Perpendicular cut-in training
AssigneeZoox, Inc.
InventorsMichael Haggblade, Benjamin Isaac Mattinson
Life of this patent
Estimating the term…
Abstract
Techniques relating to training a model for detecting that a vehicle is likely to perform a cut-in maneuver are described. Computing device(s) can receive log data associated with vehicles in an environment and can detect an event in the log data, wherein an event corresponds to a cut-in maneuver performed by a vehicle. In an example, the computing device(s) can generate training data based at least in part on converting a portion of the log data that corresponds to the event into a top-down representation of the environment and inputting the training data into a model, wherein the model is trained to output an indication of whether another vehicle is likely to perform another cut-in maneuver.
Claims20
A method comprising:
- receiving log data associated with vehicles in an environment;
- detecting an event in the log data, wherein an event corresponds to a cut-in maneuver performed by a vehicle;
- determining a period of time associated with the event;
- accessing a portion of the log data that corresponds to the period of time;
- generating training data based at least in part on converting the portion of the log data that corresponds to the period of time into a top-down multi-channel image representation of the environment;
- inputting the training data into a model;
- determining, by the model, an output comprising an indication that another vehicle is likely to perform another cut-in maneuver;
- determining a difference between the output and an expected output associated with the training data;
- altering one or more parameters of the model based on the difference;
- and
- transmitting the model to an autonomous vehicle configured to be controlled by another output of the model.
A system comprising:
- one or more processors;
- and
- one or more non-transitory computer-readable media storing instructions, that when executed by the one or more processors, cause the system to perform operations comprising:
- receiving log data associated with vehicles in an environment;
- detecting an event in the log data, wherein an event corresponds to a cut-in maneuver performed by a vehicle;
- generating training data based at least in part on converting a portion of the log data that corresponds to the event into a top-down multi-channel image representation of the environment;
- inputting the training data into a model;
- and
- transmitting the model to an autonomous vehicle configured to be controlled by another output of the model.
One or more non-transitory computer-readable media storing instructions, that when executed by one or more processors, cause the one or more processors to perform operations comprising:
- receiving log data associated with vehicles in an environment;
- detecting an event in the log data, wherein an event corresponds to a cut-in maneuver performed by a vehicle;
- generating training data based at least in part on converting a portion of the log data that corresponds to the event into a top-down multi-channel image representation of the environment;
- and
- inputting the training data into a model, wherein the model is trained to output an indication of whether another vehicle is likely to perform another cut-in maneuver.
Dependencies are inferred from the claim text.