Master Thesis Defense - August 7, 2020: Ridesharing Using Adaptive Waiting Time
Friday, August 7, 2020 at 10:00 a.m.
You are invited to attend the following M.A.Sc. (Quality Systems Engineering) thesis examination.
Dr. N. Bouguila, Chair
Dr. J.Y. Yu, Supervisor
Dr. A. Awasthi, CIISE Examiner
Dr. D. Terekhov, External Examiner (MIAE)
The culture of sharing by the advances in communication technologies has entered a new era, and ever since, sharing instead of ownership has been sharply increasing in individuals’ behaviors. Particularly in transportation, concepts of sharing a ride in either carpooling or ridesharing have been adopting for 70 years. During the past fifteen years, the revolution in communication devices has formed the online version of ridesharing that responds to transportation needs shortly. Ridesharing is considered to be a strategy to mitigate congestion and air pollution by increasing the occupancy rate of vehicles in the road network. An online ridesharing framework is an end-to- end framework that manages the request and matches the passengers to accomplish their rides together.
In this thesis, we studied the online ridesharing problem and proposed an end-to-end framework to handle the passengers. In our end-to-end framework, we design the objective with respect to the passenger’s perspective. We assume that all the passengers tend to share their ride to reduce their transportation costs using vehicles. When two passengers get matched to accomplish their ride together, they accept a deviation from their shortest path to make sharing possible. To minimize the information provided by passengers, we define scheduling flexibility using a system-wide fixed flexibility factor , which indicates the tolerable increase in travel duration, proportional to the shortest path duration. For a trip, scheduling flexibility is the amount of time that the system has to divide between the detour from the shortest path and the waiting time to find a proper match. To split the scheduling flexibility between the detour and the waiting time, we introduce the concept of adaptive waiting time, which is the key enabler of our framework to provide a quality match for passengers. For a passenger, the optimal waiting time with respect to minimizes the expected travel cost. In this work, we use future demand to calculate the expected travel cost.
We carefully design a simulation to observe the ability of our framework to match the passengers. The trip cost and trip duration are approximated using Gradient boosting trees, and we simplify the NYC road network as a grid network. The proposed approach works for 24 hours to handle 356049 ride requests on a rectangle with an area equal to 44 km2. We analyze several metrics to indicate the quality of the matching process. The simulation results show that by using our approach, 75.2% of the passengers can share their ride by increasing the trip duration for 4.334 minutes on average, and it leads to reducing the total cost by 12% and reducing the total traveled distance by 14.29%.