Assessing ride quality for autonomous vehicles
The disclosure relates to training models for identification of events likely to cause discomfort to passengers of autonomous vehicles and for assessment of overall ride quality of autonomous vehicle rides. For instance, ride data may be associated with a ride quality value indicative of a level of discomfort and/or a first overall ride quality value indicating an overall ride quality provided by the passenger for the first ride. This ride data may be used to train a model such that the model is configured to, in response to receiving ride data for a second ride as input, output a list of events likely to cause discomfort to a passenger during the second ride and/or such that the model is configured to, in response to receiving second ride data for a second ride as input, output a second overall ride quality value for the second ride.
1. A method of training a model for identification of events likely to cause discomfort to passengers of autonomous vehicles, the method comprising:
receiving, by one or more server computing devices, first ride data identifying a first output from a planner system and a second output from a perception system, the first ride data being associated with a ride quality value indicating a level of discomfort identified by a passenger of an autonomous vehicle during a first ride, wherein the first ride data further includes information about locations and types of road users other than the autonomous vehicle; and
training, by the one or more server computing devices, the model using the first ride data and any associated ride quality values such that the model is configured to, in response to receiving second ride data for a second ride as input, output a list of events that occur in the second ride that are likely to cause discomfort to one or more passengers during the second ride.
2. The method of claim 1 , further comprising;
inputting the second ride data for the second ride into the model in order to determine the list of events; and
flagging the second ride data for review based on the list of events.
3. The method of claim 1 , wherein the second ride data corresponds to simulated ride data for a simulated ride generated using autonomous vehicle control software.
4. The method of claim 1 , wherein the ride quality value is further associated with a point in time during the first ride.
5. The method of claim 1 , wherein the first ride data further includes context from a map used to control the autonomous vehicle during the first ride.
6. The method of claim 1 , wherein the first ride data further includes acceleration information for the first ride.
7. The method of claim 6 , wherein the first ride data further includes derivative of acceleration information for the autonomous vehicle for the first ride.
8. A method of training a model for assessment of overall ride quality of autonomous vehicle rides, the method comprising:
receiving, by one or more server computing devices, first ride data identifying a first output from a planner system and a second output from, a perception system, the first ride data being associated with a first overall ride quality value indicating an overall ride quality provided by a passenger of an autonomous vehicle for a first ride, wherein the first ride data further includes information about locations and types, of road users other than the autonomous vehicle; and
training, by the one or more server computing devices, the model using the first ride data and first overall ride quality value such that the model is configured to, in response to receiving second ride data for a second ride as input, output a second overall ride quality value for the second ride.
9. The method of claim 8 , wherein the second ride data corresponds to simulated ride data for a simulated ride generated using autonomous vehicle control software.
10. The method of claim 8 , wherein the first ride data further includes context from a map used to control the autonomous vehicle during the first ride.
11. The method of claim 8 , wherein the first ride data further includes acceleration information for the autonomous vehicle.
12. The method of claim 8 , further comprising receiving, output from a second model identifying a list of events likely to cause discomfort to a passenger during the first ride, and wherein the training is further based on the list of events.
13. The method of claim 8 , farther comprising converting a first ride quality value to the first overall ride quality value, and the first ride quality value corresponds to a star rating for the first ride provided by the passenger.
14. The method of claim 8 , wherein the model is a convolutional neural net over time in order to enable the model recognize relationships between events and when those events occur during a ride and how those relationships affect overall ride quality.
15. The method of claim 8 , further comprising:
inputting the second ride data for the second ride into the model in order to determine the second overall ride quality value; and
flagging the second ride data for review based he second overallride quality value.
16. The method of claim 8 , further comprising:
inputting the second ride data for the second ride into the model in order to determine the second overall ride quality value; and
using the second overall ride quality value to assess ride quality of a plurality of miles driven by autonomous vehicles.
17. The method of claim 8 , further comprising:
inputting the second ride data for the second ride into the model in order to determine the second overall ride quality value; and
using the second overall ride quality value to assess ride quality of a certaintype of maneuver.
18. The method of claim 17 , wherein the certain type of maneuver includes left or right turns.
19. The method of claim 17 , wherein the certain type of maneuver includes lane changes.