Method for Representing a Vehicle Environment with Position Points

Abstract

The invention relates to a method for representing a vehicle environment for a vehicle with a sensor system in order to detect the environment. For this purpose, the vehicle environment is described with a predetermined fixed set of position points (particles).

Description

The invention relates to the technical field of representing structures in the environment of a vehicle as a data base for driver assistance systems with machine perception.

For driver assistance systems which are based on sensor systems for detecting the environment, the modeling and representation of the vehicle environment is of great importance. One option of representation is an occupancy grid, in which the vehicle environment is divided into equidistant grid cells and each grid cell is provided with details such as occupied or unoccupied. An alternative approach is the representation in a dense environment representation. Here, an occupancy information is obtained via a defined area in the environment of the vehicle from the sensor data and is entered into an occupancy map. Such a representation allows a direct evaluation of the available maneuver space.

It is the object of the present invention to indicate a method for representing a vehicle environment.

A method and apparatus for representing a vehicle environment (2) for a vehicle (1) with a sensor system (5) is indicated for detecting the environment, the vehicle environment (2) being described with a predetermined fixed set of position points (in this application equivalent with particles). The environment representation as a fixed set of particles has the advantage of an always constant data volume which is required for storing and transferring the environment model. This applies in particular when compression methods for reducing the data to be transferred are used, which avoid the redundant transmission of cell groups with the same value. Here, the data volume to be transferred is not constant over sequenced packages. This leads to problems in a close-to-production design of communication channels, in particular in case of time-defined solutions, such as e.g. FlexRay. In particular, the proposed invention thus is advantageous in a transmission of the environment data in a vehicle, e.g. from a first evaluation unit, which calculates an environment representation, to a second unit, which is embodied e.g. as a control device for a driver assistance function. Moreover, it is advantageous that management and access of this data structure are carried out on the software side very efficiently. The fixed set of position points is further used especially dense exactly at those places, where extensive structural descriptions are necessary. This leads to a highly efficient use of the memory reserved for the representation and of the bandwidth for its transmission.

In a preferred embodiment of the invention a position point (particle) is provided with a freely definable number of attributes, which represent a characteristic of the vehicle environment at the position of the position point. An attribute is a position information indicating the position of the position point relative to a host vehicle. The position can be indicated e.g. by distance, angle, spatial coordinates or the like.

In a positive embodiment of the invention at least one position point (particle) but preferably a plurality of position points/all position points is provided with at least one further attribute, which represents a characteristic of the vehicle environment at the position of the position point. The further attribute indicates e.g. a height above the ground or a value for a traversability). For example, a height or depth of a ground wave, the depth of a road ditch, a construction fence (not traversable), a solid road marking (not traversable), a broken road marking (traversable), an object on or near the roadway, e.g. a can of Coke (traversable), another vehicle (not traversable) can be specified as an attribute to the position point.

In a preferred embodiment of the method the position points are stochastically distributed in a specifiable detection range—if there is no information on an environment structure (4), what is the case in particular with a restart of the detection system (5). Such a distribution (3) is shown as an example in FIG. 1.

In particular, the method provides that the particles are arranged in the detection range based on the data of the sensor system for detecting the environment (2), the distribution of the position points in the detection range being carried out subject to recognized structures in the vehicle environment. Structures (4), which are not or only partially traversable are represented with a high density of position points. Such a distribution is shown as an example in FIG. 3.

Preferably, the sensor system (5) comprises at least one radar sensor for detecting the environment. Here, the position points are arranged subject to a reflected radar radiation, in particular as a function of the amplitude or energy of the reflected radar radiation.

The radar sensor detects an amount of energy that can be located, which can be used in particular as an indicator of the existence and solidity of typical environment structures such as construction walls or construction warning poles. A simple interpretation of the sensor data, i.e. the specific sensor model (6), means in this case a distribution of the available set of position points analogous to the reflected energy and according to the position of the reflection.

Preferably, the method is applied when the sensor system (5) for detecting the environment comprises a plurality of different types of sensors. For each sensor and each sensor type a specific sensor model (6) is provided for adjusting the distribution of the position points. The sensor model (6) represents the recognized structures via a corresponding distribution adjustment of the particles in the environment representation. To enter the data on the structure (4) of the environment into the environment representation, a specific sensor model (6) must exist for each sensor, which takes into account its detection possibilities and capacities. A sensor model for a radar sensor has been described above as an example.

In FIG. 2 it is schematically shown that one sensor model is created for each sensor and thereafter a plurality of sensor models (6) is merged.

A simple method for merging (fusioning) of the particles of several sensors is to accumulate in cycles all particles obtained by the different sensors (sensor models) in a common representation. A cycle is a predetermined duration of time. Preferably, the position points are updated in each cycle. This method can be used in particular for the merging of the particles of several sensors, whose visual fields do not or only slightly overlap, to maintain a high accuracy of the merged representation. The total number of particles is constant in this case and equal to the sum of the numbers of particles of the merged sensors. With strongly overlapping visual fields a simple accumulation leads to an increased consumption of resources as compared with the particle representation of an individual sensor. To limit the consumption of resources, a representation with a reduced number of particles can be selected for the merger. The distribution of these particles is adjusted in the merger such that they approximate as well as possible the cumulative distribution of the particle representations to be merged.

The reduction of the number of particles can be carried out for example by providing the particles for the reduction additionally to their parameters with a weighting factor. Here, for a particle of the merged total representation it applies that e.g. the higher the weight, the more particles are in the sensor representation in its environment. For reduced total representation a new particle set is created by randomly drawing from the initial representation (sensor representation) new particles until the predetermined set is reached, the occurrence probability of a particle in the reduced representation being proportional to the weight in the initial representation.

Furthermore, the weighting factor can depend on the state variable or variables derived therefrom, e.g. a higher weighting factor with a high gradient of the course of the state variable.

The state variables of a particle of the reduced representation (e.g. occupancy probability or height) can be determined from the adjacent particles by interpolation (e.g. constant, linear, quadratic).

LIST OF REFERENCE NUMERALS

• 1 Host vehicle
• 2 area of environment presentation
• 3 stochastic distribution of the position of points
• 4 structures
• 5 sensors
• 6 sensor models
• 7 merged particle representation

Claims

1. A method for representing a vehicle environment for a vehicle with a sensor system for sensing the environment characterized in that the vehicle environment is described with a predetermined fixed set of position points (particles).

2. A method according to claim 1, characterized in that a position point is provided with a freely definable number of attributes, wherein an attribute is a position information, which indicates the position to a host vehicle.

3. A method according to claim 2, characterized in that a position point is provided with at least one further attribute, wherein this attribute indicates a height above the ground or a value for a traversability.

4. A method according to claim 1, characterized in that if there is no information on the environment structure, particularly at the start of the method, the position points are stochastically distributed in a specifiable detection range.

5. A method according to claim 1, characterized in that the particles are arranged in a detection range based on data of the sensor system for detecting the environment, wherein a distribution of the position points is effected as a function of recognized structures in the vehicle environment.

6. A method according to claim 5, wherein the sensor system for detecting the environment comprises at least one radar sensor, characterized in that the particles are arranged in the detection range subject to a locally reflected radar radiation, in particular as a function of the amplitude or energy of the reflected radar radiation.

7. A method according to claim 6, characterized in that in a region, in which the amplitude or energy of the reflected radar radiation has a high value, more particles are arranged than in a region, in which the energy of the reflected radar radiation has a low value.

8. A method according to claim 1, wherein the sensor system for detecting the environment comprises a plurality of different types of sensors characterized in that for each sensor and each sensor type a specific sensor model is provided for adjusting the distribution of the position points, wherein the distribution of the position points is effected subject to structures, in particular raised structures, in the vehicle environment.

9. A method according to claim 8, characterized in that the position points of the different types of sensors are merged by accumulating in cycles the position points of each sensor in a common merged representation.

10. A method according to claim 9, characterized in that the number of position points of the merged representation is limited by selecting a lower number of position points for the merged total representation than for the accumulated sensor representations and in particular this reduced number of position points is adjusted such that it approximates the accumulated representation at its best.

11. A method according to claim 10, characterized in that a reduction in the number of position points is achieved by providing each position point of the initial representation (sensor representation) with a weight, and in thatthe weight is determined subject to one or both of the following parameters