Abstract: A computer-implemented method is provided for creating a velocity grid using a simulated environment for use by an autonomous vehicle. The method may include simulating a road scenario with simulated objects. The method may also include recording image data collected from a camera sensor, the image data comprising a first 2D image frame comprising the simulated objects made up of a plurality of pixels. The method may also include identifying a first 3D point on the first simulated object in a 3D view of the simulated road scenario, wherein the first 3D point corresponds to the first pixel in the first 2D image frame. The method may also include generating a velocity of the first point based upon a velocity of the first simulated object, and projecting the velocity back into the first 2D image frame.

Abstract: A computer-implemented method is provided for training a machine-learning (ML) algorithm that contributes to piloting an autonomous vehicle using a velocity grid generated from a simulation. The method may include simulating a scene. The scene includes a simulated autonomous vehicle and at least one simulated object moving in the scene. The method may also include predicting a velocity of at least one point on the simulated object as it moves in the simulated scene using a ML algorithm. The method may also include comparing the predicted velocity of the at least one point on the simulated object with velocities in the velocity grid generated from a simulation. The method may further include adjusting the ML algorithm to more accurately predict velocity of at least one point on the simulated object as it moves in the scene based on a difference in the predicted velocity compared to the velocity grid, which yields a trained ML algorithm.

Abstract: A computed-implemented method is provided for creating a velocity grid using a simulated environment for use by an autonomous vehicle. The method may include simulating a road scenario with simulated objects, collecting first LiDAR data from simulated LiDAR sensors in the simulated road scenario, wherein the collected first LiDAR data comprises a first plurality of points that are representative of a first simulated object at a first 3D location and a first time. The method may also include transforming the first plurality of points from a simulated-scene frame-of-reference to a first simulated object frame-of-reference, and simulating the first simulated object to move from the first 3D location to a second 3D location within the simulated road scenario between the first time and a second time.

Abstract: The disclosed technology provides solutions for generating accurate virtual representations of real-world environments. A process of the disclosed technology can include steps for: processing image sensor data to identify one or more images of road paint; identifying a geographic location that is associated with the one or more images of road paint; and generating, within a simulated environment corresponding to the geographic location, at least one road paint mesh object that is based on the one or more images of road paint.

Abstract: A computer-implemented method is provided for creating a velocity grid of a simulated object in a simulated environment for use by an autonomous vehicle. The method may include simulating a road scenario with simulated objects. The method may also include tracking a point representative of a portion of one of the simulated objects in the simulated road scenario, wherein the point representative of the portion of one of the simulated objects moves from a first location to a second location in a time period, wherein the velocity of the point is generated from the first location to the second location. The method may further include storing the velocity of the point representative of the portion of one of the simulated objects in a velocity grid for the simulated objects in a memory device in an electrical communication with a processor.

Abstract: The disclosed technology provides solutions for generating accurate virtual representations of real-world environments. A process of the disclosed technology can include steps for: receiving map data corresponding to a geographic region, receiving image data for at least a portion of the geographic region that includes one or more images of road paint, and rendering one or more road paint mesh objects corresponding to the one or more images of the road paint, wherein a shape of the one or more road paint mesh objects is based on the map data. In some aspects, the process can further include generating a virtual environment corresponding to the geographic region that includes the one or more road paint mesh objects.

Abstract: Systems, methods, and computer-readable media are disclosed for a quick evaluation system for three-dimensional (3D) assets used in simulations for an autonomous vehicle (AV). A disclosed method comprises receiving drive data recorded in a physical environment by a vehicle having a first sensor; simulating the first sensor associated with a virtual autonomous vehicle in a virtual environment of a 3D scene including an object that at least partially corresponds to the physical environment; evaluating simulated data based on the simulation of the first sensor using a machine learning (ML) model; comparing evaluation data recorded during the evaluation of the simulation of the first sensor using the ML model to the drive data recorded in the physical environment; and generating a report based on a comparison of the evaluation data to a portion of the drive data to determine metrics associated with the object in the virtual environment.

Abstract: The disclosed technology provides solutions for generating accurate virtual representations of real-world environments. A process of the disclosed technology can include steps for: receiving map data corresponding to a geographic region, receiving image data for at least a portion of the geographic region that includes one or more images of road paint, and rendering one or more road paint mesh objects corresponding to the one or more images of the road paint, wherein a shape of the one or more road paint mesh objects is based on the map data. In some aspects, the process can further include generating a virtual environment corresponding to the geographic region that includes the one or more road paint mesh objects.

Abstract: The disclosure provides a method for deriving a surface absorption property of a surface material from LIDAR data. The method may include receiving cloud points from a LIDAR sensor. The cloud points are characterized by intensity values corresponding to reflections off the surface material of a real-world object and respective 3D coordinates. The method may also include constructing surface segments of the surface material by building triangles from three neighboring cloud points, determining the normal direction for each triangle from each cloud point, and deriving the surface absorption property for the segment of the surface material based upon the average intensity values for the three neighboring cloud points, a known angle of incidence, and the determined normal direction for each of the respective triangle. The method may further include aggregating the derived surface absorption property across a collection of respective triangles to result in the surface absorption property for a type of surface.

Abstract: A classification model can be trained by using an augmented scene. The augmented scene may be generated by placing augmentation objects into a simulated scene that is a virtual representation of a real-world scene. The augmentation objects are virtual objects representing a category for which a reference classification model has a poor performance. The reference classification model may be a simulated model trained by using a simulated scene or a real-world model trained by using a real-world scene. A training set, which includes simulated sensor data of the augmentation objects and labels of the augmentation objects, can be used to train the augmented model. The augmented model can be used by an AV to classify objects in the surrounding environment of the AV. The augmented model can have a better accuracy in classifying objects in the category than the reference classification model.

Abstract: Systems and methods for dynamic sensor data adaptation using a deep learning loop are provided. A method includes classifying, using a discriminator model, a first object from first sensor data associated with a first sensing condition, wherein the discriminator model is trained for a second sensing condition different from the first sensing condition; generating, using a generator model in response to the discriminator model failing to classify the first object, second sensor data representing a second object comprising at least a modified element of the first object; classifying, using the discriminator model, the second object from the second sensor data; and adapting, based at least in part on a difference between the first object and the second object in response to the discriminator model successfully classifying the second object, a machine learning model associated with object classification for the first sensing condition.

Abstract: Techniques for simulating LIDAR data are described. In one embodiment, a method for simulating LIDAR data may include retrieving a simulated scene that simulates a real-world scene, the simulated scene including at least one target object having a reflectivity r and located at a range R from a LIDAR sensor, the LIDAR sensor having at least one intrinsic parameter; generating a probability of detection (Pd) drop-off function for the LIDAR sensor, wherein the Pd drop-off function is related to r, R, and the at least one intrinsic parameter; for each data point comprising a ray emitted by the LIDAR sensor that hits the target object, generating a Pd value using the Pd drop-off function; and determining based on the Pd value whether to drop the data point.

Abstract: A computer-implemented method is provided for simulating sensor data acquisition. The method may include receiving simulated sensor data corresponding with a synthetic object in a simulated three-dimensional (3D) environment. The method may also include identifying an object type corresponding with the synthetic object based on a location of the synthetic object in the simulated 3D environment. The method may further include associating an intensity value with the simulated sensor data based on the object type for the synthetic object.

Abstract: The subject disclosure relates to techniques for generating localized atmospheric phenomena in a simulated environment. A process of the disclosed technology can include generating a plurality of volumetric sequences, generating a corresponding plurality of sequence slices for each of the plurality of volumetric sequences, and compiling the plurality of volumetric sequences to generate a synthetic localized atmospheric event.

Abstract: A computer-implemented method is provided for training a machine-learning (ML) algorithm that contributes to piloting an autonomous vehicle using a velocity grid generated from a simulation. The method may include simulating a scene. The scene includes a simulated autonomous vehicle and at least one simulated object moving in the scene. The method may also include predicting a velocity of at least one point on the simulated object as it moves in the simulated scene using a ML algorithm. The method may also include comparing the predicted velocity of the at least one point on the simulated object with velocities in the velocity grid generated from a simulation. The method may further include adjusting the ML algorithm to more accurately predict velocity of at least one point on the simulated object as it moves in the scene based on a difference in the predicted velocity compared to the velocity grid, which yields a trained ML algorithm.

Abstract: A computed-implemented method is provided for creating a velocity grid using a simulated environment for use by an autonomous vehicle. The method may include simulating a road scenario with simulated objects, collecting first LiDAR data from simulated LiDAR sensors in the simulated road scenario, wherein the collected first LiDAR data comprises a first plurality of points that are representative of a first simulated object at a first 3D location and a first time. The method may also include transforming the first plurality of points from a simulated-scene frame-of-reference to a first simulated object frame-of-reference, and simulating the first simulated object to move from the first 3D location to a second 3D location within the simulated road scenario between the first time and a second time.

Abstract: The subject disclosure relates to techniques for integrating synthetic plants into a virtual scene. A process of the disclosed technology can include receiving a digital asset comprising synthetic foliage, processing the digital asset to modify at least one parameter associated with the synthetic foliage to generate a modified digital asset, acquiring synthetic sensor data corresponding with the modified digital asset, and calculating a classification score for the modified digital asset based on the synthetic sensor data.

Abstract: A computer-implemented method is provided for creating a velocity grid using a simulated environment for use by an autonomous vehicle. The method may include simulating a road scenario with simulated objects. The method may also include recording image data collected from a camera sensor, the image data comprising a first 2D image frame comprising the simulated objects made up of a plurality of pixels. The method may also include identifying a first 3D point on the first simulated object in a 3D view of the simulated road scenario, wherein the first 3D point corresponds to the first pixel in the first 2D image frame. The method may also include generating a velocity of the first point based upon a velocity of the first simulated object, and projecting the velocity back into the first 2D image frame.

Abstract: A computer-implemented method is provided for creating a velocity grid of a simulated object in a simulated environment for use by an autonomous vehicle. The method may include simulating a road scenario with simulated objects. The method may also include tracking a point representative of a portion of one of the simulated objects in the simulated road scenario, wherein the point representative of the portion of one of the simulated objects moves from a first location to a second location in a time period, wherein the velocity of the point is generated from the first location to the second location. The method may further include storing the velocity of the point representative of the portion of one of the simulated objects in a velocity grid for the simulated objects in a memory device in an electrical communication with a processor.

Abstract: A glove interface object, including: a flex sensor configured to generate flex sensor data identifying a flex of a finger portion of the glove interface object; a trackable object configured to be illuminated during interactivity, the trackable object being positioned at a wrist portion of the glove interface object; a communications module configured to transmit the flex sensor data to a computing device for processing to determine a finger position pose of the glove interface object, applied for rendering a virtual hand in a view of a virtual environment on a head-mounted display (HMD), the virtual hand rendered based on the identified finger position pose, the computing device identifying the trackable object from captured image data to track a location of the glove interface object; wherein the virtual hand is rendered at a location in the virtual environment that is substantially defined by the location of the glove interface object.