Method for representing a vehicle's environment in an occupancy grid

Abstract

Method for representing a vehicle's environment in an occupancy grid for a driver assistance function, wherein the data are acquired using at least one environment sensing system, where a likely course traveled by the vehicle is determined using vehicle sensors, in particular for detecting steering angle, direction of travel of the vehicle and / or indicator activity. characterized by the fact that A search is performed in the occupancy grid for occupied grid cells that define a free space, within a search window along the likely route traveled.

Description

[0001] The invention relates to the technical field of representing a vehicle environment based on sensor systems for environmental detection for a driver assistance system.

[0002] State-of-the-art driver assistance functions for longitudinal control include adaptive cruise control (ACC) and emergency braking systems, as well as collision warning, collision mitigation, and collision prevention systems. These systems are based on capturing the environment in front of the vehicle for situation analysis, generating a list of objects and tracking them over time to identify relevant objects and, if applicable, their properties.

[0003] Furthermore, assistance functions for monitoring blind spots or assisting with lane changes are known, which rely on sensors monitoring the lateral and rear areas of the vehicle. Here, too, an object list is generated, and the objects are tracked over a period of time and evaluated to determine relevant objects. Grid-based environment models, so-called occupancy grids, are also known. In an occupancy grid, the vehicle's surroundings are divided into cells, and a classification of "accessible" or "occupied" is stored for each cell. The advantage of this representation is that it provides a driver assistance system with clearance information. Clearance means that a vehicle can move within this area without danger, i.e., without collision or leaving the roadway.

[0004] Extended occupancy grids can provide a description of moving objects, for example, by assigning a velocity to the cells. Occupancy grids are primarily used for functions that react to the static environment, such as lane edge estimation. For functions that react to dynamic objects, objects are extracted into an object list, which is then used for object-based situation analysis and control.

[0005] From DE 10 2006 056 835 A1 a method for grid-based processing of sensor signals is known.

[0006] From DE 199 35 123 A1 a vehicle DBF radar device is known which is attached to a vehicle and is designed to detect an object in the surroundings.

[0007] From DE 103 41 128 A1 a device for detecting the instantaneous distance of a motor vehicle from an obstacle is known.

[0008] From DE 10 2009 007 885 A1 a method for detecting vehicles merging into or out of the same lane is known.

[0009] The object of the present invention is to provide an improved environment model for driver assistance functions.

[0010] The problem is solved by the characteristics of independent claims.

[0011] According to the invention, a method for representing a vehicle's environment in an occupancy grid for a driver assistance function is provided. Environmental data is acquired using at least one environmental sensing system, and a likely course traveled by the vehicle is determined based on vehicle sensors, in particular for detecting steering angle, vehicle direction of travel, and / or turn signal activity. A search is performed in the occupancy grid for occupied grid cells that define a free space within a search window along the likely course traveled. In a preferred embodiment of the invention, in addition to the information "occupied" or "accessible," at least one further piece of information is stored indicating whether the cell is "moving" or "stationary." Such information includes, for example, distance, speed, or acceleration.

[0012] This method allows assistance functions to be implemented directly on an occupancy grid, i.e., without segmenting and extracting objects, managing a list of objects, or tracking objects over time. This enables a unified environmental model for both established longitudinal traffic assistance functions and novel functions based on dense environmental information. The parallel management of multiple environmental models is eliminated.

[0013] The principle of the procedure is in Fig. The model depicts a vehicle moving ahead in a grid using moving cells. The ego-vehicle searches the grid along its likely path for boundaries of the free space, specifically for moving cells—that is, cells containing information about a non-zero velocity or acceleration. Once a boundary is found, the necessary quantities or states for the functions, such as distance, velocity, or acceleration, can be determined from the cells.

[0014] In a preferred embodiment of the invention, for at least two adjacent, occupied, and moving grid cells, information about the kinematic state of an object represented by these cells is determined by taking into account the information about the kinematic state from the individual grid cells. The quantities correlated to the represented object, such as distance, velocity, and acceleration, can be determined using various methods, for example, by averaging the individual grid values ​​or by a minimum / maximum search, where the extreme value for the represented object is used.

[0015] In a particular embodiment of the invention, course prediction is improved by taking occupied cells into account during the search. In this context, a further subdivision of the static areas of the occupancy grid into "hard" boundaries, e.g., raised objects such as guardrails, which are considered non-traversable cells, is possible. Fig. to be depicted, and “soft” boundaries, e.g. lane markings, which are not recommended cells in Fig. It is useful to be able to map the area. If the search reaches a boundary that marks a non-recommended or impassable cell, the search direction is changed accordingly and the search continues. A suitable initial direction is one that points back into the unobstructed area. Furthermore, classic trajectory prediction, e.g., using steering angle or yaw rate, information from a digital map (e.g., as part of a navigation system), or information about the driver's intention, such as activated turn signals or detected lane-change intentions, can be used to determine the search direction. With good trajectory prediction, a non-recommended area can be crossed during the search, e.g., if the driver's lane-change intention is detected.

[0016] In a further embodiment of the invention, the width of the search window is based on the width of the vehicle; in particular, the width increases with increasing distance from the vehicle in order to compensate for measurement inaccuracies.

[0017] In a positive embodiment of the invention, the search window and / or the immediate vicinity of the search window are used to search for grid cells containing information about a lateral speed in order to detect vehicles entering or exiting the lane. The time until the ego-vehicle reaches the position of the search window can be calculated using the ego-vehicle's state, e.g., speed, acceleration, and / or steering angle.

[0018] Steering angle acceleration can be estimated, and thus, if necessary, a prediction of the future position of the relevant moving cells can be made. For example, lane-changers and lane-changers can be detected using lateral velocity.

[0019] In a further embodiment of the invention, the search can be performed taking into account several preceding objects. The search along the course prediction continues after moving cells have been detected. For this purpose, the information already determined about the moving cells is stored. If further moving cells are found, the stored states are compared with the newly determined ones, and a suitable rule (e.g., minimum / maximum) is used to decide whether the new states are adopted or the old states are retained.

[0020] In a particular embodiment of the invention, the driver assistance function is designed for blind spot monitoring and / or lane change assistance. A high-resolution sensor system is provided for detecting an area in front of the vehicle, and a relatively low-resolution sensor system is provided for detecting an area to the side and behind the vehicle. The representation of the vehicle's surroundings in a grid is based on the data from both sensor systems. The data from the two sensor systems in the grid are fused, and if the low-resolution sensor system detects a moving object and the high-resolution sensor system previously detected a stationary object at that position, the data from the high-resolution sensor system takes precedence. This embodiment is illustrated by… Fig. Figure 3 clarifies this. The search for relevant objects—that is, occupied grid cells—is performed in a search window that corresponds to the area to be monitored for the respective function. For blind spot monitoring, a monitoring area to the side and behind the vehicle is of interest, as shown in Fig. Figure 3 illustrates this. The length of the area is determined by the sensor system's range for environmental detection and / or the functional requirements. The same methods (averaging / minimum / maximum, minimum number of relevant cells) used in the front area can be applied to determine the state of a detected object and suppress erroneous measurements.

[0021] This method is particularly advantageous near structures such as guardrails. By fusing the typically more powerful front sensors with the less powerful side / rear sensors in the detection grid, information about static structures, represented by occupied cells in the grid, is available and confirmed by the front sensors. If a moving object is detected by the less powerful side / rear sensors at a location where a static object was previously detected by the more powerful front sensors, the measurement from the front sensor is considered more reliable and is used, thus suppressing the erroneous measurement from the side / rear sensors.

Claims

[1] Method for representing a vehicle's environment in an occupancy grid for a driver assistance function, wherein the data are acquired using at least one environment sensing system, where a likely course traveled by the vehicle is determined using vehicle sensors, in particular for detecting steering angle, direction of travel of the vehicle and / or indicator activity. characterized by , that A search is performed in the occupancy grid for occupied grid cells that define a free space, within a search window along the likely route traveled. [2] Method according to claim 1 characterized by , that at least two adjacent occupied grid cells form a complex, where information about the kinematic state of an object represented by these cells is determined taking into account the information about a kinematic state from the individual grid cells. [3] Method according to claim 1 or 2, characterized by , that occupied cells are taken into account for determining the likely course traveled, whereby at least some kinematic information is stored in these cells that suggests a stationary cell. [4] Method according to claims 1 to 3, characterized by , that the width of the search window is based on the width of the vehicle. [5] Method according to any of the preceding claims , characterized by , that the search window and / or the immediate vicinity of the search window is used to search for grid cells containing information about a lateral speed in order to detect vehicles entering or leaving the lane. [6] Method according to one of the preceding claims, wherein the driver assistance function is configured for monitoring a blind spot or for lane change assistance and a high-resolution sensor system is available for detecting an area in front of the vehicle and a relatively low-resolution sensor system for detecting an area to the side behind the vehicle, and the representation of the vehicle's surroundings in an occupancy grid is based on the data from these sensor systems characterized by , that the data from the two sensor systems are fused in the occupancy grid and, in the case that the low-resolution sensor system detects a moving object and the high-resolution sensor system previously detected a stationary object at that position, the data from the high-resolution sensor system takes precedence. [7] Driver assistance system comprising at least a sensor system for detecting a vehicle environment and an evaluation or control unit for controlling a driver assistance function, which has an electronic memory on which a method according to one of the preceding claims is stored, and a processor for executing a method according to one of the preceding claims.

Images