Simulated driving error models

1. A system comprising:

one or more processors; and

one or more computer-readable media storing computer-executable instructions that, when executed, cause the one or more processors to perform operations comprising:

receiving log data associated with an autonomous vehicle traversing an environment;

receiving ground truth data associated with the environment

determining, based at least in part on a signal from a planning system of the autonomous vehicle or a driving corridor, that an object represented in at least one of the log data or ground truth data is a relevant object;

determining, for the relevant object, a difference between a ground truth attribute associated with the ground truth data and an attribute from the log data;

determining, based at least in part on the difference, an error model;

determining, based at least in part on the error model, a response of an autonomous vehicle controller to a simulation scenario,

wherein determining the response of the autonomous vehicle controller to the simulation scenario comprises instantiating, based at least in part on the object being a relevant object, a simulated object corresponding to the object in the simulation scenario; and

transmitting, based at least in part on the response of the autonomous vehicle controller to the simulation scenario, the autonomous vehicle controller to the autonomous vehicle to control the autonomous vehicle.

2. The system of claim 1 , wherein the attribute of the relevant object comprises at least one of

a length;

a width;

a height;

a velocity;

a trajectory;

a location;

an orientation;

a classification;

a classification confidence level;

segmentation information;

time as false positive; or

time as false negative.

3. The system of claim 1 , the operations further comprising:

associating the difference with at least one of:

a distance between the autonomous vehicle and the relevant object;

a classification;

a planner interaction type,

a weather condition;

a lighting condition; or

a level of occlusion of the relevant object.

4. The system of claim 1 , wherein determining the difference between the ground truth attribute and the attribute further comprises:

receiving, from a perception system of the autonomous vehicle, perception data associated with the attribute from the log data; and

determining the attribute from the log data based on the perception data.

5. The system of claim 1 , wherein determining the response of the autonomous vehicle controller to the simulation scenario further comprises:

perturbing an attribute of the simulated object based at least in part on the error model.

6. The system of claim 1 , wherein the object is a first object, the operations further comprising:

determining, based at least in part on a signal from a planning system of the autonomous vehicle or a driving corridor, that a second object represented in at least one of the log data or ground truth data is an irrelevant object; and

wherein instantiating the simulated object corresponding to the object in the simulation scenario comprises instantiating, based at least in part on the first object being the relevant object and the second object being the irrelevant object, the simulated object corresponding to the first object in the simulation scenario.

7. A method comprising:

receiving log data associated with a vehicle traversing an environment;

receiving ground truth data associated with the environment;

determining a correspondence between the ground truth data and the log data;

determining, based at least in part on at least one of the log data or the ground truth data, a relevant object in the environment;

determining, based at least in part on the correspondence and the relevant object, an error model;

determining, based at least in part on the error model, a response of an autonomous vehicle controller to a simulation scenario,

wherein determining the response of the autonomous vehicle controller to the simulation scenario comprises instantiating a simulated object corresponding to the relevant object in the simulation scenario; and

transmitting, based at least in part on the response of the autonomous vehicle controller to the simulation scenario, the autonomous vehicle controller to the vehicle to control the vehicle.

8. The method of claim 7 , wherein determining the relevant object further comprises at least one of:

receiving, from a planning system of the vehicle, a signal identifying an object as the relevant object; or

determining that an object is located within a driving corridor determined by the vehicle.

9. The method of claim 7 , further comprising:

associating the correspondence with at least:

a distance between the vehicle and the relevant object;

a classification;

a planner interaction type,

a weather condition;

a lighting condition; or

a level of occlusion of the relevant object; and

wherein the error model indicates, for a particular distance between the vehicle and the relevant object, the classification, the planner interaction type, the weather condition, the lighting condition, or the level of occlusion of the relevant object, a variance of an attribute of the relevant object.

10. The method of claim 7 , further comprising:

receiving, from a perception system of the vehicle, perception data associated with an attribute of the relevant object; and

determining the attribute based on the perception data.

11. The method of claim 7 , wherein the correspondence comprises an attribute of an object wherein the attribute comprises at least one of

a length;

a width;

a height;

a velocity;

a trajectory;

a location;

an orientation;

a classification;

a classification confidence level;

segmentation information;

time as false positive; or

time as false negative.

12. The method of claim 7 , wherein determining the response of the autonomous vehicle controller to the simulation scenario further comprises:

perturbing an attribute of the simulated object based at least in part on the error model.

13. The method of claim 12 , wherein perturbing the attribute comprises updating the attribute based at least in part on a distribution associated with the error model.

14. The method of claim 7 , wherein determining the error model further comprising:

determining a first object represented in an attribute of the ground truth data;

determining a second object represented in an attribute of the log data;

determining, as the correspondence, a difference between the first object and the second object; and

determining, based at least in part on the correspondence, the error model.

15. The method of claim 7 , wherein determining the correspondence further comprises:

determining, as a false positive, when an object is only represented in the log data; or

determining, as a false negative, when an object is only represented in the ground truth data.

16. One or more non-transitory computer-readable media storing instructions executable by a processor, wherein the instructions, when executed, cause the processor to perform operations comprising:

receiving log data associated with a vehicle traversing an environment;

receiving ground truth data associated with the environment;

determining a correspondence between the ground truth data and the log data;

determining, based at least in part on at least one of the log data or the ground truth data, a relevant object in the environment;

determining, based at least in part on the correspondence and the relevant object, an error model;

determining, based at least in part on the error model, a response of an autonomous vehicle controller to a simulation scenario,

wherein determining the response of the autonomous vehicle controller to the simulation scenario comprises instantiating a simulated object corresponding to the relevant object in the simulation scenario; and

transmitting, based at least in part on the response of the autonomous vehicle controller to the simulation scenario, the autonomous vehicle controller to the vehicle to control the vehicle.

17. The one or more non-transitory computer-readable media of claim 16 , wherein determining the relevant object further comprises at least one of:

receiving, from a planning system of the vehicle, a signal identifying an object as the relevant object; or

determining that an object is located within a driving corridor determined by the vehicle.

18. The one or more non-transitory computer-readable media of claim 16 , wherein determining the response of the autonomous vehicle controller to the simulation scenario further comprises:

perturbing an attribute of the simulated object based at least in part on the error model.

19. The one or more non-transitory computer-readable media of claim 16 wherein determining the error model further comprises:

determining a first object represented in an attribute of the ground truth data;

determining a second object represented in an attribute of the log data;

determining, as the correspondence, a difference between the first object and the second object; and

determining, based at least in part on the correspondence, the error model.

20. The one or more non-transitory computer-readable media of claim 16 , wherein determining the correspondence further comprises:

determining, as a false positive, when an object is only represented in the log data; or

determining, as a false negative, when an object is only represented in the ground truth data.