Subjects of the present disclosure are methods for training deep learning models, methods for predicting a map matching parameter, methods for updating a digital road map, and a computer program making it possible to implement the methods and devices for updating a digital road map. The general principle is based on the use of machine learning. Accordingly, a statistical deep learning model is trained according to a “supervised” machine learning scheme. Thereafter, the pretrained statistical deep learning model is used to predict a map matching parameter on the basis of a measurement of geographic coordinates and of an identifier of the position sensor that has performed the measurement of geographic coordinates. Finally, the map matching parameter can be used to update a digital road map.

A map management device which can operate an autonomous mobile body also in a region where persons exist is provided. The map management device includes a dynamic map information generating unit which generates dynamic map information in which an obstacle is reflected on the basis of operation information on a facility obtained from a building facility management device and location information on a person obtained from a security camera or a motion detector, as map information to be used for controlling the autonomous mobile body with respect to static map information configured in advance with two-dimensional or three-dimensional grids for each floor of a building. According to this configuration, it is possible to operate the autonomous mobile body also in a region where persons exist.

A computer system including one or more processors programmed or configured to receive image data associated with an image of one or more roads, where the one or more roads comprise one or more lanes, determine a lane classification of the one or more lanes based on the image data associated with the image of the one or more roads, and provide lane classification data associated with the lane classification of the one or more lanes.

An end-to-end system for data generation, map creation using the generated data, and localization to the created map is disclosed. Mapstreams—or streams of sensor data, perception outputs from deep neural networks (DNNs), and/or relative trajectory data—corresponding to any number of drives by any number of vehicles may be generated and uploaded to the cloud. The mapstreams may be used to generate map data—and ultimately a fused high definition (HD) map—that represents data generated over a plurality of drives. When localizing to the fused HD map, individual localization results may be generated based on comparisons of real-time data from a sensor modality to map data corresponding to the same sensor modality. This process may be repeated for any number of sensor modalities and the results may be fused together to determine a final fused localization result.

Provided are a method, performed by an electronic device, of controlling a vehicle, and an electronic device for the same. A method, performed by an electronic device, of controlling a vehicle includes: transmitting, to an external server communicatively connected to the vehicle, as profile information of the vehicle, sensor information regarding at least one sensor mounted on the vehicle, communication efficiency information of the vehicle, and driving information of the vehicle; receiving, from the external server, precise map data related to at least one map layer selected based on the profile information of the vehicle from among a plurality of map layers that are combined to form a precise map and distinguished according to attributes thereof; and controlling the vehicle to perform autonomous driving by using the received at least one precise map data.

In various examples, a method to manage map data includes storing a map of a geographic area using an immutable tree. The immutable tree comprises a plurality of nodes stored using a distributed hash table. The plurality of nodes include a plurality of map tiles. At least two map tiles of the plurality of map tiles cover different geographic subregions of the geographic area of the map. The method includes hosting one or more binary large objects (BLOBs) that correspond to the plurality of map tiles in an origin data plane. The method includes making the one or more BLOBs available for distribution to one or more client devices using a content delivery network (CDN).

A method, apparatus and computer program product are provided for learning to generate maps from raw geospatial observations from sensors traveling within an environment. Methods may include: receiving a plurality of sequences of geospatial observations from discrete trajectories; aligning the discrete trajectories generating aligned geospatial observations; concatenating the aligned geospatial observations; performing attentional clustering on the concatenated, aligned geospatial observations to obtain a set of entities with feature dimensionality; processing the set of entities through an iterative attention model incorporating a Gated Recurrent Unit gating pattern to obtain attentional layer outputs; generating, from one or more Set Transformers, a feature set of map object geometries based, at least in part, on the attentional layer outputs; updating a map geometry based on the feature set from the Set Transformers generating an updated map geometry; and provide for navigational assistance or at least semi-autonomous vehicle control based on the updated map geometry.

Systems and methods for aggregating and integrating distributed grid element inputs are disclosed. A data platform is provided for a distribution power grid. The data platform provides a crowd-sourced gaming system for identifying grid elements and determining dynamic electric power topology. The data platform also provides an interactive interface for displaying a view of a certain area with identified grid elements. The data platform communicatively connects to the identified grid elements, collects data from the identified grid elements, and manages the distribution power grid.

Disclosed herein is a technique for the generating and provision of digital map data that is safe and reliable. The technique enables the verification of the digital map data in a map-client using a simple and efficient data structure to check the correctness of the map data before in-vehicle delivery to components that rely on this map data.

An end-to-end system for data generation, map creation using the generated data, and localization to the created map is disclosed. Mapstreams—or streams of sensor data, perception outputs from deep neural networks (DNNs), and/or relative trajectory data—corresponding to any number of drives by any number of vehicles may be generated and uploaded to the cloud. The mapstreams may be used to generate map data—and ultimately a fused high definition (HD) map—that represents data generated over a plurality of drives. When localizing to the fused HD map, individual localization results may be generated based on comparisons of real-time data from a sensor modality to map data corresponding to the same sensor modality. This process may be repeated for any number of sensor modalities and the results may be fused together to determine a final fused localization result.