Granted Patent B2

US 12,353,997 · App. 17/434,855 · Granted Jul 8, 2025

Inventors: Yuka Ariki (Tokyo, JP); Takuya Narihiri (Tokyo, JP)

Assignee: SONY GROUP CORPORATION

G06N3/08G06F16/90335G01C21/3461

Abstract

Claims (93)

1. An information processing apparatus, comprises:

a learning unit configured to:

receive an environment map associated with a position of a search subject, wherein the position is in an environment associated with the search subject;

receive an internal state of the search subject, wherein the learning unit includes:

a first convolutional neural network configured to output a first feature value based on the environment map; and

a second convolutional neural network configured to output a second feature value based on the internal state of the search subject:

concatenate the first feature value and the second feature value;

determine a concatenated value based on the concatenation of the first feature value and the second feature value;

determine a loss associated with the concatenated value;

perform a learning process based on the determined loss;

learn a heuristic function based on the learning process, wherein the heuristic function is associated with a path search; and

output a heuristic map, wherein

the heuristic function is represented as a two-dimensional image or an image higher than the two-dimensional image,

the internal state of the search subject includes a first element, and

the first element has a degree of freedom different from the position of the search subject in the environment.

2. The information processing apparatus according to claim 1 , wherein

the internal state of the search subject further includes a second element, and

the second element has a nonholonomic degree of freedom.

3. The information processing apparatus according to claim 1 , wherein

the internal state of the search subject further includes a state of a joint angle, and

the state of the joint angle is associated with the search subject.

4. The information processing apparatus according to claim 1 , wherein

the internal state of the search subject further includes a steering state, and

the steering state is associated with the search subject.

5. The information processing apparatus according to claim 1 , wherein

the internal state of the search subject further includes a speed, and

the speed is associated with the search subject.

6. The information processing apparatus according to claim 1 , wherein

the internal state of the search subject further includes a state of fuel, and

the state of fuel is associated with the search subject.

7. The information processing apparatus according to claim 1 , wherein

the internal state of the search subject further includes a state of a load, and

the state of the load is associated with the search subject.

8. The information processing apparatus according to claim 1 , wherein

the internal state of the search subject further includes a state of a speech, and

the state of the speech is associated with the search subject.

9. The information processing apparatus according to claim 1 , wherein

the first feature value is represented as a first image, and

a dimension of the first image and a dimension of the environment map is same.

10. The information processing apparatus according to claim 1 , wherein

the environment map includes information on an obstacle in the environment, and

the first feature value is based on the environment map and a goal position.

11. The information processing apparatus according to claim 10 , wherein the first convolutional neural network is further configured to:

receive a cost map, wherein an extraction of the cost map is based on the environment map and the goal position; and

output the first feature value based on the cost map.

12. The information processing apparatus according to claim 11 , wherein the cost map includes at least one of

a two-dimensional map or a map higher than the two-dimensional map,

information associated with an obstacle that is in one of the two-dimensional map or the map higher than the two-dimensional map, or

path information in one of the two-dimensional map or the map higher than the two-dimensional map.

13. The information processing apparatus according to claim 12 , wherein the path information includes information on a cost to the goal position.

14. The information processing apparatus according to claim 1 , wherein

the path search includes a speech plan and an action plan,

the action plan is associated with the search subject, and

the search subject is in interaction with a target person.

15. The information processing apparatus according to claim 14 , wherein

the environment map includes an emotion of the target person, and

the emotion is represented as an obstacle.

16. The information processing apparatus according to claim 1 , wherein each of the first convolutional neural network and the second convolutional neural network is a fully convolutional network.

17. An information processing apparatus, comprises:

learning unit configured to:

receive an environment map associated with a position of a search subject, wherein the position is in an environment associated with the search subject;

receive an internal state of the search subject;

determine a first feature value based on the environment map;

determine a second feature value based on the internal state of the search subject, wherein

the internal state is represented as a two-dimensional image or an image higher than the two-dimensional image,

the internal state of the search subject includes at least one a specific element, and

the specific element has a degree of freedom different from the position of the search subject in the environment; and

learn a heuristic function based on the first feature value and the second feature value; and

a search unit configured to perform a path search based on the learned heuristic function.

18. An information processing method, comprising:

determining, by a processor, an environment map associated with a position of a search subject, wherein the position is in an environment associated with the search subject;

determining, by the processor, an internal state of the search subject;

outputting, by a first convolutional neural network, a first feature value based on the environment map;

outputting, by a second convolutional neural network, a second feature value based on the internal state of the search subject;

concatenating, by the processor, the first feature value and the second feature value;

determining, by the processor, a concatenated value based on the concatenation of the first feature value and the second feature value;

determining, by the processor, a loss associated with the concatenated value;

performing, by the processor, a learning process based on the determined loss;

learning, by the processor, a heuristic function based on the learning process, wherein the heuristic function is associated with a path search; and

outputting, by the processor, a heuristic map, wherein

the heuristic function is represented as a two-dimensional image or an image higher than the two-dimensional image,

the internal state of the search subject includes at least one a specific element, and

the specific element has a degree of freedom different from the position of the search subject in the environment.

19. An information processing method, comprising:

receiving an environment map associated with a position of a search subject, wherein the position is in an environment associated with the search subject;

receiving an internal state of the search subject;

determining a first feature value based on the environment map;

determining a second feature value based on the internal state of the search subject, wherein

the internal state is represented as a two-dimensional image or an image higher than the two-dimensional image,

the internal state of the search subject includes a specific element, and

the specific element has a degree of freedom different from the position of the search subject in the environment;

learning a heuristic function based on the first feature value and the second feature value; and

performing a path search based on the learned heuristic function.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (10)