1

Evacuation Route Planning: Scalable Approaches

Shashi ShekharMcKnight Distinguished University Professor

Director, AHPCRCUniversity of Minnesotashekhar@cs.umn.edu

March 8th, 2006

Presentation at ITS Minnesota

Plans are of little importance, but planning is essential -- Winston Churchill Plans are nothing; planning is everything.-- Dwight D. Eisenhower

2

Why Evacuation Planning? Lack of effective evacuation plans Traffic congestions on all highways Great confusions and chaos

"We packed up Morgan City residents to evacuate in the a.m. on the day that Andrew hit coastal Louisiana, but in early afternoon the majority came back home. The traffic was so bad that they couldn't get through Lafayette." Mayor Tim Mott, Morgan City, Louisiana ( http://i49south.com/hurricane.htm )

( FEMA.gov)

Hurricane Rita evacuees from Houston clog I-45.

( www.washingtonpost.com)

Hurricane AndrewFlorida and Louisiana, 1992

Hurricane RitaGulf Coast, 2005

( National Weather Services)

( National Weather Services)

3

Homeland Defense & Evacuation Planning

Preparation of response to a chem-bio attack

Plan evacuation routes and schedules

Help public officials to make important decisions

Guide affected population to safetyBase Map Weather

Data

Plume Dispersion

Demographics Information

Transportation Networks

( Images from www.fortune.com )

4

Problem Statement

Given• Transportation network with capacity constraints• Initial number of people to be evacuated and their initial locations • Evacuation destinations

Output• Routes to be taken and scheduling of people on each route

Objective• Minimize total time needed for evacuation• Minimize computational overhead

Constraints• Capacity constraints: evacuation plan meets capacity of the network

5

Limitations of Related Work

Linear Programming Approach

- Optimal solution for evacuation plan

- e.g. EVACNET (U. of Florida), Hoppe and Tardos (Cornell University).

Limitation: - High computational complexity

- Cannot apply to large transportation networks

Capacity-ignorant Approach - Simple shortest path computation

- e.g. EXIT89(National Fire Protection Association)

Limitation: - Do not consider capacity constraints

- Very poor solution quality

Number of Nodes 50 500 5,000 50,000

EVACNET Running Time 0.1 min 2.5 min 108 min > 5 days

6

Related Works: Linear Programming Approach

Step 1: Convert evacuation network into time-expanded network with user provided time upper bound T.

G : evacuation network

Step 2: Use time-expanded network GT as a flow network and solve it using LP min. cost flow solver (e.g. NETFLO).

GT : time-expanded network (T=4)

(Source of network figures: H. Hamacher, S. Tjandra, Mathmatical Modeling of Evacuation Problems: A state of the art. Pedestrian and Evacuation Dynamics, pages 227-266, 2002.)

7

Proposed Approach• Existing methods can not handle large urban scenarios

• Communities use manually produced evacuation plans

• Key Ideas in Proposed Approach

• Generalize shortest path algorithms

• Honor road capacity constraints

• Capacity Constrained Route Planning (CCRP)

8

Performance Evaluation

Experiment 1: Effect of People Distribution (Source node ratio)

Results: Source node ratio ranges from 30% to 100% with fixed occupancy ratio at 30%

Figure 1 Quality of solution Figure 2 Running time

• SRCCP produces solution closer to optimal when source node ratio goes higher

• MRCCP produces close-to-optimal solution with half of running time of optimal algorithm

• Distribution of people does not affect running time of proposed algorithms when total number of people is fixed

0

100

200

300

400

500

600

30 50 70 90 100

Source Node Ratio (%)

To

tal E

va

cuat

ion

Tim

e

SRCCP

MRCCP

EVACNET

0

10

20

30

40

50

60

70

30 50 70 90 100

Source Node Ratio (%)

Ru

nn

ing

Tim

e (

se

co

nd

)

SRCCP

MRCCP

EVACNET

9

A Real Scenario

Nuclear Power Plants in Minnesota

Twin Cities

10

Minnesota Nuclear Power Plant Evacuation

Affected Cities

Monticello Power Plant

Evacuation Destination

AHPCRC

11

Monticello Emergency Planning Zone

Monticello EPZSubarea Population2 4,675 5N 3,9945E 9,6455S 6,7495W 2,23610N 39110E 1,78510SE 1,39010S 4,616 10SW 3,40810W 2,35410NW 707Total 41,950

Estimate EPZ evacuation time: Summer/Winter (good weather):     3 hours, 30 minutesWinter (adverse weather):   5 hours, 40 minutes

Emergency Planning Zone (EPZ) is a 10-mile radius around the plant divided into sub areas.

Data source: Minnesota DPS & DHS Web site: http://www.dps.state.mn.us

http://www.dhs.state.mn.us

12

Destination

Monticello Power Plant

Handcrafted Existing Evacuation Routes

13

A Real Scenario – Overlay of New Plan Routes

Source cities

Destination

Monticello Power Plant

Routes used only by old plan

Routes used only by result plan of capacity constrained routing

Routes used by both plans

Congestion is likely in old plan near evacuation destination due to capacity constraints. Our plan has richer routes near destination to reduce congestion and total evacuation time.

Twin Cities

Total evacuation time:

- Existing Routes: 268 min.

- New Routes: 162 min.

14

Project 2: Metro Evacution Planning (2005)

Agency Roles

Identify Stakeholders

Establish Steering

Committee

Perform Inventory of

Similar Efforts and Look at

Federal Requirements Finalize

Project Objectives

Regional Coordination

and Information

Sharing

Metro Evacuation Plan

Stakeholder Interviews and

Workshops

Evacuation Route

Modeling

Evacuation Routes and Traffic Mgt.

Strategies

Issues and Needs

Final Plan

Preparedness Process

15

Road Networks

1. TP+ (Tranplan) road network for Twin Cities Metro Area

Source: Met Council TP+ dataset

Summary: - Contain freeway and arterial roads with road capacity, travel time,

road type, area type, number of lanes, etc.- Contain virtual nodes as population centroids for each TAZ.

Limitation: No local roads (for pedestrian routes)

2. MnDOT Basemap

Source: MnDOT Basemap website (http://www.dot.state.mn.us/tda/basemap)

Summary: Contain all highway, arterial and local roads.

Limitation: No road capacity or travel time.

16

Demographic Datasets

1. Night time population

• Census 2000 data for Twin Cities Metro Area

• Source: Met Council Datafinder (http://www.datafinder.org)

• Summary: Census 2000 population and employment data for each TAZ.

• Limitation: Data is 5 years old; day-time population is different.

2. Day-time Population

• Employment Origin-Destination Dataset (Minnesota 2002)

• Source: MN Dept. of Employment and Economic Development - Contain work origin-destination matrix for each Census block.- Need to aggregate data to TAZ level to obtain: Employment Flow-Out: # of people leave each TAZ for work. Employment Flow-In: # of people enter each TAZ for work.

• Limitation: Coarse geo-coding => Omits 10% of workers

• Does not include all travelers (e.g. students, shoppers, visitors).

17

Scenarios

• Sources:• Prioritized list of vulnerable facilities and locations in Twin Cities

• Federal list of scenarios – Subset requiring evacuation

• Input from advisory board and stakeholder workshop

• Selected Scenarios

1. Downtown 1

2. Downtown 2

3. University

4. Mall

5. Suburban Facility

• Scenario Specification for Evacuation Route Planning• Explicit

• Source = < Location Center, Footprint Circles >

• Destination = choice ( fixed locations OR outside a destination circle)

• Implicit (Estimates from databases) with user overrides

• Transportation Network – connectivity, capacities, travel times

• Demand: Population estimates, mode preferences

18

Source Radius (mile)

# of TAZs Night Population

(Census 2000)

Employment (Census 2000)

Employment Flow-Out (2002)

Employment Flow-In (2002)

1 11 18,373 36,373 4,851 34,897

2 53 93,151 213,911 37,081 201,143

Scenario 3: University

19

Evacuation time:1 hr 48 min

Evacuation Zone:Source Radius: 1 mileDest. Radius: 1 mile

Number of Evacuees: 19,649 (Estimated Daytime)

Transportation mode: single occupancy vehicles

Evacuation Zone

Destinations nodes w/ evacuee assignment

MnDOT basemap

TP+ network

Evacuation routes on TP+ network

50

Test Scenario: State Fair Playground

20

Evacuation Zone:Source Radius: 10 mileDest. Radius: 10 mile

Number of Evacuees: 1.37 Million (Est. Daytime)

Transportation mode: single occupancy vehicles

Evacuation Zone

MnDOT basemap

TP+ network

Scalability Test : Large Scenario

21

Evacuation Routes for Large Scenario

22

Common Usages for the Tool

• Compare options– Ex.: transportation modes

• Walking may be better than driving for 1-mile scenarios

– Ex.: Day-time and Night-time needs• Population is quite different

• Identify bottleneck areas and links– Ex.: Large enclosed malls with sparse transportation network– Ex.: Bay bridge (San Francisco),

• Designing / refining transportation networks– Address evacuation bottlenecks– A quality of service for evacuation, e.g. 4 hour evacuation time

23

Driving vs. Pedestrian Modes

Driving 100% Walking 100%

Population Overwritten

24

Daytime vs. Nighttime Population

Night Time Result

Day Time Result

25

Limitations of the Tool• Evacuation time estimates

– Approximate and optimistic– Assumptions about available capacity, speed, demand, etc.

• Quality of input data– Population and road network database age!

• Ex.: Rosemount scenario – an old bridge in the roadmap!

– Data availability• Pedestrian routes (links, capacities and speed)

• On-line editing capabilities– Taking out a link is not supported yet!

26

Conclusions• Evacuation Planning is critical for homeland defense

• Existing methods can not handle large urban scenarios

• Communities use manually produced evacuation plans

• New CCRP algorithms

• Can produce evacuation plans for large urban area

• Reduce total time to evacuate!

• Improves current evacuation plans

• Next Steps:

• More scenarios: contra-flow, downtowns e.g. Washington DC

• Dual use – improve traffic flow, e.g. July 4th weekend

• Fault tolerant evacuation plans, e.g. electric power failure

27

Acknowledgements

• Organizations • AHPCRC, Army Research Lab.

• Dr. Raju Namburu• CTS, MnDOT, Federal Highway Authority• National Science Foundation

• Congresspersons and Staff • Rep. Martin Olav Sabo

• Staff persons Marjorie Duske• Rep. James L. Oberstar• Senator Mark Dayton

28