illidanlab / simulator Goto Github PK

Hello, how does the simulation experiment sample from the sample? How is the specific process achieved? How did the total number of drivers and the number of orders increase?

Hello, is the code you released complete? In the process of "mapped"_ matrix_ int.pkl "File import, is this the file you have pre trained before? If I make up a simulated data, I use run_ How should I deal with this fake data format when running ca2c.py without error? We look forward to your reply!

What are your comments on the hypothesis that takes the distribution of order's origin to imitate the distribution of travel demand event, and similarly takes the distribution of order's destination to imitate the distribution of travel supply event?

When building the hexagonal grid world, what is your consideration of the factors that determine the size of each grid? what is the size of the inner diameter of the hexagonal grid used in your experiment?

Any advice on where to mount, train, and test RL-based order dispatch algorithms on your simulator?

I found that when a driver received an order for a destination that wasn't a valid grid, instead of removing the driver, he was set to offline. The whole simulation program seems to have no way to reduce the driver except for reset. So the number of drivers in the valid grids will only increase, not decrease.

Hi Kaixiang,

If I want to add a new type of event "charing" into the simulator, do you have any suggestions about this modification?

thank you so much!

Thanks for sharing your work in git-hub :)
Can you provide sample from requests dataset.

Converting a quadrilateral grid to a 2-D matrix is easy to understand.

I would like to know how Hexagonal grid maps are converted to 2D matrices (It would be better if there was a script to handle it).

Looking forward to your reply !

Hello, can you tell me the data input and output of your simulator? Or can you give me a fake D data form? thank you

in this paper, for any agent i where s_t^i =[s_t, g_j]
here, s_t^i is a small state matrix or something else?

hi, can you help with using this and fleet routing on a map

It seems that the Open Gaia Dataset didn't provide the the status of drivers. From your view, is there a workaround to acquire driver status , such as on-line, off-line, etc., besides seeking help from DiDi AI Lab?

It seems like the generated time of each real-order in real-order data is between 0 - 143, how can I import data of different day in to the simulator to let it train different day's data in different episode?