Modeling, Design and Optimization of Demand Responsive Services, such as Feeders, ADA Paratransit, Innovative Transit/Logistic Services, Car Sharing, Ride Sharing and Ride Haling. Continuous Approximation Models; Valid Inequalities and Logic Constraints; Scheduling Algorithms.
- Applications – Vehicle Routing Problems:
Within the general context of developing a smarter and sustainable urban environment, I have been working on these transportation problems faced today by the majority of large cities.- Transit for low density areas: Urban sprawl is one of the most evident phenomena characterizing the development of urban areas in the last few decades. Today’s modern cities are progressively losing their conventional centralized identity for a more spread out model. In this framework, conventional transit services begin to struggle and are relegated to a marginal and inefficient role, causing modern urban areas to suffer from severe congestion and pollution problems. Within this context, my research interest and objective are to look for innovative transit design solutions which would efficiently provide a feasible and sustainable transportation alternative. I have particularly paid attention to the design and optimization of Feeder Services as a potential solution to the “first/last mile” problem.
- ADA Paratransit: These services have been experiencing a tremendous growth in the last decades and their demand is expected to further expand. However, these essential services do not seem to be able to operate efficiently, heavily and increasingly relying on subsidies to maintain their large operations. Paratransit operations are a typical practical application for the well known NP-Hard Pickup and Delivery Problem with Time Windows (PDPTW), a variant of the Vehicle Routing Problem (VRP), which can also be expanded to Multiple Depots (MDVRP or MDPDPTW). I am interested in the design, optimization and operations of these services.
- Ride Sharing/Ride Haling/Car Sharing: These services have the potential to change the future, in particular when paired with the rise of Autonomous Vehicle Technologies. The math models associated with them are complex and attract the attention of the optimization research community.
- Port Operations: I’ve been looking at formulating and solving the dispatching problem of the newer generation of Automatic Container Terminals with High Performance Tandem Lift Quay Cranes. Scheduling and operating these systems is particularly challenging and research has not yet addressed the problem properly.
- Methodological Approaches:
I typically apply the following fundamental research approaches to the more practical and applied topics described above:- Continuous Approximations: I am inclined to model complex systems using continuous approximations whenever reasonable and justifiable. The main purpose of this approach is to obtain analytical insights of complex decision problems with as little information as possible. These approximate models are easier for humans to comprehend and they provide elegant, handy but also potentially powerful tools to help solving many complicated decision problems. The challenge resides in recognizing when the tradeoff between modeling approximations and usefulness of the results is acceptable.
- Artificial Intelligence – Algorithm Development: Computers are extremely fast, but they are also “brainless”. Artificial Intelligence is a fascinating discipline, since it tries to close this gap by providing computers with artificial brains, which allow them to make complicated human like decisions very fast. In this context, I am interested in the development of algorithms and heuristics, especially within the scheduling context, to solve complex and time consuming problems in reasonable time as fast as possible. Real time decision making can be significantly improved by finding and adopting smart and creative solutions which are able to be flexible, adaptive and demand responsive.
- Optimization Modeling and Cuts Development: Most optimization problems involving integer variables are too hard to solve, because of their NP-completeness. Research aiming to fine tune their formulation adding effective constraints is significant. A constraint is classified as valid if it reduces the size of the relaxed feasible region, ideally making it the convex hull of the integer feasible solutions. Another category of constraints are the so called “logic cuts”. Their purpose is to reduce the feasible region by eliminating dominated integer feasible solutions and they can be indeed very effective in significantly shrinking the feasible region and considerably reducing the CPU time in the search for optimality.
- Multi Objectives Decision Analysis: Lastly, I am interested in analyzing and modeling the relationships among multiple objectives to be achieved while dealing with a single decision process. Often, poor decisions are simply taken because there is not enough time for analyzing the problem in depth or because the thought process followed by decision makers is too simplified and fundamentally flawed. The challenge is to develop models that could help speed up the decision process by breaking down the whole problem in smaller pieces in which decisions can be made much quicker and in a much more efficient way.
- Simulation: I use simulation extensively usually as a test tool for running experiments.