A Genetic Algorithm Based Microscopic Simulation To Develop The Evacuation Plan For Multi-institutional Centers Fengxiang Qiao, Ph.D., Assistant Professor, Texas Southern University Ruixin Ge, M.S., Assistant Transportation Planner, KOA Corporation, California, USA Lei Yu, Ph.D., P.E., Dean and Professor of Texas Southern University Presented at the Intelligent Transportation Society of America’s 20 th Meeting and Exposition, Houston, TX, U.S.A., May 3-5, 2009
Transportation Evacuation Large Scale Evacuation –Hurricane –Radiological incidents Small Scale Evacuation –Terrorists’ bombing threat –Toxic material leakage –Can cause equally severe consequences as the large scale emergencies if they take place in an area with a high-density population
Models Evacuation Planning and Operation SLOSH (Sea, Lake, and Overland Surges from Hurricanes) HURREVAC (HURRicane EVACuation) HAZUS-MH (HAZards US Multi-Hazards) CATS/JACE (Consequence Assessment Tool Set/Joint Assessment of Catastrophic Events), and ETIS (Evacuation Traffic Information Systems) Source: U.S. DOT and Department of Homeland Security
Analytical Tools For Transportation Modeling and Analysis NETVAC (NETwork emergency eVACuation, 6) MASSVAC (MASS eVACuation, 8), and OREMS (Oak Ridge Evacuation Modeling System, 9) Source: U.S. DOT and Department of Homeland Security
Typical Evacuation Plans For Large Scale Emergencies –Contraflow plan developed in response to evacuation difficulties caused by hurricane Katrina in New Orleans –Traffic signal plan for the Hampton Roads region of Virginia to facilitate the movement of large numbers of vehicles in advance of a storm For Relatively Small-scale Emergencies –Major focuses are on the simulation of traffic within buildings and evacuation by elevators –Less attention on evacuation in small & dense area
Research Objectives To Build Up a Microscopic Simulation Framework that Helps to Develop a Transportation Evacuation Plan for Dense Multi-Institutional Center (MIC)
Proposed Framework Identifying Study Road Network Selecting Microscopic Simulation Model Collecting Field Data Coding Simulation Network Defining Modeling Scenarios Calibrating Traffic Simulation Model Validating Traffic Simulation Model Executing Network Simulation; and Evaluating Scenarios and Optimizing Evacuation Plan
Framework of Developing Evacuation Plans Identifying the Study Roadway Network Selecting Traffic Simulation Model Collecting the Field Data Coding the Simulation Network for the Study Area Defining Model Scenarios Calibrating the Simulation Model Executing the Network Simulation Outputting Evaluation Meeting Criterion? END Modify the Network NO YES
Case Study: Texas Medical Center More Than Five Million Patient Visits Annually and One of The Highest Densities of Clinical Facilities and Basic Science and Translational Research of any Location 44 Medicine-related Institutions, 13 Hospitals, and Two Medical Schools, With Nearly 100,000 People Working or Studying in the Area A Very Typical Grouped Institutional Center With High Density
Map of TMC VISSIM Network
Three Types of Data Collected Traffic Volume and Capacity of Each Garage Vehicle Instantaneous Speed –Using GPS in the testing vehicle that followings the average traffic flows Signal timing schema and other traffic control related information
Simulation in VISSIM Flight view of a typical intersection In-car view showing instant vehicle running and roadway message in the box
Speed Collectors in VISSIM
Parameter Calibration in VISSIM Waiting time before diffusion, expressed as x 1 ; Minimum headway, x 2 ; Maximum deceleration, x 3 ; per distance, x 4 ; Accepted deceleration, x 5 ; Maximum look ahead distance, x 6 ; Average standstill distance, x 7 ; Additive part of desired safety distance, x 8 ; Multiple part of desired safety distance, x 9 ; and Distance of standing at 30 mph, x 10. The parameters can be calibrated using real traffic data
Parameter Calibration in VISSIM GPS field collected speed and VISSIM calibrated speed under different generation of gene during peak hour Genetic Algorithm is used in parameter calibration
Evacuation Plans Plan 1 –All garages be cleared in one hour –Inbound traffic volumes are controlled, only allowing emergency vehicles to enter –No traffic control was optimized Plan 2 –Optimization of signal timing –Traffic assignment is considered in one hour base Plan 3 –Optimization of signal timing –Traffic assignments are in every ¼ hour
Network Performance of Different Evacuation Plan Plan 1Plan 2Plan 3 Average speed [mi/hr] Total delay time [hr] Average delay time per vehicle [s] Total stopped delay [hr] Average stopped delay per vehicle [s] Average number of stops per vehicles
Number of Vehicles Evacuated from Garages Garage # Plan 1: 1hrPlan 2: 1hrPlan 3: 1/4hr Plan 3: 1/2hr Plan 3: 3/4hr Plan 3: 1hr Total:
Evacuation Time Related to Percentage of Vehicles for Plan 3 Percentage of VehiclesTime 0%0:00:00 10%0:06:02 20%0:12:21 30%0:18:40 40%0:25:00 50%0:31:19 60%0:37:38 70%0:43:58 80%0:50:17 90%0:56:36 100%1:02:56
Clearance Time of Number of Evacuation Vehicles
Conclusion Proposed a Framework for Developing a Transportation Evacuation Plan for Dense Multi-Institutional Centers. Microscopic Traffic Simulation VISSIM and the Optimization Algorithm Genetic Algorithm are Used Case Study Show that the Framework is Powerful and Practical