Generating and fusing traffic scenarios for automated driving systems
Automated driving systems intended for use in a vehicle can be simulated and validated using automatically generated simulation scenarios. A large number of simulation scenarios can be generated by augmenting a plurality of recorded scenarios with one or more extracted data streams. The extracted data streams can correspond to objects such as other vehicles, pedestrians, cyclists, or other obstacles or barriers. The extracted data streams can be generated automatically by identifying similar recorded scenarios and isolating differences between the similar recorded scenarios.
1. A method comprising:
extracting data from a first scenario or a second scenario, the data corresponding to a difference between the first scenario recorded by one or more first vehicle sensors and the second scenario recorded by one or more second vehicle sensors, wherein the first scenario shares at least one characteristic with the second scenario;
injecting the extracted data into a third scenario recorded by one or more third vehicle sensors to generate a fourth scenario; and
simulating automated vehicle driving software with the fourth scenario.
2. The method of claim 1 , further comprising:
automatically comparing a plurality of scenarios to identify the first scenario and the second scenario sharing the at least one characteristic.
3. The method of claim 1 , wherein the at least one shared characteristic is shared global position system (GPS) coordinates.
4. The method of claim 1 , wherein extracting the data comprises:
subtracting data recorded by a type of vehicle sensor recording in the first scenario from data recorded by a corresponding type of vehicle sensor recording in the second scenario.
5. The method of claim 1 , wherein extracting the data comprises:
normalizing the first scenario and the second scenario before extracting the data or normalizing the extracted data after extracting the data.
6. The method of claim 1 , wherein extracting the data further comprises:
aligning the first scenario and the second scenario to account for differences due to vehicle speed or vehicle position.
7. The method of claim 1 , wherein extracting the data further comprises:
identifying an object; and
extracting a data stream corresponding to the identified object.
8. The method of claim 7 , wherein the object is one of a pedestrian, animal, vehicle or cyclist.
9. The method of claim 7 , wherein the data includes dynamics of the extracted object.
10. The method of claim 1 , wherein the data is a data stream from one or more vehicle sensors.
11. The method of claim 1 , further comprising:
storing the extracted data and corresponding metadata in memory.
12. The method of claim 11 , wherein the metadata includes whether a collision occurred or not in the first scenario or the second scenario from which the data is extracted.
13. The method of claim 1 , further comprising:
transforming the extracted data or third scenario prior to injecting the extracted data into the third scenario.
14. The method of claim 1 , further comprising
automatically generating a plurality of augmented scenarios, each of the plurality of augmented scenarios created by injecting one or more of a plurality of data extracted from one or more recorded scenarios into one or more of a plurality of recorded scenarios; and
automatically simulating the automated vehicle driving software with the plurality of augmented scenarios.
15. The method of claim 1 , wherein extracting the data and injecting the extracted data are performed automatically.
16. A non-transitory computer-readable medium including instructions, which when executed by one or more processors, cause the one or more processors to perform a method of:
extracting data from a first scenario or a second scenario, the data corresponding to a difference between the first scenario recorded by one or more first vehicle sensors and the second scenario recorded by one or more second vehicle sensors, wherein the first scenario shares at least one characteristic with the second scenario;
injecting the extracted data into a third scenario recorded by one or more third vehicle sensors to generate a fourth scenario; and
simulating automated vehicle driving software with the fourth scenario.
17. A system comprising:
a first database comprising a plurality of recorded scenarios;
a second database comprising a plurality of extracted data; and
one or more processors coupled to the first and second databases and configured to:
automatically analyze the plurality of scenarios in the first database to identify scenarios sharing at least one characteristic;
extract data corresponding to one or more differences between the identified scenarios;
store the extracted data in the second database; and
generate a plurality of augmented scenarios based on the plurality of scenarios and the plurality of extracted data.
18. The system of claim 17 , the one or more processors further configured to:
simulate automated driving software with the plurality of augmented scenarios.