Tsunami Evacuation using Reinforcement Learning

This document explains the structure of the repository. For an explanation on the use of the code see the Manual

root directory

main.py => file to set a case of SARSA simulation
sarsa.py => SARSA class and functions (needs to be cleaned)

evac_plots.py => to plot evacuation curves from each epoch
make_video.py => to create a video (AVI) file of a particular epoch (statematrix or policy)

mc.py => Monte Carlo method (needs to be revised)

data directory

agentsdb.csv => input data of population (currently only 'Node' field is used)

• 'Node': Starting node (int)

linksdb.csv => edges or links of a road network

• 'codeLink': the ID of the link/edge (int)
• 'node1': starting node of the edge (int)
• 'node2': ending node of the edge (int)
• 'length': length of the edge in meters (int)
• 'width': width of the road in meters (int)

nodesdb.csv => nodes information (intersections of roads)

• number: node ID
• coord_x,coord_y: node coordinates
• evacuation: flag (0=common node; 1=evacuation point)
• reward: abs of the penalty 'reward' given at each node (-1 to account for time pressure in evacuation)

actionsdb.csv;transitionsdb.csv => created with 'pre/SetActionsandTransitions.py'

figures directory

A folder to store snapshots for video (not present in this repository)

other directory

An informal Python Notebook for various calculations (weights, etc)

pre directory

• CensusAndBuildingDatabase is a folder with original census data
• Household_database is a folder with original census data (integrated but not in used at the moment)
• Population_database is a folder with original census data
• defPathsFromNodes.py => a function to calculate the next node for a pre-determined shortest path run.
• DetectionShelters.py => to detect evacuation points from the network
• lib_importOSM => to transform OSM data into suitable format