Methods, systems, and apparatus, including computer programs encoded on computer storage media, for using learnable robotic control plans. One of the methods comprises obtaining a learnable robotic control plan comprising data defining a state machine that includes a plurality of states and a plurality of transitions between states, wherein: one or more states are learnable states, and each learnable state comprises data defining (i) one or more learnable parameters of the learnable state and (ii) a machine learning procedure for automatically learning a respective value for each learnable parameter of the learnable state; and processing the learnable robotic control plan to generate a specific robotic control plan, comprising: obtaining data characterizing a robotic execution environment; and for each learnable state, executing, using the obtained data, the respective machine learning procedures defined by the learnable state to generate a respective value for each learnable parameter of the learnable state.

A dynamic learning method for a robot includes a training and learning mode. The training and learning mode includes the following steps: dynamically annotating a belonging and use relationship between an object and a person in a three-dimensional environment to generate an annotation library; acquiring a rule library, and establishing a new rule and a new annotation by means of an interactive demonstration behavior based on the rule library and the annotation library; and updating the new rule to the rule library and updating the new annotation to the annotation library when it is determined that the established new rule is not in conflict with rules in the rule library and the new annotation is not in conflict with annotations in the annotation library.

Embodiments of architecture, systems, and methods to develop a learning/evolving system to robotically perform and model one or more activities of a medical procedure where the medical procedure may include diagnosing a patient's medical condition(s), treating medical condition(s), and robotically diagnosing a patient's medical condition(s) and performing one or more medical procedure activities based on the diagnosis without User intervention where the activities may be performed in computer-based environment formed by the learning/evolving system, live, or a combination thereof.

Disclosed is a mobile robot configured to cut lawn in a work area. The mobile robot may include a main body, a weight sensing sensor, an obstacle sensing sensor, a blade, and a processor. The mobile robot may execute an artificial intelligence (AI) algorithm and/or a machine learning algorithm, and perform communication with other electronic devices in a 5G communication environment. As a result, it is possible to enhance user convenience.

Disclosed herein an artificial intelligence (AI) apparatus for controlling authority to use an external device based on user identification using image recognition including a memory configured to store information on a user registered in the AI apparatus and authority information indicating whether a user is capable of use at least one external device under a predetermined condition, a communicator configured to receive a first image file obtained by photographing an environment including the at least one external device, a learning processor configured to provide the first image file to an image recognition model for specifying a face of a person included in an image file and an external device to be used by the person to specify first face information of a person included in the first image file and information on a first external device to be used by the person in the first image file, and a processor configured to control use of the first external device by the first user based on the authority, by acquiring a first user corresponding to the first face information and authority information of the first user.

An autonomous personal companion executing a method including capturing data related to user behavior. Patterns of user behavior are identified in the data and classified using predefined patterns associated with corresponding predefined tags to generate a collected set of one or more tags. The collected set is compared to sets of predefined tags of a plurality of scenarios, each to one or more predefined patterns of user behavior and a corresponding set of predefined tags. A weight is assigned to each of the sets of predefined tags, wherein each weight defines a corresponding match quality between the collected set of tags and a corresponding set of predefined tags. The sets of predefined tags are sorted by weight in descending order. A matched scenario is selected for the collected set of tags that is associated with a matched set of predefined tags having a corresponding weight having the highest match quality.

Techniques are described herein for generating a dynamical systems control policy. A non-parametric family of smooth maps is defined on which vector-field learning problems can be formulated and solved using convex optimization. In some implementations, techniques described herein address the problem of generating contracting vector fields for certifying stability of the dynamical systems arising in robotics applications, e.g., designing stable movement primitives. These learning problems may utilize a set of demonstration trajectories, one or more desired equilibria (e.g., a target point), and once or more statistics including at least an average velocity and average duration of the set of demonstration trajectories. The learned contracting vector fields may induce a contraction tube around a targeted trajectory for an end effector of the robot. In some implementations, the disclosed framework may use curl-free vector-valued Reproducing Kernel Hilbert Spaces.

A computer-implemented method of machine-learning is described that includes obtaining a dataset of virtual scenes. The dataset of virtual scenes belongs to a first domain. The method further includes obtaining a test dataset of real scenes. The test dataset belongs to a second domain. The method further includes determining a third domain. The third domain is closer to the second domain than the first domain in terms of data distributions. The method further includes learning a domain-adaptive neural network based on the third domain. The domain-adaptive neural network is a neural network configured for inference of spatially reconfigurable objects in a real scene. Such a method constitutes an improved method of machine learning with a dataset of scenes including spatially reconfigurable objects.

An object manipulation apparatus according to an embodiment of the present disclosure includes a memory and a hardware processor coupled to the memory. The hardware processor is configured to: calculate, based on an image in which one or more objects to be grasped are contained, an evaluation value of a first behavior manner of grasping the one or more objects; generate information representing a second behavior manner based on the image and a plurality of evaluation values of the first behavior manner; and control actuation of grasping the object to be grasped in accordance with the information being generated.

There are provided systems, methods, and apparatus, for optimizing a policy controller to control a robotic agent that interacts with an environment to perform a robotic task. One of the methods includes optimizing the policy controller using a neural network that generates numeric embeddings of images of the environment and a demonstration sequence of demonstration images of another agent performing a version of the robotic task.