Autonomous vehicle operation feature monitoring and evaluation of effectiveness
Methods and systems for monitoring use and determining risks associated with operation of a vehicle having one or more autonomous operation features are provided. According to certain aspects, operating data may be recorded during operation of the vehicle. This may include information regarding the vehicle, the vehicle environment, use of the autonomous operation features, and/or control decisions made by the features. The control decisions may include actions the feature would have taken to control the vehicle, but which were not taken because a vehicle operator was controlling the relevant aspect of vehicle operation at the time. The operating data may be recorded in a log, which may then be used to determine risk levels associated with vehicle operation based upon risk levels associated with the autonomous operation features. The risk levels may further be used to adjust an insurance policy associated with the vehicle.
1. A computer system for evaluating operation of a vehicle having an autonomous system, comprising:
one or more processors; and
a non-transitory program memory coupled to the one or more processors and storing executable instructions that, when executed by the one or more processors, cause the computer system to:
receive an indication of occurrence of a vehicle collision involving the vehicle;
receive sensor data from one or more sensors within the vehicle indicating (i) one or more environmental conditions in which the vehicle collision occurred, (ii) a person positioned within the vehicle to operate the vehicle at the time of the vehicle collision, and (iii) one or more capabilities or features of the autonomous system;
determine one or more preferred control decisions the autonomous system could have made to control the vehicle to reduce a risk of collision or mitigate an effect of the vehicle collision immediately before or during the vehicle collision based upon analysis of the sensor data using a trained machine learning program that has been previously trained to predict preferred control decisions under a plurality of operating conditions by identifying patterns in a training data set comprising sample sensor data and sample control data;
receive control decision data indicating one or more actual control decisions the autonomous system made to control the vehicle immediately before or during the vehicle collision;
determine a degree of similarity between the one or more preferred control decisions and the one or more actual control decisions; and
assign a degree of fault for the vehicle collision to the autonomous system based upon the determined degree of similarity between the one or more preferred control decisions and the one or more actual control decisions.
2. The computer system of claim 1 , wherein the one or more preferred control decisions and the one or more actual control decisions are virtually time-stamped for comparison of such controlled and actual control decisions based upon matching virtual time stamps.
3. The computer system of claim 1 , wherein the executable instructions further cause the computer system to train the machine learning program to determine control decisions that could be preferably made by autonomous systems to reduce risks of collisions or mitigate effects of collisions based upon data related to capabilities of the autonomous systems.
4. The computer system of claim 3 , wherein the executable instructions further cause the computer system to train the machine learning program to determine the control decisions that could be preferably made by the autonomous systems based upon: environmental conditions, road conditions, construction conditions, or traffic conditions.
5. The computer system of claim 3 , wherein the executable instructions further cause the computer system to train the machine learning program to determine the control decisions that could be preferably made by the autonomous systems based upon: levels of pedestrian traffic.
6. The computer system of claim 1 , wherein the sensor data indicating (i) one or more environmental conditions in which the vehicle collision occurred or (ii) the person positioned within the vehicle to operate the vehicle at the time of the vehicle collision includes at least one of camera image data, radar unit data, or infrared data.
7. The computer system of claim 1 , wherein the executable instructions further cause the computer system to adjust a risk level or model parameter associated with the autonomous vehicle or the autonomous system based upon the one or more actual control decisions made by the autonomous system.
8. A tangible, non-transitory computer-readable medium storing executable instructions for evaluating operation of a vehicle having an autonomous system that, when executed by at least one processor of a computer system, cause the computer system to:
receive an indication of occurrence of a vehicle collision involving the vehicle;
receive sensor data from one or more sensors within the vehicle indicating (i) one or more environmental conditions in which the vehicle collision occurred, (ii) a person positioned within the vehicle to operate the vehicle at the time of the vehicle collision, and (iii) one or more capabilities or features of the autonomous system;
determine one or more preferred control decisions the autonomous system could have made to control the vehicle to reduce a risk of collision or mitigate an effect of the vehicle collision immediately before or during the vehicle collision based upon analysis of the sensor data using a trained machine learning program that has been previously trained to predict preferred control decisions under a plurality of operating conditions by identifying patterns in a training data set comprising sample sensor data and sample control data;
receive control decision data indicating one or more actual control decisions the autonomous system made to control the vehicle immediately before or during the vehicle collision;
determine a degree of similarity between the one or more preferred control decisions and the one or more actual control decisions; and
assign a degree of fault for the vehicle collision to the autonomous system based upon the determined degree of similarity between the one or more preferred control decisions and the one or more actual control decisions.
9. The tangible, non-transitory computer-readable medium of claim 8 , wherein the one or more preferred control decisions and the one or more actual control decisions are virtually time-stamped for comparison of such controlled and actual control decisions based upon matching virtual time stamps.
10. The tangible, non-transitory computer-readable medium of claim 8 , wherein the executable instructions further cause the computer system to train the machine learning program to determine control decisions that could be preferably made to reduce risks of collisions or mitigate effects of collisions by autonomous systems based upon data related to capabilities of the autonomous systems.
11. The tangible, non-transitory computer-readable medium of claim 10 , wherein the executable instructions further cause the computer system to train the machine learning program to determine the control decisions that could be preferably made by the autonomous systems based upon: environmental conditions, road conditions, construction conditions, or traffic conditions.
12. The tangible, non-transitory computer-readable medium of claim 10 , wherein the executable instructions further cause the computer system to train the machine learning program to determine the control decisions that could be preferably made by the autonomous systems based upon: levels of pedestrian traffic.
13. The tangible, non-transitory computer-readable medium of claim 8 , wherein the executable instructions further cause the computer system to adjust a risk level or model parameter associated with the autonomous vehicle or the autonomous system based upon the one or more actual control decisions made by the autonomous system.
14. A computer-implemented method of evaluating operation of a vehicle having an autonomous system, the method comprising:
receiving, at one or more processors, an indication of occurrence of a vehicle collision involving the vehicle;
receiving, at the one or more processors, sensor data from one or more sensors within the vehicle indicating (i) one or more environmental conditions in which the vehicle collision occurred, (ii) a person positioned within the vehicle to operate the vehicle at the time of the vehicle collision, and (iii) one or more capabilities or features of the autonomous system;
determining, by the one or more processors, one or more preferred control decisions the autonomous system could have made to control the vehicle to reduce a risk of collision or mitigate an effect of the vehicle collision immediately before or during the vehicle collision based upon analysis of the sensor data using a trained machine learning program that has been previously trained to predict preferred control decisions under a plurality of operating conditions by identifying patterns in a training data set comprising sample sensor data and sample control data;
receiving, by the one or more processors, control decision data indicating one or more actual control decisions the autonomous system made to control the vehicle immediately before or during the vehicle collision;
determining, by the one or more processors, a degree of similarity between the one or more preferred control decisions and the one or more actual control decisions; and
assigning, by the one or more processors, a degree of fault for the vehicle collision to the autonomous system based upon the determined degree of similarity between the one or more preferred control decisions and the one or more actual control decisions.
15. The computer-implemented method of claim 14 , wherein the one or more preferred control decisions and the one or more actual control decisions are virtually time-stamped for comparison of such controlled and actual control decisions based upon matching virtual time stamps.
16. The computer-implemented method of claim 14 , further comprising:
training, by the one or more processors, the machine learning program to determine control decisions that could be preferably made to reduce risks of collisions or mitigate effects of collisions by autonomous systems based upon data related to capabilities of the autonomous systems.
17. The computer-implemented method of claim 16 , wherein training the machine learning program further comprises training the machine learning program to determine the control decisions that could be preferably made by the autonomous systems based upon: environmental conditions, road conditions, construction conditions, or traffic conditions.
18. The computer-implemented method of claim 16 , wherein training the machine learning program further comprises training the machine learning program to determine the control decisions that could be preferably made by the autonomous systems based upon: levels of pedestrian traffic.
19. The computer-implemented method of claim 14 , wherein the one or more actual control decisions include (i) a control decision to accelerate or to decelerate, (ii) an indication of a rate of acceleration or deceleration, or (iii) a control decision to change lanes or to turn the autonomous vehicle.
20. The computer-implemented method of claim 14 , further comprising:
adjusting, by the one or more processors, a risk level or model parameter associated with the autonomous vehicle or the autonomous system based upon the one or more actual control decisions made by the autonomous system.