Systems and methods for the simultaneous localization and mapping of autonomous vehicle systems are provided. A method includes receiving a plurality of input image frames from the plurality of asynchronous image devices triggered at different times to capture the plurality of input image frames. The method includes identifying reference image frame(s) corresponding to a respective input image frame by matching the field of view of the respective input image frame to the fields of view of the reference image frame(s). The method includes determining association(s) between the respective input image frame and three-dimensional map point(s) based on a comparison of the respective input image frame to the one or more reference image frames. The method includes generating an estimated pose for the autonomous vehicle the one or more three-dimensional map points. The method includes updating a continuous-time motion model of the autonomous vehicle based on the estimated pose.

A legged robot motion control method, apparatus, and device, and a storage medium. The method includes: acquiring center of mass state data corresponding to a spatial path starting point and spatial path ending point of a motion path; determining a candidate foothold of each foot in the motion path based on the spatial path starting point and the spatial path ending point; determining a variation relationship between a center of mass position variation coefficient and a foot contact force based on the center of mass state data; screening out, under restrictions of a constraint set, a target center of mass position variation coefficient and target foothold that satisfy the variation relationship; determining a target motion control parameter according to the target center of mass position variation coefficient and the target foothold; and controlling a legged robot based on the target motion control parameter to move according to the motion path.

An image processing method may be applied to a movable platform, and the movable platform may comprise a first vision sensor and a second vision sensor. The method may include obtaining a first localized image of the first vision sensor within an overlapping visual range, obtaining a second localized image of the second vision sensor within the overlapping visual range; acquiring an image captured by the first vision sensor at a first moment and an image captured at a second moment, the first vision sensor being positioned in space at the first moment differently than at the second moment; and determining a relative positional relationship between an object in the space where the movable platform is located and the movable platform based on the first localized image, the second localized image, the image captured at the first moment and the image captured at the second moment.

In order to make it possible to control, according to a content of action of a moving body, an information amount of depth information so that the information amount of the depth information becomes appropriate, this invention includes: an obtaining section (11) that obtains a content of action of a moving body; and a control section (12) that controls, according to the content of the action of the moving body, an information amount of depth information obtained from a sensor.

Techniques are provided for monocular 3D object detection from an image semantics network. An image semantics network (ISN) is a single stage, single image object detection network that is based on single shot detection (SSD). In an embodiment, the ISN augments the SSD outputs to provide encoded 3D properties of the object along with a 2D bounding box and classification scores. For each priorbox, a 3D bounding box is generated for the object using the dimensions and location of the priorbox, the encoded 3D properties and camera intrinsic parameters.

Systems and methods are provided for crowdsourcing a sparse map for autonomous vehicle navigation. In one implementation, a non-transitory computer-readable medium may include a sparse map for autonomous vehicle navigation along a road segment. The sparse map may include at least one line representation of a road surface feature extending along the road segment, each line representation representing a path along the road segment substantially corresponding with the road surface feature, and wherein the road surface feature is identified through image analysis of a plurality of images acquired as one or more vehicles traverse the road segment and a plurality of landmarks associated with the road segment.

In a robot device that identifies an obstacle on the basis of detection information of a sensor, highly accurate robot control by correct obstacle identification is realized without erroneously recognizing a leg, an arm, or the like of the robot itself as an obstacle. A self-region filter processing unit removes object information corresponding to a component of a robot device from object information included in detection information of a visual sensor, a map image generation unit generates map data based on object information from which the object information corresponding to the component of the robot device has been removed, and a robot control unit controls the robot device on the basis of the generated map data. The self-region filter processing unit calculates variable filter regions of different sizes according to the motion speed of the movable part of the robot device, and executes processing of removing the object information in the variable filter regions from the detection information of the visual sensor.

Vehicles, systems, and methods for navigating or tracking the navigation of a materials handling vehicle along a surface that may include a camera and vehicle functions to match two-dimensional image information from camera data associated with the input image of overhead features with a plurality of global target locations of a warehouse map to generate a plurality of candidate optical targets, an optical target associated with each global target location and a code; filter the targets to determine a candidate optical target; decode the target to identify the associated code; identify an optical target associated with the identified code; determine a camera metric relative to the identified optical target and the position and orientation of the identified optical target in the warehouse map; calculate a vehicle pose based on the camera metric; and navigate the materials handling vehicle utilizing the vehicle pose.

A method for performing a task of a cleaning robot is provided. The method according to an embodiment includes generating a navigation map for driving the cleaning robot using a result of at least one sensor detecting a task area in which an object is arranged, obtaining recognition information of the object by applying an image of the object captured by at least one camera to a trained artificial intelligence model, generating a semantic map indicating environment of the task area by mapping an area of the object included in the navigation map with the recognition information of the object, and performing a task of the cleaning robot based on a control command of a user using the semantic map. An example of the trained artificial intelligence model may be a deep-learning neural network model in which a plurality of network nodes having weighted values are disposed in different layers and exchange data according to a convolution relationship, but the disclosure is not limited thereto.

Navigation system (300) for land, air, marine or submarine vehicle (302), comprising a remote control workstation (301) with Manual control mode (310), Mission Planning mode (330) and tactical control mode (360) initiating command-and-control options; a navigation module (100) retrofittably disposed on the vehicle (302); a plurality of perception sensors (318) disposed on the vehicle (302); the system (300) receives manual, electrical, radio and audio commands of human operator (305) in the manual control (310) and mission planning mode (330) and converts them to dataset for training a navigation model having a navigational algorithm. The perception sensors (318) generate dataset for self-learning in real time in manual control mode (310), mission control mode (330) and tactical control mode (360); the navigational system (300) autonomously navigates with presence of communication network (390) and in absence of communication network (390).