The present invention relates to grid and voxel positioning methods based on a laser observation direction, a robot, and a chip. The grid positioning method comprises: selecting two preset intersection points that are in a same preset grid in which a laser point is located; then, in the two preset intersection points, setting the preset intersection point farthest from an observation point as a first preset intersection point, and setting the preset intersection point closest to the same observation point as a second preset intersection point; and according to a magnitude relationship between a ratio of a second preset distance to a first preset distance and a preset proportion coefficient, determining a target grid hit by the laser point in the direction of an observation ray, so as to form a latest hit grid position of the laser point within the two-dimensional grid map.

A method and a system for rhythmic motion control of a robot based on a neural oscillator, including: acquiring a current state of the robot, and a phase and a frequency generated by the neural oscillator; and obtaining a control instruction according to the acquired current state, phase and frequency and a preset reinforcement learning network so as to control the robot. The preset reinforcement learning network includes an action space, a pattern formation network and the neural oscillator. A control structure designed by the present disclosure, which is composed of the neural oscillator and the pattern formation network, can ensure formation of an expected rhythmic motion behavior; and meanwhile, a designed action space for joint position increment can effectively accelerate the training process of rhythmic motion reinforcement learning, and solve a problem that design of the reward function is time-consuming and difficult in learning with existing model-free reinforcement learning.

A robot control method, including: determining a target operation map according to target position information of a current operation of a robot; determining, based on a pre-established map connection relationship, a transfer position that the robot moves to; controlling the robot to move to the transfer position and switching maps, until the robot moves to the target operation map; and controlling the robot to move to an operation destination according to map information of the target operation map and the target position information.

Systems and methods are described for instructing performance of an action by a mobile robot based on transformed data. A system may obtain a site model in a first data format and sensor data in a second data format. The site model and/or the sensor data may be annotated. The system may transform the site model and the sensor data to generate transformed data in a third data format. The system may provide the transformed data to a computing system. For example, the system may provide the transformed data to a machine learning model. Based on the output of the computing system, the system may identify an action and instruct performance of the action by a mobile robot.

Techniques for determining robotic step timing and sequencing using reinforcement learning are provided. In one aspect, a method includes receiving a target trajectory for a robot and receiving a state of the robot. The method further includes generating, using a neural network, a set of gait timing parameters for the robot based, at least in part, on the state of the robot and the target trajectory and controlling movement of the robot based on the set of gait timing parameters.

A robot includes: a main body on which a person is carried; a handle to be gripped by the person on the main body; a moving device configured to move the main body; an operation terminal attachable to and detachable from the robot and configured to receive an input of a command related to an operation of the robot; and a control device configured to control the robot in accordance with the command received from the operation terminal. In a state where the operation terminal is attached to the robot, the operation terminal is disposed at a position where the operation terminal is operated while the handle is gripped by the person.

A method and a system for rhythmic motion control of a robot based on a neural oscillator, including: acquiring a current state of the robot, and a phase and a frequency generated by the neural oscillator; and obtaining a control instruction according to the acquired current state, phase and frequency and a preset reinforcement learning network so as to control the robot. The preset reinforcement learning network includes an action space, a pattern formation network and the neural oscillator. A control structure designed by the present disclosure, which is composed of the neural oscillator and the pattern formation network, can ensure formation of an expected rhythmic motion behavior; and meanwhile, a designed action space for joint position increment can effectively accelerate the training process of rhythmic motion reinforcement learning, and solve a problem that design of the reward function is time-consuming and difficult in learning with existing model-free reinforcement learning.

A robot control method, and a computer-readable storage medium and a wheel-legged biped robot using the same are provided. The method includes: determining a kinetic model of the wheel-legged biped robot; determining, using the kinetic model, a sliding surface of the wheel-legged biped robot; determining, according to the sliding surface, a double power reaching law and a sliding mode control law of the wheel-legged biped robot; and controlling, according to the sliding surface, the double power reaching law and the sliding mode control law, the wheel-legged biped robot. Through the above-mentioned method, the adaptability of the wheel-legged biped robot to uncertain external disturbances can be enhanced, thereby improving its robustness to effectively maintain its balance even in the environment with complex terrain.

Provided is a data processing unit that analyzes detection information of a visual sensor and determines a movement route of the moving apparatus, and the data processing unit generates traveling surface shape data such as three-dimensional point cloud data that enables analysis of a shape of a traveling surface of the moving apparatus. The data processing unit selects a generation target region of the traveling surface shape data on the basis of the target movement route information of the moving apparatus and a predetermined search range region, generates the traveling surface shape data in the selection region selected, and determines a movement route such as a foot placement position of the moving apparatus with reference to the generated traveling surface shape data.

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for controlling agents. In particular, an agent can be controlled using a hierarchical controller that includes a task policy neural network and a low-level controller neural network.