Method and device for training intelligent agent
A technology of agents and individuals, applied in the field of training agents, can solve problems such as complex interaction relationships between agents and the difficulty in meeting the needs of the global coordination mechanism
Active Publication Date: 2021-08-03
HUAWEI TECH CO LTD
View PDF7 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
When the number of agents is small, the effect of the global coordination mechanism is acceptable; when the number of agents is large, the interaction between agents is extremely complex, and the effect of the global coordination mechanism is difficult to meet the requirements
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0074] The technical solution in this application will be described below with reference to the accompanying drawings.
[0075] figure 1 is a multi-agent system suitable for this application.
[0076] figure 1 In , A~F represent 6 routers, and each router is equipped with a neural network. Therefore, a router is equivalent to an agent, and training an agent means training a neural network deployed on an agent. The lines between the routers represent communication lines. A to D are four edge routers, and the traffic between edge routers is called an aggregated flow. For example, the traffic from A to C is one aggregated flow, and the traffic from C to A is another aggregated flow.
[0077] Aggregated flows between multiple routers can be aggregated by N B (N B -1) OK, N B is the number of edge routers in the plurality of routers. exist figure 1 In the system shown, there are 4 edge routers, therefore, there are 12 aggregation flows in this system.
[0078] For each agg...
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
Login to View More Abstract
The invention provides a method for training agents. The method comprises the following steps: acquiring environment information of a first agent and environment information of a second agent; generating first information according to the environment information of the first intelligent agent and the environment information of the second intelligent agent; employing the first information to train the first intelligent agent, so the first intelligent agent outputs individual cognitive information and neighborhood cognitive information, wherein the neighborhood cognitive information of the first intelligent agent is consistent with the neighborhood cognitive information of the second intelligent agent. Since the neighborhood cognitive information of the target agent is the same as or similar to the neighborhood cognitive information of the neighborhood agent, the target agent obtained by training based on the neighborhood cognitive information of the target agent improves the correct cognitive degree of the target agent to the neighborhood environment; and the action generated by the finally obtained target agent can improve the collaboration effect among the multiple agents.
Description
technical field [0001] The present application relates to the field of artificial intelligence, in particular to a method and device for training an agent. Background technique [0002] Multi-agent collaboration is an application scenario in the field of artificial intelligence. For example, in a communication network containing multiple routers, each router can be regarded as an agent, each router has its own traffic scheduling strategy, and the traffic scheduling strategies of multiple routers need to be coordinated with each other, so as to use less The resource completes the traffic scheduling task. [0003] A method to solve the above problem is multi-agent reinforcement learning, which describes the goal of a specific task as a reward function, through the agent directly interacting with the environment and other agents, and automatically learning the strategy that can obtain the maximum long-term cumulative reward. And then coordinate multiple agents to solve specif...
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
Login to View More IPC IPC(8): H04L12/721H04L12/709G06N3/00G06N3/08H04L45/243
CPCH04L45/14H04L45/245G06N3/008G06N3/08H04L45/06H04L45/08G06F17/16G06N3/092G06N3/045G06N3/006G06N3/088
Inventor 毛航宇刘武龙郝建业
Owner HUAWEI TECH CO LTD