Woven Planet North America, Inc.

Lyft, Inc.

Motional Ad LLC

Woven by Toyota, U.s., Inc.

Sammy Omari

Sameer Qureshi

Juraj Kabzan

Julia Gomes

Eric Wolff

Alfred Charles Jones, Ii

Caglayan Dicle

Joseph Patrick Wendell Whinnery

Alan Agon

Nanxiang Li

Muhammad Usman

Ana Sofia Rufino Ferreira

Sang Uk Lee

Luc Vincent

Emil Constantin Praun

Romain Clément

Scott D Pendleton

Peter Ondruska

Emilie Jeanne Anne Danna

Bence Cserna

Helen Ruth Lurie

Matt Vitelli

Kevin C Gall

Hans Andersen

Tsung-Han Lin

Meng Tao

Nastaran Ghadar

Yunjian Jiang

Sasanka Nagavalli

Matthew Swaner Vitelli

Mason Lee

David Andrew Dolben

Erik Newell Gaasedelen

Moritz Niendorf

Carey Stover Nachenberg

Carlos Alberto De Magalhaes Massera Filho

Christian Alexander Stephan Buckl

Dor Shaviv

Provided are methods for generating future maneuver parameters in a planner, which include receiving at least one first set of parameters associated with one or more previous maneuvers of a vehicle and at least one second set of parameters associated with a maneuver goal of the vehicle, generating, using the first and second sets of parameters, a future maneuver parameter corresponding to a future maneuver of the vehicle, training at least one data model by comparing the generated future maneuver parameter to one or more reference maneuver parameters, generating, based on training, a corrected future maneuver parameter. The corrected future maneuver parameter includes a future maneuver of the vehicle and a correction to the future maneuver of the vehicle. The first set of parameters includes the generated corrected future maneuver parameter, which is used to correct at least one first parameter. The training is executed using the corrected first parameter.

Generating corrected future maneuver parameters in a planner

Provided are methods and systems for semantic behavior filtering for prediction improvement. A method for operating an autonomous vehicle, is provided. The method includes obtaining, by at least one processor, semantic image data associated with an environment where an autonomous vehicle is operating. The method includes determining, by the at least one processor, a set of agents in the environment based on the semantic image data. The method includes determining a set of predicted actions for at least one agent of the set of agents. The method includes determining, from the set of predicted actions, a set of secondary predicted actions for the at least one primary agent using semantic data. The method includes determining, from the set of predicted actions, a set of primary predicted actions other than secondary predicted actions The method includes generating a path for the autonomous vehicle based on the set of primary predicted actions.

Path generation based on predicted actions

Systems, methods, and non-transitory computer-readable media can detect an occurrence of a condition in an environment based on sensor data captured by a vehicle. A determination is made whether the occurrence of the condition satisfies a threshold associated with a likelihood that a behavior associated with an object in the environment will occur based on an interaction between the condition and the object, wherein the likelihood is based on prior observations of one or more objects. Subsequent to determining that the threshold is satisfied, a vehicle operation that is associated with the likelihood that the behavior associated with the object will occur is performed.

Systems and methods for configuring autonomous vehicle operation

In one embodiment, a method includes determining a connectivity model associated with a region of a road network, wherein the connectivity model was trained using vehicle traffic-pattern data comprising a first lane identifier and a second lane identifier indicating one or more lanes associated with a vehicle trajectory through the region and a traffic-light state corresponding to signal information of traffic lights in the region when a vehicle moved through the region, determining for at least one egress lane in the region based on the connectivity model a lane relationship indicating one or more ingress lanes in the region onto which a vehicle in the egress lane can move and one or more governing traffic lights selected from the traffic lights in the region that govern the egress lane, and encoding the lane relationship and the one or more governing traffic lights into a map of the region.

Leveraging traffic patterns to understand traffic rules

Provided are methods for prediction error scenario mining for machine learning methods, which can include determining a prediction error indicative of a difference between a planned decision of an autonomous vehicle and an ideal decision of the autonomous vehicle. The prediction error is associated with an error-prone scenario for which a machine learning model of an autonomous vehicle is to make planned movements. The method includes searching a scenario database for the error-prone scenario based on the prediction error. The scenario database includes a plurality of datasets representative of data received from an autonomous vehicle sensor system in which the plurality of datasets is marked with at least one attribute of the set of attributes. The method further includes obtaining the error-prone scenario from the scenario database for inputting into the machine learning model for training the machine learning model. Systems and computer program products are also provided.

Prediction error scenario mining for machine learning models

In one embodiment, a method includes receiving parameter data associated with trajectories captured using at least one sensor associated with one or more vehicles, determining a parameter distribution associated with movement parameters for the parameter data based on the trajectories, adjusting the parameter distribution for the parameter data associated with the movement parameters up to a respective threshold constraint, wherein the respective threshold constraint is based on at least one of the parameter data, generating driving simulations that are based on the adjusted parameter distribution for the parameter data, and evaluating an autonomous-driving system using the driving simulations.

Generating accurate and diverse simulations for evaluation of autonomous-driving systems

A method includes generating a parameter of a trajectory associated with a scenario using a path planner. The parameter is generated based on a training dataset. The method includes comparing the parameter of the trajectory against a validation parameter associated with a validation dataset. The validation parameter is based on human-based vehicle driving trajectory data associated with scenarios that satisfy a level of similarity with the scenario. The method further includes determining a level of similarity between the parameter associated with the scenario and the validation parameter associated with the scenarios, and, subsequent to determining that the level of similarity fails to satisfy a similarity threshold, the method concludes with providing training data associated with the scenario to the training dataset so that a subsequent parameter of a subsequent trajectory generated by the path planner and associated with the scenario satisfies the level of similarity against the validation parameter.

Active learning and validation system for vehicles

In one embodiment, a computing system accesses contextual data associated with a vehicle operated by a human driver. The contextual data is captured using one or more sensors associated with the vehicle. The system determines one or more predicted vehicle operations by processing the contextual data based at least on information associated with pre-recorded contextual data associated with a number of vehicles. The system detects one or more vehicle operations made by the human driver. The system determines that an event of interest is associated with the contextual data based on a comparison of the one or more vehicle operations made by the human driver and the one or more predicted vehicle operations. The system causes high-resolution contextual data associated with the event of interest to be stored.

Systems and methods for detecting and recording anomalous vehicle events

Systems, methods, and non-transitory computer-readable media can determine at least one potential route for navigating a vehicle within a geographic region. A score that measures a comfort level associated with the potential route can be determined, wherein the score is determined based on at least one sensor map that segments the geographic region into a grid of cells, and wherein the comfort level for the potential route is determined based at least in part on cells through which the vehicle travels while navigating along the potential route. A determination is made whether to use the potential route for navigating the vehicle based at least in part on the score.

Approaches for mapping geographic regions

In one embodiment, a method includes determining an initial cost volume associated with a plurality of potential trajectories of a vehicle in an environment based on a set of movement restrictions of the vehicle, generating a delta cost volume using the initial cost volume and environment data associated with the environment, wherein the delta cost volume is generated by determining adjustments to the initial cost volume that incorporate observed driving behavior, and scoring a trajectory of the plurality of potential trajectories for the vehicle based on the initial cost volume and the delta cost volume.

Path planning using delta cost volume generated from movement restrictions and observed driving behavior

Systems, methods, and non-transitory computer-readable media can determine sensor data collected by a fleet of vehicles while navigating a geographic region, the sensor data including sensor readings generated at least in part by a surface interaction between one or more tires of each of the fleet of vehicles and a road surface of the geographic region. A sensor map representing the geographic region can be determined. The map can segment the geographic region into a grid of cells. Instances of the collected sensor data can be associated with cells in the grid of cells. A corresponding fingerprint can be determined for one or more cells in the grid of cells based at least in part on a plurality of instances of sensor data associated with the cell.

Disclosed herein is technology for deriving a planned path for a vehicle that is operating in a geographic area. In an example embodiment, a computing system associated with the vehicle may function to (i) identify a set of path priors to use as a basis for deriving the planned path for the vehicle in the geographic area, (ii) sample path priors in the identified set of path priors and thereby producing a set of points that is representative of the identified set of path priors, (iii) fit a curve to the set of points produced by the sampling, and (iv) derive the planned path for the vehicle based on the fitted curve.

Systems and methods for deriving planned paths for vehicles using path priors

Growing community of inventors

Los Altos, CA, United States of America

Sammy Omari