cadCAD crowd agent-based modelling approaches. Originally forked from prey-predator demo cadCAD model https://github.com/cadCAD-org/demos/tree/master/demos/Agent_Based_Modeling/prey_predator_abm
- lab_odyssey.ipynb - The notebook for experimenting and visualizing
- main.py - main script derivate of notebook
- helpers.py - helper scripts for main.py
- run.py - Script for running all configurated experiments
- crowd_control_abm/ - Folder for the ABM simulation and model
- {simulation}/sim_params.py - Simulation parameters
- {simulation}/model/partial_state_update_block.py - The structure of the logic behind the model
- {simulation}/model/state_variables.py - Model initial state
- {simulation}/model/sys_params.py - Model parameters
- {simulation}/model/parts/ - Model logic
There are 2 types of agents: person and attraction. The goal is to balance the capacity of the attractions to the amount of visiting persons. Every person agent has a bucket list of attractions they want to enter. When persons are getting queued up, they will go to another attraction from their bucket list (not implemented yet). Persons staying for 3 timesteps and then leave the attraction, removing it from their bucket list.
In state_variables.py. Making use of an N x M grid, where attractions and persons are randomly plotted in. At the moment person agents will move to the nearest attraction of their bucket list. Each attraction agent has a MAX_ATTRACTION_CAPACITY set in state_variables.py. When full, person agents are queued in line.
Run main.py or using the Jupyter notebook. Simulation results given in about 20 sec, but visualization with animated plotly lib takes about 15-20 min for 300 timesteps with ATTRACTION_COUNT = 5 and PERSON_COUNT = 200.
- Stochastic movement of person agents via probabilities
- Monte Carlo simulations and parameter sweeps
- Person class agents modeling different behaviours (family, youngsters, elderly etc. )
- Tokenization of incentives via Ocean Protocol-like attraction tokens, art NFTs or likewise