Deep strategy learning method facing complex missions in large-scale environment

A technology for policy learning and complex tasks, applied in the field of policy search reinforcement learning algorithms for continuous state action spaces, can solve problems such as the limitations of reinforcement learning, and achieve the effects of strong generalization ability, improved automation, and good technical support

Active Publication Date: 2016-11-09
深圳市安软科技股份有限公司
View PDF5 Cites 27 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In the face of large-scale dynamic environments, it is difficult for experts to provide accurate state feature representation for the input of reinforcement learning systems
Therefore, the manual design of state variables limits the practical application of reinforcement learning
[0007] (2) Facing the limitations of complex tasks
However, there is no complete theoretical solution that can effectively solve complex tasks in large-scale environments

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Deep strategy learning method facing complex missions in large-scale environment
  • Deep strategy learning method facing complex missions in large-scale environment
  • Deep strategy learning method facing complex missions in large-scale environment

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0035] The present invention will be further described in detail below in conjunction with the accompanying drawings and through specific embodiments. The following embodiments are only descriptive, not restrictive, and cannot limit the protection scope of the present invention.

[0036] In the implementation process of the reinforcement learning scheme for complex tasks in a large-scale environment described in the present invention, the interaction process between the agent and the environment is modeled as a Markov decision process (MDP), which can use a quadruple to represent (S, A, P T , P I , r, γ): where S represents the continuous state space, A represents the continuous action space, and P T (s t+1 |s t , a t ) means that in the current state s t take action a t then transition to the next state s t+1 The state transition probability density function of P I (s 1 ) is the initial state probability density function of the agent, r(s t , a t ,s t+1 ) represen...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a deep strategy learning method facing complex missions in a large-scale environment. The deep strategy learning method uses a deep nerve network to describe a state variable sensed by an intelligent agent, constructs a strategy model having a deep recursion type model, uses a strategy searching study algorithm to search an optimal parameter, and trains a nerve network until convergence. The deep strategy learning method uses a highly abstract and distributed expression capability of the deep nerve network to express the state variable sensed by the intelligent agent and constructs a strategy model having the deep recursion type structure, and is a first complete and intensive learning scheme which can systematically solve a complex decision problem in the large-scale environment.

Description

technical field [0001] The invention belongs to the field of machine learning, and mainly relates to reinforcement learning algorithms, in particular to strategy search reinforcement learning algorithms for continuous state action spaces. Background technique [0002] Reinforcement learning (RL) is an important learning method in the field of machine learning. It mainly studies how the agent makes better decisions according to the environment at that time. one of the research areas of interest. [0003] Reinforcement learning describes the process of continuous decision-making and control of agents to achieve tasks. It does not require prior knowledge like supervised learning, nor does it require experts to give accurate reference standards, but acquires knowledge by interacting with the environment. Action selection is performed autonomously, and finally an optimal action selection strategy suitable for the current state is found, and the maximum cumulative reward of the e...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06N3/08
CPCG06N3/08
Inventor 赵婷婷杨巨成赵希任德华陈亚瑞房珊珊
Owner 深圳市安软科技股份有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap