Mobility-Activity Processing Module for Demand-Responsive Dispatching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems fail to scale mobility trace data to cover entire city populations and efficiently facilitate on-demand transportation applications, as they are limited to collaborative users who share their data, and lack the ability to explain spatio-temporal variations in individual movements and demographics.
Innovation Solution
A method and system that generate clusters based on mobility-activity patterns of collaborative individuals, assign non-collaborative individuals to these clusters using combinatorial optimization, and update mobility-activity models with an approximation function learning from observed OD demand, creating a demand-responsive transportation system that allocates vehicles efficiently based on real-time data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If mobility trace data is collected only from collaborative users who share their data, then data privacy is protected, but the coverage and reliability of mobility information is insufficient for entire city populations
Solution Approach 1:
The patent introduces mobility-activity models as intermediaries that bridge collaborative and non-collaborative users. These models infer mobility patterns of non-collaborative users based on data from collaborative users, enabling city-wide coverage while preserving individual privacy. The models act as mediators that translate limited collaborative data into comprehensive population-level mobility information.
Solution Approach 2:
The patent creates synthetic mobility traces for non-collaborative users by copying and adapting patterns from collaborative users through mobility-activity models. Instead of requiring actual data from every individual, the system generates representative mobility copies that capture essential movement patterns, thereby scaling coverage to entire city populations without compromising privacy.
2Measurement precision
If traditional classification models based on historical data are used, then individual activity patterns can be identified, but the models cannot scale to cover entire city populations
Solution Approach 1:
The patent segments the population into collaborative and non-collaborative users, and further segments mobility patterns into distinct activity clusters (residential, commercial, industrial, etc.). This segmentation allows the system to maintain high measurement precision for individual patterns while scaling to city-wide coverage by processing segments independently and aggregating results.
Solution Approach 2:
The patent uses lightweight, computationally efficient mobility-activity models that can be rapidly instantiated and discarded for different user segments. These models are designed to be resource-efficient, allowing the system to create and manage numerous individualized mobility representations across entire city populations without prohibitive computational costs.
3Measurement precision
If combinatorial optimization is used to assign non-collaborative individuals to clusters, then allocation accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent implements dynamic cluster assignments where non-collaborative users are assigned to mobility-activity clusters based on real-time optimization criteria. The system dynamically adjusts cluster memberships and re-allocates users as new data becomes available, balancing computational complexity with assignment accuracy through adaptive, iterative optimization rather than static pre-computation.
Solution Approach 2:
The patent incorporates feedback mechanisms where the performance of cluster assignments is continuously evaluated and used to refine future assignments. The system learns from observed OD demand and mobility trace data, adjusting its optimization criteria and model parameters to improve accuracy while managing computational resources efficiently through experience-based refinement.
4Measurement precision
If an approximation function learns from observed OD demand and mobility trace data, then model accuracy improves over time, but data processing requirements increase
Solution Approach 1:
The patent performs preliminary processing of mobility trace data to extract essential patterns and features before main model training. By pre-processing and organizing data into structured formats during off-peak times, the system reduces the computational burden during real-time operations, thereby lowering energy requirements while maintaining model accuracy improvement through continuous learning.
Data Source
AI summary
A method for providing a demand-responsive transportation system includes receiving mobility trace data of collaborative individuals. Clusters of individuals are generated and mobility-activity models for the clusters are defined. Non-collaborative individuals are assigned to the clusters using a combinatorial optimization problem. An Origin-Destination (OD) demand is determined from the clusters. Non-collaborative individuals are re-allocated to different ones of the clusters using an approximation function that learns from an observed OD and the mobility trace data. The mobility-activity models are trained based on the re-allocation of the non-collaborative individuals to different ones of the clusters. An OD database (OD-DB) is maintained to be queried with a geographic location and time so as to receive information from the OD-DB about the current OD demand for the geographic location and time. Control actions are issued to vehicles in a fleet of the transportation system based thereon.