Abstract: Systems and methods are provided for forecasting the motion of actors within a surrounding environment of an autonomous platform. For example, a computing system of an autonomous platform can use machine-learned model(s) to generate actor-specific graphs with past motions of actors and the local map topology. The computing system can project the actor-specific graphs of all actors to a global graph. The global graph can allow the computing system to determine which actors may interact with one another by propagating information over the global graph. The computing system can distribute the interactions determined using the global graph to the individual actor-specific graphs. The computing system can then predict a motion trajectory for an actor based on the associated actor-specific graph, which captures the actor-to-actor interactions and actor-to-map relations.

Abstract: The present disclosure provides systems and methods that combine physics-based systems with machine learning to generate synthetic LiDAR data that accurately mimics a real-world LiDAR sensor system. In particular, aspects of the present disclosure combine physics-based rendering with machine-learned models such as deep neural networks to simulate both the geometry and intensity of the LiDAR sensor. As one example, a physics-based ray casting approach can be used on a three-dimensional map of an environment to generate an initial three-dimensional point cloud that mimics LiDAR data. According to an aspect of the present disclosure, a machine-learned geometry model can predict one or more adjusted depths for one or more of the points in the initial three-dimensional point cloud, thereby generating an adjusted three-dimensional point cloud which more realistically simulates real-world LiDAR data.

Abstract: Systems and methods are directed to allocating unused or otherwise under-utilized computing resources of autonomous vehicles. In one example, a computer-implemented method obtaining, by a computing system, data describing a computational status of each autonomous vehicle of one or more autonomous vehicles describing a current or forecasted computational load. The method includes determining, by the computing system, an amount of excess computational capacity of each autonomous vehicle of the one or more autonomous vehicles, the amount of excess computational capacity for each autonomous vehicle of the one or more autonomous vehicles based at least in part on the computational status of the autonomous vehicle and a total computational capacity of the autonomous vehicle. The method includes allocating, by the computing system, at least a portion of the amount of excess computational capacity of each autonomous vehicle to processing operations associated with participation in a distributed ledger.

Abstract: Aspects of the present disclosure involve systems, methods, and devices for autonomous vehicle localization using a Lidar intensity map. A system is configured to generate a map embedding using a first neural network and to generate an online Lidar intensity embedding using a second neural network. The map embedding is based on input map data comprising a Lidar intensity map, and the Lidar sweep embedding is based on online Lidar sweep data. The system is further configured to generate multiple pose candidates based on the online Lidar intensity embedding and compute a three-dimensional (3D) score map comprising a match score for each pose candidate that indicates a similarity between the pose candidate and the map embedding. The system is further configured to determine a pose of a vehicle based on the 3D score map and to control one or more operations of the vehicle based on the determined pose.

Abstract: Aspects of the present disclosure involve a vehicle computer system comprising a computer-readable storage medium storing a set of instructions, and a method for light detection and ranging (Lidar) intensity calibration. The method includes collecting a data set comprising a plurality of raw intensity values output by a channel of a Lidar unit at a particular power level from among multiple power levels at which the channel is capable of operating. The method further includes using a linear model to compute a calibration multiplier and a bias value for the particular power level of the channel. During operation of the vehicle, calibrated intensity values are determined by applying the linear model to subsequent raw intensity values output by the channel at the particular power using the determined calibration multiplier and bias value.

Abstract: Generally, the present disclosure is directed to systems and methods for streaming processing within one or more systems of an autonomy computing system. When an update for a particular object or region of interest is received by a given system, the system can control transmission of data associated with the update as well as a determination of other aspects by the given system. For example, the system can determine based on a received update for a particular aspect and a priority classification and/or interaction classification determined for that aspect whether data associated with the update should be transmitted to a subsequent system before waiting for other updates to arrive.

Abstract: Systems, methods, tangible non-transitory computer-readable media, and devices associated with object localization and generation of compressed feature representations are provided. For example, a computing system can access source data and target data. The source data can include a source representation of an environment including a source object. The target data can include a compressed target feature representation of the environment. The compressed target feature representation can be based on compression of a target feature representation of the environment produced by machine-learned models. A source feature representation can be generated based on the source representation and the machine-learned models. The machine-learned models can include machine-learned feature extraction models or machine-learned attention models. A localized state of the source object with respect to the environment can be determined based on the source feature representation and the compressed target feature representation.

Abstract: Generally, the present disclosure is directed to systems and methods that include or otherwise leverage an iterative solver as part of optimizing a motion plan for an autonomous vehicle (AV). In particular, a scenario generator within a motion planning system can include a warm start generator configured to determine an initial trajectory that respects the dynamics of the autonomous vehicle and that closely tracks a speed profile determined by a constraint solver and one or more nominal paths determined by a route selector. A decision validator can analyze speed profiles and nominal paths to identify potential inconsistencies and validate a decision before optimization and execution. An initial trajectory can be further optimized by an iterative solver to determine an optimized trajectory for execution as a motion plan for the autonomous vehicle.

Abstract: Systems and methods for determining object prioritization and predicting future object locations for an autonomous vehicle are provided. A method can include obtaining, by a computing system comprising one or more processors, state data descriptive of at least a current or past state of a plurality of objects that are perceived by an autonomous vehicle. The method can further include determining, by the computing system, a priority classification for each object in the plurality of objects based at least in part on the respective state data for each object. The method can further include determining, by the computing system, an order at which the computing system determines a predicted future state for each object based at least in part on the priority classification for each object and determining, by the computing system, the predicted future state for each object based at least in part on the determined order.

Abstract: Systems and methods for controlling an autonomous vehicle are provided. In one example embodiment, a computer-implemented method includes receiving data indicative of an operating mode of the vehicle, wherein the vehicle is configured to operate in a plurality of operating modes. The method includes determining one or more response characteristics of the vehicle based at least in part on the operating mode of the vehicle, each response characteristic indicating how the vehicle responds to a potential collision. The method includes controlling the vehicle based at least in part on the one or more response characteristics.

Abstract: Systems, methods, tangible non-transitory computer-readable media, and devices associated with sensor output segmentation are provided. For example, sensor data can be accessed. The sensor data can include sensor data returns representative of an environment detected by a sensor across the sensor's field of view. Each sensor data return can be associated with a respective bin of a plurality of bins corresponding to the field of view of the sensor. Each bin can correspond to a different portion of the sensor's field of view. Channels can be generated for each of the plurality of bins and can include data indicative of a range and an azimuth associated with a sensor data return associated with each bin. Furthermore, a semantic segment of a portion of the sensor data can be generated by inputting the channels for each bin into a machine-learned segmentation model trained to generate an output including the semantic segment.

Abstract: A sensor assembly can include a housing that includes a view pane and a mounting feature configured to replace a trailer light of a cargo trailer of a semi-trailer truck. The sensor assembly can also include a lighting element mounted within the housing to selectively generate light, and a sensor mounted within the housing and having a field of view through the view pane. The sensor assembly can also include a communication interface configured to transmit sensor data from the sensor to a control system of the self-driving tractor.

Abstract: Systems and methods are directed to an autonomy computing system of an autonomous vehicle. The autonomy computing system can include first functional circuitry configured to generate a first output associated with a first autonomous compute function of the autonomous vehicle based on sensor data using first neural networks. The autonomy computing system can include second functional circuitry configured to generate a second output associated with the first autonomous compute function of the autonomous vehicle based on the sensor data and neural networks. The autonomy computing system can include monitoring circuitry configured to determine a difference between the first output of the first functional circuitry and the second output of the second functional circuitry. The autonomy computing system can include a vehicle control system configured to generate vehicle control signals for the autonomous vehicle based on the outputs.

Abstract: A computing system can be configured to input data that describes sensor data into an object detection model and receive, as an output of the object detection model, object detection data describing features of the plurality of the actors relative to the autonomous vehicle. The computing system can generate an input sequence that describes the object detection data. The computing system can analyze the input sequence using an interaction model to produce, as an output of the interaction model, an attention embedding with respect to the plurality of actors. The computing system can be configured to input the attention embedding into a recurrent model and determine respective trajectories for the plurality of actors based on motion forecast data received as an output of the recurrent model.

Abstract: Systems and methods are directed to determining autonomous vehicle scenarios based on autonomous vehicle operation data. In one example, a computer-implemented method for determining operating scenarios for an autonomous vehicle includes obtaining, by a computing system comprising one or more computing devices, log data representing autonomous vehicle operations. The method further includes extracting, by the computing system, a plurality of attributes from the log data. The method further includes determining, by the computing system, one or more scenarios based on a combination of the attributes, wherein each scenario includes multiple scenario variations and each scenario variation comprises multiple features. The method further includes providing, by the computing system, the one or more scenarios for generating autonomous vehicle operation analytics.

Abstract: The present disclosure is directed to analyzing motion plans of autonomous vehicles. In particular, the methods, devices, and systems of the present disclosure can: receive data indicating a motion plan, of an autonomous vehicle through an environment, based at least in part on multiple different constraints of the environment; and determine, for each constraint of the multiple different constraints, a measure of an influence of the constraint on the motion plan.

Abstract: Systems and methods for vehicle-to-vehicle communications are provided. An example computer-implemented method includes obtaining from a first autonomous vehicle, by a second autonomous vehicle, a first compressed intermediate environmental representation. The first compressed intermediate environmental representation is indicative of at least a portion of an environment of the second autonomous vehicle. The method includes generating a first decompressed intermediate environmental representation by decompressing the first compressed intermediate environmental representation. The method includes determining, using one or more machine-learned models, an updated intermediate environmental representation based at least in part on the first decompressed intermediate environmental representation and a second intermediate environmental representation generated by the second autonomous vehicle.

Abstract: Systems and methods for training object detection models using adversarial examples are provided. A method includes obtaining a training scene and identifying a target object within the training scene. The method includes obtaining an adversarial object and generating a modified training scene based on the adversarial object, the target object, and the training scene. The modified training scene includes the training scene modified to include the adversarial object placed on the target object. The modified training scene is input to a machine-learned model configured to detect the training object. A detection score is determined based on whether the training object is detected, and the machine-learned model and the parameters of the adversarial object are trained based on the detection output. The machine-learned model is trained to maximize the detection output. The parameters of the adversarial object are trained to minimize the detection output.

Abstract: Systems and methods for facilitating communication with autonomous vehicles are provided. In one example embodiment, a computing system can obtain a first type of sensor data (e.g., camera image data) associated with a surrounding environment of an autonomous vehicle and/or a second type of sensor data (e.g., LIDAR data) associated with the surrounding environment of the autonomous vehicle. The computing system can generate overhead image data indicative of at least a portion of the surrounding environment of the autonomous vehicle based at least in part on the first and/or second types of sensor data. The computing system can determine one or more lane boundaries within the surrounding environment of the autonomous vehicle based at least in part on the overhead image data indicative of at least the portion of the surrounding environment of the autonomous vehicle and a machine-learned lane boundary detection model.

Abstract: A LIDAR unit includes a housing defining a cavity. The LIDAR unit further include a plurality of emitters disposed on a circuit board within the cavity. Each of the emitters emits a laser beam along a transmit path. The LIDAR system further includes a first telecentric lens assembly positioned within the cavity and along the transmit path such that the laser beam emitted from each of the plurality of emitters passes through the first telecentric lens assembly. The LIDAR further includes a second telecentric lens assembly positioned within the cavity and along a receive path such that a plurality of reflected laser beams entering the cavity pass through the second telecentric lens assembly. The first telecentric lens assembly and the second telecentric lens assembly each include a field flattening lens and at least one other lens.