Implementing a Traffic Assignment Heuristic in GIS: Exploring the Evacuation Problem

1
Implementing a Traffic Assignment Heuristic in
GISExploring the Evacuation Problem

• John W. Fell
• Masters Thesis Defense
• 7/31/06

2
Overview

• Research Questions
• Introduce Topic Area
• What is the Evacuation Planning Problem
• Applications
• Who is concerned with it?
• Literature Review
• When research became popular
• Macro / Micro simulation
• What is the CCRP?
• Data Sources
• Street Centerlines
• Intersections
• Census

• Analysis Methodology
• Capacity
• Algorithm
• Assumptions
• Research Design
• Results Discussion
• What was implemented
• What was not implemented
• Demo
• Conclusions
• Future Research
• References

3
Presidential Initiative

• Our cities must have clear and up-to-date plans
  for responding to natural disasters, disease
  outbreaks, or terrorist attack... for evacuating
  large numbers of people in an emergencyand for
  providing the food, water, and security they
  would need.
• - President Bush, Address 15 Sep. 2005

Evacuation of New Orleans after Hurricane
Katrina, 9/05 http//spectrum.ieee.org/images/
dec05/images/nhurf1.jpg
4
Research Questions

• Can an Evacuation Planning System be implemented
  in a GIS environment?
• More specifically, can the Capacity Constrained
  Route Planner Heuristic be implemented in a GIS
  environment?

5
Introduction

• Why is this research significant?
• First Evacuation Planning System introduced in
  GIS software
• Problem is combinatorially complex
• Timely
• Who is concerned with it?
• Emergency responders
• Local officials
• Those required to perform evacuation planning

6
Introduction

• What is the Evacuation Planning Problem?
• The removal of evacuees from an (EPZ) emergency
  planning zone (Sorensen et al. 1987)
• The distribution of those evacuees to safe
  destination
• Avoid
• Traffic congestion
• Loss of life
• Property damage

• Applications
• Regulations
• Nuclear Regulatory Commission
• Hurricane
• Terrorist Attack
• Chemical Spill
• Scenario

7
Literature Review

• Emergency Evacuation Methodology (Tweedie et. al.
  1986) for Black Fox Station
• Notification When officials order evacuation to
  the realization of need to evacuate
• Mobilization When evacuee knows of need to
  evacuate to when they begin exodus from EPZ
• Egress Duration of time taken to leave an
  evacuation zone
• Microsimulation
• Local Primary road capacities and travel times
• Monte Carlo algorithm - Uncertainty
• Total vehicles() at time x

8
Literature Review

• Dissipation Rate Model(Sheffi et al 1986 Houston
  1975)
• Used empirical data from true evacuation to
  correlate EPZ size and population density with
  egress time
• Two inputs
• Inverse proportion of population
• Calibrated by empirical data
• Straightforward yet too general in comparison to
  other models
• Network connectivity, spatial distribution of
  population, capacity at intersections not
  considered

9
Literature Review

• Manual Capacity Analysis(Wilbur-Smith 1975Stone
  and Webster 1980)
• Capacities ? Highway Capacity Manual
• Number of arbitrary routes within EPZ selected
  with population loaded
• Egress time ? total vehicle count / capacity of
  each route
• Bias major issue affecting results
• Route interaction, network connectivity,
  congestion ignored

10
Literature Review

• NETVACL (Sheffi et al. 1982) motivated by the
  NETSIM Traffic Simulation Model
• Macro simulation
• Models evacuees using mathematical relationships
  among flows, densities, speeds, other relevant
  evacuation simulation factors
• Outputs given for each edge at each time interval
• Gives substantial information about traffic
  conditions during simulation
• Selects routes dynamically
• Allows different parameters for delays at
  intersection
• Dynamic capacity updating

11
Literature Review

• State of the art review on mathematical models
  for Evacuation systems (Hamacher and Tjandra
  2001)
• Produce Optimal results
• Micro-Simulation Methods
• Random Choice - not familiar with surroundings
• Modified Random Choice - return to previous path
  taken
• Cellular Automata - discrete cells permit /
  restrict access

12
Literature Review

• Macro-Simulation Methods
• Require Time-Expanded Network
• Maximum Flow maximize number of evacuees
  towards destination
• Minimum Cost Flow - sends evacuees along the path
  multiple times according the total number at the
  origin (Hoppe Tardos 1994)
• Traffic Assignment Problems

Lu et. al 2005
13
Literature Review

• Traffic Assignment Problem
• The Evacuation Problem can be structured as a
  Traffic Assignment Problem (TAP) (Sheffi et al.
  1982, Hobeika and Kim 1998)
• A procedure for loading an origin-destination
  trip table onto links of a road network (Sheffi
  et al. 1982)
• The objective is to route the traffic from
  origins to destinations as quickly as possible
  (least cost solution, minimization)
• Subject to capacity constraints (Lu et al. 2005)

http//people.hofstra.edu/geotrans/eng/ch2en/meth2
en/ch2m4en.html
14
Literature Review

• Combinatorial Complexity
• N 100 supply locations, P 10 demand locations
• N!/(P!(N-P!)
• 100!/(10!(100-10!)
• 17,310,309,456,440 possible combinations
• Impossible to determine all combinations in our
  lifetime
• Need an alternative method

(Curtin 2005)
15
Literature Review

• Heuristics
• Provide good but potentially sub-optimal results
• Not as precise as optimal solution
• Tradeoff for reduced processing time
• Alternative to optimal solution which is near
  impossible to calculate

16
Literature Review

• Capacity Constrained Route Planner (CCRP) (Lu et
  al. 2005)
• Places travel time and maximum road capacity
  constraints on a network
• First in, First out (FIFO) property imposed on
  edges
• Modeled as time series (multidimensional arrays
  in memory vs. time-expanded network) (Shekhar
  2006)
• Algorithm using Dijkstra to prioritize routes by
  quickest arrival time
• Then updates network capacities according to a
  flow amount generated from minimum of
• Evacuees remaining at source
• Available Edge Capacity
• Available Node Capacity

17
Literature Review

• Capacities
• Node Constraints
• Max Node Capacity How many evacuees can fit
• Initial Node Occupancy Number of evacuees at
  location when evacuation begins
• Edge Constraints
• Max Edge Capacity Number of evacuees that will
  fit on an edge at a particular time period
• Travel Time The time it takes for a group of
  evacuees to traverse the beginning of the edge to
  the end of the edge

18
Literature Review
Capacity Formula (Highway Capacity Manual 2000)

• where
• s0 base saturation flow rate per lane (pcphpl)
  Standard 1900
• N number of peak lanes (Data Item 87)
• fw adjustment factor for lane width
• fHV adjustment factor for heavy vehicles in
  traffic stream
• fp adjustment factor for existence of parking
  activity
• fa adjustment factor for area type (1.0 if not
  in CBD 0.9 if in CBD)
• PFH Peak Hour Factor Standard 0.92 (HCM 10-8)

19
Data Sources

• Network generated by authors
• 3 sources with evacuees
• 2 destinations with unlimited capacity

Lu et. al 2005
20
Data Sources

• Tanker truck transporting Chlorine wrecks off of
  Hwy 360
• Emergency responders decide to evacuate
  neighborhood downwind of toxic cloud
• Mitigation procedures must be completed in order
  to permit re-entry (Southworth 1987)

(China.org.cn, China Daily March 31, 2005)
2 miles
1 mile
http//www.aristatek.com/Newsletter/05200620June
/The20First20Responder20Technically20Speaking.
htm
21
(No Transcript)
22
Data Sources

• Grapevine Road Network
• Purpose Base data for network analysis. Exported
  to feature class to be used in network feature
  dataset.
• Edges 91
• Edge Constraints
• Max Edge Capacity (HCM)
• Travel Time (Speed Limit / Segment Length )
• Nodes 75
• Node Constraints
• Max Node Capacity (arbitrary)
• Initial Node Occupancy (census)
• 2000 Census Data
• Block level
• Block group level
• Random component (population  Int((HighestValue
  - LowestValue 1) Rnd) LowestValue )

All maps are in NAD 83, State Plane North Central
Texas FIPS 4202.
23
Data Sources

• Grapevine Road Capacity Estimate (HCM 2000)
• For one lane roads
• 1900 1 1.033 fHV fp fa (0.92)
• 1805.684 passenger cars per hour per lane on
  grapevine roads
• For two lane roads
• 1900 2 1.033 fHV fp fa (0.92)
• 3611.368 passenger cars per hour per lane on
  grapevine roads

24
Analysis Methodology

• CCRP Implementation
• Program algorithm in ArcGIS 9.1 using VBA and
  ArcObjects with Network Analyst extension
• Pilot Study using test network (Lu et al. 2005)
• Comparison of results from their route generation
  module with the ArcGIS module
• Dijkstras Shortest Path
• Nodes are considered stops along route
• Each stop has an arrival time (time schedule)
• Available Capacity
• Dynamic update of capacity at edges and nodes
• If available capacity, evacuees can populate edge
  or node

25
Analysis Methodology
CCRP Experiment Design
Lu et. al 2005
26
Analysis Methodology
Output Evacuation Routes
Network Query Attributes of Network Features
Shortest Path Analysis GIS Display
Proposed Equivalent Design in GIS
ND Edges Junctions
Multi Dimensional Arrays
Issue
Update Capacity Component
27
Analysis Methodology

• Algorithm Flow
• Pre-process network
• Generalized shortest path search
• Assign flow as minimum of capacities on route
• Reduce capacities along route by flow

28
Demo
29
Conclusions

• Update capacity component of CCRP implemented in
  VBA for authors example network
• Network Feature dataset does not allow for
  dynamic update of capacities on edges and at
  nodes
• Shortest Path and Update Capacity components may
  work together in VBA environment
• ArcObjects for Network Analyst extension steep
  learning curve
• Documentation in infancy

30
Future Research

• Comparison of Egress Times
• Sub-optimal solution results (ArgGIS Network
  Analyst, VBA for ArcObjects)
• Optimal solution results (ILOG CPLEX
  8.1.0,Optimization Programming Language (OPL)
  Studio version 3.6.1.)
• Number of Evacuees
• Graph egress time as a function of total evacuees
• Time of day variation
• Morning peak hour (increased travel time, edge
  capacity)
• Mid day (low med occupancy)
• Late evening (high occupancy)

31
Contributions to GIScience

• Implemented capacity updating component of CCRP
  Heuristic for test network
• Encountered pitfalls to be avoided by future
  researchers

32
References

• Sheffi, Y., H. Mahmassani, et al. 1982. A
  transportation network evacuation model.
  Transportation Research Part A General 16(3)
  209-218.
• Tweedie, S. W., Rowland, J.R., Walsh, S.J.,
  Rhoten, R.P, and P.I. Hagle (1986). "A
  Methodology for Estimating Emergency Evacuation
  Times." The Social Science Journal 23(2)
  189-204.
• Hamacher, H. W. and S. A. Tjandra. 2002.
  Mathematical modeling of evacuation problems A
  state of the art. Pedestrian and Evacuation
  Dynamics. M. Schreckenberg and S. D. Sharma
  227-266.
• Hobeika, A. G. and C. Kim. 1998. Comparison of
  traffic assignments in evacuation modeling. IEEE
  Transactions On Engineering Management 45(2)
  192-198.
• Sorensen, J. H., Vogt, B. M., and Mileti, D. S.,
  1987 Evacuation an assessment of planning and
  research, Oak Ridge National Laboratory,
  ORNL-6376, Tennessee.
• Lu, Q., George, B., and Shekhar, S. 2005.
  Capacity constrained routing algorithms for
  evacuation planning a summary of results.
  Proceedings of the Ninth International Symposium
  on Spatial and Temporal Databases.
  Springer-Verlag Berlin, Heidelberg, 2005 LNCS
  3633 291307.
• Hoppe, B., and Tardos, E. (1994). Polynomial time
  algorithms for some evacuation problems.
  Proceedings of the fifth annual ACM-SIAM
  symposium on Discrete algorithms. Arlington,
  Virginia, United States, Society for Industrial
  and Applied Mathematics.

33
Code

• Heuristic