An information processing system is provided which can increase the possibility that a path along which work quality is high is set in a work area. In an acquisition step of the information processing system, boundary information is acquired, the boundary information indicating the boundary of the work area targeted by a work machine that can travel autonomously. In a setting step, a path of the work machine is set in order to minimize unreached areas from the work area on the basis of the acquired boundary information, and when the work quality in the case of using a path set by a first method does not meet a prescribed criterion, a path is set by a second method.

A method of training a machine learning, ML, algorithm to control a watercraft is described. The watercraft is a submarine or a submersible submerged in water. The method is implemented, at least in part, by a computer, comprising a processor and a memory, aboard the watercraft. The method comprises: obtaining training data including respective sets of environmental parameters and corresponding actions of a set of communicatively isolated watercraft, including a first watercraft; and training the ML algorithm comprising determining relationships between the respective sets of environmental parameters and the corresponding actions of the watercraft of the set thereof. A method of controlling a watercraft by a trained ML algorithm is also described.

In various examples, a machine may generate, using sensor data capturing one or more views of an environment, a virtual environment including a 3D representation of the environment. The machine may render, using one or more virtual sensors in the virtual environment, one or more images of the 3D representation of the environment. The machine may apply the one or more images to one or more machine learning models (MLMs) trained to generate one or more predictions corresponding to the environment. The machine may perform one or more control operations based at least on the one or more predictions generated using the one or more MLMs.

The present embodiment relates to a cloud-based robot control method for controlling a plurality of robots which are positioned in a plurality of spaces divided arbitrarily, the method comprising the steps of: generating a control base model which can be applied to the plurality of robots in a cloud server; distributing the control base model to edge servers allocated to respective spaces; upgrading the control base model in accordance with the plurality of robots of a space, in the edge server; directly transmitting the upgraded control model from the edge server to another edge server; and controlling the plurality of robots by means of the upgraded control model in the edge server. Therefore, by sharing a deep-learning model among edge servers, supporting heterogeneous robots and heterogeneous services is possible. Further, a base deep-learning model from the cloud server is tuned into a customized deep-learning model to be suitable for respective robots in the edge server, and the deep-learning model is upgraded to an adaptive deep-learning model to be suitable for a service provided by respective robots, and thus an optimized service can be provided.

An automatic implement attachment and detachment system having a ground utility robot a sensor, a computer processor, an artificial intelligence processing unit for learning, and a computer memory where the system also includes a quick hitch attachment apparatus having a body securable to the ground utility robot, at least one mateable connection part, and an implement having a connection member where the implement attachment system is configured to automatically attach and detach the implement to and from the ground utility robot.

Robotic devices may be trained by a user guiding the robot along target action trajectory using an input signal. A robotic device may comprise an adaptive controller configured to generate control signal based on one or more of the user guidance, sensory input, performance measure, and/or other information. Training may comprise a plurality of trials, wherein for a given context the user and the robot's controller may collaborate to develop an association between the context and the target action. Upon developing the association, the adaptive controller may be capable of generating the control signal and/or an action indication prior and/or in lieu of user input. The predictive control functionality attained by the controller may enable autonomous operation of robotic devices obviating a need for continuing user guidance.

In conventional robot navigation techniques learning and planning algorithms act independently without guiding each other simultaneously. A method and system for robotic navigation with simultaneous local path planning and learning is disclosed. The method discloses an approach to learn and plan simultaneously by assisting each other and improve the overall system performance. The planner acts as an actuator and helps to balance exploration and exploitation in the learning algorithm. The synergy between dynamic window approach (DWA) as a planning algorithm and a disclosed Next best Q-learning (NBQ) as a learning algorithm offers an efficient local planning algorithm. Unlike the traditional Q-learning, dimension of Q-tree in the NBQ is dynamic and does not require to define a priori.

A system and methods for operating a multi-robot system (MRS) are disclosed. An example method can include receiving at least one transportation task; determining an optimal path for executing the at least one transportation task based at least in part on: (i) one or more transportation task parameters, (ii) a shared global critic function accessible to the first robot and the at least one additional robot, and (iii) a local critic function unique to the first robot; and executing the at least one transportation task in accordance with the determined optimal path.

Systems, methods, and computer-readable media for dynamic changes to both a learned control policy in the event of a change in the environment (e.g., introduction of a new or unseen obstacle). Rather than having to implement an entirely new policy (and a new global Q table), which can delay performance of tasks by agent(s), the present embodiments allow for a reduced delay in updating local Q table(s) based on detection of a new change in the environment. Locally changing the policy allows for more efficient updating of the policy based on changes in the environment, rather than globally changing the Q table after each change. Particularly in an event with multiple changes in the environment, the present embodiments increase efficiency in updating local and global Q tables while also reducing a delay in providing new instructions to the agent(s) in completing tasks.

Methods and devices for processing image data for scene completion, including receiving an original image from an original viewpoint corresponding to a first direction, wherein the original image includes an object; obtaining a first image from a new viewpoint corresponding to a second direction different from the first direction by rotating the original image based on 3-dimensional (3D) information generated from 2-dimensional (2D) information which is obtained from the original image; determining an area within the first image for generating a second surface of the object based on depth information about a depth between the object and the background of the original image, wherein the determined area is expected to include an object area; and obtaining a second image by inputting the first image and the determined area to an artificial intelligence (AI) inpainting model, wherein the AI inpainting model generates the second surface of the object which occupies a portion of the determined area in the second image.