A Semi-Automatic Wheelchair with Navigation Based on Virtual-Real 2D Grid Maps and EEG Signals

Round 1

Reviewer 1 Report

In this paper, the authors proposed a Reinforcement Learning (RL) method to obtain optimal path planning based on a virtual two-dimensional grid graph constructed in a real environment and evaluate its performance through simulations and real environment experiments.

There are some questions/comments that need to be answered before the paper can be considered for publication.

1. the abbreviations used for the first time in the paper should clearly state the full name, e.g., "SLAM" in line 52, and a brief description of SLAM is recommended.

2. It is suggested that the authors introduce the content of the subsequent chapters at the end of the introduction, so that the reader can quickly understand each chapter of the paper.

3. It is suggested to add some content and figures to show the acquisition and processing of EEG signals.

4. Please add the hardware information of the training platform.

5. It is suggested to add control flow in figures 2 and 3 to make the control process more intuitive.

6. It is recommended to add several figures to show the real experimental environment.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

 The manuscript proposes a semi-controlled method for electric wheelchairs, which combines an RGB-D camera system, a graphical user interface, and a map of the real environment with natural landmarks to determine the optimal path planning for wheelchair navigation. This manuscript proposes an EEG-based control system for the wheelchair based on a grid map designed to allow disabled people to reach any preset destination. Using this method could improve the difficulty in the wheelchair control of people with severe disability compared to solutions using one input. The manuscript is worth publishing, but it needs to be revised before publication.

1. The manuscript uses DQNs 55 times starting from line 3, and this manuscript designs the method on the basis of DQNs, however, no citations for DQNs are given. This has caused a lot of trouble for reviewers, and it is difficult to judge the innovative points of the manuscript. Please give the correct citation of DQNs.

2. The full spelling of the English abbreviation should be given, which is also the most basic requirement of the journal, such as: EEG. The manuscript uses EEG 25 times, and I can only understand EEG as Electroencephalogram in combination with the original content.

3. More idiomatic English should be used. For example, in line 98, for wheelchair control, we think it is more appropriate to use "front, back, left and right" than "up, down, left and right".

4. The second stage of method realization is to control the wheelchair to achieve the desired goal in a real indoor environment. The manuscript should explain how the wheelchair acquires natural landmarks, such as how the used R-GBD camera system collects information, and the information collected What has been done?

5. An important point of the method proposed in the manuscript is the navigation of the wheelchair in the real world, which is much more complex than the virtual scene. However, Section 2.3 fails to explain how the real world corresponds to the 2D grid. This is a very important part of this manuscript, please give a more detailed supplement.

6. Figure 8 discusses Relu and PRelu, and then proves the superiority of PRelu. This discussion is of little significance because PRelu was originally developed on the basis of Relu and has its own advantages.

7. The manuscript introduces reinforcement learning in one paragraph in the introduction, and uses reinforcement learning in the proposed method. Can you prove the superiority of reinforcement learning over ordinary neural networks in the discuss section.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The revision is fine to the reviewer.

Author Response

Please see the attachment

Author Response File: Author Response.pdf