a data acquisition sequence will be described when urgent processing is required, such as when an accident is highly inevitable.

First, at the time of initialization 701 shown in FIG. 7, whether the application execution unit 111 has a high degree of urgency with respect to the cell information 301 of the map management unit 113 from the cell update unit 115 via the cell selection unit 112. Register a callback to determine. The contents to be registered include that the velocity of the object (the fluctuation of the distance from the sensor) is larger than a predetermined value.

Then, when the electronic control device 110 receives the object information which is the external world recognition sensor information (702), the following processing is executed. The cell update unit 115 determines whether the received object information has a high degree of urgency, that is, whether the condition registered in 701 is satisfied. As a result, if it is determined that the degree of urgency is high, callback processing is performed. That is, the application execution unit 111 is notified via the cell selection unit 112 that the degree of urgency is high. Then, the application execution unit 111 executes an emergency process. The emergency processing includes the output of a deceleration instruction to the own vehicle and the output of an alert output instruction to the driver. This is the end of the description of the data acquisition shown in FIGS. 6 and 7.

Next, the adjustment of the cell information 301 will be described. Since the cell information 301 may include an error, adjustment and update processing for this purpose will be described. Here, an example of adjusting and updating the cell information 301 of the map management unit 113 by the application execution unit 111 is shown. When the application execution unit 111 acquires the cell information 301 of the map management unit 113 via the cell selection unit 112, for example, corrects the error of the acquisition position and the time information, and updates the data of the map management unit 113: Execute as follows. The application execution unit 111 updates the data of the map management unit 113 as the latest data via the cell update unit 115.

Similar to the cell selection unit 112, the cell update unit 115 holds the ID information given to each cell and the address information on the map management unit 113 in which the corresponding cell is stored in association with each other. Therefore, when a cell update request is received from the application execution unit 111, the data of the map management unit 113 is updated based on the cell ID of the cell. At this time, when the data of the map management unit 113 is updated by the map generation unit 114 when the data is corrected by the application execution unit 111, the data is timed based on the time information of the map management unit. Store in chronological order.

Next, in this embodiment, it will be described that a part of the cell information 301 of the occupied grid map 201 is used.

When some cells of the occupied grid map 201 of FIG. 2 are grouped and treated as an area, the application execution unit 111 specifies in advance the cells included in the area. For example, when it is desired to group radial cells from the own vehicle as in the ray casting model, the application execution unit 111 identifies in advance the cells existing on the beam extending radially from the own vehicle and sets them as a group. The cell update unit 115 receives the contents set by the application execution unit 111, and assigns a common area ID to each cell information 301 stored in the map management unit 113. At this time, the cell area designated by the application execution unit 111 can be regarded as an area regardless of whether it is continuous or discontinuous with the ID assigned to the cell information 301, and its shape is also, for example, rectangular, radial, or radial. It can be specified in any shape regardless of triangle or circle.

When the corresponding area ID is specified from the application execution unit 111 and an area information acquisition request is issued to the cell selection unit 112, the cell selection unit 112 performs the following processing. The cell selection unit 112 calculates the address information on the RAM of the cell in which the corresponding cell is stored from the designated area ID, and searches the data. The corresponding cell information 301 stored in the map management unit 113 is duplicated in the RAM prepared by the application execution unit 111.

In this example, the radial region in the traveling direction of the own vehicle was extracted, but it can also be applied to the following examples. For example, when the own vehicle moves at a relatively low speed, it is assumed that the cell information 301 of the cell relatively close to the own vehicle is specified and used. In this case, it is preferable that the lower the speed (slower) of the own vehicle, the more the cell information 301 in the narrow area in front is extracted. In addition, it is also possible to extract cell information 301 of the traveling lane of the own vehicle.

According to this embodiment described above, the necessary cell information 301 is extracted and managed as a database. Therefore, in this embodiment, it is possible to reduce the data processing load and the memory consumption of the application.

110 ... Electronic control device 110, 111 ... Application execution unit, 112 ... Cell selection unit, 113 ... Map management unit, 114 ... Map generation unit, 115 ... Cell update unit, 116 ... External world recognition sensor group

Claims (8)

It is an occupied grid map management device that manages the occupied grid map around the vehicle.
A map generation unit for generating cell information of the occupied grid map using the data around the vehicle detected by the sensor, and
A map management unit that manages the generated cell information with the cell information of each cell constituting the occupied grid map as a management unit.
An occupied grid map management device including a cell selection unit that selects cell information corresponding to the cell information acquisition request from the cell information managed by the map management unit in response to the cell information acquisition request. In the occupied grid map management device according to claim 1,
The cell information is periodically generated, stored in a storage device, and stored.
The cell information includes the occupancy state of the object in the corresponding cell.
The map management unit determines whether the objects that are the cause of the occupancy are the same when the time-series change of the occupancy state of the cell information fluctuates between occupancy, non-occupancy, and occupancy. An occupied grid map management device characterized in that at least a part of cell information stored periodically is deleted. In the occupied grid map management device according to claim 2.
The map management unit is an occupied grid map management device that determines whether or not the objects are the same when the non-occupied period is larger than a predetermined threshold value and determines that the objects are not the same. In the occupied grid map management device according to claim 1,
Furthermore, it is equipped with an application execution unit that executes application software that performs automatic driving control or driving support control for the vehicle.
The application execution unit notifies the cell selection unit of the cell information acquisition request, and then notifies the cell selection unit of the cell information acquisition request.
The cell selection unit duplicates the cell information selected in response to the notified cell information acquisition request.
The application execution unit is an occupied grid map management device, characterized in that the application software is executed by using the duplicated cell information. In the occupied grid map management device according to claim 4.
Generation of the cell information,
(1) Whether the map generation unit generates the cell information using the data around the vehicle received from the sensor.
(2) Either the application execution unit generates the cell information using the data around the vehicle received from the sensor from the sensor, and notifies the map generation unit of the generated cell information. An occupied grid map management device characterized by performing. In the occupied grid map management device according to claim 1,
The cell information is an occupancy grid map management device including occupancy probability, occupancy state, relative coordinate information, and absolute coordinate information in each cell. In the occupied grid map management device according to claim 1,
The cell selection unit is an occupied grid map management device characterized in that cell information is selected according to the operating status of a vehicle. In the occupied grid map management device according to claim 7.
The cell selection unit is an occupied grid map management device, characterized in that the slower the speed, which is the operation status, the smaller the cell information is selected.

Images