A Cognitive Electronic Jamming Decision-Making Method Based on Q-Learning and Ant Colony Fusion Algorithm
Round 1
Reviewer 1 Report
This article proposed a fusion algorithm based on ant colony and Q-Learning (Ant-QL) to improve radar jamming decision-making performance. This paper has an essential and innovative topic, but there are obvious problems: fuzzy modeling of radar countermeasures, insufficient comparative experiments, and results. The specific problems and suggestions are as follows:
1. The grammar of this article needs to be carefully checked. Simple English translation makes the article difficult to read and understand.
2. What are the ten jamming states and 16 radar working states in Section 2.3? Radar countermeasure modeling is ambiguous.
3. Table 1 and Table 2 have the same names, please check.
4. The comparison algorithm in the experimental analysis is too simple. Please try to compare the proposed algorithm with similar algorithms in the last three years.
5. The experiment only discussed the convergence rounds of the proposed method and cannot directly represent the timeliness. I suggest analyzing the response time of the proposed method in the jamming decision-making process.
The grammar of this article needs to be carefully checked. Direct English translation makes the article difficult to read and understand.
Author Response
Please see the attachment. (Response to Reviewer 1)
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This paper proposes an improved method of DQN and Q-Learning algorithm, which is applied to radar jamming decision. The pheromone mechanism in the ant colony algorithm is added to the current network. This method not only updates and learns the q table in the iterative process, but also continuously optimizes the search range of the jammer to find the optimal decision scheme, and improves the convergence speed and stability of the algorithm. In addition, the improvement of the algorithm also reduces the hardware and power requirements of the jammer, so that the jammer can play a better performance.However, there are some changes that need to be made in this article, as follows:
1. This article introduces a lot of basic theory. In addition, the innovativeness of the authors is not highlighted, and almost all are applications of existing methods, such as Q-Learning, DQN, and pheromone mechanism in ant colony algorithm.
2. This paper only selects DQN and Q-Learning algorithm, why not choose the network model with better ability in recent years for experiment?
3. In the training process, the radar anti-jamming decision-making ability is not optimized. If the jamming scene is changed, can the trained multi-jammer also show strong jamming ability?
4. The communication scheme between multiple jammers in this paper is relatively simple, and the "one size fits all" method for jammers with poor interference effect may ignore the processing of a lot of key information, so that the trained agent cannot deal with various interference scenarios well.
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Round 2
Reviewer 1 Report
Most of the problems of last time were settled well.
I think you should give the names of the 16 operating states of radar and the 10 operating states of jamming in section 2.3, which is not difficult.
The author mentions references 16,46,50, etc. Please compare the proposed methods with those in recent three years of references, so as to show the superiority.
I have no suggestions.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
