EasyNav
Lightweight ROS 2 navigation system that runs as one process from one parameter file
License
- core, official plugins, NavMap and the Nav2 bridge: Apache 2.0
- stack repositories (gridmap, octomap, pointcloud, costmap, simple): GPL 3.0
- easynav_lidarslam_ros2: BSD 2-Clause
Mixed licences across the EasyNavigation organisation. The core (EasyNavigation), the official plugins (easynav_plugins), NavMap and the Nav2 bridge are Apache-2.0. The separate stack repositories easynav_gridmap_stack, easynav_octomap_stack, easynav_pointcloud_stack, easynav_costmap_stack and easynav_simple_stack are GPL-3.0; easynav_lidarslam_ros2 is BSD-2-Clause; a few repositories carry no licence file. Checked 2026-10-23.
Overview
EasyNav (EasyNavigation) is an open-source navigation system for ROS 2 from the Intelligent Robotics Lab at Universidad Rey Juan Carlos; its project lead, Francisco Martín Rico, also leads PlanSys2. It does the same job as Nav2, taking a wheeled robot from where it is to a goal while avoiding obstacles, but it is built differently: the whole system runs as one program, system_main, configured by one parameter file, and every part (map representation, localizer, planner, controller and recovery) is a plugin chosen in that file. It is not tied to a 2D costmap: stacks exist for 2D costmaps, NavMap triangulated 3D surfaces, Bonxai probabilistic voxel maps, elevation grid maps and plain point clouds. The core and the official plugins are Apache-2.0 and are released as Debian packages for Humble, Jazzy, Kilted and Lyrical, with ready-made Gazebo simulations to try it in. The source is on GitHub and the documentation covers installation, how-tos and a guide for moving a robot over from Nav2.
EasyNav runs inside a single process. A ROS 2 lifecycle node, SystemNode, composes the rest: a sensors node that gathers point clouds and laser scans, images, IMU, GNSS, odometry and 3D detections; a maps manager that holds the environment model; a localizer; a planner; a controller; and a recovery node. A GoalManager inside the system node accepts and tracks goals. There is no behaviour tree: mission logic such as patrols goes in your own node, which sends goals through a client. Each component runs at its own rate in a real-time cycle, and the controller is the single source of velocity commands, choosing each cycle between the controller's output and the recovery system's.
What EasyNav does is decided by the plugins named in its parameter file, loaded with pluginlib. The official set, in the easynav_plugins repository, covers:
- Maps managers: Costmap2D, NavMap 3D surfaces, Bonxai voxel maps, editable navigation routes, Octomap, and a simple binary map for learning. - Localizers: AMCL on a costmap, multi-hypothesis AMCL for global localization and kidnapping, AMCL against a 3D voxel map, and a UKF that fuses GPS, odometry and IMU. - Planners: A* over a costmap, a NavMap mesh or a simple map. - Controllers: Regulated Pure Pursuit (a port of Nav2's, with an optional dynamic-window extension), MPPI, model predictive control, SeReST, a VFF reactive controller and a simple PID follower. - Recovery: a diagnostic recovery system that can stop or drive the robot, hold or abort a mission, or change parameters while running.
Each component also has a dummy plugin, so an application that needs no map or no localizer (a person follower, or GPS waypoint driving) simply leaves it out.
The PlayGrounds are complete Gazebo Harmonic simulations with ready configurations for a Kobuki indoors, a Summit XL outdoors, a TIAGo and an omnidirectional base. Moving from Nav2 is documented: the same URDF, TF frames, sensor topics and map files (from map_server or SLAM Toolbox) are reused, and an easynav_nav2_bridge package serves Nav2's NavigateToPose action so existing Nav2 clients, including RViz's goal button, keep working. A safety guide explains how to run EasyNav as the non-safety part of a robot next to a certified safety channel.
Why it matters
Nav2 is the default way to make a ROS 2 robot navigate, and it is excellent, but it is many servers coordinated by behaviour trees and built around 2D costmaps. EasyNav offers a different trade-off from an active university lab: one process and one file to configure, low latency between sensing and acting, and map representations that go beyond flat floors, including 3D meshes, voxel maps and elevation maps for outdoor and multi-level terrain. Because it reuses Nav2's conventions and speaks Nav2's goal action through a bridge, a team can try it on a robot that already runs Nav2 without rebuilding anything else.
Problem it solves
Getting a wheeled robot to navigate in ROS 2 usually means configuring a stack of separate servers, a lifecycle manager and behaviour trees, and working with a 2D costmap even when the terrain is not flat. EasyNav packs navigation into one configurable process whose map representation, localizer, planner and controller are swappable plugins, so the same system can drive a vacuum-sized indoor robot on a costmap or an outdoor robot on a 3D surface map.
Key takeaways
- A ROS 2 navigation system that runs as one program, system_main, configured by one parameter file
- Map representation, localizer, planner, controller and recovery are all plugins, with dummy plugins for parts an application does not need
- Stacks for 2D costmaps, NavMap 3D triangulated surfaces, Bonxai voxel maps, elevation grid maps and point clouds
- Controllers include Regulated Pure Pursuit, MPPI and MPC; localizers include multi-hypothesis AMCL and GPS/IMU fusion
- Released as Debian packages for Humble, Jazzy, Kilted and Lyrical; Gazebo PlayGrounds for Kobuki, Summit XL, TIAGo and an omni base
- Reuses Nav2's maps and frames, and a bridge serves Nav2's NavigateToPose action so Nav2 clients keep working
Who it is for
- ROS 2 developers building wheeled robots who find Nav2 heavy to configure or need a non-2D map
- Teams with outdoor or uneven-terrain robots that need 3D surface, voxel or elevation maps
- Researchers who want to swap in their own localizer, planner or controller as a plugin
- Students learning robot navigation, starting from the simple stack and the PlayGrounds
Limitations
- Still pre-1.0 (0.5.x), and the documentation tracks the rolling branch, which can describe features not yet released
- No behaviour trees: mission logic has to be written as a separate node that sends goals
- Smaller community and fewer plugins than Nav2; the separate gridmap, octomap and pointcloud stack repositories are GPL-3.0, not Apache-2.0
- On Humble only the core is supported; the plugins and PlayGrounds target Jazzy and later, and Rolling has no Debian packages yet
- Not a certified safety component: protective and emergency stops must come from an independent safety channel