ROS2 Navigation Stack (Nav2)
The definitive, production-grade navigation framework for autonomous mobile robots. Nav2 empowers AGVs to plan paths, navigate dynamic environments, and execute complex behaviors with industrial reliability.
_hero.png)
Core Concepts
Behavior Trees
Unlike the state machines of ROS1, Nav2 uses Behavior Trees to manage complex robot logic, allowing for modular, hierarchical task execution and easier recovery strategies.
Global Planner
Calculates the optimal path from start to goal across the entire map using algorithms like A* or Dijkstra, ensuring the robot knows the general direction before moving.
Controller Server
Handles obstacle avoidance and path following locally. It generates velocity commands (Twist messages) to keep the robot on the global path while reacting to dynamic objects.
Layered Costmaps
Nav2 builds a 2D grid representation of the world. It layers static map data, inflation layers (safety padding), and live sensor data to determine traversable space.
Recovery Servers
Autonomous systems must handle failure. Nav2 includes specific behaviors like spinning, backing up, or clearing costmaps to unstuck the robot without human intervention.
Lifecycle Nodes
Every node in Nav2 (planner, controller, map server) is a managed lifecycle node, allowing for deterministic startup, shutdown, and error handling sequences across the system.
How Nav2 Drives Autonomy
The architecture of Nav2 is designed for modularity and scalability. At its heart, the system takes a goal pose and uses current sensor data (Lidar, Depth Cameras) combined with odometry to determine the safest path.
The Map Server loads the static environment, while the AMCL node localizes the robot within that map. Simultaneously, the Planner Server computes the long-term path, which is fed to the Controller Server.
The Controller (often DWA or MPPI) converts this path into immediate velocity commands for the motor drivers, constantly adjusting for new obstacles detected by the costmaps. This feedback loop runs at high frequency to ensure smooth, collision-free motion.
_diagram.png)
Real-World Applications
Intralogistics & Warehousing
Nav2 powers fleets of AMRs (Autonomous Mobile Robots) that transport pallets and bins. Using 'Keepout Zones', robots strictly adhere to safety lanes while dynamically avoiding forklifts and workers.
Hospital Delivery
In sensitive healthcare environments, Nav2's precise localization allows tug robots to deliver linens and medicine. Low-noise trajectory planners ensure silent operation in hallways.
Agricultural Robotics
GPS-aided navigation combined with Nav2 allows field robots to traverse crop rows. The stack can be configured to use "Smac Planner" for hybrid A* planning on rough terrain.
Security & Inspection
Patrol robots utilize Nav2's waypoint follower capabilities to execute routine inspection rounds, using automatic docking behaviors to recharge when battery levels drop.
Frequently Asked Questions
What is the main difference between ROS1 Navigation and ROS2 Nav2?
Nav2 is a complete rewrite, not just a port. It introduces Behavior Trees for logic control, Lifecycle Nodes for managed states (active/inactive), and a highly modular plugin architecture that makes swapping planners and controllers significantly easier than in ROS1.
Does Nav2 support non-circular (non-holonomic) robots like Ackermann steering?
Yes, Nav2 supports differential drive, omnidirectional, and Ackermann (car-like) kinematics. Specifically, the Smac Planner and Regulated Pure Pursuit Controller are designed to handle the kinematic constraints of car-like robots effectively.
What hardware is required to run Nav2 effectively?
While Nav2 can run on a Raspberry Pi 4 for simple educational robots, industrial applications typically require an x86 mini-PC (like Intel NUC) or NVIDIA Jetson device. You also need good odometry sources (wheel encoders/IMU) and environmental sensors like 2D LiDAR or Depth Cameras.
How does Nav2 handle dynamic obstacles like people?
Nav2 uses a local costmap that updates in real-time based on sensor data. The local controller (e.g., DWA or MPPI) recalculates trajectories continuously to avoid new obstacles that weren't present in the static map.
Can I use Nav2 without a pre-existing map (SLAM)?
Yes, Nav2 can be paired with SLAM (Simultaneous Localization and Mapping) packages like SLAM Toolbox. This allows the robot to explore an unknown environment, build a map, and navigate within it simultaneously.
What is the role of Behavior Trees in Nav2?
Behavior Trees replace the rigid state machines of the past. They allow you to define complex logic visually via XML, such as "Try to follow the path; if blocked, wait 5 seconds; if still blocked, back up; if that fails, call for help."
How difficult is it to tune Nav2 for a custom robot chassis?
It requires moderate effort. The most critical step is configuring the robot's footprint (shape) and tuning the velocity/acceleration limits in the controller. Improper tuning can cause oscillation or "jerky" movement.
Can Nav2 handle "Keepout Zones" or restricted areas?
Yes, Nav2 supports "Keepout Filters." You can draw zones on a separate map layer that the global planner will treat as walls, preventing the robot from ever planning a path through specific restricted areas.
Does Nav2 support GPS navigation for outdoor robots?
Yes, Nav2 enables GPS navigation through the nav2_waypoint_follower and by fusing GPS data into the localization estimate (using robot_localization). This allows for waypoint-based navigation in large outdoor spaces.
Is Nav2 suitable for multi-robot fleets?
Nav2 runs on individual robots, but it is designed to integrate with Fleet Management Systems (FMS) like Open-RMF. Open-RMF coordinates traffic between multiple Nav2 instances to prevent deadlocks and manage shared resources like elevators.