urban-activity-model

This project aims at modeling people's activities on the urban network.

Setup

Python dependencies (once only):

pip install python-igraph

Test igraph installation in your interpreter by typing:

import igraph.test
igraph.test.run_tests()

Activity model

The activity model is described by two files, for example:

networks/SmallGrid_activities.txt
networks/SmallGrid_net_times.txt

SmallGrid_net_times.txt describes a directed network of geometry

(0)--(1)--(2)
 |    |    |
(3)--(4)--(5)

multilpied by 16 time steps of one hour between 5am and 9pm the weights are the travel times for each of the 16 time slices which are (5am,6am], (6am,7am] and so on ...

SmallGrid_activities.txt describes the available activities the activity located at Node gives Reward if done between Start time and End time activities of type -1 can be done multiple times activities of type 0 or more can only be done once

Activity engine

Run activity engine on a small grid network:

python activity_engine/activity_engine.py

Description of Activity engine

The activity engine first converts the SmallGrid into a Supernetwork using

from graph_utils.txt_to_supernetwork import txt_to_supernetwork

then it creates a modifier that sets the cost of an activity already done to +inf and solves using an extension of Dijkstra's algorithm

from pathfinding.pathfinding_extended import dijkstra_extended

Example of a Supernetwork

networks/SmallGrid_net_times.txt describes a network of geometry

(0)--(1)--(2)
 |    |    |
(3)--(4)--(5)

multilpied by 16 time steps of one hour between 5am and 9pm slice 0 has weights equal to the travel times during (5am, 6am]:

(0)--(1)--(2)
 |    |    |
(3)--(4)--(5)

slice 1 has weights equal to the travel times during (6am, 7am]:

(6)--(7)---(8)
 |    |     |
(9)--(10)--(11)

and so on ..., the slices are linked by activity edges in the time dimension

networks/SmallGrid_activities.txt describes the activities staying at home is following path 0-6-12-... staying at mall located at Node 2 between 8am amd 10am means taking edge from vertex: Node + (start-1) * num_nodes = 2 + 2 * 6 = 14 to vertex: Node + end * num_nodes = 2 + 5 * 6 = 32 where start-1=2 means activity starts sometime in time slice 2 (7am,8am] end=5 means activity ends at 10am and driver is on the road at slice 5 (10am,11am]. An edge is of type -1 means that user can take edge any number of times while an edge is of type 0, 1, ... means user can only take it once. The modifier tracks activities of types 0, 1, ... which adds up to the complexity. Ff an activity can only be done once but has start times < end times, its type can be set up to -1 because the geometry of the resulting supernetwork implicitly restricts the edge to be taken at most once.