Abstract: A method, apparatus and system for artificial intelligence-based HDRL planning and control for coordinating a team of platforms includes implementing a global planning layer for determining a collective goal and determining, by applying at least one machine learning process, at least one respective platform goal to be achieved by at least one platform, implementing a platform planning layer for determining, by applying at least one machine learning process, at least one respective action to be performed by the at least one of the platforms to achieve the respective platform goal, and implementing a platform control layer for determining at least one respective function to be performed by the at least one of the platforms. In the method, apparatus and system despite the fact that information is shared between at least two of the layers, the global planning layer, the platform planning layer, and the platform control layer are trained separately.

Abstract: A method for AI-driven augmented reality mentoring includes determining semantic features of objects in at least one captured scene, determining 3D positional information of the objects, combining information regarding the identified objects with respective 3D positional information to determine at least one intermediate representation, completing the determined intermediate representation using machine learning to include additional objects or positional information of the objects not identifiable from the at least one captured scene, determining at least one task to be performed and determining steps to be performed using a knowledge database, generating at least one visual representation relating to the determined steps for performing the at least one task, determining a correct position for displaying the at least one visual representation, and displaying the at least one visual representation on the see-through display in the determined correct position as an augmented overlay to the view of the at least

Abstract: A method, apparatus and system for artificial intelligence-based HDRL planning and control for coordinating a team of platforms includes implementing a global planning layer for determining a collective goal and determining, by applying at least one machine learning process, at least one respective platform goal to be achieved by at least one platform, implementing a platform planning layer for determining, by applying at least one machine learning process, at least one respective action to be performed by the at least one of the platforms to achieve the respective platform goal, and implementing a platform control layer for determining at least one respective function to be performed by the at least one of the platforms. In the method, apparatus and system despite the fact that information is shared between at least two of the layers, the global planning layer, the platform planning layer, and the platform control layer are trained separately.

Abstract: Techniques for augmenting a reality captured by an image capture device are disclosed. In one example, a system includes an image capture device that generates a two-dimensional frame at a local pose. The system further includes a computation engine executing on one or more processors that queries, based on an estimated pose prior, a reference database of three-dimensional mapping information to obtain an estimated view of the three-dimensional mapping information at the estimated pose prior. The computation engine processes the estimated view at the estimated pose prior to generate semantically segmented sub-views of the estimated view. The computation engine correlates, based on at least one of the semantically segmented sub-views of the estimated view, the estimated view to the two-dimensional frame. Based on the correlation, the computation engine generates and outputs data for augmenting a reality represented in at least one frame captured by the image capture device.

Abstract: A method, apparatus and system for object detection in sensor data having at least two modalities using a common embedding space includes creating first modality vector representations of features of sensor data having a first modality and second modality vector representations of features of sensor data having a second modality, projecting the first and second modality vector representations into the common embedding space such that related embedded modality vectors are closer together in the common embedding space than unrelated modality vectors, combining the projected first and second modality vector representations, and determining a similarity between the combined modality vector representations and respective embedded vector representations of features of objects in the common embedding space to identify at least one object depicted by the captured sensor data. In some instances, data manipulation of the method, apparatus and system can be guided by physics properties of a sensor and/or sensor data.

Abstract: A method, apparatus and system for determining change in pose of a mobile device include determining from first ranging information received at a first and a second receiver on the mobile device from a stationary node during a first time instance, a distance from the stationary node to the first receiver and the second receiver, determining from second ranging information received at the first receiver and the second receiver from the stationary node during a second time instance, a distance from the stationary node to the first receiver and second receiver, and determining from the determined distances during the first time instance and the second time instance, how far and in which direction the first receiver and the second receiver moved between the first time instance and the second time instance to determine a change in pose of the mobile device, where a position of the stationary node is unknown.

Abstract: A method, apparatus and system for efficient navigation in a navigation space includes determining semantic features and respective 3D positional information of the semantic features for scenes of captured image content and depth-related content in the navigation space, combining information of the determined semantic features of the scene with respective 3D positional information using neural networks to determine an intermediate representation of the scene which provides information regarding positions of the semantic features in the scene and spatial relationships among the sematic features, and using the information regarding the positions of the semantic features and the spatial relationships among the sematic features in a machine learning process to provide at least one of a navigation path in the navigation space, a model of the navigation space, and an explanation of a navigation action by the single, mobile agent in the navigation space.

Abstract: A method, apparatus and system for visual localization includes extracting appearance features of an image, extracting semantic features of the image, fusing the extracted appearance features and semantic features, pooling and projecting the fused features into a semantic embedding space having been trained using fused appearance and semantic features of images having known locations, computing a similarity measure between the projected fused features and embedded, fused appearance and semantic features of images, and predicting a location of the image associated with the projected, fused features. An image can include at least one image from a plurality of modalities such as a Light Detection and Ranging image, a Radio Detection and Ranging image, or a 3D Computer Aided Design modeling image, and an image from a different sensor, such as an RGB image sensor, captured from a same geo-location, which is used to determine the semantic features of the multi-modal image.

Abstract: During GPS-denied/restricted navigation, images proximate a platform device are captured using a camera, and corresponding motion measurements of the platform device are captured using an IMU device. Features of a current frame of the images captured are extracted. Extracted features are matched and feature information between consecutive frames is tracked. The extracted features are compared to previously stored, geo-referenced visual features from a plurality of platform devices. If one of the extracted features does not match a geo-referenced visual feature, a pose is determined for the platform device using IMU measurements propagated from a previous pose and relative motion information between consecutive frames, which is determined using the tracked feature information.

Abstract: A method, machine readable medium and system for RGBD semantic segmentation of video data includes determining semantic segmentation data and depth segmentation data for less than all classes for images of each frame of a first video, determining semantic segmentation data and depth segmentation data for images of each key frame of a second video including a synchronous combination of respective frames of the RGB video and the depth-aware video in parallel to the determination of the semantic segmentation data and the depth segmentation data for each frame of the first video, temporally and geometrically aligning respective frames of the first video and the second video, and predicting semantic segmentation data and depth segmentation data for images of a subsequent frame of the first video based on the determination of the semantic segmentation data and depth segmentation data for images of a key frame of the second video.

Abstract: Techniques are disclosed for an image understanding system comprising a machine learning system that applies a machine learning model to perform image understanding of each pixel of an image, the pixel labeled with a class, to determine an estimated class to which the pixel belongs. The machine learning system determines, based on the classes with which the pixels are labeled and the estimated classes, a cross entropy loss of each class. The machine learning system determines, based on one or more region metrics, a weight for each class and applies the weight to the cross entropy loss of each class to obtain a weighted cross entropy loss. The machine learning system updates the machine learning model with the weighted cross entropy loss to improve a performance metric of the machine learning model for each class.

Abstract: A method for providing a real time, three-dimensional (3D) navigational map for platforms includes integrating at least two sources of multi-modal and multi-dimensional platform sensor information to produce a more accurate 3D navigational map. The method receives both a 3D point cloud from a first sensor on a platform with a first modality and a 2D image from a second sensor on the platform with a second modality different from the first modality, generates a semantic label and a semantic label uncertainty associated with a first space point in the 3D point cloud, generates a semantic label and a semantic label uncertainty associated with a second space point in the 2D image, and fuses the first space semantic label and the first space semantic uncertainty with the second space semantic label and the second space semantic label uncertainty to create fused 3D spatial information to enhance the 3D navigational map.

Abstract: Techniques are disclosed for improving navigation accuracy for a mobile platform. In one example, a navigation system comprises an image sensor that generates a plurality of images, each image comprising one or more features. A computation engine executing on one or more processors of the navigation system processes each image of the plurality of images to determine a semantic class of each feature of the one or more features of the image. The computation engine determines, for each feature of the one or more features of each image and based on the semantic class of the feature, whether to include the feature as a constraint in a navigation inference engine. The computation engine generates, based at least on features of the one or more features included as constraints in the navigation inference engine, navigation information. The computation engine outputs the navigation information to improve navigation accuracy for the mobile platform.

Abstract: A method, apparatus and system for visual localization includes extracting appearance features of an image, extracting semantic features of the image, fusing the extracted appearance features and semantic features, pooling and projecting the fused features into a semantic embedding space having been trained using fused appearance and semantic features of images having known locations, computing a similarity measure between the projected fused features and embedded, fused appearance and semantic features of images, and predicting a location of the image associated with the projected, fused features. An image can include at least one image from a plurality of modalities such as a Light Detection and Ranging image, a Radio Detection and Ranging image, or a 3D Computer Aided Design modeling image, and an image from a different sensor, such as an RGB image sensor, captured from a same geo-location, which is used to determine the semantic features of the multi-modal image.

Abstract: During GPS-denied/restricted navigation, images proximate a platform device are captured using a camera, and corresponding motion measurements of the platform device are captured using an IMU device. Features of a current frame of the images captured are extracted. Extracted features are matched and feature information between consecutive frames is tracked. The extracted features are compared to previously stored, geo-referenced visual features from a plurality of platform devices. If one of the extracted features does not match a geo-referenced visual feature, a pose is determined for the platform device using IMU measurements propagated from a previous pose and relative motion information between consecutive frames, which is determined using the tracked feature information.

Abstract: A method for providing a real time, three-dimensional (3D) navigational map for platforms includes integrating at least two sources of multi-modal and multi-dimensional platform sensor information to produce a more accurate 3D navigational map. The method receives both a 3D point cloud from a first sensor on a platform with a first modality and a 2D image from a second sensor on the platform with a second modality different from the first modality, generates a semantic label and a semantic label uncertainty associated with a first space point in the 3D point cloud, generates a semantic label and a semantic label uncertainty associated with a second space point in the 2D image, and fuses the first space semantic label and the first space semantic uncertainty with the second space semantic label and the second space semantic label uncertainty to create fused 3D spatial information to enhance the 3D navigational map.

Abstract: Techniques are disclosed for improving navigation accuracy for a mobile platform. In one example, a navigation system comprises an image sensor that generates a plurality of images, each image comprising one or more features. A computation engine executing on one or more processors of the navigation system processes each image of the plurality of images to determine a semantic class of each feature of the one or more features of the image. The computation engine determines, for each feature of the one or more features of each image and based on the semantic class of the feature, whether to include the feature as a constraint in a navigation inference engine. The computation engine generates, based at least on features of the one or more features included as constraints in the navigation inference engine, navigation information. The computation engine outputs the navigation information to improve navigation accuracy for the mobile platform.

Abstract: Techniques for augmenting a reality captured by an image capture device are disclosed. In one example, a system includes an image capture device that generates a two-dimensional frame at a local pose. The system further includes a computation engine executing on one or more processors that queries, based on an estimated pose prior, a reference database of three-dimensional mapping information to obtain an estimated view of the three-dimensional mapping information at the estimated pose prior. The computation engine processes the estimated view at the estimated pose prior to generate semantically segmented sub-views of the estimated view. The computation engine correlates, based on at least one of the semantically segmented sub-views of the estimated view, the estimated view to the two-dimensional frame. Based on the correlation, the computation engine generates and outputs data for augmenting a reality represented in at least one frame captured by the image capture device.

Abstract: A method and apparatus for providing three-dimensional navigation for a node comprising an inertial measurement unit for providing gyroscope, acceleration and velocity information (collectively IMU information); a ranging unit for providing distance information relative to at least one reference node; at least one visual sensor for providing images of an environment surrounding the node; a preprocessor, coupled to the inertial measurement unit, the ranging unit and the plurality of visual sensors, for generating error states for the IMU information, the distance information and the images; and an error-state predictive filter, coupled to the preprocessor, for processing the error states to produce a three-dimensional pose of the node.