Mobility Prediction Using Quantum-Inspired Density Matrix
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods using Call Detail Records (CDRs) lack a standardized and reliable procedure for accurately measuring and predicting human and non-human mobility, failing to distinguish between human and machine-to-machine communications, and do not effectively classify mobility types or predict future mobility patterns, which is crucial for optimizing transportation and social distancing efforts.
Innovation Solution
A scalable and private-by-design mobility prediction process that collects metadata from cell sites, represents events as vectors, constructs a density matrix, diagonalizes it to obtain orthonormal eigenvectors, and extrapolates coefficients to predict future or past events, while filtering devices to distinguish human from non-human mobility and classifying mobility types based on entropy measures and virtual agent simulations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If CDR metadata is used for billing purposes with dense cell tower networks, then connectivity coverage is improved, but the system cannot accurately measure human mobility patterns
Solution Approach 1:
The patent extracts only the necessary mobility-related metadata from CDR records (cell tower identification, timestamp, device identification) while filtering out billing-specific data. This extraction transforms the billing-oriented CDR system into a mobility measurement system by taking out and utilizing only the relevant spatial-temporal components.
Solution Approach 2:
The patent makes the existing CDR infrastructure multi-functional by enabling it to serve both its original billing purpose and a new mobility measurement purpose. The same cell tower network and CDR data collection mechanism are utilized for both telecommunication billing and human mobility pattern analysis, eliminating the need for separate dedicated measurement infrastructure.
2Adaptability or versatility
If case-by-case filter methods are applied to CDR data, then specific data sets can be analyzed, but the procedure lacks reliability and replicability
Solution Approach 1:
The patent segments the CDR data processing into distinct, standardized filtering stages: initial data extraction, device identification filtering, mobility event classification, and pattern recognition. Each segment performs a specific function with defined input-output requirements, making the overall procedure systematic, reliable, and replicable across different data sets.
Solution Approach 2:
The patent employs parameter-based filtering where mobility patterns are identified by changing and analyzing specific parameters such as displacement distance, time intervals between events, velocity calculations, and acceleration patterns. By defining objective parameter thresholds and transformation rules, the method achieves consistent, replicable results across different regions and time periods.
3Measurement precision
If all telecommunication events are recorded in CDR, then billing accuracy is improved, but human mobility data cannot be distinguished from machine-to-machine communications
Solution Approach 1:
The patent extracts and isolates mobility-relevant information from the complete CDR record set by identifying specific patterns characteristic of human mobility (spatial displacement, temporal patterns, velocity changes). This extraction separates human mobility signals from the broader telecommunication event data, filtering out machine-to-machine communications that lack these mobility characteristics.
Solution Approach 2:
The patent introduces an intermediary analysis layer that processes raw CDR events through mobility-specific filtering criteria. This intermediary layer acts as a mediator between the complete telecommunication record and the final mobility pattern output, applying transformation rules and filtering algorithms that convert billing-oriented data into mobility-oriented insights while discarding irrelevant M2M communication patterns.
4Area of stationary object
If cell towers are distributed based on population density, then connectivity is improved, but the tower distribution does not respond to needs for gathering mobility information
Solution Approach 1:
The patent compensates for the limitations of existing tower distribution by introducing temporal dimensionality to the data analysis. While spatial resolution is constrained by tower locations, the system achieves high mobility measurement precision by analyzing temporal patterns, time intervals, and sequential position changes across multiple time points, effectively adding a time dimension that compensates for spatial coarseness.
Solution Approach 2:
The patent applies preliminary filtering and preprocessing actions to the raw CDR data that enhance mobility measurement precision regardless of tower density. By pre-processing the data with mobility-specific filtering criteria, velocity calculations, and pattern recognition algorithms, the system extracts high-precision mobility information from the available tower distribution data before aggregation and analysis.
Data Source
AI summary
The invention relates to a scalable and private-by-design mobility prediction process permitting to predict stays, road traffic, micro-mobility and trips segmented by mode of transportation inferred from a set of cell sites and/or towers, each comprising an antenna, distributed in a certain region and operating for a certain period of time giving support to a certain number of devices, wherein at each site, metadata of every telecommunication events concurring at its coverage area along this period of time are collected, and all the data is centralized in a single or multiple set of CDR raw metadata, and permitting for a continuous localization of a device by extrapolating any trajectory in space and time of any device, including projections into the future and projections into areas with no infrastructure, characterized in that the prediction process comprises a first step consisting in representing the events registered at CDR as vectors which components are the amplitude of probabilities of occupation for each location and time, a second step of constructing a density matrix from the vectors set obtained in the first step, a third step of diagonalizing the density matrix obtaining a basis of orthonormal vectors, assuming that basis are also the eigenvectors of an underlying pseudo-Hamiltonian, a third step of obtaining, for any incomplete vector, the coefficients of this vector in the basis of the pseudo-Hamiltonian limited in the subspace where this vector is defined, a fourth step of extrapolating these coefficients to the subspace where the vector is not defined, including future time or hypothetical networks, and a fifth step of obtaining the probabilities of occupation as the square of the obtained amplitudes for gap filling, for prediction into the future or past, or for prediction in new areas.