Closed wahaha-1 closed 4 months ago
This is the result after I use it now
For comprehensive guidelines on data generation, training, and evaluation of the LPAC neural network, please refer to our detailed documentation: LPAC Neural Network Documentation.
To modify settings, please adjust the parameters in the TOML files as necessary.
For details on obtaining and using the model discussed in our publication, refer to the Evaluation
section of the same documentation.
To facilitate video generation:
RecordPlotData()
function to capture plot data at every time step, e.g., inside the while
loop (line 52) in coverage_class.pyRenderRecordedMap("dir/", "video.mp4")
function to render videos at the end of the program.To activate video generation during the LPAC NN evaluation, set GenerateVideo = true
in eval_single.toml
. See an example TOML file here: Example eval_single.toml.
Run the evaluation script eval_single_env.py
, as described here: Evaluation Scripts.
Our reference manual provides comprehensive documentation for all available classes and functions: Reference Manual.
Please note that most functions available in the C++ Core API can also be accessed using Python.
The code for extracting data from OpenStreetMap has not yet been uploaded. Please feel free to ask any further questions or request additional information as needed.
Thank you so much for your quick reply,!!!! I'll try it soon!!!! 🫣🫣🫣🫣 I will be more detailed ,thank you very much for your patience~
Under your guidance, the third and fourth problems have been successfully solved~~~expressing my gratitude to you again!!
But now there are other questions: 1. How to customize the points of interest (areas with importance 1) and specify the initial position of the robot? I have not found a file location that can be adjusted for the time being; (I am particularly interested in it. If the data from the openmap website cannot be used, it is also very useful for me to design the map that needs to be covered and specify the initial position of the robot)
2. Can I choose not to delete the generated motion trajectory pictures after generating the video?
3. I am curious about CoverageControl/python/scripts/training/train_cnn.py. Is it similar to train_lpac.py?
4. How to use the trained model file? Now I only know that the trained network will be used in the evaluation phase, but the models folder is not used. I don't quite understand this. In my understanding, training - generating model - calling model, I don't know how to reflect this last step.
5. There is a small error in the documentation
python python/scripts/evaluators/eval_single_dataset.py \ ${CoverageControl_ws}/lpac/params/eval_single_env.toml
Should be changed to:
python python/scripts/evaluators/eval_single_env.py \ ${CoverageControl_ws}/lpac/params/eval_single.toml
forgot one thing. After I used RecordPlotData()
, I did not find the generated csv file. RenderRecordedMap()
ran normally. I found four instructions about RecordPlotData
in the document, but the calls failed.
forgot one thing. After I used
RecordPlotData()
, I did not find the generated csv file.RenderRecordedMap()
ran normally. I found four instructions aboutRecordPlotData
in the document, but the calls failed.
I understand this. The RenderRecordedMap()
is based on the data of RecordPlotData(),
but it does not generate a .csv
file like this Evaluation.
Sorry to bother you again, I don't quite understand how to use your repo. Looking at your tutorial, I roughly understand that you use the python file in scripts and the configuration file in params to perform actions. but 1. How to use lpac to guide ontology program?
2. How to simulate the actual map using the data from https://www.openstreetmap.org/ as in the paper?
3. How to call the downloaded gnuplot, ffmpeg plug-in to visualize and generate videos?
Another question is that I only found the folder that saves the data under the lpac file. 4. Is there a folder elsewhere in the project that saves the data during the running of the program?
I hope there is a basic and complete guidance on the usage process. (I have read your paper and am very interested in your ideas, especially those screenshots of experimental data. I thought I could understand how your program operates by reading the paper carefully. Sorry, I still overestimate my ability. , I have no choice but to bother you again)