On April 30, Amap officially launched its "AI Companion" product tailored for navigation scenarios. The AI companion can comprehend user speech, perceive the user's real-time location, orientation, and surrounding environment, interpret street view imagery through the camera, and integrate map data to provide highly context-aware suggestions. This represents the industry’s first multi-modal travel companion designed for real-world travel situations. By integrating multi-modal perception with mapping capabilities, Amap aims to elevate its navigation service into an AI partner capable of understanding the physical world.

In daily travel, many users encounter situations where they stand at an intersection at the start of navigation unsure of which direction to take. The AI companion offers guidance that aligns more closely with the real-world environment, translating abstract directions into visible landmarks in the user’s immediate surroundings to reduce cognitive effort.

For more complex travel needs, the AI companion can interpret multiple intentions and suggest appropriate routes. In scenic areas, it can also function as a real-time tour guide, generating a complete visiting route and highlighting key attractions along the way. Additionally, the AI companion supports visual recognition scenarios. When a user points their phone at a building or historical site and asks, "Tell me about this building," the system identifies the structure in view and provides historical information, opening hours, ticket details, and navigation options to the entrance.

Behind these features lies a technology framework built by Amap for the AI companion, centered around an Agent-based system. This framework consists of four main components: spatiotemporal context awareness, multi-modal fusion understanding, spatial action capability, and complex task processing mechanisms.

Currently, users can interact with the AI companion by tapping the "Hold to Talk" button at the bottom of the navigation screen after starting walking navigation in Amap. The feature is currently in testing and is being progressively rolled out to more users, with wider availability expected in the near future.