Lyft, Inc.

Shaohui Sun

David Tse-Zhou Lu

Vinay Shet

Ying Liu

Shaojing Li

Weiyi Hou

Lina Dong

Robert Charles Kyle Pinkerton

Somesh Khandelwal

Ivan Kirigin

Ramesh Rangarajan Sarukkai

Systems, methods, and non-transitory computer-readable media can access a plurality of schema-based encodings providing a structured representation of an environment captured by one or more sensors associated with a plurality of vehicles traveling through the environment. The plurality of schema-based encodings can be clustered into one or more clusters of schema-based encodings. At least one scenario associated with the environment can be determined based at least in part on the one or more clusters of schema-based encodings.

Approaches for encoding environmental information

Systems, methods, and non-transitory computer-readable media can determine sensor data captured by at least one sensor of a vehicle while navigating an environment over a period of time. Information describing one or more agents associated with the environment during the period of time can be determined based at least in part on the captured sensor data. A schema-based encoding describing the environment during the period of time can be generated based at least in part on the determined information and a scenario schema, wherein the schema-based encoding provides a structured representation of the environment during the period of time.

In one embodiment, a method includes receiving an image associated with an object in an environment, the image being captured by sensors associated with a vehicle, generating a feature representation of the image, determining a potential ground control point associated with the object based on the feature representation of the image, determining a predetermined location reading based on the potential ground control point, calculating a differential relative to the predetermined location reading based on the potential ground control point, and determining a location of the vehicle based on the differential and the predetermined location reading based on the potential ground control point.

Deep direct localization from ground imagery and location readings

Systems, methods, and non-transitory computer-readable media can access a first set of schema-based encodings associated with a first environment, wherein a schema-based encoding provides a structured representation of an environment based on a scenario schema. First information representing scenario information associated with the first environment can be generated based at least in part on the first set of schema-based encodings. One or more attributes for the first environment can be determined based at least in part the first information.

In one embodiment, a method includes training an image localization model for determining an adjusted location reading for a vehicle based on its location reading and the associated image. The method further includes updating a location associated with the vehicle using the adjusted location reading. The training of the image localization model includes generating feature representations of a number of training images, and learning a relationship between the feature representations and differentials between a number of raw location readings and their respective reference location readings.

Growing community of inventors

Union City, CA, United States of America

Shaohui Sun