Abstract: A computer includes a processor and a memory storing instructions executable by the processor to receive a time series of vectors from a sensor, determine a weighted moving mean of the vectors, determine an inverse covariance matrix of the vectors, receive a current vector from the sensor, determine a squared Mahalanobis distance between the current vector and the weighted moving mean, and output an indicator of an anomaly with the sensor in response to the squared Mahalanobis distance exceeding a threshold. The squared Mahalanobis distance is determined by using the inverse covariance matrix.

Abstract: Operation data from one or more vehicle subsystems are input to a vehicle dynamics model. Predicted operation data of the one or more vehicle subsystems are output from the vehicle dynamics model. The operation data and the predicted operation data are input to an optimization program that is programmed to output control directives for the one or more vehicle subsystems. One or more vehicle subsystems are operated according to the output control directives.

Abstract: A method includes obtaining state information from one or more sensors of a digital twin. The method includes determining an action at the first routing control location based on the state information, where the action includes one of a pallet merging operation and a pallet splitting operation, and determining a consequence state based on the action. The method includes calculating a transient production value based on the consequence state and a transient objective function, calculating a steady state production value based on the consequence state and a steady state objective function, and selectively adjusting one or more reinforcement parameters of the reinforcement learning system based on the transient production value and the steady state production value.

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: Systems and methods for automotive shape design by combining computational fluid dynamics (CFD) and Generative Adversarial Network (GAN). CFD simulations may be performed to determine aerodynamic properties and identify a set of candidate vehicle outline shapes. Vehicle shape outlines may be provided as input to a generative adversarial network (GAN) that is trained to learn aesthetic preferences for vehicle attributes. The GAN may be used to determine, by based on the vehicle outline shape, a set of vehicle attributes. The GAN may be used to generate photo-realistic images with the vehicle shape outline and filling in additional aesthetic styles for the given outline, such as different colors, lighting, visual appearance, wheel design, aspect ratio, etc.

Abstract: The present invention extends to methods, systems, and computer program products for classifying time series image data. Aspects of the invention include encoding motion information from video frames in an eccentricity map. An eccentricity map is essentially a static image that aggregates apparent motion of objects, surfaces, and edges, from a plurality of video frames. In general, eccentricity reflects how different a data point is from the past readings of the same set of variables. Neural networks can be trained to detect and classify actions in videos from eccentricity maps. Eccentricity maps can be provided to a neural network as input. Output from the neural network can indicate if detected motion in a video is or is not classified as an action, such as, for example, a hand gesture.

Abstract: A method includes obtaining sensor data from a plurality of sensors disposed at a plurality of routing control locations of an environment, where the sensor data is indicative of a number of a plurality of pallets at the plurality of routing control locations. The method includes calculating a plurality of difference values based on the sensor data, calculating a transient production value based on the sensor data and a transient objective function, and calculating a steady state production value based on the sensor data and a steady state objective function. The method includes generating a state vector based on the plurality of difference values, the transient production value, and the steady state production value, and defining a set of routes for a set of pallets from among the plurality of pallets based on the state vector and a digital twin of the environment.

Abstract: A method includes defining a sensor layout of a digital twin based on one or more sensor parameters and one or more routing control locations of the digital twin. The method includes simulating a manufacturing routine of a plurality of pallets and a plurality of workstations based on one or more pallet parameters associated with the plurality of pallets and one or more workstation parameters associated with the plurality of workstations and calculating, for each routing control location from among the one or more routing control locations, a transient production value and a steady state production value based on the manufacturing routine. The method includes iteratively adjusting the sensor layout of the digital twin until each transient production value is less than or equal to a threshold transient production value and each steady state production value is less than or equal to a threshold steady state production value.

Abstract: A method includes obtaining state information from one or more sensors of a digital twin. The method includes determining an action at the first routing control location based on the state information, where the action includes one of a pallet merging operation and a pallet splitting operation, and determining a consequence state based on the action. The method includes calculating a transient production value based on the consequence state and a transient objective function, calculating a steady state production value based on the consequence state and a steady state objective function, and selectively adjusting one or more reinforcement parameters of the reinforcement learning system based on the transient production value and the steady state production value.

Abstract: Operation data from one or more vehicle subsystems are input to a vehicle dynamics model. Predicted operation data of the one or more vehicle subsystems are output from the vehicle dynamics model. The operation data and the predicted operation data are input to an optimization program that is programmed to output control directives for the one or more vehicle subsystems. One or more vehicle subsystems are operated according to the output control directives.

Abstract: The present disclosure describes systems and methods that include calculating, via a reinforcement learning agent (RLA) controller, a plurality of state-action values based on sensor data representing an observed state, wherein the RLA controller utilizes a deep neural network (DNN) and generating, via a fuzzy controller, a plurality of linear models mapping the plurality of state-action values to the sensor data.

Abstract: A computer, including a processor and a memory, the memory including instructions to be executed by the processor to determine a first action based on inputting sensor data to a deep reinforcement learning neural network and transform the first action to one or more first commands. One or more second commands can be determined by inputting the one or more first commands to control barrier functions and transforming the one or more second commands to a second action. A reward function can be determined by comparing the second action to the first action. The one or more second commands can be output.

Abstract: A system comprises a processor that is programmed to define a plurality of vehicle groups based on vehicle specification data and define a plurality of sub-groups for each of the vehicle groups based on environmental data and sensor data received from each of a plurality of vehicles. The processor is programmed to adjust fuel tank leak detection classifiers for the sub-groups based on ground truth data. The ground truth data include, for each of the plurality of vehicles, a leak detection status and a leak test result.

Abstract: A computer, including a processor and a memory, the memory including instructions to be executed by the processor to determine optimal vehicle actions based on a modified version of a Nash equilibrium solution to a multiple agent game, wherein the Nash equilibrium solution is modified by performing an adaptive grid search optimization technique based on calculating rewards and penalties for the agents to determine optimal vehicle actions, wherein the agents include one or more of autonomous vehicles, non-autonomous vehicles, stationary objects, and non-stationary objects including pedestrians and wherein the rewards and the penalties for the agents are determined by simulating behavior of the agents to determine possible future states for the agents to determine the optimal vehicle actions. The instructions can include further instructions to determine a vehicle path based on the optimal vehicle actions and download the vehicle path to the vehicle.

Abstract: The method of monitoring an operation includes acquiring data from sensors including images of a workspace in which the operation is to be performed, identifying a human operator and a controlled element within the workspace using the acquired images, determining whether the operation has initiated based on a known activation trigger, estimating pose of the human operator using the images, monitoring state of the controlled element based on acquired data, and determining whether an abnormality occurred based on the estimated pose, the state of the controlled element, a duration of the operation, or a combination thereof.

Abstract: The method of monitoring an operation includes acquiring data from sensors including images of a workspace in which the operation is to be performed, identifying a human operator and a controlled element within the workspace using the acquired images, determining whether the operation has initiated based on a known activation trigger, estimating pose of the human operator using the images, monitoring state of the controlled element based on acquired data, and determining whether an abnormality occurred based on the estimated pose, the state of the controlled element, a duration of the operation, or a combination thereof.

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 a time series of vectors from a sensor, determine a weighted moving mean of the vectors, determine an inverse covariance matrix of the vectors, receive a current vector from the sensor, determine a squared Mahalanobis distance between the current vector and the weighted moving mean, and output an indicator of an anomaly with the sensor in response to the squared Mahalanobis distance exceeding a threshold. The squared Mahalanobis distance is determined by using the inverse covariance matrix.

Abstract: A computer, including a processor and a memory, the memory including instructions to be executed by the processor to calibrate utility functions that determine optimal vehicle actions based on an approximate Nash equilibrium solution for multiple agents by determining a difference between model-predicted future states for the multiple agents to observed states for the multiple agents. The instructions can include further instructions to determine a vehicle path for a vehicle based on the optimal vehicle actions.

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.