AutoNavi Launches AI Agent for Driving; Li Auto First to Support Natural Language Interaction and Proactive Trip Planning

On April 22, AutoNavi officially launched its AI Agent for automotive travel, a system capable of understanding users’ natural language expressions and proactively planning complete itineraries. Li Auto is the first automaker to integrate this AI Agent, with more car brands expected to join the collaboration soon. The AI Agent can handle ambiguous or colloquial instructions. For example, if a user says, “Go to the amusement park with Transformers to get ice cream, then head to SKP-SOLANA to buy a teddy bear cake, go to a big park to play on the slides, and grab another ice cream before heading home,” the system automatically identifies specific locations (e.g., Universal Beijing Resort, Tous Les Jours at SKP-SOLANA, Chaoyang Park, etc.), logically sequences them, and plans an optimized route. In complex scenarios—such as when a user requests, “Go sailing on a circular lake and find a highly rated local Shanghainese restaurant nearby with parking”—the AI Agent uses keyword-based reasoning to infer the target location (e.g., Dishui Lake), filters restaurants meeting the criteria, and optimizes the route to avoid unnecessary detours. For electric vehicle trips, the system factors in current battery level, total trip distance, and charger availability to recommend ultra-fast charging stations at service areas, while coordinating charging and meal times to maximize efficiency. When faced with multi-turn dynamic adjustments—such as repeated changes to route preferences—the AI Agent recognizes shifts in user sentiment and intent, prioritizing smooth-flowing routes over merely the shortest ones. The system also proactively responds based on context: it automatically inserts charging stops when battery levels run low; pre-calculates detours upon detecting upcoming traffic congestion, accidents, or roadwork; and, upon recognizing semantic cues like “family day,” recommends child-friendly attractions, family-oriented restaurants, or malls equipped with baby-care facilities. Through every trip, the AI Agent continuously learns user preferences—such as a tendency toward main roads versus side streets, or willingness to spend extra time avoiding traffic jams. To ensure safety and accuracy, the system employs a dual-engine architecture combining a “Language Brain” and a “Spatial Brain”: the language model solely interprets intent and generates queries, while all geographic information—including POIs, routes, and chargers—is verified against AutoNavi’s authoritative database before output, eliminating the risk of hallucinations. Built upon the Qwen large language model, this AI Agent features continuously evolving language understanding and reasoning capabilities, enhancing its precision in responding to real-world travel needs.