Beijing (Gasgoo)- On December 4, Baidu officially launched the Apollo Open Platform 10.0, a significant upgrade designed around its Autonomous Driving Foundation Model (ADFM). This platform provides global developers and enterprises with cost-efficient, high-performance, and safer technologies to develop autonomous driving solutions tailored for diverse scenarios, advancing toward large-scale deployment.
The Apollo 10.0 platform introduces key enhancements across software and tools. At the software core, the upgraded CyberRT framework now supports large-scale, high-level autonomous driving scenarios with zero-copy communication, achieving microsecond transmission and a tenfold performance improvement.
Its application layer provides a robust, ready-to-use autonomous driving system optimized for driving, perception, positioning, and PnC (planning and control), reducing resource usage by 50%. This ensures Level-4 autonomous driving capabilities on a single NVIDIA Orin chip. The platform also includes over five new visualization tools for performance analysis, enabling developers to monitor data and conduct precise function-level optimizations.
The platform supports enhanced perception, planning, and positioning algorithms. Upgrades include a LiDAR perception framework with BEV+OCC vision capabilities and advanced planning algorithms for improved accuracy in driving and parking scenarios, reducing vehicle model adaptation time to one week. Positioning now incorporates multi-modal BEV-based matching, covering indoor, outdoor, and complex motor road scenarios.
The Apollo 10.0 also integrates with the ROS software ecosystem, lowering software reuse costs, and enables seamless batch deployment on physical machines. Hardware support has expanded to include new domain controllers, LiDAR systems, and navigation units, providing developers with flexible and rapid adaptation options.
