The numerous man-made disasters and natural catastrophes that menace major communities accentuate the need for proper planning for emergency evacuation. Transportation networks in cities evolve over long time spans in tandem with population growth and evolution of travel patterns. In emergencies, travel demand and travel patterns drastically change from the usual everyday volumes and patterns. Given that most US and Canadian cities are already congested and operating near capacity during peak periods, network performance can severely deteriorate if drastic changes in Origin-Destination (O-D) demand patterns occur during or after a disaster. Also, loss of capacity due to the disaster and associated incidents can further complicate the matter. Therefore, the primary goal when a disaster or hazardous event occurs is to coordinate, control, and possibly optimize the utilization of the existing transportation network capacity. Emergency operation management centres face multi-faceted challenges in anticipating evacuation flows and providing proactive actions to guide and coordinate the public towards safe shelters. Numerous studies have contributed to developing and testing strategies that have the potential to mitigate the consequences of emergency situations. They primarily investigate the effect of some proposed strategies that have the potential of improving the performance of the evacuation process with modelling and optimization techniques. However, most of these studies are inherently restricted to evacuating automobile traffic using a certain strategy without considering other modes of transportation. Moreover, little emphasis is given to studying the interaction between the various strategies that could be potentially synergized to expedite the evacuation process. Also, the absence of an accurate representation of the spatial and temporal distribution of the population and the failure to identify the available modes and populations that are captive to certain modes contribute to the absence of multimodal evacuation procedures. Incorporating multiple modes into emergency evacuation has the potential to expedite the evacuation process and is essential to assuring the effective evacuation of transit-captive and special-needs populations . This dissertation presents a novel multimodal optimization framework that combines vehicular traffic and mass transit for emergency evacuation. A multi-objective approach is used to optimize the multimodal evacuation problem. For automobile evacuees, an Optimal Spatio-Temporal Evacuation (OSTE) framework is presented for generating optimal demand scheduling, destination choices and route choices, simultaneously. OSTE implements Dynamic Traffic Assignment (DTA) techniques coupled with parallel distributed genetic optimization to guarantee a near global optimal solution. For transit evacuees, a Multi-Depots, Time Constrained, Pick-up and Delivery Vehicle Routing Problem (MDTCPD-VRP) framework is presented to model the use of public transit vehicles in evacuation situations. The MDTCPD-VRP implements constraint programming and local search techniques to optimize certain objective functions and satisfy a set of constraints. The OSTE and MDTCPD-VRP platforms are integrated into one framework to replicate the impact of congestion caused by traffic on transit vehicle travel times. A proof-of-concept prototype has been tested; it investigates the optimization of a multimodal evacuation of a portion of the Toronto Waterfront area. It also assesses the impact of multiple objective functions on emergency evacuation while attempting to achieve an equilibrium state between transit modes and vehicular traffic. Then, a large-scale application, including a demand estimation model from a regional travel survey, is conducted for the evacuation of the entire City of Toronto. This framework addresses many limitations of existing evacuation planning models by: 1) synergizing multiple evacuation strategies; 2) utilizing robust optimization and solution algorithms that can tackle such multi-dimensional non deterministic problem; 3) estimating the spatial and temporal distribution of evacuation demand; 4) identifying the transit-dependent population; 5) integrating multiple modes in emergency evacuation. The framework presents a significant step forward in emergency evacuation optimization.

The numerous man-made disasters and natural catastrophes that menace major communities accentuate the need for proper planning for emergency evacuation. Transportation networks in cities evolve over long time spans in tandem with population growth and evolution of travel patterns. In emergencies, travel demand and travel patterns drastically change from the usual everyday volumes and patterns. Given that most US and Canadian cities are already congested and operating near capacity during peak periods, network performance can severely deteriorate if drastic changes in Origin-Destination (O-D) demand patterns occur during or after a disaster. Also, loss of capacity due to the disaster and associated incidents can further complicate the matter. Therefore, the primary goal when a disaster or hazardous event occurs is to coordinate, control, and possibly optimize the utilization of the existing transportation network capacity. Emergency operation management centres face multi-faceted challenges in anticipating evacuation flows and providing proactive actions to guide and coordinate the public towards safe shelters. Numerous studies have contributed to developing and testing strategies that have the potential to mitigate the consequences of emergency situations. They primarily investigate the effect of some proposed strategies that have the potential of improving the performance of the evacuation process with modelling and optimization techniques. However, most of these studies are inherently restricted to evacuating automobile traffic using a certain strategy without considering other modes of transportation. Moreover, little emphasis is given to studying the interaction between the various strategies that could be potentially synergized to expedite the evacuation process. Also, the absence of an accurate representation of the spatial and temporal distribution of the population and the failure to identify the available modes and populations that are captive to certain modes contribute to the absence of multimodal evacuation procedures. Incorporating multiple modes into emergency evacuation has the potential to expedite the evacuation process and is essential to assuring the effective evacuation of transit-captive and special-needs populations . This dissertation presents a novel multimodal optimization framework that combines vehicular traffic and mass transit for emergency evacuation. A multi-objective approach is used to optimize the multimodal evacuation problem. For automobile evacuees, an Optimal Spatio-Temporal Evacuation (OSTE) framework is presented for generating optimal demand scheduling, destination choices and route choices, simultaneously. OSTE implements Dynamic Traffic Assignment (DTA) techniques coupled with parallel distributed genetic optimization to guarantee a near global optimal solution. For transit evacuees, a Multi-Depots, Time Constrained, Pick-up and Delivery Vehicle Routing Problem (MDTCPD-VRP) framework is presented to model the use of public transit vehicles in evacuation situations. The MDTCPD-VRP implements constraint programming and local search techniques to optimize certain objective functions and satisfy a set of constraints. The OSTE and MDTCPD-VRP platforms are integrated into one framework to replicate the impact of congestion caused by traffic on transit vehicle travel times. A proof-of-concept prototype has been tested; it investigates the optimization of a multimodal evacuation of a portion of the Toronto Waterfront area. It also assesses the impact of multiple objective functions on emergency evacuation while attempting to achieve an equilibrium state between transit modes and vehicular traffic. Then, a large-scale application, including a demand estimation model from a regional travel survey, is conducted for the evacuation of the entire City of Toronto. This framework addresses many limitations of existing evacuation planning models by: 1) synergizing multiple evacuation strategies; 2) utilizing robust optimization and solution algorithms that can tackle such multi-dimensional non deterministic problem; 3) estimating the spatial and temporal distribution of evacuation demand; 4) identifying the transit-dependent population; 5) integrating multiple modes in emergency evacuation. The framework presents a significant step forward in emergency evacuation optimization.

Recent natural disasters have highlighted the need to evacuate people as quickly as possible. During hurricane Rita in 2005, people were stuck in queue buildups and large scale congestions, due to improper use of capacity, planning and inadequate response to vehicle breakdown, flooding and accidents. Every minute is precious in situation of such disaster scenarios. Understanding evacuation demand loading is an essential part of any evacuation planning. One of the factors often understood to effect evacuation, but not modeled has been the effect of a previous hurricane. This has also been termed as the 'Katrina Effect', where, due to the devastation caused by hurricane Katrina, large number of people decided to evacuate during Hurricane Rita, which hit Texas three weeks after Katrina hit Louisiana. An important aspect influencing the rate of evacuation loading is Evacuation Preparation Time also referred to as 'Mobilization time' in literature. A methodology to model the effect of a recent past hurricane on the mobilization times for evacuees in an evacuation has been presented utilizing simultaneous estimation techniques. The errors for the two simultaneously estimated models were significantly correlated, confirming the idea that a previous hurricane does significantly affect evacuation during a subsequent hurricane. The results show that the home ownership, number of individuals in the household, income levels, and level/risk of surge were significant in the model explaining the mobilization times for the households. Pet ownership and number of kids in the households, known to increase the mobilization times during isolated hurricanes, were not found to be significant in the model. Evacuation operations are marred by unexpected blockages, breakdown of vehicles and sudden flooding of transportation infrastructure. A fast and accurate simulation model to incorporate flexibility into the evacuation planning procedure is required to react to such situations. Presently evacuation guidelines are prepared by the local emergency management, by testing various scenarios utilizing micro-simulation, which is extremely time consuming and do not provide flexibility to evacuation plans. To gain computational speed there is a need to move away from the level of detail of a micro-simulation to more aggregated simulation models. The Cell Transmission Model which is a mesoscopic simulation model is considered, and compared with VISSIM a microscopic simulation model. It was observed that the Cell Transmission Model was significantly faster compared to VISSIM, and was found to be accurate. The Cell Transmission model has a nice linear structure, which is utilized to construct Linear Programming Problems to determine optimal strategies. Optimization models were developed to determine strategies for optimal scheduling of evacuation orders and optimal crossover locations for contraflow operations on freeways. A new strategy termed as 'Dynamic Crossovers Strategy' is proposed to alleviate congestion due to lane blockages (due to vehicle breakdowns, incidents etc.). This research finds that the strategy of implementing dynamic crossovers in the event of lane blockages does improve evacuation operations. The optimization model provides a framework within which optimal strategies are determined quickly, without the need to test multiple scenarios using simulation. Destination networks are the cause of the main bottlenecks for evacuation routes, such aspects of transportation networks are rarely studied as part of evacuation operations. This research studies destination networks from a macroscopic perspective. Various relationships between network level macroscopic variables (Average Flow, Average Density and Average speed) over the network were studied. Utilizing these relationships, a "Network Breathing Strategy" was proposed to improve dissipation of evacuating traffic into the destination networks. The network breathing strategy is a cyclic process of allowing vehicles to enter the network till the network reaches congestion, which is followed by closure of their entry into the network until the network reaches an acceptable state. After which entrance into the network is allowed again. The intuitive motivation behind this methodology is to ensure that the network does not remain in congested conditions. The term 'Network Breathing' was coined due to the analogy seen between this strategy to the process of breathing, where vehicles are inhaled by the network (vehicles allowed in) and dissipated by the network (vehicles are not allowed in). It is shown that the network breathing improves the dissipation of vehicle into the destination network. Evacuation operations can be divided into three main levels: at the origin (region at risk), routes and destination. This research encompasses all the three aspects and proposes a framework to assess the whole system in its entirety. At the Origin the demand dictates when to schedule evacuation orders, it also dictates the capacity required on different routes. These breakthroughs will provide a framework for a real time Decision Support System which will help emergency management official make decisions faster and on the fly.