An agent-based simulation modeling framework for Mobility-as-a-Service (MaaS)

This study develops an agent-based intelligent Mobility-as-a-Service (MaaS) simulation model consisting of three types of agents (i.e., MaaS fleet unit, travelers, and central intelligent mobility assignment module) to assess mobility service assignment processes balancing conflicting entities (e.g., demand and supply) within the MaaS ecosystem. The study follows a two-level optimization method (i.e., lower, and upper levels). It employs artificial intelligence and multicriteria decision making to solve two-dimensional mobility assignment problems (demand vs. supply). The novelty of this assignment process is that it implements an intelligence module accounting for the past system performance to make a future assignment decision in favor of both sides of the operation. This study processes 24-hour trip requests extracted from Nova Scotia Travel Activity (NovaTRAC) survey data in real-time. Two scenarios (1-D and 2-D assignments) are compared using the cost criteria, such as total waiting time, empty time, and idle time. The 1-D scenario refers to mobility assignment that emphasizes the demand side only, and the 2-D scenario formulates mobility assignment by balancing the demand and supply sides of the MaaS ecosystem. Experimental results indicate that a MaaS fleet of 350 units is the most balanced fleet in both scenarios. However, the 2-D optimization method reduces the overall supply cost by 25%. Moreover, 2-D operation demonstrates a higher fleet utilization over a considerable period, whereas the 1-D design guarantees a higher fleet utilization only during the peak period. Results of this study provide us with a benchmark for assessing more complex MaaS operation scenarios which will further aid in advancing the operational MaaS ecosystem. Our findings can help policymakers implement cost-effective MaaS solutions supporting sustainable urban mobility and SDG 13: Climate Action.