Performing autonomous path navigation using deep neural networks
A method, computer readable medium, and system are disclosed for performing autonomous path navigation using deep neural networks. The method includes the steps of receiving image data at a deep neural network (DNN), determining, by the DNN, both an orientation of a vehicle with respect to a path and a lateral position of the vehicle with respect to the path, utilizing the image data, and controlling a location of the vehicle, utilizing the orientation of the vehicle with respect to the path and the lateral position of the vehicle with respect to the path.
1. A method, comprising:
calculating based, at least in part, on one or more neural networks, an orientation of a vehicle with respect to a path and a lateral position of the vehicle with respect to the path.
2. The method of claim 1 , further comprising:
using the calculated orientation and lateral position to determine a location of the vehicle.
3. The method of claim 2 , further comprising:
positioning the vehicle to the determined location based at last in part on steering directions received from a controller module, wherein the controller module converts the calculated orientation to steering directions.
4. The method of claim 1 , further comprising:
receiving, based at least in part on one or more imaging devices installed in the vehicle, image data; and
using the image data, at the one or more neural networks, to calculate the orientation of the vehicle with respect to a path and a lateral position of the vehicle with respect to the path.
5. The method of claim 4 , wherein the image data, captured by the one or more imaging devices, comprises information that indicates a current location of the vehicle on the path.
6. The method of claim 1 , wherein the one or more neural networks are trained using image data with correct labels.
7. The method of claim 1 , further comprising:
sending image data, received at the vehicle, to a remote location; and
determining, at the remote location using the one or more neural networks, based at least in part on the image data, the orientation of the vehicle with respect to the path and the lateral position of the vehicle with respect to the path.
8. The method of claim 7 , wherein the one or more neural networks, at the remote location, further determine both rotation and translation data based at least in part on the image data.
9. A system, comprising:
one or more processors to train one or more neural networks to calculate an orientation of a vehicle with respect to a path and a lateral position of the vehicle with respect to the path.
10. The system of claim 9 , where the one or more processors are further to:
train the one or more neural networks using received video data; and
wherein the video data comprises supervised data with labels of an associated position.
11. The system of claim 10 , where the one or more processors are further to:
train the one or more neural networks by implementing a loss function, wherein the loss function is calculated by:
determining a label for an image of the video data;
comparing the label to an associated correct label for the image; and
computing a difference between the label and the correct label; and
adjusting the one or more neural networks, based on the difference between the label and the correct label, using back propagation.
12. The system of claim 11 , wherein the trained one or more neural networks apply labels to incoming video data.
13. The system of claim 9 , wherein the one or more neural networks comprises a supervised classification network.
14. A processor, comprising:
one or more arithmetic logic units (ALUs) to calculate based, at least in part, on one or more neural networks, an orientation of a vehicle with respect to a path and a lateral position of the vehicle with respect to the path.
15. The processors of claim 14 , wherein the one or more ALUs are further to adjust the vehicle to a location based on the calculated orientation and lateral position.
16. The processor of claim 15 , wherein the one or more ALUs are further to:
receive an image from one or more imaging devices installed on the vehicle; and
use information from the image to determine the location for the vehicle to adjust.
17. The processor of claim 14 , wherein the calculated orientation of the vehicle with respect to the path comprises a plurality of probabilities of a direction that the vehicle is facing with respect to the path and the calculated lateral position of the vehicle with respect to the path comprises a probability of a plurality of lateral offsets with respect to a path center.
18. The processors of claim 14 , wherein the one or more ALUs are further to cause a controller module of the vehicle to convert the calculated orientation and lateral position to steering directions, wherein the steering directions comprise of steering signals and/or steering angles.
19. The processor of claim 14 , wherein the one or more ALUs are further to use the one or more neural networks to receive image data and detect whether an object is in the image data.
20. The processor of claim 19 , wherein the one or more ALUs are further to cause the vehicle to change course based at least in part on whether the object detected is greater than or equal to a predetermined threshold.
21. The processor of claim 14 , wherein the one or more ALUs are further to use the one or more neural networks to:
receive image data;
apply simultaneous localization and mapping (SLAM) to identify a location of the vehicle within a scene from the image data; and
provide information of a relative location of static objects within the scene.