Abstract: A computer, including a processor and a memory, the memory including instructions to be executed by the processor to train a neural network included in a memory augmented neural network based on one or more images and corresponding ground truth in a training dataset by transforming the one or more images to generate a plurality of one-hundred or more variations of the one or more images including variations in the ground truth and process the variations of the one or more images and store feature points corresponding to each variation of the one or more images in memory associated with the memory augmented neural network. The instructions can include further instructions to process an image acquired by a vehicle sensor with the memory augmented neural network, including comparing a feature variance set for the image acquired by the vehicle sensor to the stored processing parameters for each variation of the one or more images, to obtain an output result.

Abstract: A method includes detecting, for each of a plurality of images, a plurality of key points, where each of the plurality of images represents an object of an assembly system. The method includes generating, for each of the plurality of images, a correspondence between the plurality of key points, and generating, for each of the plurality of images, a reference region based on the correspondence between the plurality of key points. The method includes identifying, for each of the plurality of images, a reference key point among the plurality of key points based on the reference region, and determining a pose of the object based on the reference key point of each of the plurality of images and a reference pose of the object.

Abstract: A method includes detecting, for each of a plurality of images, a plurality of key points, where each of the plurality of images represents an object of an assembly system. The method includes generating, for each of the plurality of images, a correspondence between the plurality of key points, and generating, for each of the plurality of images, a reference region based on the correspondence between the plurality of key points. The method includes identifying, for each of the plurality of images, a reference key point among the plurality of key points based on the reference region, and determining a pose of the object based on the reference key point of each of the plurality of images and a reference pose of the object.

Abstract: A computer, including a processor and a memory, the memory including instructions to be executed by the processor to train a neural network included in a memory augmented neural network based on one or more images and corresponding ground truth in a training dataset by transforming the one or more images to generate a plurality of one-hundred or more variations of the one or more images including variations in the ground truth and process the variations of the one or more images and store feature points corresponding to each variation of the one or more images in memory associated with the memory augmented neural network. The instructions can include further instructions to process an image acquired by a vehicle sensor with the memory augmented neural network, including comparing a feature variance set for the image acquired by the vehicle sensor to the stored processing parameters for each variation of the one or more images, to obtain an output result.

Abstract: A computer includes a processor and a memory storing instructions executable by the processor to receive an image including a physical landmark, output a plurality of synthetic images, wherein each synthetic image is generated by simulating at least one ambient feature in the received image, generate respective feature vectors for each of the plurality of synthetic images, and actuate one or more vehicle components upon identifying the physical landmark in a second received image based on a similarity measure between the feature vectors of the synthetic images and a feature vector of the second received image, the similarity measure being one of a probability distribution difference or a statistical distance.

Abstract: Example learning systems and methods are described. In one implementation, a machine learning system accesses data associated with a particular task and accesses data associated with an overall network. The machine learning system also accesses reward data. The machine learning system then operates on the data associated with a particular task, the data associated with an overall network, and the reward data to perform the particular task.

Abstract: A computer, including a processor and a memory, the memory including instructions to be executed by the processor to determine performance of a plurality of vehicle data sources used to operate a vehicle by evaluating each vehicle data source output data and train a deep neural network to determine reliability for each of the vehicle data sources based on the performance using reinforcement learning. The instructions can further include instructions to combine output data from the vehicle data sources based on the reliability including using the deep neural network to correlate output data from one or more vehicle data sources to the performance to determine how accurately the output data from each vehicle data source corresponds to vehicle performance and operate the vehicle based on combined output data.

Abstract: A computer, including a processor and a memory, the memory including instructions to be executed by the processor to determine performance of a plurality of vehicle data sources used to operate a vehicle by evaluating each vehicle data source output data and train a deep neural network to determine reliability for each of the vehicle data sources based on the performance using reinforcement learning. The instructions can further include instructions to combine output data from the vehicle data sources based on the reliability including using the deep neural network to correlate output data from one or more vehicle data sources to the performance to determine how accurately the output data from each vehicle data source corresponds to vehicle performance and operate the vehicle based on combined output data.

Abstract: An automation system includes a manipulation system including a manipulator for moving an object to a target location, a vision system for detecting landmarks on the object and the target location, and a learning and control module. The vision system is movable. The learning and control module is configured to control a movement of the manipulator and change a field of view of the vision system independent of the movement of the manipulator.

Abstract: A method of acquiring and processing visual data is provided, which includes: directing a light of a particular color to at least one of the plurality of landmarks on an object to illuminate the at least one of the plurality of landmarks; obtaining a first image of the object when the at least one of the plurality of landmarks on the object is illuminated; and extracting coordinates of the at least one of the plurality of landmarks from the first image.

Abstract: Systems and methods for anomaly detection in accordance with embodiments of the invention are illustrated. One embodiment includes a method for training a system for detecting anomalous samples. The method draws data samples from a data distribution of true samples and an anomaly distribution and draws a latent sample from a latent space. The method further includes steps for training a generator to generate data samples based on the drawn data samples and the latent sample, and training a cyclic discriminator to distinguish between true data samples and reconstructed samples. A reconstructed sample is generated by the generator based on an encoding of a data sample. The method identifies a set of one or more true pairs, a set of one or more anomalous pairs, and a set of one or more generated pairs. The method trains a joint discriminator to distinguish true pairs from anomalous and generated pairs.

Abstract: Example learning systems and methods are described. In one implementation, a machine learning system accesses data associated with a particular task and accesses data associated with an overall network. The machine learning system also accesses reward data. The machine learning system then operates on the data associated with a particular task, the data associated with an overall network, and the reward data to perform the particular task.

Abstract: A control and learning module for controlling a robotic arm includes at least one learning module including at least one neural network. The at least one neural network is configured to receive and be trained by both state measurements based on measurements of current state and observation measurements based on observation data during an initial learning phase. The at least one learning module is further configured to be re-tuned by updated observation data for improved performance during an operations and secondary learning phase when the robotic arm is in normal operation and after the initial learning phase.

Abstract: A method of acquiring and processing visual data is provided, which includes: directing a light of a particular color to at least one of the plurality of landmarks on an object to illuminate the at least one of the plurality of landmarks; obtaining a first image of the object when the at least one of the plurality of landmarks on the object is illuminated; and extracting coordinates of the at least one of the plurality of landmarks from the first image

Abstract: An automation system includes a manipulation system including a manipulator for moving an object to a target location, a vision system for detecting landmarks on the object and the target location, and a learning and control module. The vision system is movable. The learning and control module is configured to control a movement of the manipulator and change a field of view of the vision system independent of the movement of the manipulator.

Abstract: A LiDAR device that transmits a single or multiple continuous or intermittent laser beams to the environment and detects the reflected light on one or more detectors. The LiDAR device may include a scanning mirrors array composed of a single or multiple moving mirrors capable of changing the direction of the transmitted light. The scanning mirrors array may also include sensors and actuators which can be used to precisely control or measure the position of the mirrors. The LiDAR device may also include a lens that focuses the light captured by the mirror(s) onto a single or a multitude of detectors. The device may include laser sources and detectors operating in various wavelengths. The LiDAR device may also include laser power modulation mechanisms at a single or multitude of frequencies to improve signal detection performance and remove any ambiguity in range calculation.

Abstract: A LiDAR device that transmits a single or multiple continuous or intermittent laser beams to the environment and detects the reflected light on one or more detectors. The LiDAR device may include a scanning mirrors array composed of a single or multiple moving mirrors capable of changing the direction of the transmitted light. The scanning mirrors array may also include sensors and actuators which can be used to precisely control or measure the position of the mirrors. The LiDAR device may also include a lens that focuses the light captured by the mirror(s) onto a single or a multitude of detectors. The device may include laser sources and detectors operating in various wavelengths. The LiDAR device may also include laser power modulation mechanisms at a single or multitude of frequencies to improve signal detection performance and remove any ambiguity in range calculation.

Abstract: Example tornado detection systems and methods are described. In one implementation, a method receives data from a sensor mounted to a vehicle and analyzes the received data using a deep neural network. The method determines whether a tornado is identified in the received data based on the analysis of the received data. If a tornado is identified in the received data, the method determines a trajectory of the tornado.

Abstract: Active lens assemblies based on a deformable lens adhered to an electroactive polymer (EAP) membrane based actuator, active lens systems comprising such active lenses, and methods of actuating a deformable lens are described. The active lenses can provide controllable optical power, withstand constant periods of actuation, have low power consumption, have fewer mechanically moving parts, have fast response times and/or high portability. These benefits are important for use in precision-seeking applications, such as collision avoidance systems in automobiles.

Abstract: Multi-actuator system. The system includes at least two nano positioners having different ranges and bandwidths located in cascaded serial form to contact and move an object. A control system employs data-based control design to combine the at least two nano positioners so as to apportion actuation responsibilities among the at least two nano positioners so as to compensate for their coupled dynamics while moving the object. It is preferred to provide a separate controller for controlling separately each of the at least two nano positioners. Parameters of the separate controllers may be determined by minimizing output error.