Brain Corporation

A method for localizing a mobile robot using a feature-embedded map includes detecting a power-on event of the robot at an unknown location, capturing a first image via a camera mounted on the robot, and encoding the first image into a first feature embedding vector using an image encoder. The method accesses a computer-readable map comprising a plurality of feature embedding vectors generated from images previously captured by the robot during prior navigation, where each feature embedding vector is stored at a respective physical location coordinate. The method determines similarity between the first feature embedding vector and the plurality of feature embedding vectors, identifies matching vectors exceeding a similarity threshold, and localizes the robot by correlating the unknown location with the physical location coordinates of the matching vectors. This enables autonomous navigation without QR codes, barcodes, or structural landmarks.

Systems and methods for systems and methods for real time calibration of multiple range sensors on a robot are disclosed herein. According to at least one non-limiting exemplary embodiment, methods for self-calibration are used to independently correct rotational errors in a pose of a sensor. In some instances, the self-calibration may further yield errors along at least one translational axis. According to at least one non-limiting exemplary embodiment, methods for cross-calibration are used to correct translational errors and to ensure all sensors on a robot agree on perceived locations of objects.

Systems and methods for detecting, mapping, and route planning around cliffs for robotic devices are disclosed herein. According to at least one non-limiting exemplary embodiment, a robot processing a three-dimensional range sensor scan may utilize pre-computed neighboring points to detect cliffs, navigable ramps, and holes in a plane of a map used by the robotic device to navigate.

Systems and methods for detection of features within data collected by a plurality of robots by a centralized server are disclosed herein. According to at least one non-limiting exemplary embodiment, a plurality of robots may be utilized to collect a substantial amount of feature data using one or more sensors coupled thereto, wherein use of the plurality of robots to collect the feature data yields accurate localization of the feature data and consistent acquisition of the feature data. Systems and methods disclosed herein further enable a cloud server to identify a substantial number of features within the acquired feature data for purposes of generating insights. The substantial number of features far exceed a practical number of features of which a single neural network may be trained to identify.

Systems and methods for detecting, mapping, and route planning around cliffs for robotic devices are disclosed herein. According to at least one non-limiting exemplary embodiment, a robot processing a three-dimensional range sensor scan may utilize pre-computed neighboring points to detect cliffs, navigable ramps, and holes in a plane of a map used by the robotic device to navigate.

Systems and methods for detection of features within data collected by a plurality of robots by a centralized server are disclosed herein. According to at least one non-limiting exemplary embodiment, a plurality of robots may be utilized to collect a substantial amount of feature data using one or more sensors coupled thereto, wherein use of the plurality of robots to collect the feature data yields accurate localization of the feature data and consistent acquisition of the feature data. Systems and methods disclosed herein further enable a cloud server to identify a substantial number of features within the acquired feature data for purposes of generating insights. The substantial number of features far exceed a practical number of features of which a single neural network may be trained to identify.

Systems and methods for data and confidentiality preservation for feature identification by robotic devices are disclosed herein. According to at least one non-limiting exemplary embodiment, a method for determining if an image depicts a human is disclosed. The determination may be utilized to either (i) censor the face of the human, or (ii) identify features within the image. The determination enhances feature identification by ensuring only uncensored and unobscured images are provided to one or more models for the identification to preserve human confidentiality.

Systems and methods for charging power supplies of robotic devise are disclosed herein. A charging dock includes a first surface comprising at least one axis of symmetry defining aligned with a normal approach angle of the robot, two or more charging pads configured to transfer charge and are disposed on a first outer surface of the charging dock, at least one or more slanted surfaces, wherein each of the one or more slanted surfaces are slanted with respect to the first outer surface, and at least one or more computer readable codes affixed to at least one of the one or more slanted surfaces and a flat surface of the charging dock in a position of the charging dock that is transverse to the normal approach angle.

Systems and methods for training a robot to autonomously travel a route. In one embodiment, a robot can detect an initial placement in an initialization location. Beginning from the initialization location, the robot can create a map of a navigable route and surrounding environment during a user-controlled demonstration of the navigable route. After the demonstration, the robot can later detect a second placement in the initialization location, and then autonomously navigate the navigable route. The robot can then subsequently detect errors associated with the created map. Methods and systems associated with the robot are also disclosed.

Systems and methods for localizing devices using network signatures and coverage maps measured by robots are disclosed herein. According to at least one non-limiting exemplary embodiment, a robot may generate a coverage map based on measurements collected during operation of the robot. The coverage map generated by the robot may be temporally accurate such that a device may be localized within the coverage map based on a received network signature from the device. The network signature comprising a measure of amplitudes of Wi-Fi networks and/or cellular networks at a point within an environment of the coverage map.