Abstract: Various implementations described herein generate training instances that each include corresponding training instance input that is based on corresponding sensor data of a corresponding autonomous vehicle, and that include corresponding training instance output that is based on corresponding sensor data of a corresponding additional vehicle, where the corresponding additional vehicle is captured at least in part by the corresponding sensor data of the corresponding autonomous vehicle. Various implementations train a machine learning model based on such training instances. Once trained, the machine learning model can enable processing, using the machine learning model, of sensor data from a given autonomous vehicle to predict one or more properties of a given additional vehicle that is captured at least in part by the sensor data.

Abstract: Determining classification(s) for object(s) in an environment of autonomous vehicle, and controlling the vehicle based on the determined classification(s). For example, autonomous steering, acceleration, and/or deceleration of the vehicle can be controlled based on determined pose(s) and/or classification(s) for objects in the environment. The control can be based on the pose(s) and/or classification(s) directly, and/or based on movement parameter(s), for the object(s), determined based on the pose(s) and/or classification(s). In many implementations, pose(s) and/or classification(s) of environmental object(s) are determined based on data from a phase coherent Light Detection and Ranging (LIDAR) component of the vehicle, such as a phase coherent LIDAR monopulse component and/or a frequency-modulated continuous wave (FMCW) LIDAR component.

Abstract: Determining yaw parameter(s) (e.g., at least one yaw rate) of an additional vehicle that is in addition to a vehicle being autonomously controlled, and adapting autonomous control of the vehicle based on the determined yaw parameter(s) of the additional vehicle. For example, autonomous steering, acceleration, and/or deceleration of the vehicle can be adapted based on a determined yaw rate of the additional vehicle. In many implementations, the yaw parameter(s) of the additional vehicle are determined based on data from a phase coherent Light Detection and Ranging (LIDAR) component of the vehicle, such as a phase coherent LIDAR monopulse component and/or a frequency-modulated continuous wave (FMCW) LIDAR component.

Abstract: A system and method are arranged to provide recommendations to a user of a vehicle. In one aspect, the vehicle navigates in an autonomous mode and the sensors provide information that is based on the location of the vehicle and output from sensors directed to the environment surrounding the vehicle. In further aspects, both current and previous sensor data is used to make the recommendations, as well as data based on the sensors of other vehicles.

Abstract: Aspects of the disclosure relate to generating a speed plan for an autonomous vehicle. As an example, the vehicle is maneuvered in an autonomous driving mode along a route using pre-stored map information. This information identifies a plurality of keep clear regions where the vehicle should not stop but can drive through in the autonomous driving mode. Each keep clear region of the plurality of keep clear regions is associated with a priority value. A subset of the plurality of keep clear regions is identified based on the route. A speed plan for stopping the vehicle is generated based on the priority values associated with the keep clear regions of the subset. The speed plan identifies a location for stopping the vehicle. The speed plan is used to stop the vehicle in the location.

Abstract: Autonomous vehicles use various computing systems to transport passengers from one location to another. A control computer sends messages to the various systems of the vehicle in order to maneuver the vehicle safely to the destination. The control computer may display information on an electronic display in order to allow the passenger to understand what actions the vehicle may be taking in the immediate future. Various icons and images may be used to provide this information to the passenger.

Abstract: Aspects of the disclosure relate to generating a speed plan for an autonomous vehicle. As an example, the vehicle is maneuvered in an autonomous driving mode along a route using pre-stored map information. This information identifies a plurality of keep clear regions where the vehicle should not stop but can drive through in the autonomous driving mode. Each keep clear region of the plurality of keep clear regions is associated with a priority value. A subset of the plurality of keep clear regions is identified based on the route. A speed plan for stopping the vehicle is generated based on the priority values associated with the keep clear regions of the subset. The speed plan identifies a location for stopping the vehicle. The speed plan is used to stop the vehicle in the location.

Abstract: A method and apparatus are provided for optimizing one or more object detection parameters used by an autonomous vehicle to detect objects in images. The autonomous vehicle may capture the images using one or more sensors. The autonomous vehicle may then determine object labels and their corresponding object label parameters for the detected objects. The captured images and the object label parameters may be communicated to an object identification server. The object identification server may request that one or more reviewers identify objects in the captured images. The object identification server may then compare the identification of objects by reviewers with the identification of objects by the autonomous vehicle. Depending on the results of the comparison, the object identification server may recommend or perform the optimization of one or more of the object detection parameters.

Abstract: Aspects of the present disclosure relate generally to generating elevation maps. More specifically, data points may be collected by a laser moving along a roadway and used to generate an elevation map of the roadway. The collected data points may be projected onto a two dimensional or “2D” grid. The grid may include a plurality of cells, each cell of the grid representing a geolocated second of the roadway. The data points of each cell may be evaluated to identify an elevation for the particular cell. For example, the data points in a particular cell may be filtered in various ways including occlusion, interpolation from neighboring cells, etc. The minimum value of the remaining data points within each cell may then be used as the elevation for the particular cell, and the elevation of a plurality of cells may be used to generate an elevation map of the roadway.

Abstract: A passenger in an automated vehicle may relinquish control of the vehicle to a control computer when the control computer has determined that it may maneuver the vehicle safely to a destination. The passenger may relinquish or regain control of the vehicle by applying different degrees of pressure, for example, on a steering wheel of the vehicle. The control computer may convey status information to a passenger in a variety of ways including by illuminating elements of the vehicle. The color and location of the illumination may indicate the status of the control computer, for example, whether the control computer has been armed, is ready to take control of the vehicle, or is currently controlling the vehicle.

Abstract: Autonomous vehicles use various computing systems to transport passengers from one location to another. A control computer sends messages to the various systems of the vehicle in order to maneuver the vehicle safely to the destination. The control computer may display information on an electronic display in order to allow the passenger to understand what actions the vehicle may be taking in the immediate future. Various icons and images may be used to provide this information to the passenger.

Abstract: A passenger in an automated vehicle may relinquish control of the vehicle to a control computer when the control computer has determined that it may maneuver the vehicle safely to a destination. The passenger may relinquish or regain control of the vehicle by applying different degrees of pressure, for example, on a steering wheel of the vehicle. The control computer may convey status information to a passenger in a variety of ways including by illuminating elements of the vehicle. The color and location of the illumination may indicate the status of the control computer, for example, whether the control computer has been armed, is ready to take control of the vehicle, or is currently controlling the vehicle.

Abstract: Aspects of the disclosure relate to generating a speed plan for an autonomous vehicle. As an example, the vehicle is maneuvered in an autonomous driving mode along a route using pre-stored map information. This information identifies a plurality of keep clear regions where the vehicle should not stop but can drive through in the autonomous driving mode. Each keep clear region of the plurality of keep clear regions is associated with a priority value. A subset of the plurality of keep clear regions is identified based on the route. A speed plan for stopping the vehicle is generated based on the priority values associated with the keep clear regions of the subset. The speed plan identifies a location for stopping the vehicle. The speed plan is used to stop the vehicle in the location.

Abstract: A method and apparatus are provided for optimizing one or more object detection parameters used by an autonomous vehicle to detect objects in images. The autonomous vehicle may capture the images using one or more sensors. The autonomous vehicle may then determine object labels and their corresponding object label parameters for the detected objects. The captured images and the object label parameters may be communicated to an object identification server. The object identification server may request that one or more reviewers identify objects in the captured images. The object identification server may then compare the identification of objects by reviewers with the identification of objects by the autonomous vehicle. Depending on the results of the comparison, the object identification server may recommend or perform the optimization of one or more of the object detection parameters.

Abstract: An autonomous vehicle may access portions of a map to maneuver a roadway. The map may be split into one or more levels that represent different regions in space. For example, an overpass may be represented by one level while the road below the overpass may be on a separate level. A vehicle traveling on a particular level may use map data that is associated with that level. Furthermore, if the vehicle travels through a warp zone, it may transition from the current level to a destination level and thus begin to use map data associated with the destination level.

Abstract: Models can be generated of a vehicle's view of its environment and used to maneuver the vehicle. This view need not include what objects or features the vehicle is actually seeing, but rather those areas that the vehicle is able to observe using its sensors if the sensors were completely un-occluded. For example, for each of a plurality of sensors of the object detection component, a computer may generate an individual 3D model of that sensor's field of view. Weather information is received and used to adjust one or more of the models. After this adjusting, the models may be aggregated into a comprehensive 3D model. The comprehensive model may be combined with detailed map information indicating the probability of detecting objects at different locations. The model of the vehicle's environment may be computed based on the combined comprehensive 3D model and detailed map information.

Abstract: Models can be generated of a vehicle's view of its environment and used to maneuver the vehicle. This view need not include what objects or features the vehicle is actually seeing, but rather those areas that the vehicle is able to observe using its sensors if the sensors were completely un-occluded. For example, for each of a plurality of sensors of the object detection component, a computer may generate an individual 3D model of that sensor's field of view. Weather information is received and used to adjust one or more of the models. After this adjusting, the models may be aggregated into a comprehensive 3D model. The comprehensive model may be combined with detailed map information indicating the probability of detecting objects at different locations. The model of the vehicle's environment may be computed based on the combined comprehensive 3D model and detailed map information.

Abstract: Methods and systems are disclosed for cross-validating a second sensor with a first sensor. Cross-validating the second sensor may include obtaining sensor readings from the first sensor and comparing the sensor readings from the first sensor with sensor readings obtained from the second sensor. In particular, the comparison of the sensor readings may include comparing state information about a vehicle detected by the first sensor and the second sensor. In addition, comparing the sensor readings may include obtaining a first image from the first sensor, obtaining a second image from the second sensor, and then comparing various characteristics of the images. One characteristic that may be compared are object labels applied to the vehicle detected by the first and second sensor. The first and second sensors may be different types of sensors.

Abstract: Models can be generated of a vehicle's view of its environment and used to maneuver the vehicle. This view need not include what objects or features the vehicle is actually seeing, but rather those areas that the vehicle is able to observe using its sensors if the sensors were completely un-occluded. For example, for each of a plurality of sensors of the object detection component, a computer may generate an individual 3D model of that sensor's field of view. Weather information is received and used to adjust one or more of the models. After this adjusting, the models may be aggregated into a comprehensive 3D model. The comprehensive model may be combined with detailed map information indicating the probability of detecting objects at different locations. The model of the vehicle's environment may be computed based on the combined comprehensive 3D model and detailed map information.

Abstract: Aspects of the present disclosure relate generally to generating elevation maps. More specifically, data points may be collected by a laser moving along a roadway and used to generate an elevation map of the roadway. The collected data points may be projected onto a two dimensional or “2D” grid. The grid may include a plurality of cells, each cell of the grid representing a geolocated second of the roadway. The data points of each cell may be evaluated to identify an elevation for the particular cell. For example, the data points in a particular cell may be filtered in various ways including occlusion, interpolation from neighboring cells, etc. The minimum value of the remaining data points within each cell may then be used as the elevation for the particular cell, and the elevation of a plurality of cells may be used to generate an elevation map of the roadway.