Xin Lang Cai Jing·2025-12-29 13:39Core Insights - The report presented by Baidu's Intelligent Cloud Traffic Management Industry Solution Director highlights the launch of the SaaS traffic control platform based on Baidu's traffic large model and the "Baidu Famu" intelligent agent, showcasing its evolution from version 1.0 to 3.0 and its core advantages of being "multi, fast, good, and economical" [2][23] - The platform aims to address significant pain points in traffic control, such as insufficient external perception and data shortages, while also emphasizing the urgent need for intelligent tools in many cities [4][26] Product Development and Industry Pain Points - Baidu's new intelligent agent "Baidu Famu," launched on November 13, 2025, focuses on algorithm self-evolution to find global optimal solutions, achieving over 23% efficiency improvement in AI training and inference [24][26] - The current traffic control scenarios face challenges like long deployment cycles and reliance on manual operations, highlighting the need for intelligent tools [4][26] Core Advantages and Features of the SaaS Traffic Control Platform - The SaaS traffic control platform offers four core advantages: - "Multi" refers to diverse capabilities in platforms, data, and services - "Fast" indicates a lightweight, plug-and-play product that can be deployed in 2-3 weeks - "Good" emphasizes user-friendly tools for traffic engineers and significant practical effects, with average vehicle delays reduced by over 8% - "Economical" allows for deployment without external perception investment [6][28] Key Functionalities of the SaaS Traffic Control Platform - The platform includes six core functionalities: - Countdown data for traffic lights based on Baidu navigation maps - Traffic flow restoration with over 75% accuracy - Traffic problem diagnosis capabilities - Regional signal optimization recommendation tools - Optimization effect evaluation - Automatic report generation leveraging large model capabilities [10][32] Product Iteration and Collaboration Models - The SaaS traffic control platform has undergone continuous iterations, with version 1.0 focusing on background scheme learning and traffic flow restoration, while version 2.0 integrates local sample data for improved accuracy [33][34] - The collaboration model includes four approaches tailored to different client needs, from providing full services with on-site engineers to offering platform services for cities with existing traffic engineers [36] Nationwide Implementation and Diverse Application Practices - The SaaS traffic control platform has been successfully implemented in over 40 cities, with notable results such as a more than 20% reduction in average vehicle delays in Ordos [14][36] - A case study in Beijing demonstrated the platform's value beyond traffic control optimization, achieving a 90% reduction in overall costs and significantly shortening design times [37] New Data Element Cooperation Model and Core Product Goals - The focus is on creating a high-quality traffic management industry dataset, which will support local traffic large model training and enhance internal workflows while providing real-time information to traffic participants [40][41] - The ultimate goal of the SaaS traffic control platform is to ensure every intersection receives attention and service, improving travel efficiency and user satisfaction [43]