Abstract: The present disclosure provides systems and methods that control the motion of an autonomous vehicle by rewarding or otherwise encouraging progress toward a goal, rather than simply rewarding distance travelled. In particular, the systems and methods of the present disclosure can project a candidate motion plan that describes a proposed motion path for the autonomous vehicle onto a nominal pathway to determine a projected distance associated with the candidate motion plan. The systems and methods of the present disclosure can use the projected distance to evaluate a reward function that provides a reward that is positively correlated to the magnitude of the projected distance. The motion of the vehicle can be controlled based on the reward value provided by the reward function. For example, the candidate motion plan can be selected for implementation or revised based at least in part on the determined reward value.

Abstract: The present disclosure is directed to validating motion plans for 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 of the autonomous vehicle; receive data indicating one or more inputs utilized in generating the motion plan; and determine, based at least in part on the data indicating the motion plan and the data indicating the input(s), whether execution of the motion plan by the autonomous vehicle would violate one or more predetermined constraints applicable to motion plans for the autonomous vehicle.

Abstract: Systems and methods for using parametric modeling to design scenarios for autonomous vehicle simulation are provided. In particular, a computing system can obtain data identifying a plurality of parameters, each parameter associated with a particular scenario component. The computing system can determine values associated with a first set of parameters in the plurality of parameters. The computing system can determine one or more parameter relationships, such that values associated with a second set of parameters in the plurality of parameters are determined, at least in part, based on the values associated with the first set of parameters. The computing system can initiate a simulation of a scenario based, at least in part, on the values associated with the first set of the parameters and the one or more parameter relationships. The computing system can determine whether the simulated autonomous vehicle has successfully completed the scenario.

Abstract: Systems and methods are directed to generating behavioral predictions in reaction to autonomous vehicle movement. In one example, a computer-implemented method includes obtaining, by a computing system, local scene data associated with an environment external to an autonomous vehicle, the local scene data including actor data for an actor in the environment external to the autonomous vehicle. The method includes extracting, by the computing system and from the local scene data, one or more actor prediction parameters for the actor using a machine-learned parameter extraction model. The method includes determining, by the computing system, a candidate motion plan for the autonomous vehicle. The method includes generating, by the computing system and using a machine-learned prediction model, a reactive prediction for the actor based at least in part on the one or more actor prediction parameters and the candidate motion plan.

Abstract: An on-demand transportation management system can collect vehicle fleet utilization data corresponding to human-driven vehicles (HDVs) and autonomous vehicles (AVs) operating within a given region in connection with an on-demand transportation service. The on-demand transportation management system can then establish a set of selection priorities for respective areas of the given region based on the vehicle fleet utilization data, each selection priority indicating whether a respective area of the given region is to favor HDVs or AVs for servicing transport requests.

Abstract: Systems and methods for detecting and classifying objects proximate to an autonomous vehicle can include a sensor system and a vehicle computing system. The sensor system includes at least one LIDAR system configured to transmit ranging signals relative to the autonomous vehicle and to generate LIDAR data. The vehicle computing system receives the LIDAR data from the sensor system. The vehicle computing system also determines at least a range-view representation of the LIDAR data and a top-view representation of the LIDAR data, wherein the range-view representation contains a fewer number of total data points than the top-view representation. The vehicle computing system further detects objects of interest in the range-view representation of the LIDAR data and generates a bounding shape for each of the detected objects of interest in the top-view representation of the LIDAR data.

Abstract: Systems and methods are provided for machine-learned models including convolutional neural networks that generate predictions using continuous convolution techniques. For example, the systems and methods of the present disclosure can be included in or otherwise leveraged by an autonomous vehicle. In one example, a computing system can perform, with a machine-learned convolutional neural network, one or more convolutions over input data using a continuous filter relative to a support domain associated with the input data, and receive a prediction from the machine-learned convolutional neural network. A machine-learned convolutional neural network in some examples includes at least one continuous convolution layer configured to perform convolutions over input data with a parametric continuous kernel.

Abstract: Systems and methods are provided for detecting objects of interest. A computing system can input sensor data to one or more first machine-learned models associated with detecting objects external to an autonomous vehicle. The computing system can obtain as an output of the first machine-learned models, data indicative of one or more detected objects. The computing system can determine data indicative of at least one uncertainty associated with the one or more detected objects and input the data indicative of the one or more detected objects and the data indicative of the at least one uncertainty to one or more second machine-learned models. The computing system can obtain as an output of the second machine-learned models, data indicative of at least one prediction associated with the one or more detected objects. The at least one prediction can be based at least in part on the detected objects and the uncertainty.

Abstract: Example aspects of the present disclosure describe determining, using a machine-learned model framework, a motion trajectory for an autonomous platform. The motion trajectory can be determined based at least in part on a plurality of costs based at least in part on a distribution of probabilities determined conditioned on the motion trajectory.

Abstract: The present disclosure provides systems and methods that apply neural networks such as, for example, convolutional neural networks, to sparse imagery in an improved manner. For example, the systems and methods of the present disclosure can be included in or otherwise leveraged by an autonomous vehicle. In one example, a computing system can extract one or more relevant portions from imagery, where the relevant portions are less than an entirety of the imagery. The computing system can provide the relevant portions of the imagery to a machine-learned convolutional neural network and receive at least one prediction from the machine-learned convolutional neural network based at least in part on the one or more relevant portions of the imagery. Thus, the computing system can skip performing convolutions over regions of the imagery where the imagery is sparse and/or regions of the imagery that are not relevant to the prediction being sought.

Abstract: Systems, methods, tangible non-transitory computer-readable media, and devices associated with trajectory prediction are provided. For example, trajectory data and goal path data can be accessed. The trajectory data can be associated with an object's predicted trajectory. The predicted trajectory can include waypoints associated with waypoint position uncertainty distributions that can be based on an expectation maximization technique. The goal path data can be associated with a goal path and include locations the object is predicted to travel. Solution waypoints for the object can be determined based on application of optimization techniques to the waypoints and waypoint position uncertainty distributions. The optimization techniques can include operations to maximize the probability of each of the solution waypoints. Stitched trajectory data can be generated based on the solution waypoints. The stitched trajectory data can be associated with portions of the solution waypoints and the goal path.

Abstract: A system for determining a pose of a vehicle and building maps from vehicle priors processes received ground intensity LIDAR data including intensity data for points believed to be on the ground and height information to form ground intensity LIDAR (GIL) images including pixels in 2D coordinates where each pixel contains an intensity value, a height value, and x- and y-gradients of intensity and height. The GIL images are formed by filtering aggregated ground intensity LIDAR data falling into a same spatial bin on the ground and using a registration algorithm to align two GIL images relative to one another by estimating a 6-degree-of-freedom pose with associated uncertainty that minimizes error between the two GIL images. The aligned GIL images are provided as a pose estimate to a localizer. The system may provide online localization and pose estimation, prior building, and prior to prior alignment pose estimation using image-based techniques.

Abstract: Systems and methods for controlling the motion of an autonomous are provided. In one example embodiment, a computer implemented method includes obtaining, by one or more computing devices on-board an autonomous vehicle, data associated with one or more objects that are proximate to the autonomous vehicle. The data includes a predicted path of each respective object. The method includes identifying at least one object as an object of interest based at least in part on the data associated with the object of interest. The method includes generating cost data associated with the object of interest. The method includes determining a motion plan for the autonomous vehicle based at least in part on the cost data associated with the object of interest. The method includes providing data indicative of the motion plan to one or more vehicle control systems to implement the motion plan for the autonomous vehicle.

Abstract: Systems, methods, and vehicles for taking a vehicle out-of-service are provided. In one example embodiment, a method includes obtaining, by one or more computing devices on-board an autonomous vehicle, data indicative of one or more parameters associated with the autonomous vehicle. The autonomous vehicle is configured to provide a vehicle service to one or more users of the vehicle service. The method includes determining, by the computing devices, an existence of a fault associated with the autonomous vehicle based at least in part on the one or more parameters associated with the autonomous vehicle. The method includes determining, by the computing devices, one or more actions to be performed by the autonomous vehicle based at least in part on the existence of the fault. The method includes performing, by the computing devices, one or more of the actions to take the autonomous vehicle out-of-service based at least in part on the fault.

Abstract: A computer-implemented method for determining scene-consistent motion forecasts from sensor data can include obtaining scene data including one or more actor features. The computer-implemented method can include providing the scene data to a latent prior model, the latent prior model configured to generate scene latent data in response to receipt of scene data, the scene latent data including one or more latent variables. The computer-implemented method can include obtaining the scene latent data from the latent prior model. The computer-implemented method can include sampling latent sample data from the scene latent data. The computer-implemented method can include providing the latent sample data to a decoder model, the decoder model configured to decode the latent sample data into a motion forecast including one or more predicted trajectories of the one or more actor features.

Abstract: Systems and methods for power and thermal management of autonomous vehicles are provided. In one example embodiment, a computing system includes processor(s) and one or more tangible, non-transitory, computer readable media that collectively store instructions that when executed by the processor(s) cause the computing system to perform operations. The operations include obtaining data associated with an autonomous vehicle. The operations include identifying one or more vehicle parameters associated with the autonomous vehicle based at least in part on the data associated with the autonomous vehicle. The operations include determining a modification to one or more operating characteristics of one or more systems onboard the autonomous vehicle based at least in part on the one or more vehicle parameters.

Abstract: An autonomous robot is provided. In one example embodiment, an autonomous robot can include a main body including one or more compartments. The one or more compartments can be configured to provide support for transporting an item. The autonomous robot can include a mobility assembly affixed to the main body and a sensor configured to obtain sensor data associated with a surrounding environment of the autonomous robot. The autonomous robot can include a computing system configured to plan a motion of the autonomous robot based at least in part on the sensor data. The computing system can be operably connected to the mobility assembly for controlling a motion of the autonomous robot. The autonomous robot can include a coupling assembly configured to temporarily secure the autonomous robot to an autonomous vehicle. The autonomous robot can include a power system and a ventilation system that can interface with the autonomous vehicle.

Abstract: Systems and methods for controlling an autonomous vehicle in response to vehicle instructions from a remote computing system are provided. In one example embodiment, a computer-implemented method includes controlling a first autonomous vehicle to provide a vehicle service, the first autonomous vehicle being associated with a first convoy that includes one or more second autonomous vehicles. The method includes receiving one or more communications from a remote computing system associated with a third-party entity, the one or more communications including one or more vehicle instructions. The method includes coordinating with the one or more second autonomous vehicles to determine one or more vehicle actions to perform in response to receiving the one or more vehicle instructions from the third-party entity. The method includes controlling the first autonomous vehicle to implement the one or more vehicle actions.

Abstract: Systems, methods, tangible non-transitory computer-readable media, and devices for detecting objects are provided. For example, the disclosed technology can obtain a representation of sensor data associated with an environment surrounding a vehicle. Further, the sensor data can include sensor data points. A point classification and point property estimation can be determined for each of the sensor data points and a portion of the sensor data points can be clustered into an object instance based on the point classification and point property estimation for each of the sensor data points. A collection of point classifications and point property estimations can be determined for the portion of the sensor data points clustered into the object instance. Furthermore, object instance property estimations for the object instance can be determined based on the collection of point classifications and point property estimations for the portion of the sensor data points clustered into the object instance.