Space Robotics

Sensing, Perception, and Navigation in Space Robotics

Perception and navigation for robots in unstructured, GNSS-denied environments, from planetary surfaces to natural terrain on Earth, developed with a measurement-science mindset.

The key idea is simple: a navigation algorithm is only as useful as the confidence you can put in its output. That’s why every method developed here comes with a metrological characterization of its accuracy, uncertainty, and robustness, and is tested on real hardware and field data. 

1. SLAM and multi-sensor fusion

Knowing where a robot is and what surrounds it, without GPS, is the core problem here. The work covers Simultaneous Localization and Mapping (SLAM) using cameras, LiDARs, event cameras, and inertial sensors, combined through pose graph optimization and multi-sensor fusion. The focus is on understanding how each sensor and each design choice affects the final accuracy, so that the right combination can be picked for a given mission or application.

  • Fortuna, S., Chiodini, S., Valmorbida, A., & Pertile, M., Comparative Analysis of LiDAR-SLAM Systems: A Study of a Motorized Optomechanical LiDAR and an MEMS Scanner LiDAR. Sensors, 25(17), 5352 (2025). DOI
  • Quitadamo, V., Chiodini, S., Pertile, M., Olivieri, L., & Valmorbida, A., Event-Camera-to-Body Extrinsic Calibration Using Accumulated Event Frames. IEEE MetroAeroSpace 2026, pp. 82-87. DOI
  • Chiodini, S., Trevisanuto, G., Bettanini, C., Colombatti, G., & Pertile, M., Improving Keypoints Tracking With Machine-Learned Features in Event-Camera-Based Visual Odometry. IEEE MetroAeroSpace 2024, pp. 146-151. DOI

2. Hazard mapping and terrain traversability

Rocks, slopes, soft sand: before moving, a rover needs to know where it can safely go. Occupancy-grid and continuous terrain maps are combined with deep-learning semantic segmentation to spot hazards and traversable areas, with attention to how training choices, such as data augmentation, affect the reliability of the labels.

(Left) CNN labeler (DeepLabv3+) applied to image labeling. (Right) Map displaying the generated obstacles. Tested on the ESA Katwijk Beach Planetary Rover Dataset.

  • Polato, G., Chiodini, S., Valmorbida, A., Pertile, M., Giorgi, G., & Lorenzini, E. C., Semantic Terrain Traversability Analysis Based on Deep Learning Aimed at Planetary Rover Navigation. Aerotecnica Missili & Spazio, 104(4), 325-335 (2025). DOI
  • Chiodini, S., Polato, G., Valmorbida, A., Pertile, M., Giorgi, G., Narduzzi, C., & Lorenzini, E. C., Impact of data augmentation on labelling confidence in deep learning terrain traversability analysis for unmanned ground vehicles. Acta IMEKO, 14(3) (2025). DOI
  • Fortuna, S., Chiodini, S., Valmorbida, A., & Pertile, M., ROS 2 Integrated Continuous Terrain Mapping Comparison for Autonomous Navigation in Unstructured Environments. IEEE MetroAeroSpace 2024, pp. 128-133. DOI

3. ROS 2 autonomous navigation for planetary rovers

Algorithms are only useful once they run on a real robot. Navigation stacks are built on the Robot Operating System (ROS 2) and cover localization, path tracking, and maneuvering, including rovers with multiple locomotion modes. Being ROS 2-based, they can be ported to other platforms with little effort.

LiDAR-SLAM algorithm running onboard the MORPHEUS rover for its navigation.

  • Fortuna, S., Roncagliolo, P., Graziato, D., Merlo, A., Chiodini, S., Valmorbida, A., & Pertile, M., ROS 2-based autonomous navigation strategy for a lunar rover featuring multiple locomotion modes. Acta Astronautica, 238, 985-996 (2026). DOI
  • Fortuna, S., Chiodini, S., Valmorbida, A., & Pertile, M., Enhanced Path Tracking and Maneuvering Strategies for Lunar Rovers. 75th International Astronautical Congress (IAC 2024), pp. 1957-1969. DOI

4. Measurement of wheel-terrain interaction

On granular soil, everything happens at the wheel. Dedicated experimental setups measure slip ratio, sinkage, resistive torque, and contact pressure of small rover wheels on granular terrain, providing data for locomotion design and for estimating terrain properties.

  • Chiodini, S., Quattrocchi, A., Valenti, M., Freni, F., Cannuli, A., & Montanini, R., Testbed development for the measurement of slip ratio, sinkage, and resistive torque in small rover wheel-granular terrain interaction. Measurement, 282, 121936 (2026). DOI
  • Valenti, M., Quattrocchi, A., Chiodini, S., Freni, F., Cannuli, A., & Montanini, R., Measurement of rover wheel-soil interaction parameters using piezoresistive pressure matrix sensors. Measurement Science and Technology, 37(41), 415105 (2026). DOI

5. Absolute localization for Mars rovers

Within the ExoMars program, in collaboration with ALTEC S.p.A., several methods for rover position and attitude estimation have been developed: sun-based attitude and absolute localization, landmark-based localization, skyline-based absolute localization, and WISDOM Grid Relative Localization. The latter automatically selects the best viewpoints, so that the onboard Guidance, Navigation, and Control (GNC) system can correct the rover trajectory accurately.

6. Positioning and mapping in GNSS-degraded environments on Earth

The same tools work just as well back on Earth. Current topics include GNSS positioning with adaptive weighting and multipath isolation and visual-LiDAR 3D mapping from ground vehicles, with applications in precision agriculture, forestry, and field robotics.

  • Chiodini, S., Valmorbida, A., Anese, G., Pertile, M., Giorgi, G., Narduzzi, C., & Saggin, B., Improving GNSS Positioning Accuracy through Adaptive Weighting and Multipath Isolation. Materials Research Proceedings, 69, 42-47 (2026). DOI
  • Chilin, L., Chiodini, S., Toson, F., Bettanini, C., MĹ‚ynarczyk, A., Coradini, M., Maucieri, C., Pirotti, F., & Colombatti, G., Precision 3D Mapping of Forests and Crops Using a Visual-LiDAR System Mounted on a Rover. IEEE MetroAgriFor 2025, pp. 23-28. DOI

Facilities and equipment

  • UGV rover: a mobile platform for field testing of perception, mapping, and navigation algorithms.
  • Sensor suite: mechanical and MEMS-scanner LiDARs, cameras, event cameras, inertial sensors, and GNSS receivers, for data acquisition and benchmarking.

Interested in collaborating?

Research institutions and companies are very welcome. Collaborations can take many forms, for example:

  • Contract research on perception, localization, and navigation for robotic platforms
  • Test and validation campaigns for sensors and navigation algorithms, including uncertainty assessment
  • Joint proposals for European, ESA, and national funding programs
  • Industrial PhD programs and master’s theses developed together with partner organizations

Have a problem in mind? Just get in touch.