Operation of a vehicle using motion planning with machine learning
Techniques for operation of a vehicle using machine learning with motion planning include storing, using one or more processors of a vehicle located within an environment, a plurality of constraints for operating the vehicle within the environment. One or more sensors of the vehicle receive sensor data describing the environment. The one or more processors extract a feature vector from the stored plurality of constraints and the received sensor data. The feature vector includes a first feature describing an object located within the environment. A machine learning circuit of the vehicle is used to generate a first motion segment based on the feature vector. A number of violations of the stored plurality of constraints is below a threshold. The one or more processors operate the vehicle in accordance with the generated first motion segment.
1. A method comprising:
receiving, from one or more sensors of a vehicle, sensor data describing an environment of the vehicle;
extracting, using one or more processors of the vehicle, a feature vector from a plurality of constraints for operating the vehicle within the environment and the received sensor data;
determining, using the one or more processors, a hierarchical ranking of the plurality of constraints,
wherein a rank of each constraint of the plurality of constraints corresponds to a level of navigational safety of the vehicle, and
wherein determining the hierarchical ranking of the plurality of constraints comprises determining an order of the plurality of constraints in the hierarchical ranking using a machine learning circuit of the vehicle and the extracted feature vector;
generating, using the machine learning circuit, a first motion segment based on the feature vector and an operational metric for the first motion segment, wherein the operational metric is determined based on the hierarchical ranking of the plurality of constraints and a number of violations of the plurality of constraints associated with the first motion segment; and
causing, using the one or more processors, an operation of the vehicle in accordance with the generated first motion segment.
2. The method of claim 1 , wherein determining the hierarchical ranking of the plurality of constraints comprises:
determining that a first constraint of the plurality of constraints is related to the navigational safety of the vehicle,
determining that a second constraint of the plurality of constraints is unrelated to the navigational safety of the vehicle,
assigning a first rank to the first constraint, and
assigning a second rank to the second constraint, wherein the first rank is higher than the second rank.
3. The method of claim 1 , wherein machine learning circuit is trained based on training data comprising:
a plurality of additional motion segments, and
a plurality of additional constraints for operating the vehicle within the environment.
4. The method of claim 3 , wherein training the machine learning circuit comprises:
determining, for each of the plurality of additional motion segments, whether that additional motional segment violated one or more of the plurality of additional constraints.
5. The method of claim 4 , wherein training the machine learning circuit comprises:
responsive to determining that one of the plurality of additional motion segments violated a first subset of the plurality of additional constraints, determining a hierarchical ranking of the first subset of the plurality of additional constraints.
6. The method of claim 1 , wherein the machine learning circuit is configured to:
generate a plurality of motion segments based on the feature vector,
determine a respective a score for each of the plurality of motion segments, and
select the first motion segment from among the plurality of motion segments based on the scores.
7. The method of claim 6 , wherein, for each of the plurality of motion segments, the score is indicative of a likelihood that a number of violations of the plurality of constraints for that motion segment is lower than a number of violations of the plurality of constraints for each of the other motion segments.
8. A vehicle comprising:
one or more computer processors; and
one or more non-transitory storage media storing instructions which, when executed by the one or more computer processors, cause the one or more computer processors to:
receive sensor data describing an environment of a vehicle;
extract a feature vector from a plurality of constraints for operating the vehicle within the environment and the received sensor data;
determine a hierarchical ranking of the plurality of constraints,
wherein a rank of each constraint of the plurality of constraints corresponds to a level of navigational safety of the vehicle, and
wherein determining the hierarchical ranking of the plurality of constraints comprises determining an order of each constraint of the plurality of constraints in the hierarchical ranking using a machine learning circuit of the vehicle and the extracted feature vector;
generate, using the machine learning circuit, a first motion segment based on the feature vector and an operational metric for the first motion segment, wherein the operational metric is determined based on the hierarchical ranking of the plurality of constraints and a number of violations of the plurality of constraints associated with the first motion segment; and
causing the vehicle to operate in accordance with the generated first motion segment.
9. The vehicle of claim 8 , wherein determining the hierarchical ranking of the plurality of constraints comprises:
determining that a first constraint of the plurality of constraints is related to the navigational safety of the vehicle,
determining that a second constraint of the plurality of constraints is unrelated to the navigational safety of the vehicle,
assigning a first rank to the first constraint, and
assigning a second rank to the second constraint, wherein the first rank is higher than the second rank.
10. The vehicle of claim 8 , wherein machine learning circuit is trained based on training data comprising:
a plurality of additional motion segments, and
a plurality of additional constraints for operating the vehicle within the environment.
11. The vehicle of claim 10 , wherein training the machine learning circuit comprises:
determining, for each of the plurality of additional motion segments, whether that additional motional segment violated one or more of the plurality of additional constraints.
12. The vehicle of claim 11 , wherein training the machine learning circuit comprises:
responsive to determining that one of the plurality of additional motion segments violated a first subset of the plurality of additional constraints, determining a hierarchical ranking of the first subset of the plurality of additional constraints.
13. The vehicle of claim 8 , wherein the machine learning circuit is configured to:
generate a plurality of motion segments based on the feature vector,
determine a respective a score for each of the plurality of motion segments, and
select the first motion segment from among the plurality of motion segments based on the scores.
14. The vehicle of claim 13 , wherein, for each of the plurality of motion segments, the score is indicative of a likelihood that a number of violations of the plurality of constraints for that motion segment is lower than a number of violations of the plurality of constraints for each of the other motion segments.
15. One or more non-transitory storage media storing instructions which, when executed by one or more computing devices, cause the one or more computing devices to:
receive sensor data describing an environment of a vehicle;
extract a feature vector from a plurality of constraints for operating the vehicle within the environment and the received sensor data;
determine a hierarchical ranking of the plurality of constraints,
wherein a rank of each constraint of the plurality of constraints corresponds to a level of navigational safety of the vehicle, and
wherein determining the hierarchical ranking of the plurality of constraints comprises determining an order of the plurality of constraints in the hierarchical ranking using a machine learning circuit of the vehicle and the extracted feature vector;
generate, using the machine learning circuit, a first motion segment based on the feature vector and an operational metric for the first motion segment, wherein the operational metric is determined based on the hierarchical ranking of the plurality of constraints and a number of violations of the plurality of constraints associated with the first motion segment; and
causing the vehicle to operate in accordance with the generated first motion segment.
16. The one or more non-transitory storage media of claim 15 , wherein determining the hierarchical ranking of the plurality of constraints comprises:
determining that a first constraint of the plurality of constraints is related to the navigational safety of the vehicle,
determining that a second constraint of the plurality of constraints is unrelated to the navigational safety of the vehicle,
assigning a first rank to the first constraint, and
assigning a second rank to the second constraint, wherein the first rank is higher than the second rank.
17. The one or more non-transitory storage media of claim 15 , wherein machine learning circuit is trained based on training data comprising:
a plurality of additional motion segments, and
a plurality of additional constraints for operating the vehicle within the environment.
18. The one or more non-transitory storage media of claim 17 , wherein training the machine learning circuit comprises:
determining, for each of the plurality of additional motion segments, whether that additional motional segment violated one or more of the plurality of additional constraints.
19. The one or more non-transitory storage media of claim 18 , wherein training the machine learning circuit comprises:
responsive to determining that one of the plurality of additional motion segments violated a first subset of the plurality of additional constraints, determining a hierarchical ranking of the first subset of the plurality of additional constraints.
20. The one or more non-transitory storage media of claim 15 , wherein the machine learning circuit is configured to:
generate a plurality of motion segments based on the feature vector,
determine a respective a score for each of the plurality of motion segments, and
select the first motion segment from among the plurality of motion segments based on the scores.
21. The one or more non-transitory storage media of claim 20 , wherein, for each of the plurality of motion segments, the score is indicative of a likelihood that a number of violations of the plurality of constraints for that motion segment is lower than a number of violations of the plurality of constraints for each of the other motion segments.