OPENPATH's supervisor-specialist AI handles personalized multi-stop urban trips with accessibility
LLM agents + classical algorithms tackle wheelchair accessibility and multi-stop planning in NYC.
OPENPATH introduces a novel supervisor-specialist architecture for urban trip planning. LLM agents serve as supervisors, parsing natural-language input, classifying request intent, and orchestrating execution. They hand off to specialist classical algorithms that perform route optimization over curated mobility and accessibility data. This division of labor ensures the system honors heterogeneous user preferences—like multiple stops with varying priorities—and enforces strict wheelchair accessibility requirements when requested. The design addresses a key limitation of current systems, which optimize only for travel time and cost, ignoring personalized needs and accessibility.
Beyond individual trip planning, OPENPATH doubles as a measurement instrument for city-scale accessibility analysis. Applied to New York City, the system identifies substantial gaps in ADA infrastructure and quantifies how those gaps affect job accessibility for wheelchair users. For example, it reveals that certain neighborhoods have significantly fewer accessible routes, limiting employment opportunities. The study demonstrates how LLM agentic frameworks can support equitable transportation analysis and personalized route planning at scale, offering a transparent way to assess and improve urban infrastructure for all travelers.
- Uses LLM agents as supervisors to parse natural language and orchestrate execution, while classical algorithms optimize multi-stop routes.
- Supports personalized itineraries with heterogeneous preferences and enforces end-to-end wheelchair accessibility requirements.
- Applied to NYC, revealing substantial ADA infrastructure gaps and quantifying job accessibility impacts for wheelchair users.
Why It Matters
Brings personalized, accessible multi-stop trip planning to real cities, highlighting infrastructure inequities for wheelchair users.