Systems and methods for a scenario tagger for autonomous vehicles
Systems and methods are directed to determining autonomous vehicle scenarios based on autonomous vehicle operation data. In one example, a computer-implemented method for determining operating scenarios for an autonomous vehicle includes obtaining, by a computing system comprising one or more computing devices, log data representing autonomous vehicle operations. The method further includes extracting, by the computing system, a plurality of attributes from the log data. The method further includes determining, by the computing system, one or more scenarios based on a combination of the attributes, wherein each scenario includes multiple scenario variations and each scenario variation comprises multiple features. The method further includes providing, by the computing system, the one or more scenarios for generating autonomous vehicle operation analytics.
1. A computer-implemented method for generating autonomous vehicle simulation scenarios comprising:
obtaining log data that describes, based on a plurality of attributes, a plurality of observations collected during real-world operation of one or more vehicles;
extracting, from the log data, an attribute combination comprising one or more attributes of the plurality of attributes, wherein the one or more attributes comprise a vehicle maneuver attribute indicative of an action performed by the one or more vehicles;
determining, based at least in part on the attribute combination, a scenario associated with operating an autonomous vehicle, wherein the scenario is represented by the one or more attributes of the attribute combination,
wherein the one or more attributes comprise a lane geometry attribute, wherein the lane geometry attribute is associated with one or more lanes of a travel way, wherein the scenario is represented by the one or more lanes of the travel way, and
wherein the one or more attributes comprise an object attribute, wherein the object attribute is associated with a vehicle, a pedestrian, or a bicycle, wherein the scenario is represented by the vehicle, the pedestrian, or the bicycle; and
generating a simulation scenario for testing the autonomous vehicle based at least in part on the scenario associated with operating the autonomous vehicle, wherein the simulation scenario is indicative of the one or more attributes of the attribute combination.
2. The computer-implemented method of claim 1 , further comprising: generating training data based at least in part on the scenario associated with operating the autonomous vehicle, wherein the training data is indicative of the one or more attributes of the attribute combination.
3. The computer-implemented method of claim 2 , wherein the training data comprises a portion of the log data associated with the attribute combination, the portion of the log data comprising sensor data acquired by the one or more vehicles.
4. The computer-implemented method of claim 2 , further comprising: providing the training data for training of a machine-learned model.
5. The computer-implemented method of claim 1 , wherein determining the scenario associated with operating the autonomous vehicle comprises: determining the scenario based at least in part on a geographic region associated with the log data.
6. The computer-implemented method of claim 5 , further comprising: providing the training data to train a machine-learned model to perform at least one of: (i) recognizing an object, or (ii) predicting object behavior within the geographic region.
7. The computer-implemented method of claim 1 , wherein the scenario is represented by a new combination of attributes occurring within the log data.
8. The computer-implemented method of claim 1 , wherein the one or more attributes further comprise a travel way attribute associated with one or more lanes of travel by the one or more vehicles and an actor attribute associated with an object influencing the action performed by the one or more vehicles.
9. A computing system comprising:
one or more processors; and
one or more non-transitory computer-readable media that store instructions that, when executed by the one or more processors, cause the computing system to perform operations, the operations comprising:
obtaining log data that describes, based on a plurality of attributes, a plurality of observations collected during real-world operation of one or more vehicles;
extracting, from the log data, an attribute combination comprising one or more attributes of the plurality of attributes, wherein the one or more attributes comprise a vehicle maneuver attribute indicative of an action the vehicle performs;
determining, based at least in part on the attribute combination, a scenario associated with operating an autonomous vehicle, wherein the scenario is represented by the one or more attributes of the attribute combination,
wherein the one or more attributes comprise a lane geometry attribute, wherein the lane geometry attribute is associated with one or more lanes of a travel way, wherein the scenario is represented by the one or more lanes of the travel way, and
wherein the one or more attributes comprise an object attribute, wherein the object attribute is associated with a vehicle, a pedestrian, or a bicycle, wherein the scenario is represented by the vehicle, the pedestrian, or the bicycle; and
generating a simulation scenario for testing the autonomous vehicle based at least in part on the scenario associated with operating the autonomous vehicle, wherein the simulation scenario is indicative of the one or more attributes of the attribute combination.
10. The computing system of claim 9 , further comprising: generating training data based at least in part on the scenario associated with operating the autonomous vehicle, wherein the training data is indicative of the one or more attributes of the attribute combination.
11. The computing system of claim 10 , wherein the training data comprises a portion of the log data associated with the attribute combination, the portion of the log data comprising at least one of: sensor data, map data, perception data, prediction data, motion planning data, vehicle controller data, or vehicle status data.
12. The computing system of claim 10 , further comprising: providing the training data for training of a machine-learned model.
13. The computing system of claim 12 , wherein the machine learned model is trained to perform attribute extraction.
14. The computing system of claim 12 , wherein the machine learned model is trained to perform at least one of: (i) recognizing an object, or (ii) predicting object behavior.
15. The computing system of claim 9 , wherein the one or more attributes further comprise a travel way attribute associated with one or more lanes of travel by the one or more vehicles and an actor attribute associated with an object influencing the vehicle maneuver.
16. The computing system of claim 9 , wherein the one or more attributes comprise a lane geometry attribute, wherein the lane geometry attribute is associated with at least one of: (i) an intersection or (ii) a turn, wherein the scenario is represented by at least one of: (i) the intersection or (ii) the turn.
17. The computing system of claim 9 , wherein the one or more attributes comprise an object influencing the vehicle maneuver, wherein the object comprises a static object or a dynamic object, wherein the scenario is represented by the static object or the dynamic object.
18. One or more tangible, non-transitory, computer readable media that store instructions that when executed by one or more processors cause the one or more processors to perform operations comprising:
obtaining log data that describes, based on a plurality of attributes, a plurality of observations collected during real-world operation of one or more vehicles;
extracting, from the log data, an attribute combination comprising one or more attributes of the plurality of attributes, wherein the one or more attributes comprise a vehicle maneuver attribute indicative of an action the vehicle performs;
determining, based at least in part on the attribute combination, a scenario associated with operating an autonomous vehicle, wherein the scenario is represented by the one or more attributes of the attribute combination, and wherein the attribute combination is a new attribute combination occurring in the log data,
wherein the one or more attributes comprise a lane geometry attribute, wherein the lane geometry attribute is associated with one or more lanes of a travel way, wherein the scenario is represented by the one or more lanes of the travel way, and
wherein the one or more attributes comprise an object attribute, wherein the object attribute is associated with a vehicle, a pedestrian, or a bicycle, wherein the scenario is represented by the vehicle, the pedestrian, or the bicycle; and
generating at least one of: (i) a simulation scenario for testing the autonomous vehicle based at least in part on the scenario associated with operating the autonomous vehicle, wherein the simulation scenario is indicative of the one or more attributes of the attribute combination, or (ii) training data based at least in part on the scenario associated with operating the autonomous vehicle, wherein the training data is indicative of the one or more attributes of the attribute combination.