Abstract: An autonomous vehicle detects a tailgating vehicle and uses various response mechanisms. A vehicle is identified as a tailgater based on whether its characteristics meet a variable threshold. When the autonomous vehicle is traveling at slower speeds, the threshold is defined in distance. When the autonomous vehicle is traveling at faster speeds, the threshold is defined in time. The autonomous vehicle responds to the tailgater by modifying its driving behavior. In one example, the autonomous vehicle adjusts a headway buffer (defined in time) from another vehicle in front of the autonomous vehicle. In this regard, if the tailgater is T seconds too close to the autonomous vehicle, the autonomous vehicle increases the headway buffer to the vehicle in front of it by some amount relative to T.

Abstract: Aspects of the disclosure relate to determining vehicle operation modes for a fleet of vehicles. This may include maintaining a storage system including vehicle parameters for a vehicle fleet such that each vehicle of the fleet is associated with a set of vehicle parameters. The vehicle operation mode may be determined by selecting one of a plurality of vehicle operation levels arranged in a hierarchy or by identifying a set of tags where each level or tag corresponds to a vehicle driving capability. The determined vehicle operation mode may be communicated to the given vehicle in order to cause the vehicle to operate according to the determined vehicle operation mode.

Abstract: The disclosure relates to detecting and clearing debris from a sensor window. For instance, a system may include the sensor window and one or more processors. The sensor window may include a transparent thin film conductor. The one or more processors may be configured to identify a change in capacitance using the transparent thin film conductor, detect debris on the sensor window using the change in capacitance, and activate a cleaning system in order to clear the detected debris from the sensor window.

Abstract: Aspects of the disclosure relate to training and using a model for identifying actions of objects. For instance, LIDAR sensor data frames including an object bounding box corresponding to an object as well as an action label for the bounding box may be received. Each sensor frame is associated with a timestamp and is sequenced with respect to other sensor frames. Each given sensor data frame may be projected into a camera image of the object based on the timestamp associated with the given sensor data frame in order to provide fused data. The model may be trained using the fused data such that the model is configured to, in response to receiving fused data, the model outputs an action label for each object bounding box of the fused data. This output may then be used to control a vehicle in an autonomous driving mode.

Abstract: Methods and system are provided for training and using a model to determine states of lanes of interest. For instance, image data including an image and an associated label identifying at least one traffic light, a state of the at least one traffic light, and a lane controlled by the at least one traffic light are received and used to train the model such that the model is configured to, in response to receiving an image and a lane of interest included in the image, output a lane state for the lane of interest. This model is then used by a vehicle in order to determine a state of a lane of interest. This state is then used to control the vehicle in an autonomous driving mode based on the state of the lane of interest.

Abstract: Aspects of the present disclosure relate to a vehicle for maneuvering an occupant of the vehicle to a destination autonomously as well as providing information about the vehicle and the vehicle's environment for display to the occupant.

Abstract: Aspects of the present disclosure relate to protecting the privacy of a user of a dispatching service for driverless vehicles. For example, a client computing device may send a request for a vehicle, the request identifying a location of the client computing device. Location information identifying a current location of a vehicle dispatched to the location of the client computing device is received in response to the request. A camera view is displayed on a display of the client computing device, the camera view corresponding to information in a field of view of a camera of the client computing device. When current location of the vehicle is within the field of view of the camera and a threshold distance of the location of the client device, an indicator is displayed on the display at a location corresponding to the current location of the vehicle.

Abstract: In some implementations, an image classification system of an autonomous or semi-autonomous vehicle is capable of improving multi-object classification by reducing repeated incorrect classification of objects that are considered rarely occurring objects. The system can include a common instance classifier that is trained to identify and recognize general objects (e.g., commonly occurring objects and rarely occurring objects) as belonging to specified object categories, and a rare instance classifier that is trained to compute one or more rarity scores representing likelihoods that an input image is correctly classified by the common instance classifier. The output of the rare instance classifier can be used to adjust the classification output of the common instance classifier such that the likelihood of input images being incorrectly classified is reduced.

Abstract: Systems and methods are described that relate to a light detection and ranging (LIDAR) device. The LIDAR device includes a fiber laser configured to emit light within a wavelength range, a scanning portion configured to direct the emitted light in a reciprocating manner about a first axis, and a plurality of detectors configured to sense light within the wavelength range. The device additionally includes a controller configured to receive target information, which may be indicative of an object, a position, a location, or an angle range. In response to receiving the target information, the controller may cause the rotational mount to rotate so as to adjust a pointing direction of the LIDAR. The controller is further configured to cause the LIDAR to scan a field-of-view (FOV) of the environment. The controller may determine a three-dimensional (3D) representation of the environment based on data from scanning the FOV.

Abstract: The technology relates controlling an autonomous vehicle through a multi-lane turn. In one example, data corresponding to a position of the autonomous vehicle in a lane of the multi-lane turn, a trajectory of the autonomous vehicle, and data corresponding to positions of objects in a vicinity of the autonomous vehicle may be received. A determination of whether the autonomous vehicle is positioned as a first vehicle in the lane or positioned behind another vehicle in the lane may be made based on a position of the autonomous vehicle in the lane relative to the positions of the objects. The trajectory of the autonomous vehicle through the lane may be adjusted based on whether the autonomous vehicle is positioned as a first vehicle in the lane or positioned behind another vehicle in the lane. The autonomous vehicle may be controlled based on the adjusted trajectory.

Abstract: The technology relates to route planning and performing driving operations in autonomous vehicles, such as cargo trucks, articulating buses, as well as other vehicles. A detailed kinematic model of the vehicle in evaluated in conjunction with roadgraph and other information to determine whether a route or driving operation is feasible for the vehicle. This can include evaluating a hierarchical set of driving rules and whether current driving conditions impact any of the rules. Driving trajectories and cost can be evaluated when pre-planning a route for the vehicle to follow. This can include determining an ideal trajectory for the vehicle to take a particular driving action. Pre-planned routes may be shared with a fleet of vehicles, and can be modified based on information obtained by different vehicles of the fleet.

Abstract: Aspects of the disclosure relate to a method of controlling a vehicle having an autonomous driving mode at a multi-way stop intersection. For instance, that the vehicle has come to a stop at a multi-stop intersection may be determined. At least one road user at the intersection may be identified. Based on the determination that the vehicle has come to the stop, a period of time to wait for the at least one road user to proceed through the intersection may be determined. After waiting the determined period of time, that the at least one road user has not begun to proceed through the intersection may be determined. Based on the determination that the at least one user has not begun to proceed through the intersection, the vehicle may be controlled in the autonomous driving mode in order to proceed through the intersection.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for identifying high-priority agents in the vicinity of a vehicle and, for only those agents which are high priority agents, generating data characterizing the agents using a first prediction model. In a first aspect, a system identifies multiple agents in an environment in a vicinity of a vehicle. The system generates a respective importance score for each of the agents by processing a feature representation of each agent using an importance scoring model. The importance score for an agent characterizes an estimated impact of the agent on planning decisions generated by a planning system of the vehicle which plans a future trajectory of the vehicle. The system identifies, as high-priority agents, a proper subset of the plurality of agents with the highest importance scores.

Abstract: The technology relates to controlling a vehicle in an autonomous driving mode. For instance, sensor data identifying an object in an environment of the vehicle may be received. A grid including a plurality of cells may be projected around the object. For each given one of the plurality of cells, a likelihood that the object will enter the given one within a period of time into the future is predicted. A contour is generated based on the predicted likelihoods. The vehicle is then controlled in the autonomous driving mode in order to avoid an area within the contour.

Abstract: The present application discloses embodiments that relate to a radar system. The embodiments may include a plurality of radiating waveguides each having a waveguide input. The embodiments also include an attenuation component, which can be located on a circuit board. The embodiments further include a beamforming network. The beamforming network includes a beamforming network input. The beamforming network also includes a plurality of beamforming network outputs, where each beamforming network output is coupled to one of the waveguide inputs. Additionally, the beamforming network includes a coupling port, wherein the coupling port is configured to couple at least one waveguide of the system to a component external to the waveguide.

Abstract: Aspects of the disclosure relate to parking behaviors and maneuvering a vehicle in an autonomous driving mode accordingly. For instance, a pullover location for the vehicle to stop and wait for a passenger may be identified. The vehicle may be maneuvered in the autonomous driving mode in order to pull over by pulling forward into the pullover location. Whether to maneuver the vehicle in reverse in the pullover location before or after the passenger enters the vehicle may be determined based on context for the pull over with respect to the passenger. The vehicle may be maneuvered in the autonomous driving mode in reverse based on the determination of whether to maneuver the vehicle in reverse in the pullover location before or after the passenger enters the vehicle.

Abstract: Aspects of the disclosure relate to controlling a vehicle using a simplified model of an object. In one example, sensor data including a plurality of data points corresponding to surfaces of the object in the vehicle's environment may be received from one or more sensors of the vehicle. A first model may be determined using a subset of the plurality of data points. A set of secondary data points may be identified from the plurality of data points using a point on the vehicle. The set of secondary data points may be filtered from the subset of the plurality data points to determine a second model, wherein the second model is a simplified version of the first model. The vehicle may be controlled in an autonomous driving mode based on the second model.

Abstract: Aspects of the disclosure relate to training and using a phrase recognition model to identify phrases in images. As an example, a selected phrase list may include a plurality of phrases is received. Each phrase of the plurality of phrases includes text. An initial plurality of images may be received. A training image set may be selected from the initial plurality of images by identifying the phrase-containing images that include one or more phrases from the selected phrase list. Each given phrase-containing image of the training image set may be labeled with information identifying the one or more phrases from the selected phrase list included in the given phrase-containing images. The model may be trained based on the training image set such that the model is configured to, in response to receiving an input image, output data indicating whether a phrase of the plurality of phrases is included in the input image.

Abstract: A light detection and ranging (LIDAR) device scans through a scanning zone while emitting light pulses and receives reflected signals corresponding to the light pulses. The LIDAR device scans the emitted light pulses through the scanning zone by reflecting the light pulses from an array of oscillating mirrors. The mirrors are operated by a set of electromagnets arranged to apply torque on the mirrors, and an orientation feedback system senses the orientations of the mirrors. Driving parameters for each mirror are determined based on information from the orientation feedback system. The driving parameters can be used to drive the mirrors in phase at an operating frequency despite variations in moments of inertia and resonant frequencies among the mirrors.

Abstract: The present disclosure relates to limitation of noise on light detectors using an aperture. One example implementation includes a system. The system includes a lens disposed relative to a scene. The lens focuses light from the scene. The system also includes an aperture defined within an opaque material. The system also includes a waveguide having a first side that receives light focused by the lens and transmitted through the aperture. The waveguide guides the received light toward a second side of the waveguide opposite to the first side. The waveguide has a third side extending between the first side and the second side. The system also includes an array of light detectors that intercepts and detects light propagating out of the third side of the waveguide.