Skip to content

Latest commit

 

History

History
245 lines (168 loc) · 7.29 KB

README.md

File metadata and controls

245 lines (168 loc) · 7.29 KB
Raw

Dispatch

A discrete-event simulation (DES) model for simulating deployment of EV bus fleets, using GTFS data

Installation/Compilation

Prerequisites:

  • Python 3.x, see requirements.txt for python package requirements
  • Julia 1.x, see Project.toml for Julia package requirements
  • Cmake
  • An API key from OpenMobilityData

Acquire necessary tools with:

sudo apt install cmake zlib1g-dev pybind11-dev

Clone the repo with all neccessary submodules

git clone --recurse-submodules git@github.com:nickolasclarke/dispatch.git

Set up a conda environment and install necessary packages:

conda create -n rise
conda activate rise
conda install pip
pip install -r requirements.txt

Set up Julia env and install required julia packages. In Julia's Pkg manager

activate .
instantiate

To build with protections (some loss of speed), use the following

mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=RelWithDebInfoAndAssert -DCMAKE_INSTALL_PREFIX=`pwd` ..
make
make install

To build with fewer protections (more speed), use the following

mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=`pwd` ..
make
make install

Example Usage

Data files and some intermediate products are stored in the data/ subdirectory.

#Build everything using the instructions above!

#Acquire data
./build/bin/pull_gtfs.py acquire_remote data/feeds.db

#If data has already been downloaded and you need to rebuild the database use:
#./build/bin/pull_gtfs.py acquire_local data/feeds.db

#Validate data and create inputs for model
./build/bin/pull_gtfs.py validate data/feeds.db

#Check validation status
./build/bin/pull_gtfs.py show_validation data/feeds.db

#Get OSM extents (caches for faster runs a second time)
./build/bin/pull_gtfs.py extents data/feeds.db > data/extents

#Get a Planet OSM highways file per the instructions below

#Split up global highway data padding each box with a 2 mile margin
#On our test machine this took 1:35:48 seconds (1 hour 34 minutes) and 36.3GB RAM
./build/bin/osm_splitter data/planet-highways.osm.pbf data/extents 0.0288

#Run model on a dataset
julia --project ./sim.jl ../../temp/minneapolis ../../data/minneapolis-saint-paul_minnesota.osm.pbf ../../data/depots_minneapolis.csv /z/out

#Create inputs for a particular dataset
./parse_gtfs.py data/gtfs_minneapolis.zip msp3.pickle

#Identify possible places to put chargers
#./find_chargers.py msp3.pickle

Data Flow, Data Structures

  1. parse_gtfs.py

parse_gtfs.py reads a GTFS file and converts it into an internal data structure that can be read and understood by the project.

Data Structures

The data is stored in several tables:

trips:

A trip follows a route from the beginning of a route to its end in one direction. Many different trips take place on each route each day. Trips may follows subtly different paths which are still labeled as the same route. We parse the GTFS data into the following format:

  • trip_id: A unique identifier for the trip. Taken from GTFS
  • trip_headsign: The bus's public LED readout for the trip. Taken from GTFS. Used for debugging.
  • block_id: A unique identifier indicating a string of sequential trips which are nominally served by the same bus.
  • start_arrival_time: Time at which the bus is supposed to be present at the starting stop of the trip.
  • start_departure_time: Time at which the bus is supposed to depart the starting stop of the trip. May be the same as start_arrival_time.
  • start_stop_id: Unique identifier for the stop at which the trip begins.
  • start_lat: Latitude of the stop idenicated by start_stop_id.
  • start_lng: Longitude of the stop idenicated by start_stop_id.
  • end_arrival_time: Time at which the bus is supposed to end its trip.
  • end_departure_time: Time at which the bus is supposed to depart its last stop. May be the same as end_arrival_time. Kind of unclear what this is about. TODO
  • end_stop_id: Unique identifier for the stop at which the trip ends.
  • end_lat: Latitude of the stop idenicated by end_stop_id.
  • end_lng: Longitude of the stop idenicated by end_stop_id.
  • distance: Travel distance of the trip.
  • duration: Travel time of the trip.
  • wait_time: Total time spent sitting at stops during the trip.

stops

stop_times

road_segs

seg_props

Data Acquisition

  • GTFS transit feed data acquired from OpenMobilityData
  • OSM data for all validated feeds acquired from ?

Global Road Network

Acquire a planet.osm file

See directions here and downloads here.

We got our data from this link. The file was 50GB.

Extract roads from file

First, install OSM tools:

sudo apt install osmosis

Now, extract the roads:

osmosis --read-pbf planet-latest.osm.pbf --tf accept-ways highway=* --used-node --write-pbf planet-highways.osm.pbf

Running Osmosis on the global dataset will take a while. On our 32 core, 192 GB RAM, SSD machine it took 8 hours 10 minutes. The resulting file planet-highways.osm.pbf was 17GB.

Next, build the contraction hierarchy:

./routing_preprocess planet-highways.osm.pbf planet-highways.ch

Unfortunately, our timer didn't work for this process, but it will take 12-45 hours and require 80+GB of RAM. The resulting file planet-highways.ch was 15GB.

Test Julia Routing

include("RoutingKit.jl")

router = RoutingKit.Router("/z/msp-highways.osm.pbf", "/z/msp.ch")

brooklyn_park = (lat=45.115208, lng=-93.373463)
south_st_paul = (lat=44.892850, lng=-93.051079)

time_dist = RoutingKit.getTravelTime(router, brooklyn_park.lat, brooklyn_park.lng, south_st_paul.lat, south_st_paul.lng, 3000)

XSEDE Comet

Compilation

module load gnu
mkdir build
cd build
CXX=g++ CC=gcc cmake -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=`pwd` ..
make install -j 20

Validation

srun --partition=compute --pty --nodes=1 --ntasks=24 -t 10:00:00 --wait=0 --export=ALL /bin/bash
module load gnu
conda activate rise
./build/bin/pull_gtfs.py validate data/feeds.db

OSM Splitter

screen
srun --partition=large-shared --pty --nodes=1 --ntasks=1 --mem=60GB -t 10:00:00 --wait=0 --export=ALL /bin/bash
module load gnu
conda activate rise
./build/bin/pull_gtfs.py extents data/feeds.db > data/extents
#The following takes 1:35:48 and 36.3GB RAM
./build/bin/osm_splitter data/planet-highways.osm.pbf data/extents 0.0288