The apolloauto/apollo repository is an open-source autonomous driving platform developed by Baidu, designed to accelerate the development, testing, and deployment of self-driving vehicles. Apollo provides a high-performance and flexible architecture that supports a wide range of autonomous driving scenarios, from basic GPS waypoint following to complex urban navigation and large-scale deployment. The platform is modular, allowing developers to integrate, calibrate, and configure various hardware and software components according to their specific needs.
Apollo’s evolution is marked by a series of major releases, each introducing new features and capabilities. Starting with Apollo 1.0, which enabled automatic GPS waypoint following in controlled environments, the platform has progressively expanded its scope. Apollo 1.5 added LiDAR-based perception for fixed lane cruising, improving mapping and trajectory planning. Apollo 2.0 introduced urban road driving, with capabilities for obstacle avoidance, traffic light recognition, and lane changes. Subsequent versions, such as Apollo 2.5 and 3.0, enhanced highway driving and provided a foundation for developers to build custom solutions in low-speed, closed venues.
Apollo 3.5 and 5.0 focused on complex urban scenarios, offering 360-degree visibility and advanced perception algorithms to handle dynamic environments, including residential areas and intersections. Scenario-based planning and deep learning models were integrated to improve safety and awareness. Apollo 5.5 further advanced urban driving with curb-to-curb support and new prediction models, moving closer to fully autonomous operation. Apollo 6.0 and 7.0 incorporated additional deep learning models, data pipeline services, and introduced Apollo Studio, an online development platform for simulation and reinforcement learning.
Apollo 8.0 and later versions emphasized extensibility and developer experience. The platform introduced package management for easier module organization, streamlined the perception development process, and upgraded simulation tools for debugging and validation. Apollo 9.0 reshaped the planning and control (PnC) and perception frameworks, optimized component management, and launched Dreamview Plus, a comprehensive developer tool for visualization and resource management. Enhanced LiDAR and camera detection models, incremental training, and support for 4D millimeter-wave radar were also added, alongside adaptation for ARM architectures.
Apollo 10.0 and 11.0 focused on large-scale deployment, performance optimization, cost reduction, and safety. These versions improved hardware ecosystem compatibility, lowered software development barriers, and reinforced functional safety strategies. The platform now supports comprehensive upgrades across perception, localization, planning, and development toolchains, enabling efficient end-to-end autonomous driving operating systems.
The repository includes detailed documentation covering installation, hardware integration, package management, core modules (such as CyberRT, localization, perception, prediction, planning, control), and tools for map acquisition and calibration. Apollo supports a variety of hardware, including by-wire vehicle systems, high-performance GPUs, and ARM-based devices. It is compatible with Ubuntu and leverages Docker for containerized development environments. Developers can access quick start guides for each version, submit issues, and engage with the community through various channels.
Apollo is licensed under Apache-2.0 and provides only the source code for models and algorithms, with cybersecurity strategies integrated during commercial deployment. The platform is designed to be extensible, robust, and developer-friendly, making it a leading solution for autonomous vehicle research and application.
