System and method for cut-in and/or lane changing prediction in a vehicle
A method and a system for cut-in and/or lane changing prediction in a first vehicle ( 102 ) includes an interface ( 104 ), in particular to a data bus, of the first vehicle ( 102 ) that is configured to receive a sequence of data sets. Each data set in the sequence has values that are indicative of a position or movement of a second vehicle relative the first vehicle. The system includes a model ( 108 ), in particular a recurrent neural network, that has a long short term memory that is configured to map the values from the sequence to a quantity for the sequence, in particular at least one binary label, indicative of a probability and/or confidence that a start of a cut-in and/or lane changing event by the second vehicle is detected or that no start of a cut-in and/or lane changing event by the second vehicle is detected.
1 . A system for cut-in and/or lane changing prediction in a first vehicle ( 102 ), the system comprising
an interface ( 104 ) connected to a data bus of the first vehicle ( 102 ), the interface being configured to receive a sequence of data sets, each of the data sets in the sequence having values that are indicative of a position or movement of a second vehicle relative to the first vehicle;
a model ( 108 ) with a long short term memory, the model ( 108 ) being configured to map the values from the sequence to a quantity for the sequence, wherein the quantity is indicative of a probability and/or confidence per data set that a start of a cut-in and/or lane changing event by the second vehicle is detected or that no start of the cut-in and/or lane changing event by the second vehicle is detected;
the interface ( 104 ) being configured to mirror values that originate from one side of the first vehicle to mirrored values for the other side of the first vehicle and to provide the mirrored values for mapping by the model ( 108 ); and
an adaptive cruise control system configured to control the first vehicle in response to the detection of the start of the cut-in and/or land changing event according to a mapping result of the model.
2 . The system of claim 1 , wherein the model is configured to distinguish whether the quantity indicates a true positive or a true negative depending on a threshold for the quantity.
3 . The system of claim 1 , wherein the interface ( 104 ) is configured to receive a plurality of values indicative of the position or movement of the second vehicle, to eliminate values that comprise redundant information from the plurality of values, and to provide values from the sequence without the eliminated values for mapping by the model ( 108 ).
4 . The system of claim 1 , wherein the interface is configured to receive the sequence in a fixed timeframe.
5 . The system of claim 1 , wherein the system comprises a set of cameras and the interface is configured to receive values from a set of camera signals that describe a time evolution of physical variables related to velocity, acceleration, position and/or distance of the second vehicle relative to the first vehicle or relative to at least one lane line separating a lane in which the first vehicle is located from a lane in which the second vehicle located.
6 . The system of claim 1 , wherein the interface is configured to provide each data set with a timestamp, and wherein the model is configured to determine a label at each timestamp.
7 . The system of claim 1 , wherein the interface is configured to eliminate one or more values from the plurality of received data sets that contain redundant information.
8 . A method for cut-in and/or lane changing prediction in a first vehicle, the method comprising:
receiving from a data bus of the first vehicle, a sequence of data sets, wherein each data set in the sequence comprises values from a long range radar, a camera, and/or a sensor data fusion, that are indicative of a position or movement of a second vehicle relative the first vehicle;
mirroring values that originate from one side of the first vehicle to mirrored values for the other side of the first vehicle and providing the mirrored values for mapping by a model;
mapping, with the model, the values from the sequence to a quantity for the sequence indicative of a probability and/or confidence per data set that a start of a cut-in and/or lane changing event by the second vehicle is detected or that no start of the cut-in and/or lane changing event by the second vehicle is detected; and
controlling, with an adaptive cruise control system, the first vehicle in response to the detection of the start of the cut-in and/or lane changing event according to a mapping result of the model.
9 . The method of claim 8 , further comprising distinguishing ( 204 ) whether the quantity indicates a true positive or a true negative depending on a threshold for the quantity.
10 . The method of claim 9 , further comprising receiving a plurality of values indicative of the position or movement of the plurality of values, and providing all values from the sequence for mapping by the model.
11 . The method of claim 8 , comprising:
receiving the sequence in a fixed timeframe.
12 . The method of claim 8 , further comprising capturing at least a part of the values with a set of cameras and/or receiving at last a part of the values from a set of camera signals that describe a time evolution of physical variables related to velocity, acceleration, position and/or distance of the second vehicle relative to the first vehicle or relative to at least one lane line separating a lane in which the first vehicle is located from a lane in which the second vehicle located.
13 . The method of claim 8 , further comprising:
providing each data set with a timestamp and determining a label at each timestamp.
14 . The method of claim 8 , further comprising:
eliminating one or more values from the received data sets that contain redundant information.
15 . The method of claim 8 , further comprising:
receiving the plurality of sequences of data sets; and
labelling each data set as a cut-in maneuver or not the cut-in maneuver.