Techniques for training and implementing reinforcement learning policies for robot control
AssigneeNVIDIA Corporation
InventorsBingjie TANG, Yashraj Shyam NARANG, Dieter FOX, Fabio TOZETO RAMOS
Life of this patent
Estimating the term…
Abstract
One embodiment of a method for training a machine learning model to control a robot includes causing a model of the robot to move within a simulation based on one or more outputs of the machine learning model, computing an error within the simulation, computing at least one of a reward or an observation based on the error, and updating one or more parameters of the machine learning model based on the at least one of a reward or an observation.
Claims20
A computer-implemented method for training a machine learning model, the method comprising:
- performing one or more first operations to train the machine learning model based on a first range of difficulties of a task;
- determining a success rate of the machine learning model at the task;
- and
- in response to determining that the success rate is greater than a threshold, performing one or more second operations to train the machine learning model based on a second range of difficulties of the task, wherein a lower bound of the second range of difficulties is higher than a lower bound of the first range of difficulties, and wherein an upper bound of the second range of difficulties and an upper bound of the first range of difficulties are a same upper bound.
One or more non-transitory computer-readable storage media including instructions that, when executed by at least one processor, cause the at least one processor to perform steps for training a machine learning model, the steps comprising:
- performing one or more first operations to train the machine learning model based on a first range of difficulties of a task;
- determining a success rate of the machine learning model at the task;
- and
- in response to determining that the success rate is greater than a threshold, performing one or more second operations to train the machine learning model based on a second range of difficulties of the task, wherein a lower bound of the second range of difficulties is higher than a lower bound of the first range of difficulties, and wherein an upper bound of the second range of difficulties and an upper bound of the first range of difficulties are a same upper bound.
A system, comprising:
- one or more memories storing instructions;
- and
- one or more processors that are coupled to the one or more memories and, when executing the instructions, are configured to:
- perform one or more first operations to train a machine learning model based on a first range of difficulties of a task,
- determine a success rate of the machine learning model at the task, and
- in response to determining that the success rate is greater than a threshold, perform one or more second operations to train the machine learning model based on a second range of difficulties of the task, wherein a lower bound of the second range of difficulties is higher than a lower bound of the first range of difficulties, and wherein an upper bound of the second range of difficulties and an upper bound of the first range of difficulties are a same upper bound.
Dependencies are inferred from the claim text.