Method and Device for Establishing a Trajectory for a Vehicle

• US 20150266475A1
• Filed: 03/19/2015
• Published: 09/24/2015
• Est. Priority Date: 03/20/2014
• Status: Active Grant

• Current Assignee: Bayerische Motoren Werke AG
• Original Assignee: Bayerische Motoren Werke AG
• Inventors: TANZMEISTER, Georg
• Granted Patent: US 9,248,835 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: B60W 30/0956 the prediction being responsive to traffic or environmental parameters

Abstract

A method and a device for establishing a trajectory for a vehicle is provided. The method includes the following steps: registering coordinates of objects using a sensor; calculating a cost map on the basis of the registered coordinates, wherein the cost map is subdivided into cells and a cost value is assigned to each cell, which cost value describes the presence and/or the vicinity of an object, calculating a cost space, which has a plurality of layers, wherein each layer is subdivided into cells and a cost value is assigned to each cell, wherein the cost values in the respective layers are calculated in each case for a specific orientation of the vehicle, and determining a trajectory that is as cost-effective as possible.

Claims

1. A method for establishing a trajectory for a vehicle, the method comprising the acts of:

registering coordinates of objects using a sensor;

calculating an occupancy map or cost map on the basis of the registered coordinates, wherein the occupancy map or cost map is subdivided into cells and a cell value is assigned to each cell, which cell value describes a presence and/or a vicinity of an object;

calculating a cost space, which comprises a plurality of layers, wherein each layer is subdivided into cells and a cost value is assigned to each cell, wherein the cost values in the respective layers are calculated in each case for a specific orientation of the vehicle on the basis of the cell values; and

determining a trajectory for the vehicle that is as cost-effective as possible.

2. The method according to claim 1, wherein

the cost values in a collision space are represented by one or more of the following values, either individually or in combination:

distance to the closest object,

cost value expanded by a footprint of the vehicle in the cost map,

reliability value of the coordinates of the object registered by the sensor, or

collision probability, which is determined on the basis of the registered coordinates of the objects and empirical data or empirical relationships.

3. The method according to claim 2, wherein

trajectories are monitored for possible collisions, and trajectories containing a collision are discarded.

4. The method according to claim 1, wherein

a distinction is made between static and dynamic objects when the cost map is calculated and dynamic objects are filtered out.

5. The method according to claim 3, wherein

a distinction is made between static and dynamic objects when the cost map is calculated and dynamic objects are filtered out.

6. The method according to claim 1, wherein

the cost values of the cells are designed by expansion using a footprint of the vehicle when calculating a layer of a cost space on the basis of the cost map, wherein the expansion is carried out by a dilation or by a convolution.

7. The method according to claim 5, wherein

the cost values of the cells are designed by expansion using a footprint of the vehicle when calculating a layer of a cost space on the basis of the cost map, wherein the expansion is carried out by a dilation or by a convolution.

8. The method according to claim 6, wherein

each footprint of the vehicle has an anchor when expanding or dilating the cost map to the layers of the cost space, and

either the sum of all probabilities contained in the cells of the occupancy map within the footprint or the maximum of all cost values contained in the cells of the cost map within the footprint is assigned to the cell of the layer of the cost space which corresponds to the location of the cell of the cost map at which the anchor of the footprint is situated (referred to as anchor cell).

9. The method according to claim 6, wherein

the cost map is expanded to the cost space by a convolution, wherein, for purposes of calculating the convolution, the occupancy map and the footprint are transformed into the frequency space.

10. The method according to claim 8, wherein

the cost map is expanded to the cost space by a convolution, wherein, for purposes of calculating the convolution, the occupancy map and the footprint are transformed into the frequency space.

11. The method according to claim 6, wherein

prior to expanding the cost map to the layers of the cost space, the objects are surrounded by a cost neighborhood, wherein cost values successively decreasing to the outer edge of the cost neighborhood are assigned to cells within the cost neighborhood in the cost map.

12. The method according to claim 10, wherein

prior to expanding the cost map to the layers of the cost space, the objects are surrounded by a cost neighborhood, wherein cost values successively decreasing to the outer edge of the cost neighborhood are assigned to cells within the cost neighborhood in the cost map.

13. The method according to claim 6, wherein

the expansion is carried out by a convolution and the cost values are discretized into a predetermined number of cost values, wherein the number preferably lies in the range from 2 to 10, wherein an inverse imaging function (h(n)) is used to image the convolution results onto a cost dilation.

14. The method according to claim 12, wherein

the expansion is carried out by a convolution and the cost values are discretized into a predetermined number of cost values, wherein the number preferably lies in the range from 2 to 10, wherein an inverse imaging function (h(n)) is used to image the convolution results onto a cost dilation.

15. The method according to claim 8, wherein

the anchor cell of the footprint is arranged in the center of the footprint.

16. The method according to claim 15, wherein

the footprint of the vehicle is rectangular and identical symmetries of the footprint of an occupancy map are only calculated once.

17. The method according to claim 16, wherein

the dilation or the expansion of a two-dimensional footprint of the vehicle is replaced by two dilations with, in each case, a one-dimensional footprint, wherein use is made of the vHGW method.

18. The method according to claim 14, wherein

the dilation or the expansion of a two-dimensional footprint of the vehicle is replaced by two dilations with, in each case, a one-dimensional footprint, wherein use is made of the vHGW method.

19. The method according to claim 6, wherein

the cost values of the individual cells are weighted by the length of the portion extending from the trajectory through the respective cell when determining a trajectory in the cost space.

20. A device for establishing a trajectory for a vehicle, comprising:

a sensor;

a computer, and

an interface, wherein

the computer has a stored computer program executed to;

register coordinates of objects using a sensor;

calculate an occupancy map or cost map on the basis of the registered coordinates, wherein the occupancy map or cost map is subdivided into cells and a cell value is assigned to each cell, which cell value describes a presence and/or a vicinity of an object;

calculate a cost space, which comprises a plurality of layers, wherein each layer is subdivided into cells and a cost value is assigned to each cell, wherein the cost values in the respective layers are calculated in each case for a specific orientation of the vehicle on the basis of the cell values; and

determine a trajectory for the vehicle that is as cost-effective as possible,

wherein the trajectory with the lowest possible risk of collision is providable at the interface.