跳转至

深度活动模型：用于人类出行模式合成的生成式方法

文章背景与核心概要

人类出行模式建模对于交通运输、城市规划和公共卫生至关重要，但传统方法面临着严峻的挑战。常规的深度学习模型往往忽略了家庭成员之间的语义相互依赖性，并且严重依赖受限的GPS数据流；而传统基于活动的模型（ABM）则需要严格的假设和昂贵的数据收集成本，导致其极难在数据稀缺的地区部署。

为了解决这些难题，本文作者引入了深度活动模型（Deep Activity Model），这是一个生成式的Transformer框架：它利用社会人口统计和家庭属性来合成每日活动链；采用位置模块来分配空间区域，从而生成完整且逼真的日常出行轨迹。该模型通过开源的家庭出行调查数据进行训练，并能够通过微调有效地适应加利福尼亚州、华盛顿州和墨西哥城等不同地区。

在洛杉矶县的大规模交通模拟评估中，该框架的表现与既有基线（如SCAG ABM）及现实世界的传感器观测数据（Caltrans PeMS）高度吻合，这充分证明了基于学习的合成出行在区域规划中的巨大实用价值。

📌 Summary

Human mobility modeling is vital for transportation, urban planning, and public health, but traditional methods face steep hurdles. Conventional deep learning models often ignore semantic interdependencies among households and rely on limited GPS streams, while traditional activity-based models (ABMs) require rigid assumptions and costly data collection, making them difficult to deploy in data-sparse regions.

To resolve these challenges, the authors introduce the Deep Activity Model, a generative Transformer framework that: * Leverages socio-demographic and household attributes to synthesize daily activity chains. * Employs a location module to assign spatial zones, generating complete, realistic daily mobility trajectories. * Trains on open-source household travel survey data and adapts effectively via fine-tuning to diverse regions like California, Washington, and Mexico City.

When evaluated via large-scale traffic simulations in Los Angeles County, the proposed framework closely mirrors established baselines (such as the SCAG ABM) and real-world sensor observations (Caltrans PeMS), proving the high practical utility of learning-based synthetic mobility for regional planning.

📋 Abstract

Human mobility plays a crucial role in transportation, urban planning, and public health, but current approaches face important limitations. Existing deep learning models tend to overlook the semantic interdependencies among activities and households and rely on restricted GPS data, while activity-based models depend on rigid assumptions and extensive data, making them costly and difficult to adapt to new regions, especially those with limited conventional travel data. To address these limitations, we propose a generative Transformer model that uses socio-demographic and household attributes to synthesize daily activity chains, with a location module assigning spatial zones to produce complete daily trajectories. Trained on open-source and widely available household travel survey data and then fine-tuned with local data, the model captures national activity patterns and transfers effectively to California, Washington, and Mexico City. This approach offers potential for advancing synthetic human mobility modeling and provides urban planners and policymakers with improved tools for simulating transportation systems and supporting decisions in urban development and public health. Its practical utility is demonstrated through large-scale traffic simulations in Los Angeles County. Compared to the SCAG Activity-Based Model (ABM), the proposed method produces consistent spatial demand patterns, achieving an activity location cosine similarity of 0.997 and a network-level vehicle-miles-traveled Mean Absolute Percentage Error (MAPE) of 4.97%. Compared to real-world observations from Caltrans PeMS, the simulated traffic achieves MAPE of 5.85% for traffic volume and 4.36% for speed on California's I-405 corridor, presenting the practicality of learning-based synthetic mobility for regional simulation and planning.