Inquiry About Specific Implementation Details in Your GitHub Project

jeffchen1108 commented 7 months ago

"Hello All Author's Of The Project,

I hope this message finds you well. I am a recently enrolled master's student who has just started exploring the fascinating field of visual navigation. While researching this area, I came across your impressive visual navigation project on GitHub, and I'm very intrigued by your work. I have a couple of questions regarding the implementation details.

Firstly, I was wondering on which simulator you have implemented your project? Could you please specify the simulator you used for testing and development?

Additionally, as a newcomer to the field, I am interested in implementing a similar visual navigation system in Gazebo. Do you think it's feasible to adapt your project to work in Gazebo, or are there any specific challenges that I should be aware of?

Thank you for your time, and I appreciate any insights you can provide on this matter.

Best regards, Jeff

Dominic-DallOsto commented 7 months ago

Hi Jeff,

Thanks for your interest!

We never fully implemented this project in simulation, only on real robots (Jackal and Miro) where we did our testing.

If you did want to implement this in simulation, Gazebo would likely be the easiest option due to it's ROS integration. You could then just connect the Gazebo to this teach and repeat approach through the corresponding ROS topics.

At a high level, you would need:

A suitably visually rich virtual environment for the robot to navigate in (I'd be concerned the visual processing would fail if the environment was very featureless, like walls of just one colour)
A robot model with a camera and odometry sensors, and a suitable motor controller - then you would connect the ROS topics for the sensors and motors to this teach and repeat approach
A way to initialise the robot at a similar place in the environment for both the teach stage (where you control the robot manually) and the repeat stage (where it is controlled by this approach)

Maybe it would also help to think in terms of what area you would be interested in focusing on developing. For example, if you're interested in improving the visual processing you might need to produce a more realistic looking simulation environment with different lighting effects than if you're more interested in the control aspect.

I hope that helps. If you have any more detailed questions feel free to let me know,

Dominic

On 30/01/2024 4:33 pm, jeffchen1108 wrote:

"Hello All Author's Of The Project,

I hope this message finds you well. I am a recently enrolled master's student who has just started exploring the fascinating field of visual navigation. While researching this area, I came across your impressive visual navigation project on GitHub, and I'm very intrigued by your work. I have a couple of questions regarding the implementation details.

Firstly, I was wondering on which simulator you have implemented your project? Could you please specify the simulator you used for testing and development?

Additionally, as a newcomer to the field, I am interested in implementing a similar visual navigation system in Gazebo. Do you think it's feasible to adapt your project to work in Gazebo, or are there any specific challenges that I should be aware of?

Thank you for your time, and I appreciate any insights you can provide on this matter.

Best regards, Jeff

— Reply to this email directly, view it on GitHub https://github.com/QVPR/teach-repeat/issues/40, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGM5S3AV5OREY22GRM26SULYREHENAVCNFSM6AAAAABCRO2BL2VHI2DSMVQWIX3LMV43ASLTON2WKOZSGEYDQMJVGU3DKMI. You are receiving this because you are subscribed to this thread.Message ID: @.***>

--------------9c520h62iYC8qY9aDnZ58jyq Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: 8bit

<!DOCTYPE html>

Hi Jeff,

Thanks for your interest!

We never fully implemented this project in simulation, only on real robots (Jackal and Miro) where we did our testing.

If you did want to implement this in simulation, Gazebo would likely be the easiest option due to it's ROS integration. You could then just connect the Gazebo to this teach and repeat approach through the corresponding ROS topics.

At a high level, you would need:

A suitably visually rich virtual environment for the robot to navigate in (I'd be concerned the visual processing would fail if the environment was very featureless, like walls of just one colour)
A robot model with a camera and odometry sensors, and a suitable motor controller - then you would connect the ROS topics for the sensors and motors to this teach and repeat approach
A way to initialise the robot at a similar place in the environment for both the teach stage (where you control the robot manually) and the repeat stage (where it is controlled by this approach)

Maybe it would also help to think in terms of what area you would be interested in focusing on developing. For example, if you're interested in improving the visual processing you might need to produce a more realistic looking simulation environment with different lighting effects than if you're more interested in the control aspect.

I hope that helps. If you have any more detailed questions feel free to let me know,

Dominic

On 30/01/2024 4:33 pm, jeffchen1108 wrote:

"Hello All Author's Of The Project,

I hope this message finds you well. I am a recently enrolled master's student who has just started exploring the fascinating field of visual navigation. While researching this area, I came across your impressive visual navigation project on GitHub, and I'm very intrigued by your work. I have a couple of questions regarding the implementation details.

Firstly, I was wondering on which simulator you have implemented your project? Could you please specify the simulator you used for testing and development?

Additionally, as a newcomer to the field, I am interested in implementing a similar visual navigation system in Gazebo. Do you think it's feasible to adapt your project to work in Gazebo, or are there any specific challenges that I should be aware of?

Thank you for your time, and I appreciate any insights you can provide on this matter.

Best regards,
Jeff

—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.

--------------9c520h62iYC8qY9aDnZ58jyq--

jeffchen1108 commented 7 months ago

Dear Dominic Dall'Osto,

First and foremost, I would like to extend my sincerest gratitude for taking the time to respond to my previous query and for the invaluable advice you provided. Your willingness to engage and assist is greatly appreciated.

I am writing to discuss further the possibility of simulating your visual navigation project, specifically within the Gazebo environment. The concept of your innovative approach is intriguing, and I am keen to delve into its practical applications.

Given that you mentioned not having attempted to simulate this project before, I understand that there may be significant groundwork required on my part. In this context, I am seeking clarification on whether I need to develop both the environmental setup and the autonomous vehicle from scratch for the simulation in Gazebo. Any guidance or suggestions you could offer would be most helpful.

Additionally, I aim to streamline my approach by focusing on the essential aspects of the simulation. It would be beneficial to know which files or components of the project are critical for the simulation and which may be less pertinent.

I am currently dedicating my time to thoroughly comprehend the concepts and methodologies that are central to your work. Commencing the simulation work is my next step, and while there is some concern about the feasibility of replicating the project's success in a simulated environment, I am committed to this endeavor.

I assure you of my utmost respect for your work and my eagerness to contribute to the field. Any further advice or insights you could provide would be immensely valuable and greatly influence my ability to carry out this simulation effectively.

Thank you once again for your time and consideration.

Warm regards,

Jeff

Dominic-DallOsto commented 7 months ago

Hi Jeff,

No worries.

Yeah, the first step would be to setup a simulation of a visually interesting environment with a mobile robot with camera able to move around this environment. You should be able to follow some Gazebo tutorials that will introduce you to how the program works that are also based on a wheeled mobile robot. There's also a library of prebuilt Gazebo environments that might suit your needs.

Once you have this, you'll need to modify the launch files for this approach (eg. for the teach run here - https://github.com/QVPR/teach-repeat/blob/master/launch/data_collection_miro.launch) to use the right ROS topic names from your simulation.

Thanks, Dom

On 3/02/2024 8:23 am, jeffchen1108 wrote:

Dear Dominic Dall'Osto,

First and foremost, I would like to extend my sincerest gratitude for taking the time to respond to my previous query and for the invaluable advice you provided. Your willingness to engage and assist is greatly appreciated.

I am writing to discuss further the possibility of simulating your visual navigation project, specifically within the Gazebo environment. The concept of your innovative approach is intriguing, and I am keen to delve into its practical applications.

Given that you mentioned not having attempted to simulate this project before, I understand that there may be significant groundwork required on my part. In this context, I am seeking clarification on whether I need to develop both the environmental setup and the autonomous vehicle from scratch for the simulation in Gazebo. Any guidance or suggestions you could offer would be most helpful.

Additionally, I aim to streamline my approach by focusing on the essential aspects of the simulation. It would be beneficial to know which files or components of the project are critical for the simulation and which may be less pertinent.

I am currently dedicating my time to thoroughly comprehend the concepts and methodologies that are central to your work. Commencing the simulation work is my next step, and while there is some concern about the feasibility of replicating the project's success in a simulated environment, I am committed to this endeavor.

I assure you of my utmost respect for your work and my eagerness to contribute to the field. Any further advice or insights you could provide would be immensely valuable and greatly influence my ability to carry out this simulation effectively.

Thank you once again for your time and consideration.

Warm regards,

Jeff

— Reply to this email directly, view it on GitHub https://github.com/QVPR/teach-repeat/issues/40#issuecomment-1925193516, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGM5S3E73Q2DAEJARUQLU3TYRXQWBAVCNFSM6AAAAABCRO2BL2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSMRVGE4TGNJRGY. You are receiving this because you commented.Message ID: @.***>

--------------0MolAdL8xTTMAl0hZ0Y3CXm7 Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: 8bit

<!DOCTYPE html>

Hi Jeff,

No worries.

Yeah, the first step would be to setup a simulation of a visually interesting environment with a mobile robot with camera able to move around this environment. You should be able to follow some Gazebo tutorials that will introduce you to how the program works that are also based on a wheeled mobile robot. There's also a library of prebuilt Gazebo environments that might suit your needs.

Once you have this, you'll need to modify the launch files for this approach (eg. for the teach run here - https://github.com/QVPR/teach-repeat/blob/master/launch/data_collection_miro.launch) to use the right ROS topic names from your simulation.

Thanks,
Dom

On 3/02/2024 8:23 am, jeffchen1108 wrote:

Dear Dominic Dall'Osto,

First and foremost, I would like to extend my sincerest gratitude for taking the time to respond to my previous query and for the invaluable advice you provided. Your willingness to engage and assist is greatly appreciated.

I am writing to discuss further the possibility of simulating your visual navigation project, specifically within the Gazebo environment. The concept of your innovative approach is intriguing, and I am keen to delve into its practical applications.

Given that you mentioned not having attempted to simulate this project before, I understand that there may be significant groundwork required on my part. In this context, I am seeking clarification on whether I need to develop both the environmental setup and the autonomous vehicle from scratch for the simulation in Gazebo. Any guidance or suggestions you could offer would be most helpful.

Additionally, I aim to streamline my approach by focusing on the essential aspects of the simulation. It would be beneficial to know which files or components of the project are critical for the simulation and which may be less pertinent.

I am currently dedicating my time to thoroughly comprehend the concepts and methodologies that are central to your work. Commencing the simulation work is my next step, and while there is some concern about the feasibility of replicating the project's success in a simulated environment, I am committed to this endeavor.

I assure you of my utmost respect for your work and my eagerness to contribute to the field. Any further advice or insights you could provide would be immensely valuable and greatly influence my ability to carry out this simulation effectively.

Thank you once again for your time and consideration.

Warm regards,

Jeff

—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you commented.

--------------0MolAdL8xTTMAl0hZ0Y3CXm7--

QVPR / teach-repeat

Inquiry About Specific Implementation Details in Your GitHub Project #40