Abstract: Systems, methods, and computer-readable media are disclosed for improved smart infrastructure data transfer. An example method may involve receiving, by a smart infrastructure device and from a first vehicle, first information associated with the first vehicle in a first format associated with a first communication protocol. The first information is converted from the first format into an agnostic format. An image, video, or real-time feed of an environment of the smart infrastructure device is captured. The first vehicle and a second vehicle in the image, video, or real-time feed is identified. It is determined that the second vehicle is temporarily or permanently incapable of performing a communication with the smart infrastructure device based on the image, video, or real-time feed. The image, video, or real-time feed is analyzed to generate second information associated with the second vehicle. The second information is converted into the agnostic format.

Abstract: Methods, systems, and products for generating an updated map for use with an autonomous vehicle driving operation or a simulation thereof may include obtaining first map data associated with a first map of a geographic location including a roadway, and the first map data may include at least one first lane segment. Second map data associated with a second map of the geographic location may be obtained, and the second map data may include at least one second lane segment. A plurality of non-overlapping areas may be determined based on the first lane segment(s) and the second lane segment(s). A first non-overlapping and/or a first warp point within the first non-overlapping area may be selected. The first lane segment(s) may be warped around the first warp point to increase a total overlapping area based on the based on the second lane segment(s) and the first lane segment(s) after warping.

Abstract: Provided are autonomous vehicles and methods of controlling autonomous vehicles through topological planning with bounds, including receiving map data and sensor data, expanding a topological tree by adding a plurality of nodes to represent a plurality of actions associated with the plurality of constraints, generating a bound based on a constraint in the geographic area, the bound associated with an action for navigating the autonomous vehicle relative to the at least one constraint, storing the bound in a central bound storage, linking a set of bounds of a tree node to the bound via a bound identifier, wherein the first bound is initially linked as an active bound, or alternatively, as an inactive bound after determining it is not the most restrictive bound at any sample index, and control the autonomous vehicle based on the topological tree, to navigate the plurality of constraints.

Abstract: Systems for managing access to an autonomous vehicle includes an autonomous vehicle including a plurality of storage compartments, wherein each of the plurality of storage compartments comprises a locking mechanism and at least one processor to receive data associated with an item to be positioned in a storage compartment of the plurality of storage compartments, determine that one of the plurality of storage compartments has storage capacity for the item, designate one of the plurality of storage compartments for storage of the item, activate the locking mechanism of the designated storage compartment to lock the designated storage compartment after the item is positioned in the designated storage compartment, and activate the locking mechanism of the designated storage compartment to unlock the designated storage compartment to allow removal of the item from the designated storage compartment. Methods, computer program products, and autonomous vehicles are also disclosed.

Abstract: Methods and systems for jointly estimating a pose and a shape of an object perceived by an autonomous vehicle are described. The system includes data and program code collectively defining a neural network which has been trained to jointly estimate a pose and a shape of a plurality of objects from incomplete point cloud data. The neural network includes a trained shared encoder neural network, a trained pose decoder neural network, and a trained shape decoder neural network. The method includes receiving an incomplete point cloud representation of an object, inputting the point cloud data into the trained shared encoder, outputting a code representative of the point cloud data. The method also includes generating an estimated pose and shape of the object based on the code. The pose includes at least a heading or a translation and the shape includes a denser point cloud representation of the object.

Abstract: A method of transferring data between an autonomous vehicle and a vehicle traffic control infrastructure system. The method includes receiving, by a communication device of a vehicle, a data payload for a smart traffic control infrastructure node. The method includes, by a computing device of the vehicle: determining that the vehicle is or will be within a communication range of the smart traffic control infrastructure node, determining a length of time that the vehicle will be in the communication range of the smart traffic control infrastructure node, and assembling a communication package with at least a portion of the data payload that can be transferred in the determined length of time. The method includes, by a communication device of the vehicle when the vehicle has an ad hoc communication link with the smart traffic control infrastructure node, transmitting the assembled communication package to the smart node.

Abstract: Disclosed herein are system, method, and computer program product aspects for enabling an autonomous vehicle (AV) to react to objects posing a risk to the AV. The system can monitor an object within a vicinity of the AV. A plurality of trajectories predicting paths the object will take can be generated, the plurality of trajectories being based on a plurality of inputs indicating current and past characteristics of the object. Using a learned model, a forecasted position of the object at an instance in time can be generated. An error value representing how accurate the forecasted position is versus an observed position of the object can be stored. Error values can be accumulated over a period of time. A risk factor can be assigned for the object based on the accumulated error values. A maneuver for the AV can be performed based on the risk factor.

Abstract: A method of scheduling a plurality of tasks in an autonomous vehicle system (AVS) includes, by a processor, prior to runtime of an autonomous vehicle, identifying a plurality of tasks to be implemented by the AVS of the autonomous vehicle, for each of the tasks, identifying at least one fixed parameter and at least one variable, and developing a schedule for each of the tasks. The schedule includes an event loop that minimizes an overall time for execution of the tasks. The method includes compiling the schedule into an execution plan, and saving the execution plan to a memory of the autonomous vehicle. During runtime of the autonomous vehicle, the processor receives data corresponding to the variables of the tasks, and uses the variables to implement the execution plan on the autonomous vehicle.

Abstract: Systems and methods for managing data. The methods comprise by a computing system: generating publication identifiers and version values for source data to be stored into a data warehouse; causing a plurality of fact tables in the data warehouse to be populated with the source data and the publication identifiers; causing a publication table in the data warehouse to be updated to include the publication identifiers and the version values so as to be respectively associated with resource names; receiving a query for information directed to the plurality of fact tables; retrieving the publication identifiers from the publication table, in response to the query; and obtaining source data from each said fact table of the plurality of fact tables that is associated with publication identifiers that are stored in both the fact table and the publication table.

Abstract: A system uses data captured by vehicle-mounted sensors to generate a view of a ground surface. The system does this by receiving digital image frames and associating a location and pose of the vehicle that captured the image with each digital image frame. The system will access a three dimensional (3D) ground surface estimation model of the ground surface, select a region of interest (ROI) of the ground surface, and select a vehicle pose. The system will identify digital image frames that are associated with the pose and also with a location that corresponds to the ROI. The system will generate a visual representation of the ground surface in the ROI by projecting ground data for the ROI from the ground surface estimation model to normalized 2D images that are created from the digital image frames. The system will save the visual representation to a two-dimensional grid.

Abstract: Methods, systems, and computer program products for navigating a vehicle are disclosed. The methods include extracting lane segment data associated with lane segments of a vector map that are within a region of interest, and analyzing the lane segment data and a heading of the vehicle to determine whether motion of the vehicle satisfies a condition. The condition can be associated with (i) an association between the heading of the vehicle and a direction of travel of a lane that corresponds to the current location of the vehicle and/or (ii) a minimum stopping distance to an imminent traffic control measure in the lane that corresponds to the current location of the vehicle. When the motion does not satisfy the condition, the methods include causing the vehicle to perform a motion correction.

Abstract: A removable brush head for a personal groom device, the removable brush head having a stationary portion, wherein the stationary portion is annular and includes a plurality of outwardly-extending brush bristles. The removable brush head also includes a movable portion positioned within the stationary portion, wherein the movable portion is configured to at least partially rotate relative to the stationary portion, and further wherein the movable portion includes a plurality of outwardly-extending brush bristles. Additionally, the removable brush head includes a locking collar, wherein the locking collar is slidably couplable to the stationary portion so as to axially retain the movable portion within the stationary portion while allowing the movable portion to at least partially rotate relative to the stationary portion.

Abstract: Disclosed herein are system, method, and computer program product embodiments for automated autonomous vehicle pose validation. An embodiment operates by generating a range image from a point cloud solution comprising a pose estimate for an autonomous vehicle. The embodiment queries the range image for predicted ranges and predicted class labels corresponding to lidar beams projected into the range image. The embodiment generates a vector of features from the range image. The embodiment compares a plurality of values to the vector of features using a binary classifier.

Abstract: An automated bootstrap process implemented as a simple state machine generates an initial pose for an autonomous vehicle, without reliance on human intervention. To trigger initiation of the bootstrap process automatically, the autonomous vehicle remains stationary. A GPS-derived position estimate, combined with lidar sweep data and HD map reference point cloud data, can be used to generate a pose using an iterative closest point algorithm. The bootstrap solution can then be automatically validated by a machine learning-based binary classifier trained with appropriate features. Full automation of the bootstrap process may facilitate launching a fleet service of autonomous vehicles.

Abstract: Devices, systems, and methods are provided for enhanced pointing angle validation. A device may generate a collimated beam using a light source emitting a light beam through a fiducial target in an optical instrument. The device may capture an image fiducial target using a camera, wherein the camera is mounted on a mounting datum that is orthogonal to the collimated beam. The device may determine a pointing angle associated with camera based on the captured image of the fiducial target. The device may compare a location of the fiducial target in the image to an optical center of the camera. The device may determine a validation status of camera based on the location of the fiducial target in the image.

Abstract: A system detects multiple instances of an object in a digital image by receiving a two-dimensional (2D) image that includes a plurality of instances of an object in an environment. For example, the system may receive the 2D image from a camera or other sensing modality of an autonomous vehicle (AV). The system uses a first object detection network to generate a plurality of predicted object instances in the image. The system then receives a data set that comprises depth information corresponding to the plurality of instances of the object in the environment. The data set may be received, for example, from a stereo camera of an AV, and the depth information may be in the form of a disparity map. The system may use the depth information to identify an individual instance from the plurality of predicted object instances in the image.

Abstract: Systems and methods for on-board selection of data logs for training a machine learning model are provided. The system includes an autonomous vehicle having a plurality of sensors and a processor. The processor receives a plurality of unlabeled images from the plurality of sensors, a machine learning model, and a loss function corresponding to the machine learning model. For each of the plurality of images, the processor then determines one or more predictions using the machine learning model, compute an importance function based on the loss function and the one or more predictions, and transmit that image to a remote server for updating the machine learning model when a value of the importance function is greater than a threshold.

Abstract: A method of verifying execution sequence integrity of an execution flow includes receiving, by a local monitor of an automated device monitoring system from one or more sensors of an automated device, a unique identifier for each function in a subset of an execution flow for which the local monitor is responsible for monitoring. The method includes combining the received unique identifiers to generate a combination value, applying a hashing algorithm to the combination value to generate a temporary hash value, retrieving, from a data store, a true hash value, determining whether the temporary hash value matches the true hash value, and in response to the temporary hash value not matching the true hash value, generating a fault notification. The true hash value represents a result of applying the hashing algorithm to a combination of actual unique identifiers associated with each function in the subset.

Abstract: This document describes methods by which a system determines a pickup/drop-off zone (PDZ) to which a vehicle will navigate to perform a ride service request. The system will define a PDZ that is a geometric interval that is within a lane of a road at the requested destination of the ride service request by: (i) accessing map data that includes the geometric interval; (ii) using the vehicle's length and the road's speed limit at the destination to calculate a minimum allowable length for the PDZ; (iii) setting, start point and end point boundaries for the PDZ having an intervening distance that is equal to or greater than the minimum allowable length; and (iv) positioning the PDZ in the lane at or within a threshold distance from the requested destination. The system will then generate a path to guide the vehicle to the PDZ.

Abstract: Systems and methods for generating a map. The methods comprise: performing, by a computing device, ray-casting operations to generate a 3D point cloud with a reduced number of data points associated with moving objects; generating, by the computing device, a 2D binary mask for at least one semantic label class of the 3D point cloud; determining, by the computing device, x-coordinates and y-coordinates for a 2D volume defining an object of the at least one semantic label class; identifying, by the computing device, data points in the 3D point cloud based on the 2D volume; comparing, by the computing device, z-coordinates of the identified data points to at least one threshold value selected for the at least one semantic label class; and generating, by the computing device, the map by removing data points from the 3D point cloud based on results of the comparing.