Hrl Laboratories, LLC

The Boeing Company

Gm Global Technolgoy Operations, Inc.

Theboeing Company

Kyungnam Kim

Yuri Owechko

Deepak Khosla

Yang Chen

Shinko Yuanhsien Cheng

Lei Zhang

Changsoo S Jeong

Hyukseong Kwon

Alexander L Honda

Swarup S Medasani

Jiejun Xu

Terrell Nathan Mundhenk

Soheil Kolouri

Heiko Hoffmann

Rajan Bhattacharyya

Changsoo Jeong

Mohammad Rostami

Michael J Daily

Kenji Yamada

Fredy Monterroza

Hai-Wen Chen

David W Payton

James W Wells

Kang-Yu Ni

Shankar R Rao

David J Huber

Ryan M Uhlenbrock

Dmitriy V Korchev

Zhichao Chen

Arturo Flores

Zhiqi Zhang

Tiffany J Hwu

Roland J Menassa

Jose Cruz-Albrecht

Charles E Martin

Qin Jiang

Charles W Wampler, Ii

Vincent De Sapio

Nigel D Stepp

Leandro G Barajas

Rashmi N Sundareswara

Jaehoon Choe

Jason Meltzer

Hyun (Tiffany) J Kim

Alejandro Nijamkin

Li Yang Ku

Alexander Krayner

Shay Deutsch

Benjamin Nuernberger

Michael Cao

Alexander J Gabourie

Divya P Kulkarni

Zachary Murez

Braden Mailloux

Xi Hang Cao

Nathan Mundhenk

Hyun J Kim

Deepark Khosla

A method for training a controller to control a robotic system includes: receiving a neural network of an original controller for the robotic system based on origin data samples from an origin domain and labels in a label space, the neural network including encoder and classifier parameters, the neural network being trained to: map an input data sample from the origin domain to a feature vector in a feature space using the encoder parameters; and assign a label of the label space to the input data sample using the feature vector based on the classifier parameters; updating the encoder parameters to minimize a dissimilarity, in the feature space, between: origin feature vectors computed from the origin data samples; and target feature vectors computed from target data samples from a target domain; and updating the controller with the updated encoder parameters to control the robotic system in the target domain.

Systems and methods for few-shot transfer learning

A virtual lane estimation system includes a memory device, a sensor and a computer. The memory device is configured to store a road map that corresponds to a portion of a road ahead of a vehicle. The sensor is configured to observe a plurality of trajectories of a plurality of neighboring vehicles that traverse the portion of the road. The computer is configured to initialize a recursive self-organizing map as a plurality of points arranged as a two-dimensional grid aligned with the road map, train the points in the recursive self-organizing map in response to the trajectories, generate a directed graph that contains one or more virtual lanes through the road map in response to the points trained to the trajectories, and generate a control signal that controls navigation of the vehicle through the portion of the road in response to the virtual lanes in the directed graph.

Virtual lane estimation using a recursive self-organizing map

A vehicle and a system and a method of operating the vehicle. The system includes a reasoning engine, an episodic memory, a resolver and a controller. The reasoning engine infers a plurality of possible scenarios based on a current state of an environment of the vehicle. The episodic memory determines a historical likelihood for each of the plurality of possible scenarios. The resolver selects a scenario from the plurality of possible scenarios using the historical likelihoods. The controller operates the vehicle based on the selected scenario.

System for extending functionality of hypotheses generated by symbolic/logic-based reasoning systems

A control system of the autonomous vehicle may generate multiple possible behavior control movements based on the driving goal and the assessment of the vehicle environment. In doing so, the method and system selects one of the best behavior control, among the multiple possible movements, and the selection is based on the quantitative grading of its driving behavior.

Forward modeling for behavior control of autonomous vehicles

Described is a system for adapting a deep convolutional neural network (CNN). A deep CNN is first trained on an annotated source image domain. The deep CNN is adapted to a new target image domain without requiring new annotations by determining domain agnostic features that map from the annotated source image domain and a target image domain to a joint latent space, and using the domain agnostic features to map the joint latent space to annotations for the target image domain.

Domain adaption learning system

A method of path planning for a host vehicle includes: receiving host vehicle, environmental and obstacle information; calculating one or more projected host vehicle locations; computing a projected obstacle location for each obstacle; and determining a collision potential between each projected host vehicle location and each projected obstacle location. Until a maximum number of steps is reached, and while at least one projected host vehicle location has an associated collision potential below a collision threshold, the method further includes repeating the calculating, computing and determining steps.

System, method and controller for graph-based path planning for a host vehicle

Described is a system for unsupervised domain adaptation in an autonomous learning agent. The system adapts a learned model with a set of unlabeled data from a target domain, resulting in an adapted model. The learned model was previously trained to perform a task using a set of labeled data from a source domain. The set of labeled data has a first input data distribution, and the set of unlabeled target data has a second input data distribution that is distinct from the first input data distribution. The adapted model is implemented in the autonomous learning agent, causing the autonomous learning agent to perform the task in the target domain.

System and method for unsupervised domain adaptation via sliced-wasserstein distance

A method is disclosed. The method receives a two dimensional or a three dimensional computer generated representation of an area, receives a plurality of images of the area captured by one or more video cameras, detects a first moving object in the plurality of images, generates a computer representation of the first moving object, correlates the images of the area captured by the one or more video cameras with the two dimensional or three dimensional computer generated representation of the area, and displays the computer representation of the first moving object in the two dimensional or the three dimensional computer generated representation of the area.

Contextual visual dataspaces

Described is a low power system for mobile devices that provides continuous, behavior-based security validation of mobile device applications using neuromorphic hardware. A mobile device comprises a neuromorphic hardware component that runs on the mobile device for continuously monitoring time series related to individual mobile device application behaviors, detecting and classifying pattern anomalies associated with a known malware threat in the time series related to individual mobile device application behaviors, and generating an alert related to the known malware threat. The mobile device identifies pattern anomalies in dependency relationships of mobile device inter-application and intra-applications communications, detects pattern anomalies associated with new malware threats, and isolates a mobile device application having a risk of malware above a predetermined threshold relative to a risk management policy.

System for continuous validation and threat protection of mobile applications

Described is a system for object recognition. The system generates a training image set of object images from multiple image classes. Using a training image set and annotated semantic attributes, a model is trained that maps visual features from known images to the annotated semantic attributes using joint sparse representations with respect to dictionaries of visual features and semantic attributes. The trained model is used for mapping visual features of an unseen input image to its semantic attributes. The unseen input image is classified as belonging to an image class, and a device is controlled based on the classification of the unseen input image.

Attribute aware zero shot machine vision system via joint sparse representations

A method includes receiving image data at a first tracking system. The image data represents a region in an image of a sequence of images. The method also includes generating a first tracking fingerprint based on the image data. The method further includes comparing the first tracking fingerprint and a second tracking fingerprint. The method also includes providing an output from the first tracking system to a second tracking system based on a result of the comparison of the first tracking fingerprint and the second tracking fingerprint. The output includes an instruction associated with an object model stored at the second tracking system.

System and method to improve object tracking using tracking fingerprints

An apparatus to generate a model of a surface of an object includes a data set pre-aligner configured to receive multiple sets of surface data that correspond to respective portions of a surface of an object and that include three-dimensional (3D) points. The data set pre-aligner is also configured to perform a pre-alignment of overlapping sets to generate pre-aligned sets, including performing a rotation operation on a second set of the surface data, relative to a first set of the surface data that overlaps the second set, to apply a rotation amount that is selected from among multiple discrete rotation amounts and based on a similarity metric. The apparatus includes a data set aligner configured to perform an iterative alignment of the pre-aligned sets to generate aligned sets. The apparatus also includes a 3D model generator configured to combine the aligned sets to generate a 3D model of the object.

Three-dimensional point data alignment with pre-alignment

Described is a system that can recognize novel objects that the system has never before seen. The system uses a training image set to learn a model that maps visual features from known images to semantic attributes. The learned model is used to map visual features of an unseen input image to semantic attributes. The unseen input image is classified as belonging to an image class with a class label. A device is controlled based on the class label.

Zero shot machine vision system via joint sparse representations

Methods, apparatus, and articles of manufacture are disclosed to generate a synthetic point cloud of a spacecraft. An example apparatus includes a point cloud generator to generate a first synthetic point cloud of a first simulated space vehicle based on a simulated illumination source and a simulated image sensor, where the simulated illumination source and the simulated image sensor is operatively coupled to a second simulated space vehicle at a first position, where the simulated image sensor measures a parameter of the first simulated space vehicle, where the simulated illumination source uses a first configuration. The example apparatus further includes a pose determiner to determine a first pose of the simulated illumination source based on a first difference between the first synthetic point cloud and a reference synthetic point cloud using a first registration method, and a command generator to generate a command to move the second simulated space vehicle from the first position to a second position based on the first pose.

Methods and apparatus to generate a synthetic point cloud of a spacecraft

Described is a system for multi-view embedding for object recognition. During operation, the system receives an input image and transforms raw data of objects in the image with corresponding labels into low-level features and high-level semantic representations of the labels. A trained object recognition model is generated by embedding the low-level features with multiple high-level semantic representations. The system then receives data of an unknown object and assigns a label to the unknown object using the trained object recognition model. Finally, a device can be controlled based on the label.

Multi-view embedding with soft-max based compatibility function for zero-shot learning

Growing community of inventors

Oak Park, CA, United States of America

Kyungnam Kim