Latent Demand Modeling for Transportation Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for modeling demand in transportation systems, such as public bus or train systems, face challenges in accurately correlating passenger data with time and location due to irregular route operations and long intervals between vehicle services, leading to incomplete demand analysis as a function of time of day and day of week.
Innovation Solution
A method and system using Bayesian latent modeling techniques to determine boarding count models, geographic and time-specific generalized boarding models, approximated uniform arrival models, and probability of no demand models, which generate reports on instantaneous demand and probability of no demand, allowing for informed decisions on route optimization and resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If automated passenger counters are used to measure passengers boarding or alighting, then passenger data collection is enabled, but data cannot be accurately correlated to time and place due to irregular route operations and long intervals between vehicle services
Solution Approach 1:
The patent introduces an intermediary computational model that processes automated passenger counter data along with route schedule information and time intervals. This intermediary layer reconciles the mismatch between irregular vehicle arrivals and regular time intervals, enabling accurate correlation of passenger data to specific times and locations without requiring changes to the existing passenger counting infrastructure.
Solution Approach 2:
The patent replaces the mechanical assumption of regular vehicle intervals with a computational model that uses actual schedule data and observed time intervals. Instead of assuming uniform time periods between vehicles, the system substitutes this with dynamic interval calculations based on real route operations, thereby resolving the time correlation problem.
2Productivity
If vehicle service intervals are extended to reduce operating costs, then more passengers may board due to longer waiting time accumulation, but demand analysis becomes incomplete because the increase is confounded by interval length rather than reflecting true population demand
Solution Approach 1:
The patent changes the parameter used for demand measurement from raw passenger counts to standardized demand metrics that account for time interval variations. By introducing normalization factors based on vehicle service intervals and waiting time accumulation, the system transforms incomplete demand measurements into accurate representations of true population demand, enabling valid comparisons across different service frequencies.
3Productivity
If stops are skipped when no passengers are observed boarding, then operational efficiency improves, but complete demand information is lost as these stops are ignored completely with no registration in the automated passenger counter
Solution Approach 1:
The patent applies preliminary action by using route schedule information and historical patterns to predict which stops are likely to have passengers before vehicles arrive. This allows the system to prepare for potential passenger boarding events at stops that may not have been observed in previous intervals, ensuring that demand data is captured even when no passengers are immediately visible or when vehicles skip stops under certain conditions.
Data Source
AI summary
A method and system for identifying demand in a transportation system. A boarding count model is determined based upon passenger arrival information, and a geographic and time-specific generalized boarding model is determined based upon the boarding count model as well as information related to a plurality of stops on a route in the transportation system. For each of the plurality of stops, an approximated uniform arrival model is determined based upon the generalized arrival model and a time period between arriving vehicles at a specific stop, an instantaneous demand model is determined based upon the uniform arrival model, and a probability of no demand model is determined based upon the uniform arrival model. A report including the instantaneous demand and the probability of no demand determined can be generated. Based upon the report, various operational parameters for the transportation system can be manually or automatically adjusted.