CoDiffMob

Code for the paper "Noise Matters: Diffusion Model-based Urban Mobility Generation with Collaborative Noise Priors"

Getting Started

step 1: clone the repository

step 2: create a virtual environment with conda or virtualenv

# with conda
conda create -n noise python=3.10
conda activate noise
# with virtualenv
virtualenv -p python3.10 noise
source noise/bin/activate

step 3: install the dependencies

pip install -r requirements.txt

Training the diffusion model

Entrypoint to the diffusion model training is train.py. The script takes a configuration file as input, which specifies the dataset path, hyperparameters, and other settings. The configuration file is in the YAML format, an example is provided in the configs directory.

• Modify the configuration file configs/{city}.yml to specify the dataset path and other hyperparameters.

• Run the following command to train the diffusion model

python3 train.py --config configs/{city}.yml --save /path/to/log_dir

The trained model will be saved in the log directory and you can check the training process in tensorboard by running tensorboard --logdir /path/to/log_dir. We trained our model for 200 epochs on both ISP and MME datasets.

Generating synthetic trajectories

We introduce a two-stage method to generate synthetic trajectories. The first stage is to sample location transitions based on EPR model and flow data. The second stage is to build noise prior and generate synthetic trajectories based on it by the trained diffusion model.

step 1: sample location transitions

• Obtain the transition matrix, population, and move probability as priot knowledge of the generation task.
• Run the script scripts/epr_sampling.py to sample location transitions.

python3 scripts/epr_sampling.py --num-traj {num} --save /path/to/save_dir

step 2: generate synthetic trajectories

• Construct the noise prior from the sampled location transitions
• Run the genration process by the trained diffusion model based on the noise prior

The generation process is in the generate.ipynb notebook. You can run the notebook to see the synthetic trajectories.