Composite Travel Generative Adversarial Networks for Tabular and Sequential Population Synthesis

Badu-Marfo, Godwin; Farooq, Bilal; Patterson, Zachary

doi:10.1109/tits.2022.3168232

Cited by 14 publications

(4 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This framework is flexible enough to incorporate various neural network architectures. Badu-Marfo et al ( 2022) used a GAN model in which people's demographics and trajectories were learned and synthesized with two model components [19]. The model was assessed with the trip length distribution and route segment usage.…”

Section: Machine-learning Based Methodsmentioning

confidence: 99%

Learning Daily Human Mobility with a Transformer-Based Model

Wang,

Osaragi

2024

IJGI

View full text Add to dashboard Cite

The generation and prediction of daily human mobility patterns have raised significant interest in many scientific disciplines. Using various data sources, previous studies have examined several deep learning frameworks, such as the RNN and GAN, to synthesize human movements. Transformer models have been used frequently for image analysis and language processing, while the applications of these models on human mobility are limited. In this study, we construct a transformer model, including a self-attention-based embedding component and a Generative Pre-trained Transformer component, to learn daily movements. The embedding component takes regional attributes as input and learns regional relationships to output vector representations for locations, enabling the second component to generate different mobility patterns for various scenarios. The proposed model shows satisfactory performance for generating and predicting human mobilities, superior to a Long Short-Term Memory model in terms of several aggregated statistics and sequential characteristics. Further examination indicates that the proposed model learned the spatial structure and the temporal relationship of human mobility, which generally agrees with our empirical analysis. This observation suggests that the transformer framework can be a promising model for learning and understanding human movements.

show abstract

Section: Machine-learning Based Methodsmentioning

confidence: 99%

Learning Daily Human Mobility with a Transformer-Based Model

Wang,

Osaragi

2024

IJGI

View full text Add to dashboard Cite

show abstract

“…It is also used to measure the similarity between the group distribution of the original dataset and the group distribution of the generated trajectories. The KL divergences of the following distributions between the simulated dataset and the original dataset were evaluated: the visit frequency distribution (VFD) for the entire dataset and the VFD of the top‐100 visits (VFD‐top100) (Kulkarni et al., 2018; Ouyang et al., 2018), the distribution (LD) of the lengths of the trajectories (Badu‐Marfo et al., 2022), and the distribution of feature nodes within a trajectory group (VFD‐Cluster). Note that the VFD‐Cluster metric is defined based on the Group Distribution Loss to measure the similarity of the generated trajectory clusters and the corresponding clusters in the original dataset.…”

Section: Evaluationsmentioning

confidence: 99%

“…One viable solution is to synthesize pedestrian trajectories by accounting for pedestrian movement characteristics. Although some data simulation techniques such as generative adversarial networks (GANs) (Goodfellow et al., 2020), variational autoencoder (VAE) (Kingma & Welling, 2013), and generative adversarial imitation learning (GAIL) (Ho & Ermon, 2016) have been used to generate outdoor semantic trajectory datasets (Badu‐Marfo et al., 2022; Cao & Li, 2021; Choi et al., 2021; Jeong et al., 2021; Ouyang et al., 2018), these methods cannot be readily used for indoor trajectory synthetization due to significant differences in the movement patterns of vehicles in road networks and pedestrians in indoor spaces. These outdoor trajectory simulation models often neglect the importance of maintaining topological correctness and semantic consistency, thereby limiting their application to indoor environments.…”

Section: Introductionmentioning

confidence: 99%

A deep pedestrian trajectory generator for complex indoor environments

He,

Zhang,

Wang

et al. 2024

Transactions in GIS

View full text Add to dashboard Cite

Pedestrian trajectory data, which can be used to mine pedestrian motion patterns or to model pedestrian dynamics, is crucial for indoor location‐based service studies and applications. However, researchers are faced with the challenges of data shortage and privacy restrictions when using pedestrian trajectory data. We present an Indoor Pedestrian Trajectory Generator (IPTG), which is a novel deep learning model to synthesize pedestrian trajectory data. IPTG first produces feature sequences that encode the spatial–temporal and semantic features of the walking process and then interpolates them into complete trajectories using A* and perturbation algorithms. IPTG has specially designed loss functions that preserve topological constraints and semantic characteristics. Incorporating the prior knowledge of environment constraints and pedestrian walking patterns, the IPTG model is capable of generating topologically and logically sound indoor pedestrian trajectories. We evaluated the synthesized trajectories based on multiple metrics and examined the generated trajectories qualitatively. The results show that IPTG outperforms several baselines, demonstrating its ability to generate semantically meaningful and spatiotemporally coherent trajectories.

show abstract

“…Many generalized approaches to capture nonlinearity in DCMs using ANNs have been explored in the past several years [21,22,23].…”

Section: Generalized Approach To Capture Nonlinearity Using Annsmentioning

confidence: 99%