Generating simulated edge-case driving scenarios
AssigneeCognata Ltd.
InventorsDan Atsmon, Ehud Spiegel
Life of this patent
Estimating the term…
Abstract
A system for generating simulated driving scenarios, comprising at least one hardware processor adapted for generating a plurality of simulated driving scenarios, each generated by providing a plurality of input driving objects to a machine learning model, where the machine learning model is trained using another machine learning model, trained to compute a classification indicative of a likelihood that a simulated driving scenario produced by the machine learning model comprises an interesting driving scenario.
Claims19
A system for generating simulated driving scenarios, comprising at least one hardware processor adapted for:
- generating a plurality of simulated driving scenarios, each generated by providing a plurality of input driving objects to a machine learning model, where the machine learning model is trained using another machine learning model, trained to compute a classification indicative of a likelihood that a simulated driving scenario produced by the machine learning model comprises an interesting driving scenario;
- wherein training the other machine learning model comprises using a plurality of recorded data sets, each recorded while a vehicle traverses a physical scene and comprises a recorded driving scenario and a plurality of recorded driving commands, the training is according to a difference between the plurality of recorded driving commands and a plurality of computed driving commands computed by the other machine learning model in response to the recorded driving scenario.
A system for generating simulated driving scenarios, comprising at least one hardware processor adapted for:
- generating a plurality of simulated driving scenarios, each generated by providing a plurality of input driving objects to a machine learning model, where the machine learning model is trained using another machine learning model, trained to compute a classification indicative of a likelihood that a simulated driving scenario produced by the machine learning model comprises an interesting driving scenario;
- wherein the machine learning model is further provided with a plurality of simulation parameters characteristic of at least one interesting driving scenario;
- and
- wherein the machine learning model generates at least one of the plurality of simulated driving scenarios further according to the plurality of simulation parameters;
- wherein the plurality of simulation parameters comprises a plurality of time-space-matrix distance values describing a plurality of distances, during an identified time interval, between a vehicle simulated by an autonomous driver and one or more objects of the plurality of input objects;
- wherein the plurality of time-space-matrix distance values are computed using captured driving data.
A method for generating simulated driving scenarios, comprising:
- generating a plurality of simulated driving scenarios, each generated by providing a plurality of input driving objects to a machine learning model, where the machine learning model is trained using another machine learning model, trained to compute a classification indicative of a likelihood that a simulated driving scenario produced by the machine learning model comprises an interesting driving scenario;
- wherein training the other machine learning model comprises using a plurality of recorded data sets, each recorded while a vehicle traverses a physical scene and comprises a recorded driving scenario and a plurality of recorded driving commands, the training is according to a difference between the plurality of recorded driving commands and a plurality of computed driving commands computed by the other machine learning model in response to the recorded driving scenario.
Dependencies are inferred from the claim text.