Gast ::
| Hauptseite > Publikationsdatenbank > Beyond the prompt for pedestrian traffic and crowd behavior management: PedAgent, a Large Language Model empowered Agent with an integrated simulating-analyzing-optimizing functional chain |
| Hauptseite > Publikationsdatenbank > Beyond the prompt for pedestrian traffic and crowd behavior management: PedAgent, a Large Language Model empowered Agent with an integrated simulating-analyzing-optimizing functional chain |
| Journal Article | FZJ-2026-04936 |
; ; ; ; ; ; ;
2027
Elsevier
Amsterdam [u.a.]
This record in other databases:
Please use a persistent id in citations: doi:10.1016/j.tbs.2026.101414
Abstract: Pedestrian walking constitutes the most fundamental mode of urban mobility, yet the recurrent occurrence of pedestrian traffic-related safety accidents, e.g., crowd crushes, underscores the urgent need to enhance the management and safeguarding of pedestrian facilities, particularly those characterized by high-density gatherings (e.g., urban subway stations during peak hours). Although pedestrian flow simulation models and empirically validated management policies exist, they can barely support the full-process requirements of pedestrian traffic and crowd safety management and often operate in isolation rather than forming an integrated solution, e.g, simulation software can generate individual-level trajectories but lacks the capability to derive case-sensitive crowd operation improvement strategies building upon these microscopic information, whereas empirically validated policies are typically tested under fixed boundary conditions, rendering their effectiveness uncertain in real-world pedestrian management scenarios that are highly dynamic and variable. To address these limitations, this study introduces PedAgent v1.0, a dialogue-based intelligent agent for pedestrian traffic management that integrates the reasoning capabilities of the Large Language Model (LLM)with a Retrieval-Augmented Generation (RAG) based pedestrian knowledge base and a custom developed Pedestrian Traffic Kit (PTK), thereby establishing an end-to-end functional chain of simulating-analyzing-optimizing through dedicated prompt engineering, enabling the seamless translation of users’ natural language requests into executable simulation tasks and the generation of feasible, simulation-supported, and qualitatively validated safety management strategies. Through extensive case studies, we demonstrate: i) diverse user-PedAgent interaction patterns across multiple conflict hot-zones within pedestrian facilities, highlighting PedAgent’s adaptability and generalizability, and ii)the indispensable role of PTK in enabling domain-specific simulation and analysis within the framework. Technically, PedAgent constitutes the first LLM based agent within the domain of pedestrian traffic research, significantly improving the efficiency of simulation assisted policy formulation by supporting an end-to-end management workflow that encompasses scenario specification, parameter configuration, model execution, simulation output analysis, and policy recommendation. By integrating these traditionally disjoint components, PedAgent bridges a longstanding gap between research oriented theoretical findings and practical policy formulation, thereby enhancing the intelligence and safety of pedestrian facility management.
; Clarivate Analytics Master Journal List ; Current Contents - Social and Behavioral Sciences ; Essential Science Indicators ; IF >= 5 ; JCR ; SCOPUS ; Social Sciences Citation Index
|
The record appears in these collections: |