Systems and methods for estimating the origins of abnormal driving
Systems, methods, and other embodiments described herein relate to estimating the origins of abnormal driving through observed driving patterns. In one embodiment, a method includes detecting, using parallel computations, abnormal classifications associated with a subject vehicle and nearby vehicles according to driving patterns derived from observation data, the abnormal classifications being associated with exceeding a position range in the observation data. The method also includes estimating an origin of abnormal driving with happens-before analysis according to the abnormal classifications, and the abnormal driving is associated with deviations in a traffic flow associated with the subject vehicle and the nearby vehicles. The method also includes controlling the subject vehicle using a driving command according to the origin.
1. A detection system comprising:
a processor; and
a memory storing instructions that, when executed by the processor, cause the processor to:
detect, using parallel computations, abnormal classifications associated with a subject vehicle and nearby vehicles according to driving patterns derived from observation data, the abnormal classifications being associated with exceeding a position range in the observation data;
estimate an origin of abnormal driving with happens-before analysis according to the abnormal classifications, and the abnormal driving is associated with deviations in a traffic flow associated with the subject vehicle and the nearby vehicles, and the happens-before analysis outputs interactions between the subject vehicle and the nearby vehicles by temporal organization of the abnormal classifications; and
control the subject vehicle using a driving command according to the origin.
2. The detection system of claim 1 , wherein the instructions to estimate the origin of the abnormal driving further include instructions to:
determine, using the happens-before analysis, that distracted driving by the subject vehicle preceded aggressive driving by the nearby vehicles; and
output according to the happens-before analysis that the subject vehicle is the origin of the abnormal driving.
3. The detection system of claim 2 , wherein the instructions to determine that the distracted driving by the subject vehicle preceded the aggressive driving further include instructions to predict that one of the nearby vehicles preceding the subject vehicle is moving in a motion pattern within a driving lane and a separation distance that is constant.
4. The detection system of claim 1 , wherein the instructions to detect the abnormal classifications further include instructions to analyze, by the subject vehicle, following distances between the subject vehicle and a follower vehicle and the subject vehicle and a preceding vehicle individually for the parallel computations.
5. The detection system of claim 1 , wherein the instructions to detect the abnormal classifications further include instructions to:
compare the driving patterns of the subject vehicle and the nearby vehicles in a time-series using locations derived from the observation data; and
identify abnormal events from abrupt deviations to following distances and speed between the subject vehicle and the nearby vehicles using the driving patterns in the time-series.
6. The detection system of claim 5 further including instructions to:
process the time-series by a machine learning (ML) model; and
output, by the ML model, probabilities of accuracy associated with the abnormal classifications for the subject vehicle and the nearby vehicles.
7. The detection system of claim 1 further including instructions to communicate the origin to a server for reducing collision risk, wherein the server forms a mitigation plan for automated vehicles of the nearby vehicles.
8. The detection system of claim 1 further including instructions to:
reduce noise by filtering the observation data about the subject vehicle and the nearby vehicles in a driving scenario associated with the abnormal classifications; and
derive a model of abnormal events in time with the happens-before analysis for synchronization and ordering of the abnormal classifications.
9. The detection system of claim 1 , wherein the abnormal classifications include one of aggressive driving, speeding, distracted driving, erratic trajectories, reckless driving, and off-center trajectories.
10. A non-transitory computer-readable medium comprising:
instructions that when executed by a processor cause the processor to:
detect, using parallel computations, abnormal classifications associated with a subject vehicle and nearby vehicles according to driving patterns derived from observation data, the abnormal classifications being associated with exceeding a position range in the observation data;
estimate an origin of abnormal driving with happens-before analysis according to the abnormal classifications, and the abnormal driving is associated with deviations in a traffic flow associated with the subject vehicle and the nearby vehicles, and the happens-before analysis outputs interactions between the subject vehicle and the nearby vehicles by temporal organization of the abnormal classifications; and
control the subject vehicle using a driving command according to the origin.
11. A method comprising:
detecting, using parallel computations, abnormal classifications associated with a subject vehicle and nearby vehicles according to driving patterns derived from observation data, the abnormal classifications being associated with exceeding a position range in the observation data;
estimating an origin of abnormal driving with happens-before analysis according to the abnormal classifications, and the abnormal driving is associated with deviations in a traffic flow associated with the subject vehicle and the nearby vehicles, and the happens-before analysis outputs interactions between the subject vehicle and the nearby vehicles by temporal organization of the abnormal classifications; and
controlling the subject vehicle using a driving command according to the origin.
12. The method of claim 11 , wherein estimating the origin of the abnormal driving further includes:
determining, using the happens-before analysis, that distracted driving by the subject vehicle preceded aggressive driving by the nearby vehicles; and
outputting according to the happens-before analysis that the subject vehicle is the origin of the abnormal driving.
13. The method of claim 12 , wherein determining that the distracted driving by the subject vehicle preceded the aggressive driving further includes predicting that one of the nearby vehicles preceding the subject vehicle is moving in a motion pattern within a driving lane and a separation distance that is constant.
14. The method of claim 11 , wherein detecting the abnormal classifications further includes analyzing, by the subject vehicle, following distances between the subject vehicle and a follower vehicle and the subject vehicle and a preceding vehicle individually for the parallel computations.
15. The method of claim 11 , wherein detecting the abnormal classifications further includes:
comparing the driving patterns of the subject vehicle and the nearby vehicles in a time-series using locations derived from the observation data; and
identifying abnormal events from abrupt deviations to following distances and speed between the subject vehicle and the nearby vehicles using the driving patterns in the time-series.
16. The method of claim 15 further comprising:
processing the time-series by a machine learning (ML) model; and
outputting, by the ML model, probabilities of accuracy associated with the abnormal classifications for the subject vehicle and the nearby vehicles.
17. The method of claim 11 further comprising communicating the origin to a server for reducing collision risk, wherein the server forms a mitigation plan for automated vehicles of the nearby vehicles.
18. The method of claim 11 further comprising:
reducing noise by filtering the observation data about the subject vehicle and the nearby vehicles in a driving scenario associated with the abnormal classifications; and
deriving a model of abnormal events in time with the happens-before analysis for synchronization and ordering of the abnormal classifications.
19. The method of claim 11 , wherein the nearby vehicles include a follower vehicle and preceding vehicles associated with the subject vehicle and the position range is associated with one of constant acceleration and deceleration for vehicle control.
20. The method of claim 11 , wherein the abnormal classifications include one of aggressive driving, speeding, distracted driving, erratic trajectories, reckless driving, and off-center trajectories.