A model evaluation method based on multi-agent Q-learning
Through the multi-agent Q learning model evaluation method, the problem of complex queue control of search and rescue personnel during the search and rescue process is solved. Through model evaluation, the maximum function and timely discovery of problems are ensured, and the effective evaluation of the multi-agent Q learning model is realized. optimization.
Patent Information
- Application Number
- CN202110381457.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-04-09
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2041-04-09
AI Technical Summary
The existing multi-agent Q-learning model has complex overall queue control when search and rescue personnel search and rescue in the crashed sea area, and it is difficult to evaluate which theoretical model is more practical, and lacks effective evaluation methods.
A model evaluation method based on multi-agent Q learning is proposed. By establishing an initial model, data acquisition, analysis, function demonstration, data adjustment and secondary demonstration, the performance and cost of the model are compared and evaluated.
This method can accurately evaluate the functions and parameters of the model, find problems and improve them, ensure that the model has no impact in actual application, and the evaluation process is simple and convenient.
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of multi-agent Q-learning models, and specifically to a multi-agent Q-learning model evaluation method. Background Art
[0002] At present, search and rescue personnel mostly use AUV (Autonomous Underwater Vehicle) to search and rescue in the accident area. Since the control of the overall fleet of large-scale underwater gliders is relatively complicated, we propose a theoretical model based on Multi-agent and Q-learning to perform multi-layer distributed control of the overall fleet. However, in use, it is unclear which theoretical model is more practical, and it is impossible to make a correct evaluation of the two theoretical models. Therefore, we propose a model evaluation method based on multi-agent Q-learning to solve the above problem. Summary of the invention
[0003] The purpose of the present invention is to provide a model evaluation method based on multi-agent Q learning to solve the problems raised in the above background technology.
[0004] To achieve the above object, the present invention provides the following technical solution: a model evaluation method based on multi-agent Q learning, the specific steps are as follows:
[0005] S1. Model building: Building the initial design model based on the initial data and modeling requirements;
[0006] S2. Model data collection: collect and classify the important data of each part of the initially established model, and some of the data are calculated according to the corresponding formula;
[0007] S3. Model data analysis: theoretically extrapolate the collected data and compare them one by one with the requirements of modeling data;
[0008] S4. Model function demonstration: by dynamically displaying the established model on the computer, the model is demonstrated according to the actual operation requirements;
[0009] S5. Adjust model data: make appropriate modifications to the data of some key parts within the requirements of model establishment and increase the values there;
[0010] S6. Second demonstration: The model with changed data is dynamically demonstrated on the computer again. During the second demonstration, the demonstration speed is slowed down.
[0011] S7. Evaluation results: Compare the results of the first and second demonstrations with the cost and time required to build the model, and make a corresponding evaluation.
[0012] Preferably, the adjustment model data in S5 adjusts one or more of the pressure value, the displacement, and the power value.
[0013] Preferably, the secondary demonstration in S6 and the model function demonstration in S4 are both set as 3D dynamic demonstrations, and the location of the modified data is marked in the secondary demonstration in S6.
[0014] Preferably, the data collection in S2 collects the data on the model separately and divides them into primary and secondary.
[0015] Preferably, the model building in S1 builds a virtual model according to design requirements and basic design procedures.
[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: by setting mold data collection, the present invention can understand the specific parameters of the model; by setting model data analysis, it can understand whether the model data meets the requirements of model establishment and judge whether the corresponding functions can be realized; by setting model function demonstration, it can discover corresponding problems and make timely improvements; by setting and adjusting model data, it can test the model functions under different data to maximize the functions; by setting a secondary demonstration, the modified model can be further tested to find problems in time; the device can accurately evaluate the model to avoid affecting the use in the later stage, and the evaluation is simple and convenient. DETAILED DESCRIPTION
[0017] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0018] A model evaluation method based on multi-agent Q learning, the specific steps are as follows:
[0019] S1. Model building: Building the initial design model based on the initial data and modeling requirements;
[0020] S2. Model data collection: collect and classify the important data of each part of the initially established model, and some of the data are calculated according to the corresponding formula;
[0021] S3. Model data analysis: theoretically extrapolate the collected data and compare them one by one with the requirements of modeling data;
[0022] S4. Model function demonstration: by dynamically displaying the established model on the computer, the model is demonstrated according to the actual operation requirements;
[0023] S5. Adjust model data: make appropriate modifications to the data of some key parts within the requirements of model establishment and increase the values there;
[0024] S6. Second demonstration: The model with changed data is dynamically demonstrated on the computer again. During the second demonstration, the demonstration speed is slowed down.
[0025] S7. Evaluation results: Compare the results of the first and second demonstrations with the cost and time required to build the model, and make a corresponding evaluation.
[0026] The adjustment of model data in S5, adjusting one or more of the pressure value, displacement, and power value, can maximize the model function by changing the corresponding values for ease of use. The secondary demonstration in S6 and the model function demonstration in S4 are both set as 3D dynamic demonstrations. The location of the modified data is marked in the secondary demonstration in S6. Through the secondary demonstration, the use of the model simulation after the data is modified can be intuitively understood for subsequent improvements. The data acquisition in S2 collects the data on the model separately and divides them into primary and secondary. Dividing the data into primary and secondary can quickly find important data.
[0027] The model building in S1 is to build a virtual model according to the design requirements and basic design procedures.
[0028] By setting up a model, the present invention can first detect the corresponding functions of the initially designed model and use it as a reference standard. By setting up mold data collection, the specific parameters of the model can be understood, so that the processing and production process can be learned. By setting up model data analysis, it can be understood whether the model data meets the requirements for model establishment and whether the corresponding functions can be achieved. By setting up a model function demonstration, the corresponding problems can be discovered in time and improved in time. By setting up and adjusting the model data, the model functions under different data can be tested to maximize the functions. By setting up a secondary demonstration, the modified model can be further tested to discover problems in time. The device can accurately evaluate the model to avoid affecting the use later, and the evaluation is simple and convenient.
[0029] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.
Claims
1. A model evaluation method based on multi-agent Q-learning, It is characterized in that The specific steps are as follows: S1. Model building: Building the initial design model based on the initial data and modeling requirements; S2. Model data collection: collect and classify the important data of each part of the initially established model, and some of the data are calculated according to the corresponding formula; S3. Model data analysis: theoretically extrapolate the collected data and compare them one by one with the requirements of modeling data; S4. Model function demonstration: by dynamically displaying the established model on the computer, the model is demonstrated according to the actual operation requirements; S5. Adjust model data: make appropriate modifications to the data of some key parts within the requirements of model establishment and increase the values there; S6. Second demonstration: The model with changed data is dynamically demonstrated on the computer again. During the second demonstration, the demonstration speed is slowed down. S7. Evaluation results: Compare the results of the first and second demonstrations with the cost and time required to build the model, and make a corresponding evaluation.
2. A model evaluation method based on multi-agent Q learning according to claim 1, Features: The adjustment model data in S5 adjusts one or more of the pressure value, the displacement, and the power value.
3. A model evaluation method based on multi-agent Q learning according to claim 1, Features: The secondary demonstration in S6 and the model function demonstration in S4 are both set as 3D dynamic demonstrations, and the location of the modified data is marked in the secondary demonstration in S6.
4. A model evaluation method based on multi-agent Q learning according to claim 1, Features: The data collection in S2 collects the data on the model separately and divides them into primary and secondary.
5. The model evaluation method based on multi-agent Q learning according to claim 1, Features: The model building in S1 builds a virtual model according to design requirements and basic design procedures.
