Methods for training and controlling a robot
MYUI2023004862UUndetermined Publication Date: 2025-02-11WONG JUN XUAN
0 Cites 0 Cited by
Patent Information
- Application Number
- MYUI2023004862
- Authority / Receiving Office
- MY · MY
- Patent Type
- Utility models
- Current Assignee / Owner
- Filing Date
- 2023-08-11
- Publication Date
- 2025-02-11
- Estimated Expiration
- 2043-08-11
Abstract
Embodiments relate to methods for training and controlling a robot. An exemplary method for training a robot having actuatable devices includes the steps of defining a target, an action space, an observation space and reward values, employing at least one Deep Q Network, DQN agent and at least one Deep Deterministic Policy Gradient, DDPG agent, actuating the at least one actuatable device to perform at least one action, recording the actions and outcomes, repeating the steps of actuating the at least one actuatable device and recording the action performed and outcome achieved, selecting a plurality of series of actions performed, calculating cumulative reward value and selecting at least one series having maximal cumulative reward value.
Need to check novelty before this filing date? Find Prior Art
