摘要：AbstractEnhancing bus public transportation implies making buses more attractive for passengers while improving service performance by reducing dwell time and increasing passenger flow. Thus, future buses must be designed with a focus on passengers’ needs, requirements and preferences, and their physical and psychological abilities to take advantage of new design concepts. Currently, manufacturers’ knowledge relies on experiences with buses already in service and statistical data about passengers when designing new bus concepts. However, evaluating these concepts from passenger's perspective is difficult at an early stage of the design process without access to a real vehicle. Including agent-based simulation in the design process is an inexpensive and efficient procedure to analyze new concepts (wheel-well position, number of doors, etc.) in relation to the expectations and needs of current and future passengers. In our approach, passengers are modelled as autonomous agents. Agents have the ability to move within the bus, interact with it and make complex decisions according to their preferences. Preferences are modelled as prioritized lists of goals. Since passenger preferences change depending on the bus occupancy, four preference models have been considered for <25%, <50%, <75% and >75% occupancy. Goals can be defined from observations of real passengers or synthetically (to model future passengers). The agent model implements a decision-making algorithm that quantifies the attractiveness of each available seat, standing location and door. The algorithm returns the target whose characteristics better fit passenger's preferences, considering the occupancy onboard and additional factors. A case-study that compares two bus layouts (three vs. four doors) is presented. Both layouts are evaluated in terms of passenger flow and dwell time. Results show the correlation between passengers’ decisions and layout design constraints, demonstrating that agent-based simulation can be effectively used in passenger-centered design methodologies.
