Nav2

Nav2 is the professionally-supported successor to the ROS Navigation Stack, optimized for mobile and surface robotics. It enables robots to navigate complex environments while avoiding obstacles and completing user-defined tasks.

The framework uses behavior trees to orchestrate modular servers for planning, control, and recovery behaviors. It supports holonomic, differential-drive, legged, and Ackermann robot types with both circular and arbitrarily-shaped footprints for collision checking.

Core capabilities

• Global planning: Compute optimal paths through environments using A*, Dijkstra, or custom planners
• Local control: Follow paths with DWB, Pure Pursuit, or other controller plugins
• Behavior trees: Define complex navigation logic using BehaviorTree.CPP for decision making
• State estimation: Integrate with AMCL, SLAM Toolbox, and robot_localization for positioning
• Environmental modeling: Costmap-based representations with layer plugins for sensor fusion
• Recovery behaviors: Handle failure conditions like getting stuck or sensor errors

Architecture

Nav2 follows a server-based architecture where independent modules communicate via ROS 2 actions and services:

• BT Navigator: Behavior tree orchestration and high-level decision making
• Planner Server: Global path planning algorithms
• Controller Server: Local trajectory following and obstacle avoidance
• Behavior Server: Recovery and auxiliary behaviors
• Smoother Server: Path refinement and optimization
• Route Server: Graph-based routing for structured environments
• Waypoint Follower: Multi-point mission execution

Use cases

• Warehouse AMR navigation and pallet transport
• Indoor service robotics and delivery robots
• Outdoor autonomous vehicles and agricultural robots
• Research and education in mobile robotics
• Fleet management integration for multi-robot systems

Limitations

• Requires ROS 2 expertise for configuration and tuning; steep learning curve for beginners
• Performance depends heavily on sensor quality and calibration (LIDAR, odometry, IMU)
• Complex environments with dynamic obstacles may require extensive parameter tuning
• No built-in fleet management; requires external orchestration for multi-robot coordination
• Limited support for non-planar or highly dynamic locomotion (legged robots need custom integration)
• Behavior tree XML files can become complex for sophisticated applications