How ethz-asl/mav_active_3d_planning works
Main components
• Core Planning Engine: Represent the active_3d_planning_core package as the heart of the framework. It encapsulates the planners, trajectory generators, cost/gain evaluators, and modules for sensor and map processing.
• ROS Interface: Include active_3d_planning_ros that adapts the core functionality to the ROS ecosystem. This component will handle message passing, ROS nodes, and topic/service exchanges.
• MAV-Specific Modules: Show active_3d_planning_mav which provides modules tailored to micro aerial vehicles, integrating sensor inputs and specific motion planning (e.g., specialized trajectory generators).
• Voxblox Integration: Include active_3d_planning_voxblox as the mapping component, managing the underlying occupancy or TSDF maps via voxblox, and interfacing with the core planner.
• Application Layer: Represent active_3d_planning_app_reconstruction that configures and launches the planner setups. This layer also provides simulation configurations (via RViz configurations, launch files, and experiment evaluation scripts) and interfaces to external simulators (e.g., Unreal CV).
• External Dependencies: Illustrate connections to external libraries and tools used (e.g., mav_trajectory_generation, catkin_simple, glog_catkin, eigen_catkin, and other simulation modules like rotors_simulator).
Architectural patterns and design principles
• The project appears to adopt a modular design pattern. Emphasize the separation of concerns: core logic, interfacing, specialized domain modules, and application orchestration.
• Documentation and configuration files (found in the “cfg” and “launch” directories) indicate a design that is highly configurable at runtime, so highlight the configurability loop (i.e., YAML-driven module instantiation).
• The layered approach permits the core to be used stand-alone (non-ROS) while the ROS-specific package wraps around it to support message passing and node-based execution.