Traffic perception route recommendation method based on multi-modal fusion and generative adversarial network

Abstract

The invention discloses a traffic perception route recommendation method based on multi-modal fusion and a generative adversarial network. The method comprises the following steps: preprocessing historical travel data and traffic data according to a route recommendation task; encoding the user historical portrait, the traffic flow condition and the user query information based on a multi-mode encoder; using a generative adversarial network to generate candidate routes meeting user requirements by taking the joint feature representation as input; the generated route is dynamically optimized and adjusted through real-time traffic jam information, so that the generated route can avoid a high-jam area; and comprehensively evaluating the plurality of generated candidate routes, and recommending an optimal route. The efficient route recommendation method which has the multi-modal fusion characteristic and gives consideration to generation and optimization is provided for complex scenes with personalized travel requirements and traffic dynamic changes, the personalized degree, practicability and traffic perception ability of the recommended route are remarkably improved, and higher-quality and more intelligent travel experience is provided for users.

Images