Abstract: A system includes a camera configured to capture image data of an environment, a monitoring system configured to generate a gaze sequences of a subject, and a computing device communicatively coupled to the camera and the monitoring system. The computing device is configured to receive the image data from the camera and the gaze sequences from the monitoring system, implement a machine learning model comprising a convolutional encoder-decoder neural network configured to process the image data and a side-channel configured to inject the gaze sequences into a decoder stage of the convolutional encoder-decoder neural network, generate, with the machine learning model, a gaze probability density heat map, and generate, with the machine learning model, an attended awareness heat map.

Abstract: Systems and methods for modeling and predicting scene occupancy in an environment of a robot are disclosed herein. One embodiment processes past agent-trajectory data, map data, and sensor data using one or more encoder neural networks to produce combined encoded input data; generates a weights vector for a Gaussian Mixture Model (GMM) based on the combined encoded input data; produces a volumetric spatio-temporal representation of occupancy in an environment of a robot by generating, for a plurality of modes of the GMM in accordance with the weights vector, corresponding sample probability distributions of scene occupancy based on respective means and variances of the plurality of modes, wherein the respective means and variances sample coefficients of a set of learned basis functions; and controls the operation of the robot based, at least in part, on the volumetric spatio-temporal representation of occupancy.

Abstract: Systems and methods are provided for generating a statistical parameter representing a state of a surgical procedure from sensor data. Sensor data representing a time period. is received from a sensor. Numerical features representing the time period are generated from the sensor data. Each of a plurality of long short term memory units are updated according to the plurality of numerical features via a message passing process. The long short term memory units are connected to form a graph, with a first set of the long short term memory units representing a plurality of nodes of the graph and a second set of the long short term memory units representing a plurality of hyperedges of the graph. A statistical parameter representing a state of the surgical procedure for the time period is derived from an output of one of the long short term memory units and provided to a user.

Abstract: A method includes receiving data relating to pedestrian activity at one or more locations outside of a crosswalk, analyzing the data, based on the data, identifying at least one location of the one or more locations as a constructive crosswalk, and controlling operation of an autonomous vehicle based on the at least one location of the constructive crosswalk.

Abstract: System, methods, and other embodiments described herein relate to coordinating and optimizing prediction by a model using intermediate data and vehicle-to-vehicle (V2V) communications that improves bandwidth utilization. In one embodiment, a method includes processing acquired data, by a subject vehicle using a prediction model, to execute a vehicle task associated with navigating a driving scene. The method also includes extracting processed data according to the acquired data from intermediate activations of layers within the prediction model according to a cooperation model. The method also includes quantifying relevancy of the processed data, for a target vehicle, to select a subset using the cooperation model. The method also includes communicating the subset to a selected vehicle having the prediction model for additional prediction according to the relevancy and available resources to transmit the subset.

Abstract: Systems and methods of model or prediction algorithm selection are provided. An autonomous control system may include a perception component that, based on environmental inputs regarding an object(s), a vehicle's operating characteristics, etc., outputs a current state of the vehicle's surrounding environment. This in turn, is used as input to a prediction component comprising a plurality of prediction algorithms. The prediction component outputs a set of predictions regarding the trajectory of the object(s). Accordingly, for each object, a set of trajectories at specific timesteps may be generated by the different prediction algorithms which are input to a planner component. These trajectories may then be analyzed, compared, or otherwise processed to determine which trajectory regarding the object is most accurate. The prediction algorithm or model that produced the most accurate predicted trajectory may then be used for subsequent predictions/timesteps.

Abstract: Systems and methods for learning and managing robot user interfaces are disclosed herein. One embodiment generates, based on input data including information about past interactions of a particular user with a robot and with existing HMIs of the robot, a latent space using one or more encoder neural networks, wherein the latent space is a reduced-dimensionality representation of learned behavior and characteristics of the particular user, and uses the latent space as input to train a decoder neural network associated with (1) a new HMI distinct from the existing HMIs or (2) a particular HMI among the existing HMIs to alter operation of the particular HMI. The trained first decoder neural network is deployed in the robot to control, at least in part, operation of the new HMI or the particular HMI in accordance with the learned behavior and characteristics of the particular user.

Abstract: A method for indicating occlusion information at an ego agent includes observing a spatial area from a first viewpoint of one or more first sensors associated with the ego agent. The method also includes identifying the spatial area as an occluded area in accordance with observing the spatial area from a second viewpoint of the one or more first sensors after observing the spatial area from the first viewpoint. The method further includes receiving, from a target agent, a message indicating the spatial area is occluded from one or more second sensors associated with the target agent. The method still further includes transmitting, to the target agent in accordance with receiving the message, the occlusion information indicating information associated the spatial area based on identifying the spatial area as the occluded area.

Abstract: Systems and methods of model or prediction algorithm selection are provided. An autonomous control system may include a perception component that, based on environmental inputs regarding an object(s), a vehicle's operating characteristics, etc., outputs a current state of the vehicle's surrounding environment. This in turn, is used as input to a prediction component comprising a plurality of prediction algorithms. The prediction component outputs a set of predictions regarding the trajectory of the object(s). Accordingly, for each object, a set of trajectories at specific timesteps may be generated by the different prediction algorithms which are input to a planner component. These trajectories may then be analyzed, compared, or otherwise processed to determine which trajectory regarding the object is most accurate. The prediction algorithm or model that produced the most accurate predicted trajectory may then be used for subsequent predictions/timesteps.

Abstract: A control system, computer-readable storage medium and method of preventing occlusion of and minimizing shadows on the driver's face for driver monitoring. The system includes a steering wheel, a plurality of fiberscopes arranged evenly spaced around the steering wheel, and one or more video cameras arranged at remote ends of the plurality of fiberscopes. Distal ends of the fiberscopes emerge to a surface of the steering wheel through holes that are perpendicular to an axis of rotation of the steering wheel. Each of the distal ends of the fiberscopes includes a lens. The system includes a plurality of light sources and an electronic control unit connected to the one or more video cameras and the light sources.

Abstract: A method for prioritizing occlusion information is presented. The method includes determining, at a first time period, a first sensor's view of a spatial area is occluded. The method also includes observing, at a second time period, the spatial area. The method further includes determining a level of risk associated with the spatial area based on the observation. The method still further includes prioritizing transmission of the occlusion information corresponding to the spatial area based on the determined level of risk and transmitting the occlusion information corresponding to the spatial area based on the priority.

Abstract: Systems and methods are provided for identifying a current phase of a surgical procedure. Sensor data representing a time period is received and a plurality of numerical features representing the time period are generated from the sensor data. A statistical parameter representing a plurality of stored values from a memory is generated at a sufficient statistics model. An output, representing a surgical phase associated with the time period is provided at a recurrent neural network from a set of inputs that includes the plurality of numerical features and the statistical parameter.

Abstract: Disclosed are systems and methods for training at least one policy using a framework for encoding human behaviors and preferences in a driving environment. In one example, the method includes the steps of setting parameters of rewards and a Markov Decision Process (MDP) of the at least one policy that models a simulated human driver of a simulated vehicle and an adaptive human-machine interface (HMI) system configured to interact with each other and training the at least one policy to maximize a total reward based on the parameters of the rewards of the at least one policy.

Abstract: Systems and methods are provided for an advanced driver-assistance system (ADAS) that obtains data from a plurality of sensors. In some embodiments, the system can retrieve data regarding a user’s past interactions and analyze the data with the sensor data to determine the user’s behavior. In some embodiments, the ADAS can determine whether a user is unaware of an ADAS feature based on this behavior and a prompt that recommends the ADAS feature. The user’s response to this prompt may be incorporated into the user’s behavior for future recommendations.

Abstract: Embodiments of the present disclosure comprise systems and methods that implement domain adaptation. For example, some embodiments may improve standard machine learning systems by providing a best estimation of unknown situations using existing trained models. The trained models may be adapted to use in new scenarios that might not have been identified during the training phase of the machine learning. Using this adaptive approach, the model can help the vehicle system prepare a best estimate of the environment that was not identified during training and take an appropriate action.

Abstract: Systems and methods described herein relate to estimating risk associated with a vehicular maneuver. One embodiment acquires a geometric representation of an intersection including a lane in which a vehicle is traveling and at least one other lane; discretizes the at least one other lane into a plurality of segments; determines a trajectory along which the vehicle will travel; estimates a probability density function for whether a road agent external to the vehicle is present in the respective segments; estimates a traffic-conflict probability of a traffic conflict in the respective segments conditioned on whether an external road agent is present; estimates a risk associated with the vehicle following the trajectory by integrating a product of the probability density function and the traffic-conflict probability over the at least one other lane and the plurality of segments; and controls operation of the vehicle based, at least in part, on the estimated risk.

Abstract: Systems and methods to systematically identify and collect operational vehicle data for selective transmission to a non-local storage location for further analysis and use in training autonomous and semi-autonomous vehicles are provided. The systems and methods provided overcome limitations in storage and transmission of collected data by selectively archiving only that collected driving data warranting further analysis.

Abstract: A method of providing dynamic and variable learning for an ego vehicle by determining and using most-trustworthy inputs includes determining, based on ambient conditions of an environment of the ego vehicle, a level of trustworthiness of sensor values obtained from one or more ego vehicle sensors. The method also includes determining a level of confidence in an accuracy of an output of a subsystem of the ego vehicle. The output of the subsystem is based on the sensor values. The level of confidence is based on the level of trustworthiness of the sensor values.

Abstract: A system and method for generating a predicted vehicle trajectory includes a generative adversarial network configured to receive a trajectory vector of a target vehicle and generate a set of latent state vectors using the received trajectory vector and an artificial neural network. The latent state vectors each comprise a high-level sub-vector, ZH. The GAN enforces ZH to be correlated to an annotation coding representing semantic categories of vehicle trajectories. The GAN selects a subset, from the set of latent state vectors, using farthest point sampling and generates a predicted vehicle trajectory based on the selected subset of latent state vectors.

Abstract: Systems and methods for training a neural network for estimating a trajectory of a vehicle are disclosed. In one embodiment, a system includes one or more processors, and a non-transitory computer-readable medium storing computer-readable instructions. The computer-readable instructions cause the one or more processors to receive sensor data of a plurality of examples from a plurality of vehicle sensors, input the sensor data into a sensor data neural network to generate a sensor data intermediate space and receive structured data of the plurality of examples. The computer-readable instructions cause the one or more processors to input the structured data into a structured data neural network to generate a structured data intermediate space, calculate a first loss between the sensor data intermediate space and the structured data intermediate space using a first loss function, and provide the first loss to the sensor data neural network and the structured data neural network.