Collaborative multi-robot tasks using action primitives
AssigneeIntel Corporation
InventorsVenkat Natarajan, Arjun Kg, Gagan Acharya, Amit Sudhir Baxi, Rita H. Wouhaybi, Wen-Ling Margaret Huang
Life of this patent
Estimating the term…
Abstract
Various aspects of methods, systems, and use cases include techniques for training or using a model to control a robot. A method may include identifying a set of action primitives applicable to a set of robots, receiving information corresponding to a task (e.g., a collaborative task), and determining at least one action primitive based on the received information. The method may include training a model to control operations of at least one robot of the set of robots using the received information and the at least one action primitive.
Claims25
A system for robotics machine learning training, the system comprising:
- at least one processing circuitry;
- and
- memory, including instructions for training a set of robots to perform collaborative tasks, which when executed by the at least one processing circuitry, cause the at least one processing circuitry to perform operations to:
- identify a set of action primitives applicable to the set of robots from an action primitive library, each action primitive of the set of action primitives defining a control action for a robot or a cooperative action of two or more robots;
- receive information corresponding to a collaborative task;
- determine at least one action primitive to retrieve from the action primitive library based on the received information;
- and
- train a model to control operations of at least one robot of the set of robots to complete the collaborative task using the received information and the at least one action primitive.
At least one machine-readable medium including instructions for robotics machine learning training, which when executed by processing circuitry, causes the processing circuitry to perform operations to:
- identify a set of action primitives applicable to a set of robots from an action primitive library, each action primitive of the set of action primitives defining a control action for a robot or a cooperative action of two or more robots;
- receive information corresponding to a collaborative task, including a reward;
- determine at least one action primitive to retrieve from the action primitive library based on the received information;
- and
- train a model to control operations of at least one robot of the set of robots to complete the collaborative task using the received information and the at least one action primitive.
A robot for machine learning training, the robot comprising:
- processing circuitry;
- and
- memory, including instructions for use in training a set of robots to perform collaborative tasks, which when executed by the processing circuitry, cause the processing circuitry to perform operations to:
- generate a set of action primitives applicable to the set of robots for storing in a primitives library, each action primitive of the set of action primitives defining a control action for the robot
- generate a set of interaction primitives applicable to the set of robots for storing in the primitives library, each interaction primitive of the set of interaction primitive defining a cooperative control action for the robot and at least one other robot;
- receive at least one action primitive to retrieve from the action primitive library and a reward based on information corresponding to a collaborative task, the at least one action primitive used to complete a portion of the collaborative task;
- and
- iteratively train a model with the robot to control operations of the set of robot to complete the collaborative task using the at least one action primitive and the reward.
An apparatus for robotics machine learning training, the apparatus comprising:
- means for identifying a set of action primitives applicable to the set of independent robots from an action primitive library, each action primitive of the set of action primitives defining a control action for a robot or a cooperative action involving two or more robots;
- means for receiving3 information corresponding to a robotic task, including a reward;
- means for determining at least one action primitive to retrieve from the action primitive library based on the received information;
- and
- means for training a model to control operations of at least one robot of the set of independent robots to complete the collaborative task using the received information and the at least one action primitive.
Dependencies are inferred from the claim text.