This study introduces a new travel demand management strategy for managing parking scarcity called Parking Allocation and Ride-Sharing System (PARS). In PARS, a centralized algorithm allocates parking spaces to drivers who are willing to participate in a coordinated carpool. The algorithm is used to optimize the creation of carpools going to and returning from a particular venue and simultaneously reserve parking for these carpools at the venue. An efficient mixed integer linear programming (MIP) formulation is presented and two heuristics, namely Ride Decomposition (RD) and Quick Converge (QC), are proposed and compared via internally generated experiments. Experimental results show that a commercial solver is able to solve the MIP with thousands of individuals to optimality in minutes. For larger instances, the RD and QC heuristic algorithms can solve the problem, on average, 42.23% and 86.39% faster than the commercial solver and provide solutions that are 3.61% and 3.49% from optimal, respectively. See our paper here: An integrated Ridesharing and Parking Allocation System.

A paratransit system is a mobility alternative for people with disabilities. It is considered one of the costliest transportation systems. Most of them work by enabling customers to book reservations in advance. This situation often leads to problems like no-shows and cancellations, realities that affect the system's operational efficiency. As part of this project, we used data from the Metropolitan Bus Authority of Puerto Rico to train a set of machine learning models to predict now-shows and cancellations. We identified critical predictive features and developed demand forecasting models using the trained models. See a paper where we shared the results of our work: Predicting Trip Cancellations and No-Shows in Paratransit Operations