LLM-enhanced semantic embeddings reduce bus bunching via reinforcement learning

Researchers have developed a novel method to mitigate bus bunching using reinforcement learning enhanced by semantic stop embeddings. This approach leverages a large language model (LLM) offline to create rich representations of bus stops, incorporating physical attributes, contextual information, and historical data. These embeddings are then integrated into a deep Q-learning controller, significantly reducing headway variability, bunching events, and passenger waiting times compared to traditional methods. The study also explored the transferability of these policies across different bus routes, finding that while zero-shot transfer is limited, fine-tuning can accelerate learning and improve performance. AI

IMPACT This research demonstrates how LLMs can enrich state representations for reinforcement learning agents, potentially improving control systems in various domains beyond transit.

RANK_REASON The cluster contains an academic paper detailing a new methodology. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• Daganzo
• reinforcement learning
• Semantic Stop Embedding

• paper
• product

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