Ship control method based on competitive hybrid network

A hybrid network and control method technology, applied in neural learning methods, biological neural network models, instruments, etc., can solve problems such as low ship control accuracy

Active Publication Date: 2021-02-26
HARBIN ENG UNIV
View PDF4 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0010]The purpose of this invention is to solve the problem of low control accuracy of existing sh

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
  • Ship control method based on competitive hybrid network
  • Ship control method based on competitive hybrid network
  • Ship control method based on competitive hybrid network

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment approach 1

[0039] Specific implementation mode one: the specific process of the ship control method based on the competitive hybrid network in this implementation mode is as follows:

[0040] Step 1. Establishing an individual agent network model;

[0041] Step 2. Establish a dominant hybrid network model;

[0042] Step 3. Establish a state-value hybrid network model;

[0043] Step 4: Input the individual observation history into the individual agent network model to obtain the individual advantage value function A i (τ i ,a i ) and the individual state value function V i (τ i );

[0044] The individual advantage value function A i (τ i ,a i ) is passed to the dominant mixed network model, and the dominant mixed network model outputs the joint advantage function value A tot (τ,a);

[0045] The individual state value function V i (τ i ) is passed to the state value mixed network model, and the state value mixed network model outputs the joint state value mixed value V tot (τ...

specific Embodiment approach 2

[0052] Specific embodiment two: the difference between this embodiment and specific embodiment one is: the individual agent network model is established in the step one; the specific process is:

[0053] The individual agent network model includes n agent network structures (the number of n depends on the specific application scenario);

[0054] The network structure of each agent is the same, consisting of Group1 and Group2;

[0055] Group1 includes an input layer, a ReLU activation layer, a GRU unit and the first linear network layer;

[0056] The output end of the input layer is connected to the input end of the ReLU activation layer, the output end of the ReLU activation layer is connected to the input end of the GRU unit, and the output end of the GRU unit is connected to the first linear network layer;

[0057] These structures are connected in series.

[0058] The input layer accepts the observation history of the agent, and the observation history consists of the loc...

specific Embodiment approach 3

[0064] Specific implementation mode three: the difference between this implementation mode and specific implementation mode one or two is that: in the step two, a dominant hybrid network model is established; specifically:

[0065] The advantage hybrid network model consists of a parameter network structure and a super network structure;

[0066] The parameter network structure includes a fourth linear network layer and a fifth linear network layer, the fourth linear network layer and the fifth linear network layer are connected in parallel, and the inputs of the fourth linear network layer and the fifth linear network layer are global observations;

[0067] The fourth linear network layer processes the input global observation value to obtain an output value, which is used as the parameter matrix of the first layer network in the hypernetwork structure;

[0068] The fifth linear network layer processes the input global observation value to obtain the output value as the param...

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 relates to a ship control method, in particular to a ship control method based on a competitive hybrid network. The objective of the invention is to solve the problem of low control precision of an existing ship in a complex environment. The method comprises the following steps: 1, establishing an individual agent network model; 2, establishing a dominant hybrid network model; 3, establishing a state value hybrid network model; 4, inputting the individual observation history into the individual intelligent agent network model to obtain an individual dominant value function and anindividual state value function, transmitting the individual dominant value function to a dominant hybrid network model, and outputting a joint dominant function value by the dominant hybrid networkmodel; transmitting the individual state value function to a state value hybrid network model, and outputting a joint state value hybrid value by the state value hybrid network model; and adding the joint dominance function value and the joint state value mixed value to obtain a joint action value function. The method is applied to the field of ship control.

Description

technical field [0001] The invention relates to a ship control method. Background technique [0002] Many problems in real life can be attributed to cooperative multi-agent problems, such as traffic planning, robot control and automatic driving. However, there are many problems in directly applying the multi-intelligence algorithm to this real situation. For example, in reality, the field of view of the agent should only be local, and as the number of agents increases, the action space of the model will grow explosively, and How to judge the contribution of each agent in each task. These problems make the performance of multi-agent algorithms unsatisfactory in practical tasks. [0003] The initial solution to these problems is that each agent has a network to learn its own strategy alone, that is, independent learners (IQL). But this approach leads to no communication between agents, and the agent will regard other agents as part of the environment. This instability will ...

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
IPC IPC(8): G06F30/27G06N3/04G06N3/08G06N20/00
CPCG06F30/27G06N3/08G06N20/00G06N3/045
Inventor 王红滨谢晓东何鸣王念滨周连科崔琎
Owner HARBIN ENG UNIV
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