Back to results
US 10,832,093 B1Granted
Tuning simulated data for optimized neural network activation
AssigneeZoox, Inc.
InventorEkaterina Hristova Taralova
Life of this patent
Estimating the term…
Abstract
Techniques described herein are directed to comparing, using a machine-trained model, neural network activations associated with data representing a simulated environment and activations associated with data representing real environment to determine whether the simulated environment is causes similar responses by the neural network, e.g., a detector. If the simulated environment and the real environment do not activate the same way (e.g., the variation between neural network activations of real and simulated data meets or exceeds a threshold), techniques described herein are directed to modifying parameters of the simulated environment to generate a modified simulated environment that more closely resembles the real environment.
Claims19
A computer-implemented method comprising:
- receiving first data and second data associated with a real environment;
- inputting, as first input, at least a portion of the first data into an artificial neural network, the artificial neural network comprising a plurality of neural network layers;
- receiving, as a first intermediate output, a first plurality of activations associated with a selected layer of the artificial neural network;
- inputting, as second input, at least a portion of the second data into the artificial neural network;
- receiving, as a second intermediate output, a second plurality of activations from the selected layer;
- determining, based at least in part on the first intermediate output and the second intermediate output, a first similarity score;
- training a model based at least in part on the first similarity score;
- receiving third data and fourth data, the third data associated with the real environment and the fourth data associated with a corresponding simulated environment;
- inputting, as third input, at least a portion of the third data into the artificial neural network;
- receiving, as a third intermediate output, a third plurality of activations from another selected layer of the artificial neural network;
- inputting, as fourth input, at least a portion of the fourth data into the artificial neural network;
- receiving, as a fourth intermediate output, a fourth plurality of activations from the other selected layer;
- determining, based at least in part on the third intermediate output and the fourth intermediate output and the model, a second similarity score;
- based at least in part on the second similarity score, modifying a parameter of the simulated environment to generate a modified simulated environment;
- one or more of testing or validating an algorithm to be used onboard an autonomous vehicle using the modified simulated environment;
- and
- transmitting the algorithm to the autonomous vehicle in the real environment based at least in part on the testing or the validating,
- wherein the algorithm is configured to control the autonomous vehicle in the real environment.
15 more elements
A system comprising:
- a processor;
- a non-transitory computer-readable medium that, when executed by the processor, cause the processor to perform operations comprising:
- receiving first data associated with a real environment;
- receiving second data associated with a simulated environment;
- inputting, as first input, at least a portion of the first data into an artificial neural network, the artificial neural network comprising a plurality of neural network layers;
- receiving, as a first intermediate output, a first plurality of activations associated with a selected layer of the artificial neural network;
- inputting, as second input, at least a portion of the second data into the artificial neural network;
- receiving, as a second intermediate output, a second plurality of activations from the selected layer;
- determining, based at least in part on the first intermediate output and the second intermediate output, a first similarity score using a machine-trained model;
- based at least in part on the first similarity score, modifying a parameter of the simulated environment to generate a modified simulated environment;
- one or more of testing or validating an algorithm to be used onboard an autonomous vehicle using the modified simulated environment;
- and
- transmitting the algorithm to the autonomous vehicle in the real environment based at least in part on the testing or the validating,
- wherein the algorithm is configured to control the autonomous vehicle in the real environment.
11 more elements
One or more non-transitory computer-readable media that, when executed by a processor, cause the processor to perform operations comprising:
- receiving first data associated with a real environment;
- receiving second data associated with a simulated environment;
- inputting, as first input, at least a portion of the first data into an artificial neural network, the artificial neural network comprising a plurality of neural network layers;
- receiving, as a first intermediate output, a first plurality of activations associated with a selected layer of the artificial neural network;
- inputting, as second input, at least a portion of the second data into the artificial neural network;
- receiving, as a second intermediate output, a second plurality of activations from the selected layer;
- determining, based at least in part on the first intermediate output and the second intermediate output, a first similarity score using a machine-trained model;
- based at least in part on the first similarity score, modifying a parameter of the simulated environment to generate a modified simulated environment;
- one or more of testing or validating an algorithm to be used onboard an autonomous vehicle using the modified simulated environment;
- and
- transmitting the algorithm to the autonomous vehicle in the real environment based at least in part on the testing or the validating,
- wherein the algorithm is configured to control the autonomous vehicle in the real environment.
9 more elements
Dependencies are inferred from the claim text.