Forward modeling for behavior control of autonomous vehicles

1. A method for controlling an autonomous vehicle, comprising:

receiving, by a controller of the autonomous vehicle, a sensor input from a plurality of sensors of the autonomous vehicle;

determining a plurality of possible planned movements of the autonomous vehicle in a future using a plurality of autonomous driving techniques and the sensor input from the plurality of sensors, wherein the plurality of autonomous driving techniques includes a rule-based model and a machine-learning model, the plurality of possible planned movements includes a first possible planned movement determined using the rule-based model, and the plurality of possible planned movements includes a second possible planned movement determined using the machine-learning model;

grading each of the plurality of possible planned movements to obtain a plurality of scores each corresponding to one of the plurality of possible planned movements, wherein the plurality of scores includes a highest score, the plurality of scores includes a first score corresponding to the first possible planned movement determined using the rule-based model, the plurality of scores includes a second score corresponding to the second possible planned movement determined using the machine-learning model, and grading each of the plurality of possible planned movements includes:

grading the first possible planned movement that was determined using the rule-based model to obtain the first score;

grading the second possible planned movement that was determined using the machine-learning model to obtain the second score;

comparing the first score of the first possible planned movement that was determined using the rule-base model with the second score of the second possible planned movement that was determined using the machine-learning model to determine which of the plurality of autonomous driving techniques results in the highest score;

selecting one of the plurality of possible planned movements that corresponds with the highest score of the plurality of scores;

determining a predicted movement of at least one other vehicle based on the selected one of the plurality of possible planned movements;

determining a plurality of possible reactive movements of the autonomous vehicle in the future based on the predicted movement of the at least one other vehicle;

modifying the plurality of possible planned movements to include the plurality of possible reactive movements to obtain a plurality of modified planned movements in the future;

regrading each of the plurality of modified planned movements to obtain a plurality of updated scores each corresponding to one of the plurality of modified planned movements, wherein the plurality of updated scores includes a highest updated score;

selecting one of the plurality of modified planned movements that corresponds with the highest updated score of the plurality of scores; and

commanding, by the controller, the autonomous vehicle to move according to the selected one of the plurality of modified planned movements with the highest updated score.

2. The method of claim 1 , wherein the machine-learning model is a classification and tree regression (CART) model.

3. The method of claim 1 , wherein grading each of the possible planned movements includes determining a speed of the autonomous vehicle for each of the plurality of possible planned movements, a distance from the autonomous vehicle to another object for each of the plurality of possible planned movements, a presence of a stop sign for each of the plurality of possible planned movements, a distance from the autonomous vehicle to the stop sign for each of the plurality of possible planned movements, a presence of a pedestrian for each of the plurality of possible planned movements, and a distance from the autonomous vehicle to the pedestrian for each of the plurality of possible planned movements.

4. The method of claim 3 , wherein grading each of the possible planned movements includes assigning a partial score to each of the speed of the autonomous vehicle for each of the plurality of possible planned movements, the distance from the autonomous vehicle to another object for each of the plurality of possible planned movements, the presence of the stop sign for each of the plurality of possible planned movements, the distance from the autonomous vehicle to the stop sign for each of the plurality of possible planned movements, the presence of a pedestrian for each of the plurality of possible planned movements, and the distance from the autonomous vehicle to the pedestrian for each of the plurality of possible planned movements in order to obtain a plurality of partial movement scores for each of the plurality of possible planned movements.

5. The method of claim 4 , wherein each of the plurality of scores is equal to a sum of all of the plurality of partial movement scores.

6. The method of claim 5 , wherein regrading each of the plurality of possible planned movements includes determining an updated speed of the autonomous vehicle for each of the plurality of modified planned movements, an updated distance from the autonomous vehicle to another object for each of the plurality of modified planned movements, an updated presence of a stop sign for each of the plurality of modified planned movements, an updated distance from the autonomous vehicle to the stop sign for each of the plurality of modified planned movements, an updated presence of a pedestrian for each of the plurality of modified planned movements, and an updated distance from the autonomous vehicle to the pedestrian for each of the plurality of modified planned movements.

7. The method of claim 6 , wherein regrading each of the possible modified movements includes assigning an updated partial score to each of the updated speed of the autonomous vehicle for each of the plurality of modified planned movements, the updated distance from the autonomous vehicle to another object for each of the plurality of modified planned movements, the updated presence of the stop sign for each of the plurality of modified planned movements, the updated distance from the autonomous vehicle to the stop sign for each of the plurality of modified planned movements, the updated presence of a pedestrian for each of the plurality of modified planned movements, and the updated distance from the autonomous vehicle to the pedestrian for each of the plurality of modified planned movements in order to obtain a plurality of partial modified scores for each of the plurality of modified planned movements.

8. The method of claim 1 , wherein the future is 3.5 seconds ahead of a current time.

9. The method of claim 1 , wherein the predicted movement of at least one other vehicle is determined using a kinematic prediction model.

10. A control system for an autonomous vehicle, comprising:

a plurality of sensors;

a controller in communication with the plurality of sensors, wherein the controller is programmed to:

receive input from a plurality of sensors of the autonomous vehicle;

determine a plurality of possible planned movements of the autonomous vehicle in a future using a plurality of autonomous driving techniques and input from the plurality of sensors, wherein the plurality of autonomous driving techniques includes a rule-based model and a machine-learning model, the plurality of possible planned movements includes a first possible planned movement determined using the rule-based model, and the plurality of possible planned movements includes a second possible planned movement determined using the machine-learning model;

grade each of the plurality of possible planned movements to obtain a plurality of scores each corresponding to one of the plurality of possible planned movements, wherein the plurality of scores includes a highest score, the plurality of scores includes a first score corresponding to the first possible planned movement determined using the rule-based model, the plurality of scores includes a second score corresponding to the second possible planned movement determined using the machine-learning model, and the controller is programmed to:

grade the first possible planned movement that was determined using the rule-based model to obtain the first score;

grade the second possible planned movement that was determined using the machine-learning model to obtain the second score;

compare the first score of the first possible planned movement that was determined using the rule-base model with the second score of the second possible planned movement that was determined using the machine-learning model to determine which of the plurality of autonomous driving techniques results in the highest score;

select one of the plurality of the possible planned movements that corresponds with the highest score of the plurality of scores;

determine a predicted movement of at least one other vehicle based on the selected one of the plurality of possible planned movements;

determine a plurality of possible reactive movements of the autonomous vehicle in the future based on the predicted movement of the at least one other vehicle;

modify the plurality of possible planned movements to include the plurality of possible reactive movements to obtain a plurality of modified planned movements in the future;

regrade each of the plurality of modified planned movements to obtain a plurality of updated scores each corresponding to one of the plurality of modified planned movements, wherein the plurality of updated scores includes a highest updated score; and

select one of the plurality of modified planned movements that corresponds with the highest updated score of the plurality of scores; and

command the autonomous vehicle to move according to the selected one of the plurality of modified planned movements with the highest updated score.

11. The control system of claim 10 , wherein machine-learning model is a classification and tree regression (CART) model.

12. The control system of claim 10 , wherein the controller grades each of the possible planned movements by determining a speed of the autonomous vehicle for each of the plurality of possible planned movements, a distance from the autonomous vehicle to another object for each of the plurality of possible planned movements, a presence of a stop sign for each of the plurality of possible planned movements, a distance from the autonomous vehicle to the stop sign for each of the plurality of possible planned movements, a presence of a pedestrian for each of the plurality of possible planned movements, and a distance from the autonomous vehicle to the pedestrian for each of the plurality of possible planned movements.

13. The control system of claim 12 , wherein the controller grades each of the possible planned movements by assigning a partial score to each of the speed of the autonomous vehicle for each of the plurality of possible planned movements, the distance from the autonomous vehicle to another object for each of the plurality of possible planned movements, the presence of the stop sign for each of the plurality of possible planned movements, the distance from the autonomous vehicle to the stop sign for each of the plurality of possible planned movements, the presence of a pedestrian for each of the plurality of possible planned movements, and the distance from the autonomous vehicle to the pedestrian for each of the plurality of possible planned movements in order to obtain a plurality of partial movement scores for each of the plurality of possible planned movements.

14. The control system of claim 13 , wherein each of the plurality of scores is equal to a sum of all of the plurality of partial movement scores.

15. The control system of claim 14 , wherein regrading each of the plurality of possible planned movements includes determining an updated speed of the autonomous vehicle for each of the plurality of modified planned movements, an updated distance from the autonomous vehicle to another object for each of the plurality of modified planned movements, an updated presence of a stop sign for each of the plurality of modified planned movements, an updated distance from the autonomous vehicle to the stop sign for each of the plurality of modified planned movements, an updated presence of a pedestrian for each of the plurality of modified planned movements, and an updated distance from the autonomous vehicle to the pedestrian for each of the plurality of modified planned movements.

16. The control system of claim 15 , wherein regrading each of the possible modified movements includes assigning an updated partial score to each of the updated speed of the autonomous vehicle for each of the plurality of modified planned movements, the updated distance from the autonomous vehicle to another object for each of the plurality of modified planned movements, the updated presence of the stop sign for each of the plurality of modified planned movements, the updated distance from the autonomous vehicle to the stop sign for each of the plurality of modified planned movements, the updated presence of a pedestrian for each of the plurality of modified planned movements, and the updated distance from the autonomous vehicle to the pedestrian for each of the plurality of modified planned movements in order to obtain a plurality of partial modified scores for each of the plurality of modified planned movements.

17. The control system of claim 10 , wherein the future is 3.5 seconds ahead of a current time.

18. The control system of claim 10 , wherein the predicted movement of at least one other vehicle is determined using a kinematic prediction model.