Training and testing automated driving models
Methods and systems for deploying an automated driving model include testing an automated driving model with a set of scenarios that include a single obstacle. The set of scenarios is pruned to include only those scenarios that do not result in a collision, but that may result in a collision with the addition of another obstacle. New scenarios are added to the set of scenarios to include scenarios that have an additional obstacle to form an updated set of scenarios. The automated driving model is tested with the updated set of scenarios. The automated driving model is deployed to a vehicle to control operation of one or more systems in the vehicle.
1. A computer-implemented method for deploying an automated driving model, comprising:
testing an automated driving model with a set of scenarios that include a single obstacle;
pruning the set of scenarios to include only those scenarios that do not result in a collision, but that may result in a collision with the addition of another obstacle;
adding new scenarios to the set of scenarios to include scenarios that have an additional obstacle to form an updated set of scenarios;
testing the automated driving model with the updated set of scenarios; and
deploying the automated driving model to a vehicle to control operation of one or more systems in the vehicle.
2. The computer-implemented method of claim 1 , further comprising repeating said steps of pruning, adding, and testing with the updated set of scenarios until a number of obstacles reaches a target number of obstacles.
3. The computer-implemented method of claim 1 , wherein pruning the set of scenarios comprises removing any scenarios that had an outcome where the vehicle is always safe and any scenarios that had an outcome where the vehicle collides with an obstacle.
4. The computer-implemented method of claim 1 , wherein each scenario has one or more attributes selected from the group consisting of the obstacle traveling out of the scenario boundaries, sensing the obstacle by the vehicle, the obstacle colliding with the vehicle, the obstacle moving into the vehicle's path, and the vehicle reaching a scenario goal position.
5. The computer-implemented method of claim 1 , wherein testing the automated driving model comprises determining whether a vehicle that is operated by the automated driving model can reach a scenario goal position within a scenario timeframe and without colliding with an obstacle.
6. The computer-implemented method of claim 5 , wherein testing the automated driving model comprises extrapolating obstacle positions based on obstacle starting positions and one or more obstacle parameters.
7. The computer-implemented method of claim 5 , wherein testing the automated driving model comprises determining a vehicle's possible range of paths as determined by one or more vehicle characteristics.
8. The computer-implemented method of claim 5 , wherein testing the automated driving model comprises limiting an ability of the vehicle to detect obstacles to only those objects that are within a sensing distance.
9. The computer-implemented method of claim 5 , wherein determining the vehicle that is operated by the automated driving model can reach the scenario goal position comprises using the automated driving model and simulated sensor information to direct motions for the vehicle within a scenario boundary.
10. A non-transitory computer readable storage medium comprising a computer readable program for deploying an automated driving model, wherein the computer readable program when executed on a computer causes the computer to perform the steps of:
testing an automated driving model with a set of scenarios that include a single obstacle;
pruning the set of scenarios to include only those scenarios that do not result in a collision, but that may result in a collision with the addition of another obstacle;
adding new scenarios to the set of scenarios to include scenarios that have an additional obstacle to form an updated set of scenarios;
testing the automated driving model with the updated set of scenarios; and
deploying the automated driving model to a vehicle to control operation of one or more systems in the vehicle.
11. The computer readable storage medium of claim 10 , further comprising repeating said steps of pruning, adding, and testing with the updated set of scenarios until a number of obstacles reaches a target number of obstacles.
12. A system for deploying an automated driving model, comprising:
a testing module configured to test an automated driving model with a set of scenarios that include a single obstacle;
a pruning module configured to prune the set of scenarios to include only those scenarios that do not result in a collision, but that may result in a collision with the addition of another obstacle, and to add new scenarios to the set of scenarios to include scenarios that have an additional obstacle to form an updated set of scenarios, wherein the testing module is further configured to test the automated driving model with the updated set of scenarios; and
a network adapter configured to deploy the automated driving model to a vehicle to control operation of one or more systems in the vehicle.
13. The system of claim 12 , wherein the testing and pruning module are configured to repeat the pruning, adding, and testing of updated scenarios until a number of obstacles reaches a target number of obstacles.
14. The system of claim 12 , wherein the pruning module is further configured to remove any scenarios that had an outcome where the vehicle is always safe and any scenarios that had an outcome where the vehicle collides with an obstacle.
15. The system of claim 12 , wherein each scenario has one or more attributes selected from the group consisting of the obstacle traveling out of the scenario boundaries, sensing the obstacle by the vehicle, the obstacle colliding with the vehicle, the obstacle moving into the vehicle's path, and the vehicle reaching a scenario goal position.
16. The system of claim 12 , wherein the testing module is further configured to determine whether a vehicle that is operated by the automated driving model can reach a scenario goal position within a scenario timeframe and without colliding with an obstacle.
17. The system of claim 16 , wherein the testing module is further configured to extrapolate obstacle positions based on obstacle starting positions and one or more obstacle parameters.
18. The system of claim 16 , wherein the testing module is further configured to determine a vehicle's possible range of paths as determined by one or more vehicle characteristics.
19. The system of claim 16 , wherein the testing module is further configured to limit an ability of the vehicle to detect obstacles to only those objects that are within a sensing distance.
20. The system of claim 16 , wherein the testing module is further configured to use the automated driving model and simulated sensor information to direct motions for the vehicle within a scenario boundary.