Travel survey simulation method and device, electronic equipment and readable storage medium

CN122264079APending Publication Date: 2026-06-23TSINGHUA UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
TSINGHUA UNIVERSITY
Filing Date
2026-02-03
Publication Date
2026-06-23

AI Technical Summary

Technical Problem

Existing travel survey methods lack logical reasoning support, logical consistency, and cross-domain generalization ability, and cannot meet the needs of modern cities for high-precision, real-time travel data.

Method used

By acquiring user profiles and date conditions, a pre-trained reasoning-enhanced language model is used. A trip chain reasoning sample set is constructed by combining real travel survey data. The pre-trained large language model is then fine-tuned with basic and enhanced supervision to generate travel reasoning chains and structured travel tables. A two-stage fine-tuning mechanism is adopted to improve the model's generalization ability.

Benefits of technology

The generated travel data possesses the diversity of natural language expression and the explicitness of travel decision-making logic, significantly improving spatiotemporal consistency and behavioral interpretability. It solves the problem of insufficient generalization ability in cross-city migration and cold start scenarios, and realizes low-cost, highly reliable, and interpretable travel data generation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122264079A_ABST
    Figure CN122264079A_ABST
Patent Text Reader

Abstract

The application provides a travel survey simulation method and device, electronic equipment and readable storage medium, wherein the method comprises: obtaining a user portrait and a date condition of a target individual; the user portrait comprises at least one of age, gender, occupation, family structure, income level and vehicle ownership; based on a pre-trained inference enhanced language model, a travel inference chain and a structured travel table are generated according to the user portrait and the date condition; the inference enhanced language model is obtained by performing basic supervision fine-tuning and enhanced supervision fine-tuning on a pre-trained large language model by using a trip chain inference sample set constructed by using real travel survey data. The method significantly improves the spatio-temporal consistency and behavior explainability, effectively solves the problem of insufficient generalization ability of traditional methods in cross-city migration and cold start scene, and realizes low-cost, high-credibility, explainable and easy-to-deploy travel data generation.
Need to check novelty before this filing date? Find Prior Art