Reliable odometry in highly dynamic environments remains challenging when it relies on ICP-based registration: ICP assumes near-static scenes and degrades in repetitive or low-texture geometry. We introduce Dynamic-ICP, a Doppler-aware registration framework. The method (i) estimates ego translational velocity from per-point Doppler velocity via robust regression and builds a velocity filter, (ii) clusters dynamic objects and reconstructs object-wise translational velocities from ego-compensated radial measurements, (iii) predicts dynamic points with a constant-velocity model, and (iv) aligns scans using a compact objective that combines point-to-plane geometry residual with a translation-invariant, rotation-only Doppler residual. The approach requires no external sensors or sensor-vehicle calibration and operates directly on FMCW LiDAR range and Doppler velocities. We evaluate Dynamic-ICP on three real-world datasets-HeRCULES, HeLiPR, AevaScenes-focusing on highly dynamic scenes. Dynamic-ICP consistently improves rotational stability and translation accuracy over the state-of-the-art methods.

Material awareness can improve robotic navigation and interaction, particularly in conditions where cameras and LiDAR degrade. We present a lightweight mmWave radar material classification pipeline designed for ultra-low-power edge devices (TI IWRL6432), using compact range-bin intensity descriptors and a Multilayer Perceptron (MLP) for real-time inference. While the classifier reaches a macro-F1 of 94.2% under the nominal training geometry, we observe a pronounced performance drop under realistic geometry shifts, including sensor height changes and small tilt angles. These perturbations induce systematic intensity scaling and angle-dependent radar cross section (RCS) effects, pushing features out of distribution and reducing macro-F1 to around 68.5%. We analyze these failure modes and outline practical directions for improving robustness with normalization, geometry augmentation, and motion-aware features.

Karst aquifers provide critical freshwater resources but pose significant hazards due to their complex and poorly understood subsurface geometry. Mapping these environments is challenging because sonar data from underwater exploration is sparse and noisy, while navigation estimates suffer from drift limiting standard 3D reconstruction methods. We present a pipeline for reconstructing underwater karst conduits from a sonar profiler. We combine a continuous-time SLAM approach to correct trajectory drift with a novel two-stage deep learning method for surface reconstruction, producing an immersive and navigable 3D mesh for hydrogeological analysis.

Space resource acquisition and utilization, commonly referred to as Space Mining, represent critical pathways for enabling sustained human exploration and unlocking commercial opportunities in space. These resources mainly include helium-3, water, mineral resources on the Moon and Mars, and abundant mineral deposits on asteroids. Due to the harsh conditions of space, communication delays, and high launch costs, the development of autonomous robotic systems is critical to achieving efficient, cost-effective space mining. This paper provides a comprehensive overview of space mining robotics and associated technologies. First, we review the background of space mining, including international policies, commercial entities, and recent advancements. We define a systematic six-stage architecture for space mining: Exploration is initiated by (1) remote sensing for target identification and (2) precise in situ robotic detection; Sampling progresses from (3) single-robot small-scale sampling to (4) multi-robot large-scale excavation; and Extraction integrates (5) autonomous resource extraction and (6) final integration into in situ construction or terrestrial transport. Additionally, we review and curate existing resources for space mining research, including real-world mission data, terrestrial analog datasets, and high-fidelity simulation environments. Finally, we identify critical open challenges in autonomous space mining and delineate a strategic research roadmap to bridge current technological gaps, fostering the transition toward a sustainable off-world economy. To track ongoing developments in space mining, we maintain an updated project page: https://github.com/OpenSpace-Lab/Space-Mining-with-Robotics-List.

Automation of healthcare workflows and devices demands safe and trustworthy robotic behavior, particularly in environments shared with patients and medical staff. For ceiling-mounted imaging robots, the key challenge lies in perceiving and monitoring the 3D workspace to plan safe, collision-free motions around people and equipment. Beyond simple obstacle avoidance, semantic understanding is essential to distinguish between object types — such as patients, walking aids, or medical tools — and to adapt motion behavior accordingly. We address this challenge with a semantic-aware obstacle tracking and avoidance pipeline that extends prior 2D semantic navigation concepts into full 3D space. The approach combines 2D semantic segmentation with depth projection to estimate object positions and dimensions in real time from RGB-D data. These detections are fused in a tracking module to build a continuous, semantic world model from which class-dependent safety margins are derived. The resulting information enables adaptive motion planning that increases distance from high-risk objects (e.g., persons) or reduces velocity when close interaction is required. Experiments on a real ceiling-mounted robot in laboratory scenarios demonstrate the system’s ability to enhance safety, predictability, and contextual awareness during automated healthcare procedures.

Autonomous medical systems must meet stringent hygiene and safety requirements while operating reliably in dynamic clinical environments. This paper presents a collision avoidance system based on the fusion of two mm-wave radar technologies - frequency modulated continuous wave (FMCW) and pulsed coherent radar (PCR). The system is fully integrated behind sealed covers of medical devices and enables unobtrusive and hygienic use without compromising functionality. We demonstrate that both radar types provide robust detection of dynamic obstacles, even through layers of disinfectants, blood and polycarbonate materials. A fail-safe system architecture based on redundant sensor paths and dual microcontrollers ensures reliable operation under fault conditions. Experimental validation on a robotic X-ray system confirms the responsiveness, accuracy and suitability of the system for clinical integration. The results show that radar fusion offers a promising path to hygienic and certified safe motion planning in healthcare robotics.

The exploration of lunar caves is a critical aspect of the space exploration program of the European Space Agency (ESA). To facilitate this mission, the DAEDALUS study investigated a novel spherical robot design in 2021. The proposed robot uses a unique telescopic linear rod mechanism to generate rotation and hence locomotion. This drive mechanism requires a dedicated control scheme to ensure both locomotion and simultaneously stabilization of the robot. The overall task of following a curved trajectory is also a problem that cannot be solved by simple algorithms. In this work, we introduce, calculate, and simulate a solution for these tasks, the Virtual Pose Instruction Plane (VPIP). The VPIP breaks the problem of multiple independent controllable rods down to two controllable parameters (roll and pitch of the plane), which control the linear motion velocity, balance and ultimately curvature motion of the robot. Initial simulations show that both speed and cornering can be controlled by the VPIP.

Spherical robots offer unique advantages for mapping applications in hazardous or confined environments, thanks to their protective shells and omnidirectional mobility. This work presents two complementary spherical mapping systems: a lightweight, non-actuated design and an actuated variant featuring internal pendulum-driven locomotion. Both systems are equipped with a Livox Mid-360 solid-state LiDAR sensor and run LiDAR-Inertial Odometry (LIO) algorithms on resource-constrained hardware. We assess the mapping accuracy of these systems by comparing the resulting 3D point-clouds from the LIO algorithms to a ground truth map. The results indicate that the performance of state-of-the-art LIO algorithms deteriorates due to the high dynamic movement introduced by the spherical locomotion, leading to globally inconsistent maps and sometimes unrecoverable drift.

Every year, excavators cause several milion euros in damage. One of the reasons for this is the lack of routing plans on construction sites. The "Digitalization and Smart Services in Civil Engineering"project is investigating the extent to which the visualization of pipelines can be used with the aid of a laser projector. A Livox laser scanner is used to take the surroundings into account in the projection. To be able to display the projection undistorted, the algorithm performs a ray tracing process to adjust the projection. However, this requires extrinsic calibration between the laser scanner and projector. The experiments show that the method presented here successfuly calibrates the projector and laser scanner and that an accurate projection is possible with a root mean square error of less than 40 mm. The projections in the DigSmart project may deviate by a maximum of 100 mm.

Many mobile mapping systems utilize the onboard identification of the ground plane from 3D Light Detection and Ranging (LiDAR) data, e.g., for Simultaneous Localization and Mapping (SLAM) purposes. This enables autonomy in different environments for numerous systems such as Unmanned Aerial Vehicles (UAV) or self-driving cars. However, there is no research concerned with spherical systems on this topic. To this end, we study the real-time on-board identification of the ground plane on spherical mobile mapping systems in this work. These systems provide unique challenges such as poor ground coverage due to the weak incidence angles of the laser beams, fast rotations, and largely varying sensor orientations. We implement multiple algorithms based on point-cloud pre-processing, plane fitting, and plane detection, to address these challenges and compare them in terms of runtime, accuracy, and robustness. For this purpose we create our own datasets - including ground truth data available from a terrestrial laser scanner (TLS) - featuring different environments, slopes, and rotation speeds. Our analysis proves the real-time capability and suggests that a geometrical pre-processing approach, followed by a Random Sample Consensus (RANSAC) based plane detection algorithm outperforms the other tested approaches. Furthermore, the evaluation reveals challenges regarding the unusual locomotion mechanism and suggests that, in order to reduce the median angle error of the plane normal below 9., we need to utilize an on-board globally consistent map, instead of only individual LiDAR frames.

Precise localization of mobile robots is key to many (semi)-autonomous operations, such as planetary exploration. In situations, where Global Navigation Satellite Systems (GNSS) is unavailable, Ultra-Wideband (UWB) technology is a common replacement. This typically relies on the positions of distributed anchors to be known beforehand. In this work we expand on a system, that remotely distributes the anchors, which means the position of them is unknown. The robot is equipped with three UWB tags, which perform Two-Way-Ranging (TWR) distance measurements with all the anchors. These distances are used to determine the relative position of the anchors to the robot. These positions are interpreted as landmarks in an Extended Kalman Filter (EKF)-Simultaneous Localization and Mapping (SLAM) algorithm, which combines them with the wheel odometry of the robot. Our experiments show promising results in a setup with four anchors, performing considerably better than the wheel odometry on its own. The system is also capable of operating through an outage of the UWB anchors. After such an outage the pose of the robot is corrected in multiple experiments, though not to the same standard as before the outage.

Future planetary exploration missions will rely heavily on efficient human–robot interaction to ensure astronaut safety and maximize scientific return. In this context, digital twins offer a promising tool for planning, simulating, and optimizing extravehicular activities. This study presents the development and evaluation of a digital twin for the AMADEE-24 analog Mars mission, organized by the Austrian Space Forum and conducted in Armenia in March 2024. Alternative local positioning methods were evaluated to enhance the system’s utility in Global Navigation Satellite System (GNSS)-denied environments. The digital twin integrates telemetry from the Aouda space suit simulators, inertial measurement unit motion capture (IMU-MoCap), and sensor data from the Intuitive Rover Operation and Collecting Samples (iROCS) rover. All nine experiment runs were reconstructed successfully by the developed digital twin. A comparative analysis of localization methods found that Simultaneous Localization and Mapping (SLAM)-based rover positioning and IMU-MoCap localization of the astronaut matched Global Positioning System (GPS) performance. Adaptive Cluster Detection showed significantly higher deviations compared to the previous GNSS alternatives. However, the IMU-MoCap method was limited by discontinuous segment-wise measurements, which required intermittent GPS recalibration. Despite these limitations, the results highlight the potential of alternative localization techniques for digital twin integration.

Localization of an autonomous mobile robot during planetary exploration is challenging due to the unknown terrain, the difficult lighting conditions and the lack of any global reference such as satellite navigation systems. We present a novel approach for robot localization based on ultra-wideband (UWB) technology. The robot sets up its own reference coordinate system by distributing UWB anchor nodes in the environment via a rocket-propelled launcher system. This allows the creation of a localization space in which UWB measurements are employed to supplement traditional SLAM-based techniques. The system was developed for our involvement in the ESA-ESRIC challenge 2021 and the AMADEE-24, an analog Mars simulation in Armenia by the Austrian Space Forum (ÖWF).

Spherical mobile mapping systems are not thoroughly studied in terms of inertial pose estimation filtering. The underlying inherent rolling motion introduces high angular velocities and aggressive system dynamics around all principal axes. This motion profile also needs different modeling compared to state-of-the-art competitors, which heavily focus on more rotationally-restricted systems such as UAV, handheld, or cars. In this work we compare our previously proposed "Delta-filter", which was heavily motivated by the sensors inability to provide covariance estimations, with a Kalman-filter design using a covariance model. Both filters fuse two 6-DoF pose estimators with a motion model in real-time, however the designs are theoretically suitable for an arbitrary number of estimators. We evaluate the trajectories against ground truth pose measurement from an OptiTrackTMmotion capturing system. Furthermore, as our spherical systems are equipped with laser-scanners, we evaluate the resulting point clouds against ground truth maps available from a Riegl VZ400 terrestrial laser-scanner (TLS). Our source code and datasets can be found on github (Arzberger, 2023).

In [1], Algorithm 1 on p. 5007 should appear this way. Algorithm 1: Attitude Estimation From Weights in Case of Symmetries.

In this article, we present SceneFactory, a workflow-centric and unified framework for incremental scene modeling that conveniently supports a wide range of applications, such as (unposed and/or uncalibrated) multiview depth estimation, LiDAR completion, (dense) RGB-D/RGB-LiDAR (RGB-L)/Mono/Depth-only reconstruction, and simultaneous localization and mapping (SLAM). The workflow-centric design uses multiple blocks as the basis for constructing different production lines. The supported applications, i.e., productions avoid redundancy in their designs. Thus, the focus is placed on each block itself for independent expansion. To support all input combinations, our implementation consists of four building blocks that form SceneFactory: first, tracking, second, flexion, third, depth estimation, and fourth, scene reconstruction. The tracking block is based on Mono SLAM and is extended to support RGB-D and RGB-L inputs. Flexion is used to convert the depth image (untrackable) into a trackable image. For general-purpose depth estimation, we propose an unposed and uncalibrated multiview depth estimation model (U-2-MVD) to estimate dense geometry. U-2-MVD exploits dense bundle adjustment to solve for poses, intrinsics, and inverse depth. A semantic-aware ScaleCov step is then introduced to complete the multiview depth. Relying on U-2-MVD, SceneFactory both supports user-friendly 3-D creation (with just images) and bridges the applications of Dense RGB-D and Dense Mono. For high-quality surface and color reconstruction, we propose dual-purpose multiresolutional neural points for the first surface accessible surface color field design, where we introduce improved point rasterization for point cloud-based surface query. We implement and experiment with SceneFactory to demonstrate its broad applicability and high flexibility. Its quality also competes or exceeds the tightly-coupled state of the art approaches in all tasks.

Light detection and ranging (lidar) is valuable during non-cooperative space rendezvous scenarios. By processing the 3D point clouds, it is possible to provide a navigation solution, consisting of an estimate of the relative pose of the approached spacecraft. To enable a safe rendezvous, the pose estimation has to be precise, but also robust if the output is used as a primary navigation solution. Navigation has to be performed in real-time, and onboard computing hardware has a reduced processing capability. Therefore, the real-time requirement is a main driver of the design. Additionally, a spacecraft often has a symmetrical shape. In this case, the pose estimation method has to account for the fact that multiple attitudes represent the same configuration. This work investigates the use of a point-based neural network, or 3D neural network, for the pose estimation task. This network is integrated in a full pose estimation pipeline, where every component is optimized to achieve real-time requirements on a representative onboard computing hardware. After preprocessing, the neural network produces a relative position and attitude estimation in a single-stage, where the attitude estimation considers the symmetries of the spacecraft. Furthermore, a high-fidelity lidar simulator is used, which enables to generate an extensive synthetic dataset. The method is trained and optimized solely on synthetic data. After training, the pose estimation is evaluated on real lidar data acquired at a hardware- in-the-loop rendezvous facility. Results highlight that the method is accurate and robust, without a loss in performance when evaluated on real data. Finally, the flight-readiness is demonstrated by runtime evaluations on an onboard computer candidate, showing that the method is suited for real-time processing.

With computer technology advancing in both software and hardware, the benefits of embodied intelligence are becoming increasingly evident. This robust interactive learning model enables artificial intelligence (AI) to be more flexibly deployed across diverse fields. In recent years, the development of multi-modal large language models (LLMs) has further accelerated the progress of AI, prompting extensive research on how to leverage these advancements to enhance the field of autonomous driving. This perspective believes that embodied intelligence can significantly enhance the application of LLMs, analyzing the new opportunities brought to the mining industry, and emphasizing the potential of their integration to revolutionize various aspects of the field. Meanwhile, This perspective also examines the challenges of deploying embodied agents in mining, while emphasizing their promising future and offering insights into potential research and development avenues.