Skip to content
RobothecaBeta
Log in

NeuPAN

  • Software
  • Navigation and mapping
  • Library

NeuPAN is an end-to-end, real-time, map-free motion planner for robots that maps obstacle points directly to control actions using model-based learning.

Source on GitHub: github.com/hanruihua/NeuPAN (opens in a new tab). Facts read at commit d225c38 (opens in a new tab).

Licenses

Code
GPL-3.0-only (opens in a new tab)

Read by Robotheca from LICENSE (opens in a new tab) (GPL-3.0-only) at commit d225c38 (opens in a new tab), on 22 Oct 2026.

Commercial use
Yes

Copyleft: Modified source must be published.

What it does

  • Maps 2D obstacle points directly to control actions, avoiding separate perception, mapping and planning modules.
  • Uses point-level collision avoidance constraints inside an MPC-style optimization for safe navigation in cluttered or unknown spaces.
  • Requires a pretrained DUNE model per robot geometry, trainable on random points, and reusable across environments.
  • Supports differential drive, Ackermann and omnidirectional kinematics; other kinematics need changes to the constraints.
  • Can be fed by 3D lidar projected to a plane, or by 2D points extracted from a camera, and can take paths from global planners like A*.
  • Includes IR-SIM examples and a ROS/ROS2 wrapper for simulation and real-robot deployment.

Written by Robotheca from the project's README (opens in a new tab) at commit d225c38.

Getting started

Difficulty
Advanced
Hardware it needs, as the project states it
Requires Python >= 3.10. A GPU can optionally be used to train the DUNE model.

Maintenance

Status
Active: a commit in the last 90 days
Newest commit
21 Oct 2026 on main
Latest release
neupan v1.3 (5 Feb 2026)

Checked on GitHub on 22 Oct 2026.

Links

Robots that can run it

No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.

Robotheca lists software as a reference: nothing is hosted or run here, and every fact links to where it was read.