Modeling of Real-Time Vehicle Routing Algorithms Applied to Disaster Management in Urban Traffic Networks

elfinoverwroughtNetworking and Communications

Jul 18, 2012 (4 years and 11 months ago)

275 views

California-Catalonia Program for Engineering Innovation 2008-2009 Progress Report




Modeling of Real-Time Vehicle Routing Algorithms Applied to Disaster
Management in Urban Traffic Networks



TEAM MEMBERS

Principal Investigators

HSSoE Professor R. Jayakrishnan
Department of Civil and Environmental Engineering
University of California at Irvine SST 557
Irvine, CA 92697-3600
rjayakri@uci.edu

Catalonia Prof. Laia Pages
Professora Associada
Department d’Enginyeria Mechanica
ETSEIAT, EDIFICI TR45
Universitat Politecnica de Catalunya
Carrer Colom, 11
08222 Terrassa, Spain
Laia.Pages@upc.edu

Collaborators

Mr. Alejandro Bayas
MS Student


Institute of Transportation Studies at UCI
University of California, Irvine
AIRB Ste. 4000
Irvine, CA 92697



Industrial partners



Dr. Mithilesh Jha
EarthTech Inc.
21064 Bake Parkway Ste 200
Lake Forest, CA 92630
Ph: 562-951-2000






California-Catalonia Program for Engineering Innovation 2008-2009 Progress Report



EXECUTIVE SUMMARY

This research is to enhance a modeling framework and software system developed earlier at UCI
and UPC for mass passenger transport systems for application to the type of disaster scenarios in the
US context, specifically in California, and to thus demonstrate the viability of the model for future
applications in the US and Europe. The earlier research started from theoretical vehicle routing
algorithms developed at UCI and applied those to the City of Barcelona for modeling-based design
of real-time dispatched passenger feeder services (shuttle vans, taxis, etc) to the subway mass
transport system in the city. The research yielded a complete modeling system for “City Logistics”
which can now be made more elaborate in scope, including different classes of dispatch vehicles
such as Emergency Medical Service (EMS) and Police vehicles. This research brings this
framework back to California for modeling-based design for specific disaster management plans.
The work involves (a) developing a theoretical component that formulates optimizations schemes
for a broader routing problem, (b) enhancing the simulation software system with additional new
algorithms for disaster management vehicle routing, and c) demonstrating the unique effectiveness
of the modeling systems in the US context through application to a southern California network.

During the first 9 months of the project, the work on enhancing the Mass Transport vehicle routing
methodology for disaster conditions has been initiated. The Aimsun software’s applicability to
California networks was studied as part of the MS Thesis of Alejandro Bayas, which dealt with
Aimsun’s modeling of specific geometric conditions. Disaster scenarios to be used for the studies
in this project have been completed as part of another MS thesis at UC Irvine, that focused on the
disruptions caused by earthquake-induced failure of highway bridge links, specifically in the Los
Angeles area (near the interchange of I-10 and I-405 freeways).

WORK COMPLETED

The basic theoretical problem in this research is rather simple: how to optimally route n vehicles in
a pick up and delivery system to transport m passenger vehicles and l fleet-dispatch vehicles with
their own individual origin-destination travel desires in real-time, as disaster situations arise. We
call the above problem “Real Time Passenger and Dispatch Vehicle Routing Problem”
(RTPDVRP). This is an extension of the Mass Transport Vehicle Routing Problem (MTVRP) that
was studied in earlier UCI-UPC project. The problem has direct significance in emergency
situations that arise in networks such as in southern California which are subject to potential
earthquake disruptions.

The research in this project has so far developed a set of earthquake disruption scenarios, with
elaborate attention paid to specific bridge failures in the Los Angeles Area. The severity of
earthquakes is varied in an associated project, yielding different network link failure conditions
based on the structural properties of the bridges. The methodology to simulate earthquake and
develop the scenarios for traffic simulation are shown in Figure 1 and the candidate simulation
network to be used for Aimsun simulations under this project is shown in Figure 2.

The next step was developing schemes that will be flexible enough to consider innovative
emergency response vehicle dispatch schemes. This includes developing vehicle movement
networks on the fly and finding the best locations for transfer of passengers and people needing
medical attention between different vehicle classes. This yields a real-time network design problem
which is difficult in scope but can be solved along the same lines as in our previous research that
California-Catalonia Program for Engineering Innovation 2008-2009 Progress Report


applied the routing schemes for Barcelona. The global optimization proposed in this research
follows from the earlier Cal-Cat project where it is decomposed in three steps as shown in Figure 3.


Figure 1. Methodology developed to generate Earthquake Disruptions in the Traffic System

Figure 2. West Los Angeles network model where earthquake disaster scenarios are developed
(Freeways I-405 & I-10 shown. Various links fail based on earthquake severity)
Scenario
Earthquake
Attenuation
Function
Geometry (Nodes,
Links, Bridges)
Bridge Fragility
Parameters
Initial TRF
Functions
Estimate Site
Ground Motion
Simulate Bridge
Damage States
Disable Damaged
Links
Prepare
Mesoscopic
Model
Network to simulate Travel
Scenarios under Earthquake
Disruptions
Reduce Base
P&A (damage)
Start
Trip Distribution
(Exp. Gravity
Model)
Decision: If the
Base Ps & As are
reduced by >1%,
distribute the
reduced Ps & As.
Else, use Base
OD matrix.
Perform
Subarea
Analysis
Base
P&A
Establish
Base Travel
Times
Decision: If any network
links have been disabled (i.e.
the network is damaged),
then calculate Average Delay
Per Vehicle.
California-Catalonia Program for Engineering Innovation 2008-2009 Progress Report



Figure 3. Solution Scheme for Mass Transport Vehicle Routing Problem

The approach in the research formulates relaxed versions of MTVRP in linear formulations to
handle the nonlinearity of the cost functions by iterative solutions where simulations show the costs,
which corresponds to steps 1 and 2 shown in Fig.3 above. In the third step, known vehicle routing
heuristics are used to route the vehicles using a cost function, where cumulative travel time of
passengers in the system is taken into account. In the enhanced model for combined routing of
mass transport and personal commuter vehicles with EMS and police vehicles, the above
hierarchical optimization is augmented by specifically keeping track of the vehicle classes in the
optimization framework, which would effectively incorporate parallel steps within the level 2 and 3
in figure 1. This work has been initiated and will be completed during the research visits by the PI
and co-PI during the summer of 2009.

As part of the MS research of Alejandro Bayas, the application of the Aimsun Simulation Model to
California conditions was investigated in detail. A simple case study of specific types of lane-drop
conditions studied is shown in Figure 4. For this reason the next figure recreates the network,
changing the longitude of segments so that it is easy to appreciate.

Section 1 2 3 4 5 end








Centroid 210 Centroid 212
Figure 4: Aimsun Network Rpresentation For a Lane-drop Geometric Study (Bayas, 2009)

Specific vehicle characteristics were also calibrated, to be used for the Aimsun simulations. An
example is shown in Fig. 5. Illustrative results from simulations under DTA (dynamic traffic
assignment with vehicle routing) are shown in Fig. 6, even though the test network did not offer the
rerouting possibilities of a full network with multiple path options under earthquake routing of fleet
vehicles. The preparatory work with the smaller candidate network has not yet been extended to the
larger networks of study, such as shown in Figure 2. This was due to some delays in procuring the
license for the full version of the AIMSUN software. This work will be completed during the
summer of 2009, using the simulation capabilities developed at UPC Terrassa, with the direct
involvement of Dr. Laia Pages.
STEP 1 NETWORK AGGREGATION
Posi tion of Centroids and
distribution of zones
STEP 2 MTNDP (Mass Transport
Network Design Problem)
Set the coverage densiti es and the
servi ce rates in the aggregated
net work
STEP 3 LMTVRP (Local Mass
Transport Vehicle Routing Problem)
Rout ing of vehicles at each
zone independently in the
det ai led network
STATIC
Known de mand
DYNAMIC
Unknown demand
STATIC
Known demand
STEP 1 NETWORK AGGREGATION
Posi tion of Centroids and
distribution of zones
STEP 2 MTNDP (Mass Transport
Network Design Problem)
Set the coverage densiti es and the
servi ce rates in the aggregated
net work
STEP 3 LMTVRP (Local Mass
Transport Vehicle Routing Problem)
Rout ing of vehicles at each
zone independently in the
det ai led network
STATIC
Known de mand
DYNAMIC
Unknown demand
STATIC
Known demand

California-Catalonia Program for Engineering Innovation 2008-2009 Progress Report



Figure 5: Characteristics of a specific vehicle type modeled in Aimsun (Bayas, 2009)

Figure 6. Example output statistics from Aimsun simulations under DTA (Bayas, 2009)
(“Dynamic Traffic Assignment” refers to Real-time Vehicle Rerouting)
California-Catalonia Program for Engineering Innovation 2008-2009 Progress Report


BENEFITS OF THE ACADEMIC/INDUSTRIAL PARTNERSHIP FOR CALIFORNIA
AND CATALONIA

On the academic side, the research has been beneficial to both the PI and the co-PI, and have led
to publications in simulation area. The graduate student (Balsells Fellow) supported by this
project, Mr. Alejandro Bayas, completed his Masters thesis tackling a portion of this research
project and will be continuing his academic participation in the work after returning to
Barcelona.

The industrial partner, Earth Tech Systems, has stayed interested and involved in this work and
as the model applicability to the specific disaster management is demonstrated in this project, it
is expected that the industrial partner will be actively involved in transferring the knowledge-
base for further uses of similar techniques in California.



TRAVEL

Due to certain passport and visa problems, Prof. Jayakrishnan has not been able to visit UPC
Terrassa yet, unlike originally planned. The visit is now planned to complete the remaining
work, during the summer of 2009.

PUBLICATIONS


One MS Thesis has been published, directly supported by this research project.

1) Alejandro Bayas (2009) “Study of Variable Speed-Limit Impacts on Travel Time, Fuel
Consumption and Emissions in a Lane-drop Scenario,” MS Thesis in Civil Engineering, Univ.
of California, Irvine, Apr. 2009

The following two publications include work that is motivated by this research project -

2) Pierre Auza, R. Jayakrishnan, Masanobu Shinozuka, (2010) “Using Mesoscopic Simulation
in a Seismic Risk Analysis Framework Applied to a Los Angeles Network,” Under review for
publication in the Transportation Research Record Journal and presentation at the Annual
Meeting of the Transportation research Board, Washington, DC, January 2009

3) Xavier Costa Sanfeliu, Francesc Astal Coma, Laia Pages, et al. (2009) “Estudi Simulacio De
Xarxes De Urbanes De Transport Sota Situacions De Canvis De La Zarza Viaria (Tall De
Carrers, Emergencias, Tancament De Carrers Al Transit, Etc.) Mitjancant El Programa
AIMSUN,” PFC Report, Escola Tecnica Superior d’Enginyeries, UPC. June 2009.

APPENDICES

None.