US-20260125053-A1 - Lane Change Classification of a Vehicle

US20260125053A1US 20260125053 A1US20260125053 A1US 20260125053A1US-20260125053-A1

Abstract

A method for lane change classification of a vehicle has the steps of: acquiring sensor data representing a movement of a vehicle on a traffic lane over a predetermined time period; determining, based on the sensor data, a lateral velocity of the vehicle and a lateral distance of the vehicle relative to a boundary of the traffic lane; and assigning the movement of the vehicle to a lane changing class or a lane keeping class based on a classification function. The lane changing class is associated with the vehicle departing from the lane. The lane keeping class is associated with the vehicle staying on the lane. The classification function is defined in a lateral velocity dimension of the vehicle and a lateral distance dimension of the vehicle, and the classification function represents a boundary between the lane changing class and the lane keeping class.

Inventors

Ramin ZOHOURI

Assignees

BAYERISCHE MOTOREN WERKE AKTIENGESELLSCHAFT

Dates

Publication Date: 20260507
Application Date: 20251105
Priority Date: 20241106

Claims (11)

1 . A method for lane change classification of a vehicle, the method comprising the steps of: acquiring sensor data representing a movement of a vehicle on a traffic lane over a predetermined time period; determining, on the basis of the sensor data, a lateral velocity of the vehicle and a lateral distance of the vehicle relative to a boundary of the traffic lane; and assigning the movement of the vehicle to a lane changing class or a lane keeping class on the basis of a classification function, the lane changing class being associated with the vehicle departing from the traffic lane, and the lane keeping class being associated with the vehicle staying in the traffic lane, wherein the classification function is defined in a lateral velocity dimension of the vehicle and a lateral distance dimension of the vehicle, and wherein the classification function represents a boundary between the lane changing class and the lane keeping class.
2 . The method of claim 1 , wherein the classification function is defined by at least one polynomial function.
3 . The method of claim 1 , wherein the classification function is defined sectionally by a plurality of polynomial functions.
4 . The method of claim 1 , wherein assigning the movement of the vehicle to the lane changing class or the lane keeping class comprises: determining a probability value representing a probability of the vehicle departing from the traffic lane or staying in the traffic lane, respectively, wherein the probability value is determined on the basis of a distance between at least one movement data point and the classification function, the at least one movement data point being dependent on the lateral velocity and the lateral distance of the vehicle.
5 . The method of claim 1 , wherein the method is performed by a computer system of an ego vehicle travelling on an ego traffic lane separated from the traffic lane by the boundary.
6 . The method of claim 5 , wherein an assisted or automated driving function of the ego vehicle is controlled based on the assigning of the movement of the vehicle to the lane changing class or the lane keeping class.
7 . The method of claim 1 , wherein the classification function is defined only in the lateral velocity dimension and the lateral distance dimension.
8 . The Method of claim 1 , wherein the predetermined time period represented by the sensor data is at least 0 . 4 seconds.
9 . A system, comprising: a plurality of computer hardware components configured to carry out the acts of: acquiring sensor data representing a movement of a vehicle on a traffic lane over a predetermined time period; determining, on the basis of the sensor data, a lateral velocity of the vehicle and a lateral distance of the vehicle relative to a boundary of the traffic lane; and assigning the movement of the vehicle to a lane changing class or a lane keeping class on the basis of a classification function, the lane changing class being associated with the vehicle departing from the traffic lane, and the lane keeping class being associated with the vehicle staying in the traffic lane, wherein the classification function is defined in a lateral velocity dimension of the vehicle and a lateral distance dimension of the vehicle, and wherein the classification function represents a boundary between the lane changing class and the lane keeping class.
10 . A vehicle comprising the computer system of claim 9 .
11 . A computer product comprising a non-transitory computer readable medium having stored thereon program code which, when executed by a computer system, causes the acts of: acquiring sensor data representing a movement of a vehicle on a traffic lane over a predetermined time period; determining, on the basis of the sensor data, a lateral velocity of the vehicle and a lateral distance of the vehicle relative to a boundary of the traffic lane; and assigning the movement of the vehicle to a lane changing class or a lane keeping class on the basis of a classification function, the lane changing class being associated with the vehicle departing from the traffic lane, and the lane keeping class being associated with the vehicle staying in the traffic lane, wherein the classification function is defined in a lateral velocity dimension of the vehicle and a lateral distance dimension of the vehicle, and wherein the classification function represents a boundary between the lane changing class and the lane keeping class.

Description

CROSS REFERENCE TO RELATED APPLICATION This application claims priority under 35 U.S.C. §119 from European Patent Application No. EP 24211 019.5, filed November 6, 2024, the entire disclosure of which is herein expressly incorporated by reference. BACKGROUND AND SUMMARY The present disclosure relates to methods for lane change classification of vehicles. Lane change classification based on vehicle sensor data can be used in different types of assisted or automatic driving functions, e.g., lane keeping assistance, active lane guidance, adaptive cruise control, or automated lane change. Although regular lane change maneuvers can be detected with high reliability it has been found that certain vehicle movements within the lane can often be falsely interpreted as a lane change. In other words, a vehicle, which only appears to be performing a lane change (but actually stays in the lane), can be classified as false positive with respect to performing a lane change. For example, a vehicle may travel very close to the lane boundary but in fact does not perform a lane change. Due to the movement of the vehicle towards the lane boundary, a system may falsely classify this vehicle as performing a change from the current lane to the adjacent lane. This scenario can be particularly critical when the vehicle appears to cut into the lane of another vehicle traveling with higher velocity (so called cut-in maneuver). As a result, an ego vehicle, which relies on the lane change classification of a target vehicle, may automatically brake unexpectedly although this is in fact not necessary. The scenario can also be dangerous for other traffic participants, which do not expect a sudden braking event. Therefore, false positive classification of lane changes can potentially cause accidents, especially in heavy traffic situations including traffic jams. It is an object of the invention to provide a reliable method for lane change classification of a vehicle, in particular for minimizing the risk of false positive classifications of lane changes. In one aspect, the present disclosure is directed to a method for lane change classification of a vehicle, the method comprising the following steps carried out by computer hardware components: acquiring sensor data representing a movement of a vehicle on a traffic lane over a predetermined time period; determining, on the basis of the sensor data, a lateral velocity of the vehicle and a lateral distance of the vehicle relative to a boundary of the traffic lane; assigning the movement of the vehicle to a lane changing class or a lane keeping class on the basis of a classification function, the lane changing class being associated with the vehicle departing from the lane, the lane keeping class being associated with the vehicle staying on the lane, wherein the classification function is defined in a lateral velocity dimension of the vehicle and a lateral distance dimension of the vehicle, and the classification function representing a boundary between the lane changing class and the lane keeping class. It has been found that the method provides lane change classification with high reliability. Particularly, the rate of false positives is reduced, thereby avoiding potentially dangerous traffic scenarios in connection with automated driving systems, which rely on the lane change classification result. The classification function itself is defined in two dimensions for describing the state of the vehicle in the lateral direction, i.e. oblique to the direction of travelling according to the traffic lane. The lateral velocity dimension represents the velocity of the vehicle in the lateral direction, wherein the lateral distance dimension represents the distance of the vehicle to a boundary of the traffic lane. The boundary is preferably between two adjacent lanes, but it can in principle also be an outer boundary. Furthermore, the boundary can be marked on the lane, which is preferred, but this in not mandatory. The two lateral vehicle parameters, namely lateral velocity and lateral distance are interrelated insofar that lateral velocity represents the rate of change of lateral distance. However, the distance itself provides more information than the velocity alone. To this end, it has been found that the particular combination is effective not only for providing a compact data base, but also for defining the classification function in the same parameter space. In an embodiment, the classification function is defined by at least one polynomial function, preferably sectionally by a plurality of polynomial functions. This type of function is relatively easy to train and can be stored in a compact representation. The possible complexity of the classification boundary is also sufficient for robust classification. It is also less prone to overfitting than more complex function types. In one specific embodiment, the function can be defined by at least one spline function, for example by interpolating