Uatc, LLC

Carnegie Mellon University

Deere & Company

Uber Technologies, Inc.

Carl Knox Wellington

Carlos Vallespi-Gonzalez

Anthony Joseph Stentz

Colin Jeffrey Green

Adam Milstein

Herman Herman

Aaron J Bruns

James Joseph Phelan

John M Hageman

Cristian Dima

Zachary Pezzementi

Joan Campoy

Victor S Sierra

Hanke Boesch

Cason Robert Male

Aaron C Courville

Scott Klaus Boehmke

David Patrick Rice

Michael Bode

Narek Melik-Barkhudarov

Shivam Gautam

Ankit Laddha

Gregory P Meyer

Eric Meyhofer

Peter Brueckner

Dillon Collins

Randy Warner

Anurag Kumar

David Driscoll

Shreyash Pandey

Jake Scott Charland

Systems and methods for detecting objects and predicting their motion are provided. In particular, a computing system can obtain a plurality of sensor sweeps. The computing system can determine movement data associated with movement of the autonomous vehicle. For each sensor sweep, the computing system can generate an image associated with the sensor sweep. The computing system can extract, using the respective image as input to one or more machine-learned models, feature data from the respective image. The computing system can transform the feature data into a coordinate frame associated with a next time step. The computing system can generate a fused image. The computing system can generate a final fused image. The computing system can predict, based, at least in part, on the final fused representation of the plurality of sensors sweeps from the plurality of sensor sweeps, movement associated with the feature data at one or more time steps in the future.

Systems and methods for object detection and motion prediction by fusing multiple sensor sweeps into a range view representation

A traffic signal analysis system for an autonomous vehicle (AV) can receive image data from one or more cameras, the image data including an upcoming traffic signaling system located at an intersection. The system can determine an action for the AV through the intersection and access a matching signal map specific to the upcoming traffic signaling system. Using the matching signal map, the system can generate a signal template for the upcoming traffic signaling system and determine a first subset and a second subset of the plurality of traffic signal faces that apply to the action. The system can dynamically analyze the first subset and the second subset to determine a state of the upcoming traffic signaling system for the action, and generate an output for the AV indicating the state of the upcoming traffic signaling system for the action.

Traffic signal analysis system

A method and non-transitory computer-readable medium capture an image of bulk grain and apply a feature extractor to the image to determine a feature of the bulk grain in the image. For each of a plurality of different sampling locations in the image, based upon the feature of the bulk grain at the sampling location, a determination is made regarding a classification score for the presence of a classification of material at the sampling location. A quality of the bulk grain of the image is determined based upon an aggregation of the classification scores for the presence of the classification of material at the sampling locations.

Grain quality monitoring

Systems and methods for implementing a low-latency braking action for an autonomous vehicle are provided. A computing system can include a vehicle autonomy system comprising one or more processors configured to determine a motion plan for an autonomous vehicle based at least in part on sensor data from one or more sensors of the autonomous vehicle. The computing system can further include a low-latency braking system comprising one or more processors configured to determine that the autonomous vehicle has a likelihood of colliding with an object in a surrounding environment based at least in part on a previously-determined motion plan obtained from the vehicle autonomy system. In response to determining that the autonomous vehicle has a likelihood of colliding with the object in the surrounding environment, the low-latency braking system can further be configured to implement a braking action for the autonomous vehicle.

Systems and methods for low-latency braking action for an autonomous vehicle

A self-driving vehicle (SDV) can operate by analyzing sensor data to autonomously control acceleration, braking, and steering systems of the SDV along a current route. The SDV includes a number of sensors generating the sensor data and a control system to detect conditions relating to the operation of the SDV, such as vehicle speed and local weather, select a set of sensors based on the detected conditions, and prioritize the sensor data generated from the selected set of sensors to control aspects relating to the operation of the SDV.

Dynamic sensor selection for self-driving vehicles

An image capture device can include one or more image sensors and one or more image processors. The image sensor(s) can be configured to detect incoming light provided incident to a surface of each image sensor, each image sensor configured to provide full image frames of image capture data at a first resolution. The image processor(s) can be coupled to the image sensor(s) and configured to receive the image capture data at the first resolution, downsample the image capture data outside one or more regions of interest, and provide a digital image output having a second resolution within the one or more regions of interest and a third resolution outside the one or more regions of interest, wherein the third resolution is less than the second resolution.

Image capture device with customizable regions of interest

A traffic signal analysis system can receive image data from one or more cameras of an autonomous vehicle, where the image data includes a traffic signaling system located at an intersection. The traffic signal analysis system can determine a pass-through action for the autonomous vehicle through the intersection, and access a matching signal map that includes characteristic information indicating properties of the traffic signaling system. Based on the characteristic information and the image data, the traffic signal analysis system can identify a state of the traffic signaling system for the pass-through action, and generate an output for the autonomous vehicle indicating the state of the traffic signaling system for the pass-through action.

The disclosed terrain model is a generative, probabilistic approach to modeling terrain that exploits the 3D spatial structure inherent in outdoor domains and an array of noisy but abundant sensor data to simultaneously estimate ground height, vegetation height and classify obstacles and other areas of interest, even in dense non-penetrable vegetation. Joint inference of ground height, class height and class identity over the whole model results in more accurate estimation of each quantity. Vertical spatial constraints are imposed on voxels within a column via a hidden semi-Markov model. Horizontal spatial constraints are enforced on neighboring columns of voxels via two interacting Markov random fields and a latent variable. Because of the rules governing abstracts, this abstract should not be used to construe the claims.

System and method for generating a terrain model for autonomous navigation in vegetation

A method and system for detecting an obstacle comprises a terrain estimator for estimating a local terrain surface map based on at least one of range data points, color data, and infrared data gathered by electromagnetic perception focused in front of a vehicle. The map is composed of a series of terrain cells. An analyzer estimates at least one of predicted roll data, predicted pitch data, predicted ground clearance data, and predicted friction coefficient data based on the estimated terrain map for respective terrain cells and vehicular constrain data. A local planner determines predicted vehicle control data for terrain cells within the terrain along a planned path of the vehicle. One or more vehicle sensors sense at least one of actual roll data, actual pitch data, actual ground clearance data, and actual friction coefficient data for the terrain cells when the vehicle is coextensively positioned with the corresponding terrain cell. A learning module adjusts at least one of the terrain map estimation and the control data determination based on the sensed actual data.

Method and system for estimating navigability of terrain

Growing community of inventors

Pittsburgh, PA, United States of America

Carl Knox Wellington