Abstract: Example embodiments relate to crosstalk reduction for light detection and ranging (lidar) devices using wavelength locking. An example embodiment includes a lidar device. The lidar device includes a first light emitter configured to emit a first light signal and a second light emitter configured to emit a second light signal. The lidar device also includes a first light guide and a second light guide. In addition, the lidar device includes a first light detector and a second light detector. Further, the lidar device includes a first wavelength-locking mechanism configured to use a portion of the first light signal to maintain a wavelength of the first light signal and a second wavelength-locking mechanism configured to use a portion of the second light signal to maintain a wavelength of the second light signal. The wavelengths of the first light signal and the second light signal are different.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for determining occupancies of surrounding agents. One of the methods includes obtaining scene data characterizing an environment at a current time point; processing a first network input generated from the scene data using a first neural network to generate an intermediate output; obtaining an identification of a future time point that is after the current time point; and generating, from the intermediate output and the future time point, an occupancy output, wherein the occupancy output comprises respective occupancy probabilities for each of a plurality of locations in the environment, wherein the respective occupancy probability for each location characterizes a likelihood that one or more agents will occupy the location at the future time point.

Abstract: Methods, computer systems, and apparatus, including computer programs encoded on computer storage media, for processing a perspective view range image generated from sensor measurements of an environment. The perspective view range image includes a plurality of pixels arranged in a two-dimensional grid and including, for each pixel, (i) features of one or more sensor measurements at a location in the environment corresponding to the pixel and (ii) geometry information comprising range features characterizing a range of the location in the environment corresponding to the pixel relative to the one or more sensors. The system processes the perspective view range image using a first neural network to generate an output feature representation. The first neural network comprises a first perspective point-set aggregation layer comprising a geometry-dependent kernel.

Abstract: Aspects of the disclosure relate to controlling a vehicle having an autonomous driving mode or an autonomous vehicle. For instance, a polygon representative of the shape and location of a first object may be received. A polyline contour representation of a portion of a polygon representative of the shape and location of a second object may be received. The polyline contour representation may be in half-plane coordinates and including a plurality of vertices and line segments. Coordinates of the polygon representative of the first object may be converted to the half-plane coordinate system. A collision location between the polyline contour representation and the polygon representative of the first object may be determined using the converted coordinates. The autonomous vehicle may be controlled in the autonomous driving mode to avoid a collision based on the collision location.

Abstract: The present disclosure relates to optical systems and methods for their manufacture. An example method includes coupling a first surface of a light-emitter substrate to a reference surface of a carrier substrate. The method also includes registering a mold structure with respect to the reference surface of the carrier substrate. Furthermore, the method includes using the mold structure to form an optical material over at least a portion of the light-emitter substrate. The optical material is shaped according to a shape of the mold structure and includes at least one registration feature. The method also includes coupling an optical lens element to the optical material such that the optical lens element is registered to the at least one registration feature.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing inputs using slice-based dynamic neural networks. One of the methods includes receiving a new input for processing by a neural network that includes a first conditional neural network layer that has a set of shared parameters and a respective set of slice parameters for each of a plurality of slices. One or more slices to which the new input belongs are identified. The new input is processed to generate a network output, including: receiving a layer input to the first conditional neural network layer; and processing the layer input using the set of shared parameters, the respective one or more sets of slice parameters for the identified one or more slices, but not the respective sets of slice parameters for any other slices to which the new input does not belong.

Abstract: The technology relates to maneuvering a vehicle prior to making a turn from a current lane of the vehicle. As an example, a route that includes making the turn from the lane is identified. An area of the lane prior to the turn having a lane width of at least a predetermined size is also identified. Sensor data identifying an object within the lane is received from a perception system of a vehicle. Characteristics of the object, including a location of the object relative to the vehicle, are identified from the sensor data. The vehicle is then maneuvered through the area prior to making the turn using the identified characteristics.

Abstract: Aspects of the disclosure relate to training a machine learning model on a distributed computing system. The model can be trained using selected processors of the training platform. The distributed system automatically modifies the model for instantiation on each processor, adjusts an input pipeline to accommodate the capabilities of selected processors, and coordinates the training between those processors. Simultaneous processing at each stage can be scaled to reduce or eliminate bottlenecks in the distributed system. In addition, autonomous monitoring and re-allocating of resources can further reduce or eliminate bottlenecks. The training results may be aggregated by the distributed system, and a final model may then be transmitted to a user device.

Abstract: The technology relates to a seamless interface between an autonomous vehicle service provider and one or more ridesharing, ride-hailing or delivery partner companies, in order to provide timely, efficient rider or delivery services and support. A partner trip application programming interface (API) provides robust features to the ridesharing or ride-hailing partner companies and also supports delivery of meals, groceries, packages or other cargo. For instance, the API may be employed with delivery partners (e.g., food, package or bulk cargo delivery), and/or for concierge services (e.g., a hotel or medical provider or an e-commerce specialist that coordinates deliveries). The technology supports both ad hoc and scheduled trips. Agents for different partners may schedule trips to or from specific stores or other (geofenced) locations, and observe and manage trips for an entire enterprise. Permissions can be tailored for each agent.

Abstract: One example method involves obtaining a plurality of scans of a field-of-view (FOV) of a light detection and ranging (LIDAR) device disposed inside a housing. Obtaining each scan of the plurality of scans comprises: transmitting, through a plurality of sections of the housing, a plurality of light pulses emitted from the LIDAR device in different directions toward the housing; and detecting a plurality of returning light pulses comprising reflected portions of the transmitted plurality of light pulses that are reflected back toward the LIDAR device. The method also involves detecting an obstruction that at least partially occludes the LIDAR device from scanning the FOV through the housing based on the plurality of scans.

Abstract: The technology relates to controlling a vehicle in an autonomous driving mode. In one instance, sensor data identifying an object in an environment of the vehicle may be received. A first path of a first trajectory where the vehicle will pass the object may be determined. A function is used to determining a first maximum speed of the vehicle based on a predetermined minimum lateral clearance between the object and the vehicle. The first maximum speed may be used to determine whether an actual lateral clearance between the object and the vehicle will meet the predetermined minimum lateral clearance. The determination of whether the actual lateral clearance will meet the predetermined minimum lateral clearance may be used to generate a first speed plan for the first trajectory. The vehicle may be controlled in the autonomous driving mode according to the first trajectory including the first speed plan and the first path.

Abstract: The present disclosure relates to contaminant detection systems and related optical systems and methods. An example contaminant detection system includes an optical coupler configured to couple light into and/or out of an optical element. The contaminant detection system also includes a plurality of light-emitter devices configured to emit emission light toward the optical coupler. The contaminant detection system additionally includes a plurality of detector devices configured to detect at least a portion of the emission light by way of the optical element and the optical coupler. The plurality of detector devices is also configured to provide detector signals indicative of a presence of a contaminant on the optical element.

Abstract: Methods, systems, and apparatus, for a stray-light testing station. In one aspect, the stray-light testing station includes an illumination assembly including a spatially extended light source and one or more optical elements arranged to direct a beam of light from the spatially extended light source along an optical path to an optical receiver assembly including a lens receptacle configured to receive a lens module and position the lens module in the optical path downstream from the parabolic mirror so that the lens module focuses the beam of light from the spatially extended light source to an image plane, and a moveable frame supporting the optical receiver assembly including one or more adjustable alignment stages to position the optical receiver assembly relative to the illumination assembly such that the optical path of the illumination assembly is within a field of view of the optical receiver assembly.

Abstract: A system may include a first sensor of a first type and a second sensor of a second different type and having a detector. A field of view of the second sensor may be formed by a plurality of regions of interest (ROIs) defined by the detector. Control circuitry of the system may be configured to perform operations including obtaining, from the first sensor, first sensor data representing an environment, and determining, based on the first sensor data, information associated with a feature of interest within the environment. The operations may also include determining, based on the information, a particular ROI that corresponds to an expected position of the feature at a later time, obtaining a plurality of ROI sensor data from the particular ROI instead of obtaining full-resolution sensor data, and analyzing the plurality of ROI sensor data to determine one or more attributes of the feature.

Abstract: Methods, computer systems, and apparatus, including computer programs encoded on computer storage media, for performing a conditional behavior prediction for one or more agents. The system obtains context data characterizing an environment. The context data includes data characterizing a plurality of agents, including a query agent and one or more target agents, in the environment at a current time point. The system further obtains data identifying a planned future trajectory for the query agent after the current time point, and for each target agent in the set, processes the context data and the data identifying the planned future trajectory using a first neural network to generate a conditional trajectory prediction output that defines a conditional probability distribution over possible future trajectories of the target agent after the current time point given that the query agent follows the planned future trajectory for the query agent after the current time point.

Abstract: A computing system may operate a LIDAR device to emit and detect light pulses in accordance with a time sequence including standard detection period(s) that establish a nominal detection range for the LIDAR device and extended detection period(s) having durations longer than those of the standard detection period(s). The system may then make a determination that the LIDAR detected return light pulse(s) during extended detection period(s) that correspond to particular emitted light pulse(s). Responsively, the computing system may determine that the detected return light pulse(s) have detection times relative to corresponding emission times of particular emitted light pulse(s) that are indicative of one or more ranges. Given this, the computing system may make a further determination of whether or not the one or more ranges indicate that an object is positioned outside of the nominal detection range, and may then engage in object detection in accordance with the further determination.

Abstract: Aspects of the disclosure provide for controlling an autonomous vehicle. For instance, a first probability distribution may be generated for the vehicle at a first future point in time using a generative model for predicting expected behaviors of objects and a set of characteristics for the vehicle at an initial time expected to be perceived by an observer. Planning system software of the vehicle may be used to generate a trajectory for the vehicle to follow. A second probability distribution may be generated for a second future point in time using the generative model based on the trajectory and a set of characteristics for the vehicle at the first future point expected to be perceived by the observer. A surprise assessment may be generated by comparing the first probability distribution to the second probability distribution. The vehicle may be controlled based on the surprise assessment.

Abstract: The technology relates to assisting large self-driving vehicles, such as cargo vehicles, as they maneuver towards and/or park at a destination facility. This may include a given vehicle transitioning between different autonomous driving modes. Such a vehicles may be permitted to drive in a fully autonomous mode on certain roadways for the majority of a trip, but may need to change to a partially autonomous mode on other roadways or when entering or leaving a destination facility such as a warehouse, depot or service center. Large vehicles such as cargo truck may have limited room to maneuver in and park at the destination, which may also prevent operation in a fully autonomous mode. Here, information from the destination facility and/or a remote assistance service can be employed to aid in real-time semi-autonomous maneuvering.

Abstract: The technology relates to assisting large self-driving vehicles, such as cargo vehicles, as they maneuver towards and/or park at a destination facility. This may include a given vehicle transitioning between different autonomous driving modes. Such a vehicles may be permitted to drive in a fully autonomous mode on certain roadways for the majority of a trip, but may need to change to a partially autonomous mode on other roadways or when entering or leaving a destination facility such as a warehouse, depot or service center. Large vehicles such as cargo truck may have limited room to maneuver in and park at the destination, which may also prevent operation in a fully autonomous mode. Here, information from the destination facility and/or a remote assistance service can be employed to aid in real-time semi-autonomous maneuvering.