Generating autonomous vehicle simulation data from logged data
Logged data from an autonomous vehicle is processed to generate augmented data. The augmented data describes an actor in an environment of the autonomous vehicle, the actor having an associated actor type and an actor motion behavior characteristic. The augmented data may be varied to create different sets of augmented data. The sets of augmented data can be used to create one or more simulation scenarios that in turn are used to produce machine learning models to control the operation of autonomous vehicles.
1. A method for generating simulation data to be used in training a machine learning model for an autonomous vehicle, the method comprising:
receiving sensor data from one or more sensors of a vehicle;
generating augmented data from the sensor data, the augmented data describing an actor in an environment of the vehicle;
generating a variation of the augmented data; and
generating a simulation scenario, the simulation scenario generated using the variation of the augmented data.
2. The method of claim 1 , wherein generating the variation of the augmented data includes changing one or more actor characteristics associated with the actor.
3. The method of claim 1 , wherein the actor is associated with an actor characteristic, and the actor characteristic includes one from a group of actor velocity, actor type, actor size, actor geometric shape, actor reflectivity, actor color, actor path, lateral offset of motion of the actor, longitudinal offset of motion of the actor, and actor behavior response.
4. The method of claim 1 , wherein the generating a variation of the augmented data includes deleting the actor, adding an additional actor, or both.
5. The method of claim 1 , wherein generating the variation of the augmented data comprises:
generating a set of augmented data samples; and
wherein the generating the simulation scenario using the variation of the augmented data includes generating the simulation scenario from the set of augmented data samples.
6. The method of claim 1 , wherein generating the variation of the augmented data includes generating a plurality of sets of augmented data samples, and wherein generating the simulation scenario using the variation of the augmented data includes generating a plurality of simulation scenarios each one corresponding to one set of the plurality of sets of augmented data samples.
7. The method of claim 1 , further comprising:
executing a simulation based on the simulation scenario to generate a simulated output;
providing the simulation scenario as a training input to the machine learning model to generate a predicted output of the machine learning model; and
updating the machine learning model based on a difference between the predicted output and the simulated output of the simulation scenario.
8. The method of claim 1 , wherein the simulation scenario describes motion behavior of a simulated vehicle and at least one simulated actor.
9. The method of claim 1 , wherein the actor includes one from a group of another vehicle, a bicycle, a scooter, traffic, a pedestrian, an animal, machinery, debris, a static object, and a non-human moveable object.
10. A system comprising one or more processors and memory operably coupled with the one or more processors, wherein the memory stores instructions that, in response to execution of the instructions by the one or more processors, cause the one or more processors to perform operations comprising:
receiving sensor data from a sensor of a vehicle;
generating augmented data from the sensor data, the augmented data describing an actor in an environment of the vehicle;
generating a variation of the augmented data; and
generating a simulation scenario, the simulation scenario generated using the variation of the augmented data.
11. The system of claim 10 , wherein generating the variation of the augmented data includes changing one or more actor characteristics associated with the actor.
12. The system of claim 10 , wherein the actor is associated with an actor characteristic, and the actor characteristic includes one from a group of actor velocity, actor type, actor size, actor geometric shape, actor reflectivity, actor color, actor path, lateral offset of motion of the actor, longitudinal offset of motion, and actor behavior response.
13. The system of claim 10 , wherein the generating the variation of the augmented data includes deleting the actor, adding an additional actor, or both.
14. The system of claim 10 , wherein generating the variation of the augmented data comprises:
generating a set of augmented data samples; and
wherein the generating the simulation scenario using the variation of the augmented data includes generating the simulation scenario from the set of augmented data samples.
15. The system of claim 10 , wherein generating the variation of the augmented data includes generating a plurality of sets of augmented data samples, and wherein generating the simulation scenario using the variation of the augmented data includes generating a plurality of simulation scenarios each one corresponding to one set of the plurality of sets of augmented data samples.
16. The system of claim 10 , wherein the operations further comprise:
executing a simulation based on the simulation scenario to generate a simulated output;
providing the simulation scenario as a training input to a machine learning model to generate a predicted output of the machine learning model; and
updating one or more weights in the machine learning model based on a difference between the predicted output and the simulated output of the simulation scenario.
17. The system of claim 10 , wherein the actor includes one from a group of another vehicle, a bicycle, a scooter, traffic, a pedestrian, an animal, machinery, a static object, and a non-human moveable object.
18. The system of claim 10 , wherein the simulation scenario describes motion behavior of a simulated vehicle and at least one simulated actor.
19. A non-transitory computer readable storage medium storing computer instructions executable by one or more processors to perform a method of generating simulation data for an autonomous vehicle, the method comprising:
receiving sensor data from one or more sensors of a vehicle;
generating augmented data from the sensor data, the augmented data describing an actor in an environment of the vehicle;
generating a variation of the augmented data; and
generating a simulation scenario, the simulation scenario generated using the variation of the augmented data.
20. The non-transitory computer readable storage medium of claim 19 , wherein generating the variation of the augmented data comprises:
generating a set of augmented data samples; and
wherein the generating the simulation scenario using the variation of the augmented data includes generating the simulation scenario from the set of augmented data samples.