AutoNavi officially releases the Automotive Travel AI Agent, capable of understanding ambiguous semantics and performing precise searches

Mars Finance message, April 23 — AutoNavi officially released an Automotive Travel AI Agent. Unlike traditional “navigation + voice assistant,” this system no longer simply executes user commands; instead, it can understand user needs and complete end-to-end trip planning. As introduced, the core capabilities of AutoNavi’s Automotive Travel AI Agent can be summarized as: fuzzy-semantic precise search, chaining complex itineraries into one-sentence links, dynamic spatial route reasoning, and multi-round, dialogue-style route editing. For example, if a user says: “First go to an amusement park with Transformers, then go to the World Trade Plaza to buy a little bear cake, then go to a big park with a slide, then have an ice cream, and then go home.” The itinerary planned by AutoNavi’s Automotive Travel AI Agent is: “Beijing Universal Resort — Dole Day (World Trade Plaza South Street Store) — Chaoyang Park — iGELATO Italian Handcrafted Ice Cream (Chaoyang Park Store) — Home.”

In terms of the technical architecture, AutoNavi’s Automotive Travel AI Agent adopts a “language brain + spatial brain” dual-engine design. The large language model (with the Qianwen model as the underlying foundation) only handles understanding user intent and generating query requests. All geographic information (POI, routes, charging-station status, etc.) must be obtained from AutoNavi’s database and verified by the spatial brain to ensure results are truly reliable and accurate, preventing “hallucinations” from affecting travel safety.