LLMs used to generate realistic urban travel patterns from GPS data

Researchers have developed a new method called HTP for generating realistic urban trajectories using large language models (LLMs). This approach first generates travel patterns and then synthesizes GPS points, addressing limitations of existing methods that struggle with capturing travel patterns and generating fixed-length trajectories. HTP utilizes a trajectory-specific residual quantization variational autoencoder to convert GPS data into travel pattern tokens, which are then integrated into an LLM's vocabulary for sequence generation. Experiments show HTP significantly outperforms baseline methods in generation quality. AI

IMPACT This method could improve urban planning and smart city applications by enabling the generation of realistic, privacy-preserving trajectory data.

RANK_REASON The cluster contains a research paper detailing a new method for trajectory generation using LLMs. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• HTP
• LLMs
• RQ-VAE

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →