Dynamic Electric Dial-a-Ride on a Fixed Circuit
Dynamic Electric Dial-a-Ride on a Fixed Circuit
Rotem Zohar,M.Sc. student at the department of Industrial Advisor: Dr. Mor Kaspi
Abstract:
Advancements in autonomous vehicle technology are revolutionizing urban mobility. Shared mobility services utilizing electric autonomous vehicles offer significant benefits, including cost efficiency, enhanced safety, and improved service reliability. However, regulatory constraints currently restrict these services to controlled environments, such as campuses, airports, and predefined urban routes. These services often operate on fixed circular routes, which, from an operational planning perspective, present both unique challenges and opportunities. In this study, we present and analyze the Dynamic Electric Dial-a-Ride Problem on Fixed Circuits (eDARP-FC). This problem involves making dynamic decisions about when and where vehicles should stop to pick up and drop off passengers, as well as determining optimal recharging times and durations. These decisions must comply with various operational constraints, including vehicle capacity, battery management, and service regulations. The objective is to minimize rejected requests and operational costs. We examined two approaches to address the problem: independent and sequence-based. The independent approach handles request assignments and charging decisions separately, using simple, scalable, and easy-to-implement rule-based mechanisms. However, this lack of synchronization often leads to inefficiencies, as charging decisions have long-term impacts on vehicle availability, emphasizing the need for a more integrated strategy. In contrast, the sequence-based approach leverages the structured nature of fixed circuits by organizing requests into feasible sequences. This coordinated strategy enables the proactive rejection of unserviceable requests, effectively balancing system constraints and reducing the number of abandoned requests by approximately 20%. Building on these insights, we applied reinforcement learning (RL) to dynamically refine the selection of operational rules rather than directly managing low-level assignments and charging decisions. The RL model learns to adaptively switch between predefined rules for request assignment and charging, considering critical factors such as low-battery vehicles and specific time-sensitive conditions within the system. Given the cascading effects of charging decisions, ensuring smooth transitions between rules is crucial for maintaining system stability. To address this, we formulated a dedicated optimization problem. Preliminary results suggest that our RL approach has considerable potential to improve overall system performance.
Bio:
Rotem Zohar is a M.Sc. student in the department of industrial engineering in Tel Aviv university. specializing in operations research and intelligent transportation systems. Her research focuses on optimizing on-demand transportation services using electric autonomous shuttles operating on designated circuits. Rotem has presented work at the European Conference on Operational Research (EURO) and the Transportation Science and Logistics (TSL) Workshop. In addition to her research, she serves as a Teaching Assistant for a simulation course, helping students apply theoretical concepts to real-world problems. Rotem holds a BSc in Industrial Engineering and Management from Tel Aviv University and acquired technical expertise while serving in the Israeli Defense Forces' 8200 unit.
