Table 1

Model of the red side

Table 2

Model of the blue side

Fig. 1

Sketch map of the game terrain

Table 3

Terrain features and constraints

Fig. 2

Overall framework of intelligent decision-making based on deep reinforcement learning

Fig. 3

Deployment of red entities

Fig. 4

Feature analysis network structure based on deep neural networks

Fig. 5

Actor network framework of the decision agent

Fig. 6

Critic network framework of the decision agent

Table 4

Deployment of blue entities

Fig. 7

Deployment of both red and blue side

Fig. 8

Reward curve in the training of the single scene

Fig. 9

Entity loss curve in the training of the single scene

Fig. 10

Entity loss curve in the evaluation of the single scene

Table 5

The blue deployment of multi attack formations and multi wave intervals

Fig. 11

Complex scene reward curve based on multi attack formations and multi wave intervals

Fig. 12

Entity loss curve in the training of the complex scenes based on multi attack formations and multi wave intervals

Fig. 13

Statistical distribution chart of red entity loss in the complex scenes before and after training based on multi attack formations and multi wave intervals

Fig. 14

Statistical distribution chart of blue entity loss in the complex scenes before and after training based on multi attack formations and multi wave intervals

Fig. 15

Display of 10 scenes

Table 6

Blue prior attack order

Fig. 16

Average loss probability of red and blue entities and the survival probability of the red command vehicle in 10 typical scenes based on multi attack strategy

Table 7

Blue deployment of multi attack routes

Fig. 17

Average loss probability of red and blue entities and the survival probability of the red command vehicle in 10 typical scenes based on multi attack routes

Fig. 18

Statistical distribution chart of the number of red and blue loss entities of the unique trained model in 10 typical scenarios

Fig. 19

Statistical distribution chart of the number of red entity losses in the trained model and random model under unknown attack strategy (attack nearby)

Fig. 20

Statistical distribution chart of the number of blue entity losses in the trained model and random model under unknown attack strategy (attack nearby)

Fig. 21

Probability of red and blue entity loss and the survival probability of red command vehicle in the trained model and random model under unknown attack strategy (attack nearby)

Fig. 22

Comparison of the number of red and blue entity losses in the model under unknown wave interval