MarsLab

Open-source · ROS2-native · NVIDIA Isaac Sim

MarsLab

A Martian Rover Simulator for
Planetary Rover Autonomous Navigation

Hoyun Kim1, Beomsu Kim1, and Giseop Kim1

1Department of Robotics and Mechatronics Engineering, DGIST

Paper GitHub Results Interactive demo

Abstract

Future Mars missions will require rover autonomy that can operate across unstructured terrain, changing illumination, atmospheric dust, and limited communication. Simulation is a practical way to study these conditions before deployment, but existing Mars-relevant resources differ in scope, including mission-oriented simulators, fixed analog datasets, task-specific environments, and open robotics interfaces.

We present MarsLab, an open-source, ROS2-native Mars rover simulator for autonomy and navigation algorithm development. MarsLab combines HiRISE-derived and procedural terrain with customizable rock, crater, solar-illumination, and atmospheric-dust settings, and runs a Perseverance-class rover model in NVIDIA Isaac Sim. The runtime publishes RGB, depth, RGB-D point clouds, LiDAR, IMU, wheel odometry, and ground-truth pose data through standard ROS2 topics.

We demonstrate MarsLab with SLAM benchmarks across sensing modalities, dust levels, scene geometry, and route length, and with visual place recognition benchmarks over repeated Mars Base traversals under illumination and dust changes. Controlled scene variation and shared ground-truth trajectories enable trajectory-level and image-level evaluation within the same simulator.

Explore MarsLab in your browser

Free-camera terrain exploration, synchronized rover sensors, and actual SLAM experiment replay. No installation required.

Open the interactive lab in a full window ↗

Customizable Mars environments

Replay the same experiment across terrain geometry, landmarks, illumination, and atmospheric dust.

Jezero Plain
Main Crater
Grand Canyon
Mars Base

Solar geometry

z = 90 degrees minus e

Elevation e determines zenith angle z. Azimuth changes shadow direction.

Azimuth 70° → 280° · elevation 32°

Atmospheric dust optical depth, τ

I(z) = I0 exp[−τ m(z)]

Increasing τ attenuates the direct solar beam; m(z) is relative optical airmass.

Continuous rendering: τ = 0.3 → 6.0 · τ = 0.30

Geometry is unchanged; the renderer also accounts for diffuse sky illumination.

[28] J. Appelbaum and D. J. Flood, “Solar radiation on Mars,” Solar Energy, vol. 45, no. 6, pp. 353–363, 1990.

Rover traverse · Grand Canyon
Rover traverse · Main Crater

SLAM benchmark

Absolute Trajectory Error RMSE after Umeyama SE(3) alignment. Lower is better.

LiDAR remains sub-metre while RGB degrades 7.9× under dense dust.

Camera methods are evaluated at τ = 0.5 and τ = 6.0. MOLA uses LiDAR, which is unaffected by the radiance-only dust model.

0.13–0.38 mLiDAR SLAM across all scenes
7.9×ORB-SLAM degradation on Mars Base
Tracking lostMonocular SLAM in Grand Canyon

Visual place recognition

Repeated Mars Base traversals test place retrieval under illumination and dust changes.

Daytime Mars Base query image Retrieved dark image of the corresponding Mars Base location
Day / Dark. Correct top-1 retrieval across a strong illumination change.

BoQ · Day/Dark91.44%

Recall@1 · τ = 0.05

BoQ · Dust94.65%

Recall@1 · τ = 0.05 → 0.60

BoQ ranks first in both protocols. NetVLAD follows, while AnyLoc recovers strongly by R@10.

Paper and citation

MarsLab provides repeatable Mars autonomy experiments through standard ROS2 interfaces.

@inproceedings{kim2026marslab,
  title={MarsLab: A Martian Rover Simulator for Planetary Rover Autonomous Navigation},
  author={Kim, Hoyun and Kim, Beomsu and Kim, Giseop},
  booktitle={International Conference on Space Robotics (iSpaRo)},
  year={2026}
}
Download paper

Isaac Sim 5.1 · ROS2 Jazzy · HiRISE-derived terrain