Abel Gawel

Hermann Blum, Marcus G. Müller, Abel Gawel et al.

In order to operate in human environments, a robot's semantic perception has to overcome open-world challenges such as novel objects and domain gaps. Autonomous deployment to such environments therefore requires robots to update their knowledge and learn without supervision. We investigate how a robot can autonomously discover novel semantic classes and improve accuracy on known classes when exploring an unknown environment. To this end, we develop a general framework for mapping and clustering that we then use to generate a self-supervised learning signal to update a semantic segmentation model. In particular, we show how clustering parameters can be optimized during deployment and that fusion of multiple observation modalities improves novel object discovery compared to prior work. Models, data, and implementations can be found at https://github.com/hermannsblum/scim

Marcus G Müller, Wout Boerdijk, Maximilian Durner et al.

Terrain understanding is fundamental for mobile robots operating in unstructured outdoor environments. Existing vision-based traversability estimation methods rely on robot-specific annotations or semantic class mappings, limiting transferability across platforms and requiring costly re-annotation when robot capabilities change, while standard semantic segmentation methods only focus on specific predefined classes, which do not capture the variety of terrains. In this work, we propose a transformer-based architecture that jointly performs class-specific semantic segmentation and class-agnostic terrain segmentation within a unified network, called Trinity. Terrain regions are segmented based solely on visual appearance, without predefined semantic labels or robot-dependent traversability scores. This formulation enables the learning of robot-agnostic visual terrain priors that can be combined with robot-specific experience for downstream tasks such as traversability estimation, visual odometry, and mission planning. To enable large-scale training with diverse terrain appearances, we extend the OAISYS simulator and introduce RUGDSynth, a synthetic dataset inspired by RUGD with class-agnostic terrain samples. Furthermore, we present the EXTerra Dataset, providing real-world images annotated with both class-specific and class-agnostic terrain labels. Experiments demonstrate the feasibility of the proposed task and the effectiveness of our joint segmentation approach in complex outdoor environments. Code and datasets will be released with this publication (after review).

Andrea Boscolo Camiletto, Alfredo Bochicchio, Alexander Liniger et al.

Efficient relocalization is essential for intelligent vehicles when GPS reception is insufficient or sensor-based localization fails. Recent advances in Bird's-Eye-View (BEV) segmentation allow for accurate estimation of local scene appearance and in turn, can benefit the relocalization of the vehicle. However, one downside of BEV methods is the heavy computation required to leverage the geometric constraints. This paper presents U-BEV, a U-Net inspired architecture that extends the current state-of-the-art by allowing the BEV to reason about the scene on multiple height layers before flattening the BEV features. We show that this extension boosts the performance of the U-BEV by up to 4.11 IoU. Additionally, we combine the encoded neural BEV with a differentiable template matcher to perform relocalization on neural SD-map data. The model is fully end-to-end trainable and outperforms transformer-based BEV methods of similar computational complexity by 1.7 to 2.8 mIoU and BEV-based relocalization by over 26% Recall Accuracy on the nuScenes dataset.

Teng Foong Lam, Hermann Blum, Roland Siegwart et al.

High accuracy 3D surface information is required for many construction robotics tasks such as automated cement polishing or robotic plaster spraying. However, consumer-grade depth cameras currently found in the market are not accurate enough for these tasks where millimeter (mm)-level accuracy is required. This paper presents SL Sensor, a structured light sensing solution capable of producing high fidelity point clouds at 5 Hz by leveraging on phase shifting profilometry (PSP) codification techniques. The SL Sensor was compared with to two commercial depth cameras - the Azure Kinect and RealSense L515. Experiments showed that the SL Sensor surpasses the two devices in both precision and accuracy for indoor surface reconstruction applications. Furthermore, to demonstrate SL Sensor's ability to be a structured light sensing research platform for robotic applications, a motion compensation strategy was developed that allows the SL Sensor to operate during linear motion when traditional PSP methods only work when the sensor is static. Field experiments show that the SL Sensor is able to produce highly detailed reconstructions of spray plastered surfaces. The robot operating system (ROS)-based software and a sample hardware build of the SL Sensor are made open-source with the objective to make structured light sensing more accessible to the construction robotics community. All documentation and code is available at https://github.com/ethz-asl/sl_sensor/ .

Stamatios Georgoulis, Weining Ren, Alfredo Bochicchio et al.

Rapid and reliable identification of dynamic scene parts, also known as motion segmentation, is a key challenge for mobile sensors. Contemporary RGB camera-based methods rely on modeling camera and scene properties however, are often under-constrained and fall short in unknown categories. Event cameras have the potential to overcome these limitations, but corresponding methods have only been demonstrated in smaller-scale indoor environments with simplified dynamic objects. This work presents an event-based method for class-agnostic motion segmentation that can successfully be deployed across complex large-scale outdoor environments too. To this end, we introduce a novel divide-and-conquer pipeline that combines: (a) ego-motion compensated events, computed via a scene understanding module that predicts monocular depth and camera pose as auxiliary tasks, and (b) optical flow from a dedicated optical flow module. These intermediate representations are then fed into a segmentation module that predicts motion segmentation masks. A novel transformer-based temporal attention module in the segmentation module builds correlations across adjacent 'frames' to get temporally consistent segmentation masks. Our method sets the new state-of-the-art on the classic EV-IMO benchmark (indoors), where we achieve improvements of 2.19 moving object IoU (2.22 mIoU) and 4.52 point IoU respectively, as well as on a newly-generated motion segmentation and tracking benchmark (outdoors) based on the DSEC event dataset, termed DSEC-MOTS, where we show improvement of 12.91 moving object IoU.

Hermann Blum, Francesco Milano, René Zurbrügg et al.

We propose a novel robotic system that can improve its perception during deployment. Contrary to the established approach of learning semantics from large datasets and deploying fixed models, we propose a framework in which semantic models are continuously updated on the robot to adapt to the deployment environments. By combining continual learning with self-supervision, our robotic system learns online during deployment without external supervision. We conduct real-world experiments with robots localising in 3D floorplans. Our experiments show how the robot's semantic perception improves during deployment and how this translates into improved localisation, even across drastically different environments. We further study the risk of catastrophic forgetting that such a continuous learning setting poses. We find memory replay an effective measure to reduce forgetting and show how the robotic system can improve even when switching between different environments. On average, our system improves by 60% in segmentation and 10% in localisation accuracy compared to deployment of a fixed model, and it maintains this improvement while adapting to further environments.

Lukas Bernreiter, Abel Gawel, Hannes Sommer et al.

In this paper, we present a semantic mapping approach with multiple hypothesis tracking for data association. As semantic information has the potential to overcome ambiguity in measurements and place recognition, it forms an eminent modality for autonomous systems. This is particularly evident in urban scenarios with several similar looking surroundings. Nevertheless, it requires the handling of a non-Gaussian and discrete random variable coming from object detectors. Previous methods facilitate semantic information for global localization and data association to reduce the instance ambiguity between the landmarks. However, many of these approaches do not deal with the creation of complete globally consistent representations of the environment and typically do not scale well. We utilize multiple hypothesis trees to derive a probabilistic data association for semantic measurements by means of position, instance and class to create a semantic representation. We propose an optimized mapping method and make use of a pose graph to derive a novel semantic SLAM solution. Furthermore, we show that semantic covisibility graphs allow for a precise place recognition in urban environments. We verify our approach using real-world outdoor dataset and demonstrate an average drift reduction of 33 % w.r.t. the raw odometry source. Moreover, our approach produces 55 % less hypotheses on average than a regular multiple hypotheses approach.

Hermann Blum, Julian Stiefel, Cesar Cadena et al.

This paper presents a localization system for mobile robots enabling precise localization in inaccurate building models. The approach leverages local referencing to counteract inherent deviations between as-planned and as-built data for locally accurate registration. We further fuse a novel image-based robust outlier detector with LiDAR data to reject a wide range of outlier measurements from clutter, dynamic objects, and sensor failures. We evaluate the proposed approach on a mobile robot in a challenging real world building construction site. It consistently outperforms the traditional ICP-based alingment, reducing localization error by at least 30%.

Leiv Andresen, Adrian Brandemuehl, Alex Hönger et al.

This paper presents the perception, mapping, and planning pipeline implemented on an autonomous race car. It was developed by the 2019 AMZ driverless team for the Formula Student Germany (FSG) 2019 driverless competition, where it won 1st place overall. The presented solution combines early fusion of camera and LiDAR data, a layered mapping approach, and a planning approach that uses Bayesian filtering to achieve high-speed driving on unknown race tracks while creating accurate maps. We benchmark the method against our team's previous solution, which won FSG 2018, and show improved accuracy when driving at the same speeds. Furthermore, the new pipeline makes it possible to reliably raise the maximum driving speed in unknown environments from 3~m/s to 12~m/s while still mapping with an acceptable RMSE of 0.29~m.

Abel Gawel, Hermann Blum, Johannes Pankert et al.

We present a fully-integrated sensing and control system which enables mobile manipulator robots to execute building tasks with millimeter-scale accuracy on building construction sites. The approach leverages multi-modal sensing capabilities for state estimation, tight integration with digital building models, and integrated trajectory planning and whole-body motion control. A novel method for high-accuracy localization updates relative to the known building structure is proposed. The approach is implemented on a real platform and tested under realistic construction conditions. We show that the system can achieve sub-cm end-effector positioning accuracy during fully autonomous operation using solely on-board sensing.

Nicolas Marchal, Charlotte Moraldo, Roland Siegwart et al.

Deep learning has enabled remarkable advances in scene understanding, particularly in semantic segmentation tasks. Yet, current state of the art approaches are limited to a closed set of classes, and fail when facing novel elements, also known as out of distribution (OoD) data. This is a problem as autonomous agents will inevitably come across a wide range of objects, all of which cannot be included during training. We propose a novel method to distinguish any object (foreground) from empty building structure (background) in indoor environments. We use normalizing flow to estimate the probability distribution of high-dimensional background descriptors. Foreground objects are therefore detected as areas in an image for which the descriptors are unlikely given the background distribution. As our method does not explicitly learn the representation of individual objects, its performance generalizes well outside of the training examples. Our model results in an innovative solution to reliably segment foreground from background in indoor scenes, which opens the way to a safer deployment of robots in human environments.

Luigi Freda, Mario Gianni, Fiora Pirri et al.

Teams of UGVs patrolling harsh and complex 3D environments can experience interference and spatial conflicts with one another. Neglecting the occurrence of these events crucially hinders both soundness and reliability of a patrolling process. This work presents a distributed multi-robot patrolling technique, which uses a two-level coordination strategy to minimize and explicitly manage the occurrence of conflicts and interference. The first level guides the agents to single out exclusive target nodes on a topological map. This target selection relies on a shared idleness representation and a coordination mechanism preventing topological conflicts. The second level hosts coordination strategies based on a metric representation of space and is supported by a 3D SLAM system. Here, each robot path planner negotiates spatial conflicts by applying a multi-robot traversability function. Continuous interactions between these two levels ensure coordination and conflicts resolution. Both simulations and real-world experiments are presented to validate the performances of the proposed patrolling strategy in 3D environments. Results show this is a promising solution for managing spatial conflicts and preventing deadlocks.

Juraj Kabzan, Miguel de la Iglesia Valls, Victor Reijgwart et al.

This paper presents the algorithms and system architecture of an autonomous racecar. The introduced vehicle is powered by a software stack designed for robustness, reliability, and extensibility. In order to autonomously race around a previously unknown track, the proposed solution combines state of the art techniques from different fields of robotics. Specifically, perception, estimation, and control are incorporated into one high-performance autonomous racecar. This complex robotic system, developed by AMZ Driverless and ETH Zurich, finished 1st overall at each competition we attended: Formula Student Germany 2017, Formula Student Italy 2018 and Formula Student Germany 2018. We discuss the findings and learnings from these competitions and present an experimental evaluation of each module of our solution.

Jiadong Guo, Paulo V. K. Borges, Chanoh Park et al.

Place recognition is a challenging problem in mobile robotics, especially in unstructured environments or under viewpoint and illumination changes. Most LiDAR-based methods rely on geometrical features to overcome such challenges, as generally scene geometry is invariant to these changes, but tend to affect camera-based solutions significantly. Compared to cameras, however, LiDARs lack the strong and descriptive appearance information that imaging can provide. To combine the benefits of geometry and appearance, we propose coupling the conventional geometric information from the LiDAR with its calibrated intensity return. This strategy extracts extremely useful information in the form of a new descriptor design, coined ISHOT, outperforming popular state-of-art geometric-only descriptors by significant margin in our local descriptor evaluation. To complete the framework, we furthermore develop a probabilistic keypoint voting place recognition algorithm, leveraging the new descriptor and yielding sublinear place recognition performance. The efficacy of our approach is validated in challenging global localization experiments in large-scale built-up and unstructured environments.

Nikhil Bharadwaj Gosala, Andreas Bühler, Manish Prajapat et al.

In autonomous racing, vehicles operate close to the limits of handling and a sensor failure can have critical consequences. To limit the impact of such failures, this paper presents the redundant perception and state estimation approaches developed for an autonomous race car. Redundancy in perception is achieved by estimating the color and position of the track delimiting objects using two sensor modalities independently. Specifically, learning-based approaches are used to generate color and pose estimates, from LiDAR and camera data respectively. The redundant perception inputs are fused by a particle filter based SLAM algorithm that operates in real-time. Velocity is estimated using slip dynamics, with reliability being ensured through a probabilistic failure detection algorithm. The sub-modules are extensively evaluated in real-world racing conditions using the autonomous race car "gotthard driverless", achieving lateral accelerations up to 1.7G and a top speed of 90km/h.

Fadri Furrer, Tonci Novkovic, Marius Fehr et al.

Collecting 3D object datasets involves a large amount of manual work and is time consuming. Getting complete models of objects either requires a 3D scanner that covers all the surfaces of an object or one needs to rotate it to completely observe it. We present a system that incrementally builds a database of objects as a mobile agent traverses a scene. Our approach requires no prior knowledge of the shapes present in the scene. Object-like segments are extracted from a global segmentation map, which is built online using the input of segmented RGB-D images. These segments are stored in a database, matched among each other, and merged with other previously observed instances. This allows us to create and improve object models on the fly and to use these merged models to reconstruct also unobserved parts of the scene. The database contains each (potentially merged) object model only once, together with a set of poses where it was observed. We evaluate our pipeline with one public dataset, and on a newly created Google Tango dataset containing four indoor scenes with some of the objects appearing multiple times, both within and across scenes.

Hermann Blum, Abel Gawel, Roland Siegwart et al.

Sensor fusion is a fundamental process in robotic systems as it extends the perceptual range and increases robustness in real-world operations. Current multi-sensor deep learning based semantic segmentation approaches do not provide robustness to under-performing classes in one modality, or require a specific architecture with access to the full aligned multi-sensor training data. In this work, we analyze statistical fusion approaches for semantic segmentation that overcome these drawbacks while keeping a competitive performance. The studied approaches are modular by construction, allowing to have different training sets per modality and only a much smaller subset is needed to calibrate the statistical models. We evaluate a range of statistical fusion approaches and report their performance against state-of-the-art baselines on both real-world and simulated data. In our experiments, the approach improves performance in IoU over the best single modality segmentation results by up to 5%. We make all implementations and configurations publicly available.

Abel Gawel, Yukai Lin, Théodore Koutros et al.

Rapid deployment and operation are key requirements in time critical application, such as Search and Rescue (SaR). Efficiently teleoperated ground robots can support first-responders in such situations. However, first-person view teleoperation is sub-optimal in difficult terrains, while a third-person perspective can drastically increase teleoperation performance. Here, we propose a Micro Aerial Vehicle (MAV)-based system that can autonomously provide third-person perspective to ground robots. While our approach is based on local visual servoing, it further leverages the global localization of several ground robots to seamlessly transfer between these ground robots in GPS-denied environments. Therewith one MAV can support multiple ground robots on a demand basis. Furthermore, our system enables different visual detection regimes, and enhanced operability, and return-home functionality. We evaluate our system in real-world SaR scenarios.

Takahiro Miki, Marija Popovic, Abel Gawel et al.

Many robotic applications, such as search-and-rescue, require multiple agents to search for and perform actions on targets. However, such missions present several challenges, including cooperative exploration, task selection and allocation, time limitations, and computational complexity. To address this, we propose a decentralized multi-agent decision-making framework for the search and action problem with time constraints. The main idea is to treat time as an allocated budget in a setting where each agent action incurs a time cost and yields a certain reward. Our approach leverages probabilistic reasoning to make near-optimal decisions leading to maximized reward. We evaluate our method in the search, pick, and place scenario of the Mohamed Bin Zayed International Robotics Challenge (MBZIRC), by using a probability density map and reward prediction function to assess actions. Extensive simulations show that our algorithm outperforms benchmark strategies, and we demonstrate system integration in a Gazebo-based environment, validating the framework's readiness for field application.

Abel Gawel, Carlo Del Don, Roland Siegwart et al.

Global registration of multi-view robot data is a challenging task. Appearance-based global localization approaches often fail under drastic view-point changes, as representations have limited view-point invariance. This work is based on the idea that human-made environments contain rich semantics which can be used to disambiguate global localization. Here, we present X-View, a Multi-View Semantic Global Localization system. X-View leverages semantic graph descriptor matching for global localization, enabling localization under drastically different view-points. While the approach is general in terms of the semantic input data, we present and evaluate an implementation on visual data. We demonstrate the system in experiments on the publicly available SYNTHIA dataset, on a realistic urban dataset recorded with a simulator, and on real-world StreetView data. Our findings show that X-View is able to globally localize aerial-to-ground, and ground-to-ground robot data of drastically different view-points. Our approach achieves an accuracy of up to 85 % on global localizations in the multi-view case, while the benchmarked baseline appearance-based methods reach up to 75 %.

Abel Gawel, Renaud Dubé, Hartmut Surmann et al.

Global registration of heterogeneous ground and aerial mapping data is a challenging task. This is especially difficult in disaster response scenarios when we have no prior information on the environment and cannot assume the regular order of man-made environments or meaningful semantic cues. In this work we extensively evaluate different approaches to globally register UGV generated 3D point-cloud data from LiDAR sensors with UAV generated point-cloud maps from vision sensors. The approaches are realizations of different selections for: a) local features: key-points or segments; b) descriptors: FPFH, SHOT, or ESF; and c) transformation estimations: RANSAC or FGR. Additionally, we compare the results against standard approaches like applying ICP after a good prior transformation has been given. The evaluation criteria include the distance which a UGV needs to travel to successfully localize, the registration error, and the computational cost. In this context, we report our findings on effectively performing the task on two new Search and Rescue datasets. Our results have the potential to help the community take informed decisions when registering point-cloud maps from ground robots to those from aerial robots.

Abel Gawel, Mina Kamel, Tonci Novkovic et al.

Autonomous delivery of goods using a MAV is a difficult problem, as it poses high demand on the MAV's control, perception and manipulation capabilities. This problem is especially challenging if the exact shape, location and configuration of the objects are unknown. In this paper, we report our findings during the development and evaluation of a fully integrated system that is energy efficient and enables MAVs to pick up and deliver objects with partly ferrous surface of varying shapes and weights. This is achieved by using a novel combination of an electro-permanent magnetic gripper with a passively compliant structure and integration with detection, control and servo positioning algorithms. The system's ability to grasp stationary and moving objects was tested, as well as its ability to cope with different shapes of the object and external disturbances. We show that such a system can be successfully deployed in scenarios where an object with partly ferrous parts needs to be gripped and placed in a predetermined location.