Baidu Apollo EM Motion Planner (1807.08048v1)

Abstract: In this manuscript, we introduce a real-time motion planning system based on the Baidu Apollo (open source) autonomous driving platform. The developed system aims to address the industrial level-4 motion planning problem while considering safety, comfort and scalability. The system covers multilane and single-lane autonomous driving in a hierarchical manner: (1) The top layer of the system is a multilane strategy that handles lane-change scenarios by comparing lane-level trajectories computed in parallel. (2) Inside the lane-level trajectory generator, it iteratively solves path and speed optimization based on a Frenet frame. (3) For path and speed optimization, a combination of dynamic programming and spline-based quadratic programming is proposed to construct a scalable and easy-to-tune framework to handle traffic rules, obstacle decisions and smoothness simultaneously. The planner is scalable to both highway and lower-speed city driving scenarios. We also demonstrate the algorithm through scenario illustrations and on-road test results. The system described in this manuscript has been deployed to dozens of Baidu Apollo autonomous driving vehicles since Apollo v1.5 was announced in September 2017. As of May 16th, 2018, the system has been tested under 3,380 hours and approximately 68,000 kilometers (42,253 miles) of closed-loop autonomous driving under various urban scenarios. The algorithm described in this manuscript is available at https://github.com/ApolloAuto/apollo/tree/master/modules/planning.

• The paper presents a hierarchical motion planning system that integrates multilane strategies with decoupled path-speed optimization for efficient autonomous driving.
• The paper introduces a dynamic programming and spline-based quadratic programming framework to ensure smooth, safe trajectories while adhering to traffic regulations.
• The paper demonstrates the planner’s robustness through extensive real-world testing over 68,000 km, highlighting its scalability and reliability in urban environments.

Overview of the Baidu Apollo EM Motion Planner

The paper presents the development of a real-time motion planning system integrated into the Baidu Apollo autonomous driving platform, focusing on achieving level-4 autonomy with a strong emphasis on safety, comfort, and scalability. The authors outline a hierarchical motion planning approach that includes both multilane and single-lane strategies.

Hierarchical Planning Architecture

The planner operates in a hierarchical structure. The top layer is responsible for multilane strategies, which involve managing lane-change scenarios by calculating and comparing potential trajectories across multiple lanes. Within each lane, the system iteratively optimizes path and speed using a Frenet coordinate system.

Optimization Techniques

The paper introduces a novel approach combining dynamic programming and spline-based quadratic programming for path and speed optimization. This framework accommodates traffic rules and obstacle avoidance while maintaining smoothness in the trajectory planning. The path and speed are optimized iteratively, enhancing the planner's adaptability to dynamic traffic conditions by refining trajectories through multiple planning cycles.

Multi-lane Strategy

The multilane strategy is crucial for managing lane changes efficiently and safely. The system uses parallel processing to evaluate lane-level trajectories, ensuring that the most viable trajectory is selected based on safety and cost metrics. This method enhances the planner’s capability to handle both passive and active lane changes, which are essential elements in autonomous navigation.

Path-Speed Iteration

A significant contribution is the decoupled path-speed approach, which overcomes the limitations of direct 3D optimization in the Frenet frame by separately optimizing path and speed. This separation allows greater flexibility and efficiency, despite the non-optimality associated with decoupling when facing dynamic obstacles. An iterative process further refines these decoupled solutions, handling dynamic obstacles adeptly.

Decision-Making and Traffic Regulations

The planner distinguishes between non-negotiable traffic regulations and negotiable on-road decisions, integrating these within the optimization routine. This distinction allows the planner to maintain a clear focus on safety while also pursuing optimal trajectory planning, thus balancing compliance with practical driving needs.

Performance Metrics

The planner has been deployed in Apollo autonomous vehicles, undergoing extensive testing over more than 68,000 kilometers under various urban traffic scenarios. This extensive empirical evaluation underscores the system’s robustness and reliability in real-world conditions.

Implications and Future Directions

The research presented in this paper has significant implications for autonomous driving technology. By demonstrating a scalable and efficient motion planning model, it advances the practical deployment of autonomous vehicles. The iterative optimization strategy and hierarchical planning framework could inspire further research in adaptive AI-driven traffic management systems, possibly extending the current model to incorporate learning mechanisms for enhanced adaptability and decision-making precision in complex traffic environments.

The paper provides valuable insights into the development of real-time motion planning in autonomous systems, indicating a promising future in achieving fully autonomous navigation in urban contexts. Continued research could explore the integration of machine learning algorithms to further enhance the adaptability and intelligence of such systems.