Lyft, Inc.

Woven Planet North America, Inc.

Woven by Toyota, U.s., Inc.

Sameer Qureshi

David Tse-Zhou Lu

Sammy Omari

Emilie Jeanne Anne Danna

Asif Haque

Vinay Shet

Ying Liu

Gerard Joyce

Mark Douglas Huyler

Luc Vincent

Kumar Hemachandra Chellapilla

Michael Jared Benisch

Alexis Weill

Alan Agon

Emil Constantin Praun

Taggart Matthiesen

Logan Green

Matt Vitelli

Shaojing Li

Sasanka Nagavalli

Muhammad Usman

Mason Lee

Victoria Mary Elizabeth Bellotti

Arturo Sanchez Corral

Jennie Arielle Braunstein

Moritz Niendorf

Examples disclosed herein may involve a computing system that is operable to (i) identify, within a given period of operation of a vehicle having an associated sensor system for capturing sensor data, one or more times when the vehicle was driving in a lane having substantially-straight lane geometry, (ii) for each identified time, determine a respective measure of a lateral offset between the vehicle's associated sensor system and a lateral reference point of the vehicle, and (iii) based on the respective measure of the lateral offset that is determined for each of the one or more identified times, determine the lateral offset between the vehicle's associated sensor system and the lateral reference point of the vehicle.

Detecting positioning of a sensor system associated with a vehicle

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

Systems, methods, and non-transitory computer-readable media can determine a first utility metric associated with a region and first autonomous vehicle eligibility criteria, wherein the first utility metric is determined based on a first plurality of rides and a subset of the first plurality of rides that can be successfully executed within the region based on the first autonomous vehicle eligibility criteria. A second utility metric associated with the region and second autonomous vehicle eligibility criteria can be determined, wherein the second utility metric is determined based on a second plurality of rides and a subset of the second plurality of rides that can be successfully executed within the region based on the second autonomous vehicle eligibility criteria. An autonomous vehicle associated with the first autonomous vehicle eligibility criteria can be selected to drive in the region based on a comparison of the first utility metric and the second utility metric.

Systems and methods for autonomous vehicle performance evaluation

A method includes generating, while a vehicle is operating in an autonomous-driving mode, a planned trajectory associated with a computing system of the vehicle based on first sensor data capturing an environment of the vehicle. The method further includes, while the vehicle is operating according to the planned trajectory, receiving a disengagement instruction associated that causes the vehicle to disengage from operating in the autonomous-driving mode and switch to operating in a disengagement mode. Subsequent to the vehicle operating in the disengagement mode, the method further includes capturing second sensor data and generating a simulation of the environment. The simulation is based on sensor data associated with the environment and the planned trajectory. Additionally, subsequent to the vehicle operating in the disengagement mode, the method concludes with evaluating a performance of an autonomy system based on the simulation, and providing feedback based on the evaluation.

Vehicle disengagement simulation and evaluation

Systems, methods, and non-transitory computer-readable media can receive transportation information associated with a transportation request, the transportation information comprising a pick up location and a drop off location. A first route associated with the transportation request and a non-autonomous vehicle can be determined. A second route associated with the transportation request and an autonomous vehicle can be determined based on an operating design domain (ODD) associated with one or more autonomous vehicles in a fleet of vehicles. At least one performance metric associated with the second route can be determined. The second route can be selected based at least in part on the at least one performance metric and a comparison of the first route and the second route. An autonomous vehicle from the fleet of vehicles can be assigned to the transportation request based on selection of the second route.

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

Systems, methods, and non-transitory computer-readable media can determine a road segment. A set of features associated with the road segment can be determined based at least in part on data captured by one or more sensors of a vehicle. A level of similarity between the road segment and each of a set of road segment types can be determined by comparing the set of features to features associated with each of the set of road segment types. The road segment can be classified as a road segment type based on the level of similarity. Scenario information associated with the road segment can be determined based on the classified road segment type.

Road segment similarity determination

In particular embodiments, a computing system may determine a measured driving characteristic of a driving control system based on observations of vehicles driven by the driving control system. The system may determine a difference between the measured driving characteristic and a target driving characteristic, which is based on objectivations of one or more manually controlled vehicles. The system may determine an evaluation objective for the driving control system. The system may determine a weight function for the evaluation objective. The system may determine a score for the driving control system with respect to the evaluation objective by weighting the difference between the measured driving characteristic and the target driving characteristic using the weight function. The system may apply, based on the score, an adjustment to the driving control system to reduce a difference between a subsequently measured driving characteristic of the driving control system and the target driving characteristic.

Evaluating driving control systems for elegant driving

Examples disclosed herein may involve (i) obtaining data for one or more data variables related to autonomous operation of a vehicle in a test environment being facilitated by the vehicle's autonomy system, (ii) based on the obtained data, evaluating one or more predefined fault rules, each of which comprises (a) one or more predefined criteria related to the one or more data variables and (b) a predefined fault that is to be injected into the autonomy system when the one or more predefined criteria are determined to be satisfied, (iii) based on the evaluation, injecting a predefined fault into the autonomy system, and (iv) capturing data indicative of a response by a response mechanism of the vehicle to the vehicle autonomously operating in accordance with the injected fault.

Systems and methods for injecting faults into an autonomy system

In one embodiment, a method by a computing system of a vehicle includes determining an environment of the vehicle. The method includes generating, based on the environment, multiple proposed vehicle actions with associated operational data. The method includes determining a comfort level for each proposed vehicle action by processing the environment and operational data using a model for predicting comfort levels of vehicle actions. The model is trained using records of performed vehicle actions. The record for each performed vehicle action includes environment data, operational data, and a perceived passenger comfort level for the performed vehicle action. The method includes selecting a vehicle action from the proposed vehicle actions based on the determined comfort level. The method includes causing the vehicle to perform the selected vehicle action.

Subjective route comfort modeling and prediction

Systems, methods, and non-transitory computer-readable media can receive disengagement information associated with one or more autonomous vehicles, the disengagement information identifying a plurality of disengagements of an autonomy system during operation of the one or more autonomous vehicles. Each disengagement of the plurality of disengagements can be categorized based on a plurality of categories, wherein a first category of the plurality of categories is associated with disengagement that would not have led to a negative outcome. A performance metric associated with the one or more autonomous vehicles can be determined based on the categorizing each disengagement of the plurality of disengagements. Autonomous vehicle performance of the one or more autonomous vehicles can be evaluated based on the performance metric.

Systems, methods, and non-transitory computer-readable media can receive data captured by one or more sensors associated with a vehicle. One or more objects in an environment of the vehicle can be identified based on the data captured by the one or more sensors. A position estimate of the vehicle can be generated within a known map based on one or more positional inferences pertaining to the vehicle, the one or more positional inferences pertaining to the vehicle being determined based on the one or more objects or features identified in the environment of the vehicle.

Systems and methods for estimating vehicle position based on contextual sensor information

Systems, methods, and non-transitory computer-readable media can obtain information describing a static map of a geographic location, wherein the static map is determined based at least in part on a plurality of three-dimensional representations of the geographic location captured by one or more sensors of one or more vehicles. At least one training example that includes visual features and a corresponding label can be generated based on an unsupervised process for generating training examples, wherein the visual features are extracted based on the static map and at least one three-dimensional representation of the geographic location. At least one machine learning model can be trained to distinguish between static objects and non-static objects in visual data based on the at least one training example, wherein the at least one machine learning model is trained based on an unsupervised learning process.

Dynamic object detection model based on static map collection data

In one embodiment, a computing system of a vehicle may capture, using one or more sensors of the vehicle, sensor data associated with a first vehicle of interest. The computing system may identify one or more features associated with the first vehicle of interest based on the sensor data. The computing system may determine a driving behavior model associated with the first vehicle of interest based on the one or more features of the first vehicle of interest. The computing system may predict a driving behavior of the first vehicle of interest based on at least the determined driving behavior model. The computing system may determine a vehicle operation for the vehicle based on at least the predicted driving behavior of the first vehicle of interest.

Growing community of inventors

Sunnyvale, CA, United States of America

Sameer Qureshi