Parked car classification based on a velocity estimation
A method controlling an ego vehicle in an environment includes determining, via a flow model of a parked vehicle recognition system, a flow between a first representation of the environment and a second representation of the environment. The method also includes determining, via a velocity model of the parked vehicle recognition system, a velocity of a vehicle in the environment based on the flow. The method further includes determining, via a parked vehicle classification model of the parked vehicle recognition system, the vehicle is parked based on the velocity of the vehicle and one or more of features associated with the vehicle and/or the environment. The method still further includes planning a trajectory of the ego vehicle based on determining the vehicle is parked.
1. A method controlling an ego vehicle in an environment, comprising:
determining, via a flow model of a parked vehicle recognition system, a flow between a first representation of the environment and a second representation of the environment, the first representation and the second representation being captured by one or more sensors associated with the ego vehicle;
determining, via a velocity model of the parked vehicle recognition system, a velocity of a vehicle in the environment based on the flow;
determining, via a parked vehicle classification model of the parked vehicle recognition system, the vehicle is parked based on the velocity of the vehicle and one or more of features associated with the vehicle and/or the environment, the one or more features being extracted from one or both of the first representation or the second representation of the environment via a feature extractor associated with the parked vehicle classification model;
planning a trajectory of the ego vehicle based on determining the vehicle is parked; and
controlling the ego vehicle to follow the trajectory based on the environment.
2. The method of claim 1 , wherein the one or more features include vehicle estimated features.
3. The method of claim 2 , wherein the vehicle estimated features include one or more of an object type of the vehicle, a first distance from a center point of the vehicle to a road boundary, an estimated absolute speed of the vehicle, a second distance to a nearest intersection from the vehicle, a map location type, a free lane ratio, or an edge distance between an edge of the vehicle and the road boundary.
4. The method of claim 2 , wherein the velocity is concatenated with the vehicle estimated features.
5. The method of claim 1 , further comprising: obtaining the first representation via a first LiDAR sweep performed at a first time period via the one or more sensors; and obtaining the second representation via a second LiDAR sweep performed at a second time period via the one or more sensors.
6. The method of claim 1 , wherein the flow model is associated with a flow loss, the velocity model is associated with a velocity loss, and the parked vehicle classification model is associated with a parked vehicle loss.
7. The method of claim 6 , further comprising training the parked vehicle recognition system end-to-end to minimize a sum of the flow loss, the velocity loss, and the parked vehicle loss.
8. An apparatus for controlling an ego vehicle in an environment the apparatus comprising:
at least one processor; and
at least one memory coupled with the at least one processor and storing instructions operable, when executed by the at least one processor, to cause the apparatus:
to determine, via a flow model of a parked vehicle recognition system, a flow between a first representation of the environment and a second representation of the environment, the first representation and the second representation being captured by one or more sensors associated with the ego vehicle;
to determine, via a velocity model of the parked vehicle recognition system, a velocity of a vehicle in the environment based on the flow;
to determine, via a parked vehicle classification model of the parked vehicle recognition system, the vehicle is parked based on the velocity of the vehicle and one or more of features associated with the vehicle and/or the environment, the one or more features being extracted from one or both of the first representation or the second representation of the environment via a feature extractor associated with the parked vehicle classification model;
to plan a trajectory of the ego vehicle based on determining the vehicle is parked; and
to control the ego vehicle to follow the trajectory based on the environment.
9. The apparatus of claim 8 , wherein the one or more features include vehicle estimated features.
10. The apparatus of claim 9 , wherein the vehicle estimated features include one or more of an object type of the vehicle, a first distance from a center point of the vehicle to a road boundary, an estimated absolute speed of the vehicle, a second distance to a nearest intersection from the vehicle, a map location type, a free lane ratio, or an edge distance between an edge of the vehicle and the road boundary.
11. The apparatus of claim 9 , wherein the velocity is concatenated with the vehicle estimated features.
12. The apparatus of claim 8 , wherein execution of the instructions further cause the apparatus to: obtain the first representation via a first LiDAR sweep performed at a first time period via the one or more sensors; and obtain the second representation via a second LiDAR sweep performed at a second time period via the one or more sensors.
13. The apparatus of claim 8 , wherein the flow model is associated with a flow loss, the velocity model is associated with a velocity loss, and the parked vehicle classification model is associated with a parked vehicle loss.
14. The apparatus of claim 13 , wherein execution of the instructions further cause the apparatus to train the parked vehicle recognition system end-to-end to minimize a sum of the flow loss, the velocity loss, and the parked vehicle loss.
15. A non-transitory computer-readable medium having program code recorded thereon for controlling an ego vehicle in an environment the program code executed by at least one processor and comprising:
program code to determine, via a flow model of a parked vehicle recognition system, a flow between a first representation of the environment and a second representation of the environment, the first representation and the second representation being captured by one or more sensors associated with the ego vehicle;
program code to determine, via a velocity model of the parked vehicle recognition system, a velocity of a vehicle in the environment based on the flow;
program code to determine, via a parked vehicle classification model of the parked vehicle recognition system, the vehicle is parked based on the velocity of the vehicle and one or more of features associated with the vehicle and/or the environment, the one or more features being extracted from one or both of the first representation or the second representation of the environment via a feature extractor associated with the parked vehicle classification model;
program code to plan a trajectory of the ego vehicle based on determining the vehicle is parked; and
program code to control the ego vehicle to follow the trajectory based on the environment.
16. The non-transitory computer-readable medium of claim 15 , wherein the one or more features include vehicle estimated features.
17. The non-transitory computer-readable medium of claim 16 , wherein the vehicle estimated features include one or more of an object type of the vehicle, a first distance from a center point of the vehicle to a road boundary, an estimated absolute speed of the vehicle, a second distance to a nearest intersection from the vehicle, a map location type, a free lane ratio, or an edge distance between an edge of the vehicle and the road boundary.
18. The non-transitory computer-readable medium of claim 16 , wherein the velocity is concatenated with the vehicle estimated features.
19. The non-transitory computer-readable medium of claim 15 , wherein the program code further includes: program code to obtain the first representation via a first LiDAR sweep performed at a first time period via the one or more sensors; and program code to obtain the second representation via a second LiDAR sweep performed at a second time period via the one or more sensors.
20. The non-transitory computer-readable medium of claim 15 , wherein: the flow model is associated with a flow loss, the velocity model is associated with a velocity loss, and the parked vehicle classification model is associated with a parked vehicle loss; and the program code further includes program code to train the parked vehicle recognition system end-to-end to minimize a sum of the flow loss, the velocity loss, and the parked vehicle loss.