dataandcrowd / seoultrafficabm

TRAPSim - Traffic Simulation in NetLogo

Photographed by Matthias Ripp

What is this model?

We built an agent-based model a traffic simulation for Central Seoul to understand the coupled problems of emissions, behaviour, and the estimated exposure to PM₁₀ for groups of drivers and subway commuters.

Pulished Papers

• Original Journal Article: https://www.sciencedirect.com/science/article/pii/S0198971522001387
• Summary version in the ESSA Conference: https://link.springer.com/chapter/10.1007/978-3-030-92843-8_26
• Methods: https://www.sciencedirect.com/science/article/pii/S2215016122000577

Brief Background

Seoul has a population of more than 10 million with an addition of 8 million commuters flood in from adjacent cities everyday. Traffic congestion has been so common in Seoul, which resulted in hazardous levels of air pollution. During the cold and dry seasons (<0°C), more and more people take public transportation or private vehicles, which will intensify air pollution. This can cause further problems to commuters' health.

Studies have proven the impact of exhaust emissions (particularly diesel engines) to human health. With tougher regulations, vehicle manufactuers now produce vehicles that would less harm the environment. However, less attention as been paid to non-exhaust emission (NEE). NEE components including tyre, brake wear, and road resuspension are also threateners to the atmosphere because the particles whether large or small can cause breathing problems and eventually lung disorder.

NEE is mainly generated by the stop-and-go behaviours of drivers, and is likely to occur at congested areas. The UK's Air Quality Expert Group has also raised the severeness of NEE to human health in their 2019 Report.

We assume that traffic is the gratest harm to human respiratory health, and even with the advent of EVs, NEE will still maintain as a problem.

What is ABM and why are you employing the method in your research

Agent-based modelling (ABM) is a microscopic computtational method that simulates actions and interactions between agents (Wilensky and Rand, 2015; Crooks et al., 2019).

This method is very useful because unique individuals can be created and these individuals are given micro rules that is close to reality. We can also test different scenarios, either prospective or retrospective, to understand the causes that might have happened if we did 'this' or possible projections that could happen if we did 'that'. We call this a 'what-if' scenario.

What are you testing?

For our study, we test whether reducing traffic can alleviate the pollution levels and whether taking a polluted but quicker path or less polluted but longer path makes a difference to pedestrians' exposure levels.

Where to find details?

You can go to the Wiki tab (https://github.com/mrsensible/SeoultrafficABM/wiki) for more information. The details include:

• Change logs
• Model interface/purpose
• Non-exhaust emission (NEE)
• A* algorithm: a pathfinding algorithm for vehicles
• Sensitivity analysis and Calibration
• Spatial Output
• Scenario Forecasting

Contacts

If you want to discover more, feel free to contact me via

• personal email: [email protected]
• work email: [email protected]
• twitter DM: @hyesop

seoultrafficabm's Issues

Emission

Code to integrate array HPC jobs

library(data.table)
library(feather)
library(tidyverse)
library(janitor)

tbl_feather <- 
  list.files(pattern = "*.feather") %>% 
  purrr::map_df(~read_feather(.)) %>% 
  clean_names() %>% 
  select(-c(random_seed, siminputrow))

tbl_feather$run_number <-  rep(1:20, each = 510956)

tbl_feather %>% 
  select(health_loss, step, 
         e_heatlh = health_of_one_of_employees,
         d_heatlh = health_of_one_of_cars_with_not_random_car,
         drivers_p, walkers_p) %>% 
  group_by(health_loss,step) %>% 
  summarise_all(list(mean = mean, sd = sd)) %>% 
  rename_at(vars(ends_with("_mean")), 
            funs(str_replace(., "_mean", ""))) -> tbl_mean

write_feather(tbl_mean, "CBD_HealthLoss.feather")

Vehicle attributes and Grades

Air Quality

Road Name

Traffic flow

Sensitivity: In Cone

emission_factor	poll_cone	Jongno	Sejong	Yulgok	Samil	Pirun
1	45	44.1	44.5	43.9	43.8	43.9
	60	43.8	44.4	43.8	43.7	43.7
	90	43.5	44.4	43.8	43.6	43.7
5	45	58.4	55.7	58.7	58.9	60.1
	60	59.3	56.3	59	59.5	60.4
	90	60.4	56.6	59.1	59.5	60.8
10	45	73.2	71.2	77.3	77	80.5
	60	76.6	72.3	77.9	77.4	81
	90	79.6	73.1	78.5	78.1	81.8
20	45	102	100	112	113	118
	60	109	102	114	115	120
	90	116	104	115	118	120

Sensitivity: Dilution

emission_factor	countdown	Jongno	Sejong	Yulgok	Samil	Pirun
1	5	46.1	46	46.5	46.2	46.4
1	10	46.7	46.5	46.8	46.7	47
5	5	65.5	65.6	66.3	66.2	67.1
5	10	67.4	67.3	67.7	67.7	68.5
10	5	93.9	95.4	96.3	95.7	98.7
10	10	99	99.2	100	99.6	102
20	5	153	150	155	155	159
20	10	164	160	165	164	167

Traffic speed