System and method for lane association/transition association with splines

1 . A method for vehicle behavior prediction, the method comprising:

detecting a vehicle entering an unmarked road segment and having a current position in the unmarked road segment;

determining a set of exit paths available for the detected vehicle to reach different exit lanes of the unmarked road segment;

determining a heading of the detected vehicle to enter the unmarked road segment and reach the current position as a past history of the detected vehicle;

generating splines representing predicted paths between the different exit lanes of the unmarked road segment and the current position of the detected vehicle;

selecting a spline having a curvature conforming with the past history of the detected vehicle;

predicting an exit path taken by the detected vehicle as the exit path from the set of exit paths corresponding to the selected spline; and

controlling an ego vehicle while following a trajectory planned for the ego vehicle to avoid the detected vehicle moving along the predicted exit path taken by the detected vehicle to direct the ego vehicle to reach an exit lane of the unmarked road segment.

2 . The method of claim 1 , in which determining heading of the detected vehicle is performed in response to the detected vehicle reaching an intersection.

3 . The method of claim 1 , in which determining the set of exit paths comprises:

generating a birds-eye-view of the unmarked road segment; and

generating the set of exit paths based on the current position and the heading of the detected vehicle relative to the different exit lanes of the unmarked road segment.

4 . The method of claim 1 , in which predicting the exit path taken by the detected vehicle comprises:

determining a curvature of the different exit lanes of the unmarked road segment in relation to the current position and the heading of the detected vehicle according to the generated splines; and

inferring the predicted exit path taken by the detected vehicle as the exit lane having the least curvature according to the generated splines.

5 . The method of claim 1 , in which predicting the exit path taken by the detected vehicle comprises:

sensing a curvature of the upcoming road segments corresponding to the different exit lanes of the unmarked road segment according to the generated splines; and

ranking the set of paths available for the detected vehicle to reach the different exit lanes of the unmarked road segment based off of the sensed curvatures of the upcoming road segments following the different exit lanes according to the generated splines.

6 . The method of claim 1 , further comprising training a model to predict the exit path taken by the detected vehicle to reach the exit lane of the unmarked road segment based on a curvature of the different exit lanes of the unmarked road segment in relation to the detected vehicle.

7 . The method of claim 1 , in which detecting comprises:

identifying the vehicle entering an intersection; and

sensing a curvature of the upcoming road segments connected to the different exit lanes of the intersection.

8 . A non-transitory computer-readable medium having program code recorded thereon for vehicle behavior prediction, the program code being executed by a processor and comprising:

program code to detect a vehicle entering an unmarked road segment and having a current position in the unmarked road segment;

program code to determine a set of exit paths available for the detected vehicle to reach different exit lanes of the unmarked road segment;

program code to determine a heading of the detected vehicle to enter the unmarked road segment and reach the current position as a past history of the detected vehicle;

program code to generate splines representing predicted paths between the different exit lanes of the unmarked road segment and the current position of the detected vehicle;

program code to select a spline having a curvature conforming with the past history of the detected vehicle;

program code to predict an exit path taken by the vehicle from the set of exit paths corresponding to the selected spline; and

program code to control an ego vehicle while following a trajectory planned for the ego vehicle to avoid the detected vehicle moving along the predicted exit path taken by the detected vehicle to direct the ego vehicle to reach an exit lane of the unmarked road segment.

9 . The non-transitory computer-readable medium of claim 8 , in which the program code to determine the current position and the heading of the detected vehicle is performed in response to the detected vehicle reaching an intersection.

10 . The non-transitory computer-readable medium of claim 8 , in which the program code to determine the set of exit paths comprises:

program code to generate a birds-eye-view of the unmarked road segment; and

program code to generate the set of exit paths based on the current position and the heading of the detected vehicle relative to the different exit lanes of the unmarked road segment.

11 . The non-transitory computer-readable medium of claim 8 , in which the program code to predict the exit path taken by the detected vehicle comprises:

program code to determine a curvature of the different exit lanes of the unmarked road segment in relation to the current position and the heading of the detected vehicle according to the generated splines; and

program code to infer the predicted exit path taken by the detected vehicle as an exit lane having the least curvature according to the generated splines.

12 . The non-transitory computer-readable medium of claim 8 , in which the program code to predict the exit path taken by the detected vehicle comprises:

program code to sense a curvature of the upcoming road segments corresponding to the different exit lanes of the unmarked road segment according to the generated splines; and

program code to rank the set of paths available for the detected vehicle to reach the different exit lanes of the unmarked road segment based off of the sensed curvatures of the upcoming roads following the different exit lanes according to the generated splines.

13 . The non-transitory computer-readable medium of claim 8 , further comprising program code to train a model to predict the exit path taken by the detected vehicle to reach an exit lane of the unmarked road segment based on a curvature of the different exit lanes of the unmarked road segment in relation to the detected vehicle.

14 . The non-transitory computer-readable medium of claim 8 , in which the program code to detect comprises:

program code to identify the vehicle entering an intersection; and

program code to sense a curvature of the upcoming road segments connected to the different exit lanes of the intersection.

15 . A system for vehicle behavior prediction, the system comprising:

a vehicle perception module to detect a vehicle entering an unmarked road segment and having a current position in the unmarked road segment;

a vehicle paths determination module to determine a set of exit paths available for the detected vehicle to reach different exit lanes of the unmarked road segment, to determine a heading of the detected vehicle to enter the unmarked road segment and reach the current position as a past history of the detected vehicle, to generate splines representing predicted paths between the different exit lanes of the unmarked road segment and the current position of the detected vehicle, and to select a spline having a curvature conforming with the past history of the detected vehicle;

a vehicle path prediction module to predict an exit path taken by the vehicle from the set of exit paths corresponding to the selected spline; and

a vehicle trajectory planner module to control an ego vehicle while following a trajectory planned for the ego vehicle to avoid the detected vehicle moving along the predicted exit path taken by the detected vehicle to direct the ego vehicle to reach an exit lane of the unmarked road segment.

16 . The system of claim 15 , in which the vehicle perception module is further to determine the current position and the heading of the detected vehicle in response to the detected vehicle reaching an intersection.

17 . The system of claim 15 , in which the vehicle paths determination module is further to generate a birds-eye-view of the unmarked road segment, and to generate the set of exit paths based on the current position and the heading of the detected vehicle relative to the different exit lanes of the unmarked road segment.

18 . The system of claim 15 , in which the vehicle path prediction module is further to determine a curvature of the different exit lanes of the unmarked road segment in relation to the current position and the heading of the detected vehicle according to the generated splines; and to infer the predicted exit path taken by the detected vehicle as an exit lane having the least curvature according to the generated splines.