Localization and mapping are key capabilities of autonomous systems. In this paper, we propose a modified Siamese network to estimate the similarity between pairs of LiDAR scans recorded by autonomous cars. This can be used to address both, loop closing for SLAM and global localization. Our approach utilizes a deep neural network exploiting different cues generated from LiDAR data. It estimates the similarity between pairs of scans using the concept of image overlap generalized to range images and furthermore provides a relative yaw angle estimate. Based on such predictions, our method is able to detect loop closures in a SLAM system or to globally localize in a given map. For loop closure detection, we use the overlap prediction as the similarity measurement to find loop closure candidates and integrate the candidate selection into an existing SLAM system to improve the mapping performance. For global localization, we propose a novel observation model using the predictions provided by OverlapNet and integrate it into a Monte-Carlo localization framework. We evaluate our approach on multiple datasets collected using different LiDAR scanners in various environments. The experimental results show that our method can effectively detect loop closures surpassing the detection performance of state-of-the-art methods and that it generalizes well to different environments. Furthermore, our method reliably localizes a vehicle in typical urban environments globally using LiDAR data collected in different seasons.

Many modern robotics applications rely on 3D maps of the environment. Due to the large memory requirements of dense 3D maps, compression techniques are often necessary to store or transmit 3D maps efficiently. In this work, we investigate the problem of compressing dense 3D point cloud maps such as those obtained from an autonomous vehicle in large outdoor environments. We tackle the problem by learning a set of local feature descriptors from which the point cloud can be reconstructed efficiently and effectively. We propose a novel deep convolutional autoencoder architecture that directly operates on the points themselves so that we avoid voxelization. Additionally, we propose a deconvolution operator to upsample point clouds, which allows us to decompress to an arbitrary density. Our experiments show that our learned compression achieves better reconstructions at the same bit rate compared to other state-of-the-art compression algorithms. We furthermore demonstrate that our approach generalizes well to different LiDAR sensors. For example, networks learned on maps generated from KITTI point clouds still achieve state-of-the-art compression results for maps generated from nuScences point clouds.

The application of autonomous robots in agriculture is gaining increasing popularity thanks to the high impact it may have on food security, sustainability, resource use efficiency, reduction of chemical treatments, and the optimization of human effort and yield. With this vision, the Flourish research project aimed to develop an adaptable robotic solution for precision farming that combines the aerial survey capabilities of small autonomous unmanned aerial vehicles (UAVs) with targeted intervention performed by multi-purpose unmanned ground vehicles (UGVs). This paper presents an overview of the scientific and technological advances and outcomes obtained in the project. We introduce multi-spectral perception algorithms and aerial and ground-based systems developed for monitoring crop density, weed pressure, crop nitrogen nutrition status, and to accurately classify and locate weeds. We then introduce the navigation and mapping systems tailored to our robots in the agricultural environment, as well as the modules for collaborative mapping. We finally present the ground intervention hardware, software solutions, and interfaces we implemented and tested in different field conditions and with different crops. We describe a real use case in which a UAV collaborates with a UGV to monitor the field and to perform selective spraying without human intervention.

Precision management of agricultural fields as well as plant breeding are central factors for keeping yields high and to provide food, feed, and fiber for our society. A key element in breeding trials but also for targeted management actions is to analyze the growth state of individual plants objectively and at a large scale. In this letter, we address the problem of analyzing crops in real agricultural fields based on camera data recorded with mobile robots and to derive information about the plant development, e.g., to monitor phenotypic traits such as growth stage. We propose a novel single-stage object detection approach that localizes crops and weeds in the field. At the same time, it detects plant-specific leaf keypoints intending to estimate leaf count at a plant level, which is a key trait for classifying the growth stage. We implemented and thoroughly tested our approach on real sugar beet fields. As our experiments show, it performs the required detections and shows superior performance with respect to a state-of-the-art two-stage approach based on Mask R-CNN.

Knowing the distance to nearby objects is crucial for autonomous cars to navigate safely in everyday traffic. In this paper, we investigate monocular depth estimation, which advanced substantially within the last years and is providing increasingly more accurate results while only requiring a single camera image as input. In line with recent work, we use an encoder-decoder structure with so-called packing layers to estimate depth values in a self-supervised fashion. We propose integrating a joint pre-training of semantic segmentation plus depth estimation on a dataset providing semantic labels. By using a separate semantic decoder that is only needed for pre-training, we can keep the network comparatively small. Our extensive experimental evaluation shows that the addition of such pre-training improves the depth estimation performance substantially. Finally, we show that we achieve competitive performance on the KITTI dataset despite using a much smaller and more efficient network.

Panoptic segmentation is the recently introduced task that tackles semantic segmentation and instance segmentation jointly [18]. In this paper, we present an extension of SemanticKITTI [1], a large-scale dataset providing dense point-wise semantic labels for all sequences of the KITTI Odometry Benchmark [10]. This extension enables training and evaluation of LiDAR-based panoptic segmentation. We provide the data and discuss the processing steps needed to enrich a given semantic annotation with temporally consistent instance information, i.e., instance information that supplements the semantic labels and identifies the same instance over sequences of LiDAR point clouds. Additionally, we present two strong baselines that combine state-of-the-art LiDAR-based semantic segmentation approaches with a state-of-the-art detector enriching the segmentation with instance information and that allow other researchers to compare their approaches against. We believe that our extension of SemanticKITTI with strong baselines enables the creation of novel algorithms for LiDAR-based panoptic segmentation as much as it has for the original semantic segmentation and semantic scene completion tasks. Data, code, and an online evaluation service using a hidden test set are publicly available at http://semantic-kitti.org.

A holistic semantic scene understanding exploiting all available sensor modalities is a core capability to master self-driving in complex everyday traffic. To this end, we present the SemanticKITTI dataset that provides point-wise semantic annotations of Velodyne HDL-64E point clouds of the KITTI Odometry Benchmark. Together with the data, we also published three benchmark tasks for semantic scene understanding covering different aspects of semantic scene understanding: (1) semantic segmentation for point-wise classification using single or multiple point clouds as input; (2) semantic scene completion for predictive reasoning on the semantics and occluded regions; and (3) panoptic segmentation combining point-wise classification and assigning individual instance identities to separate objects of the same class. In this article, we provide details on our dataset showing an unprecedented number of fully annotated point cloud sequences, more information on our labeling process to efficiently annotate such a vast amount of point clouds, and lessons learned in this process. The dataset and resources are available at http://www.semantic-kitti.org.

Conventional farming still relies on large quantities of agrochemicals for weed management which have several negative side‐effects on the environment. Autonomous robots offer the potential to reduce the amount of chemicals applied, as robots can monitor and treat each plant in the field individually and thereby circumventing the uniform chemical treatment of the whole field. Such agricultural robots need the ability to identify individual crops and weeds in the field using sensor data and must additionally select effective treatment methods based on the type of weed. For example, certain types of weeds can only be effectively treated mechanically due to their resistance to herbicides, whereas other types can be treated trough selective spraying. In this article, we present a novel system that provides the necessary information for effective plant‐specific treatment. It estimates the stem location for weeds, which enables the robots to perform precise mechanical treatment, and at the same time provides the pixel‐accurate area covered by weeds for treatment through selective spraying. The major challenge in developing such a system is the large variability in the visual appearance that occurs in different fields. Thus, an effective classification system has to robustly handle substantial environmental changes including varying weed pressure, various weed types, different growth stages, changing visual appearance of the plants and the soil. Our approach uses an end‐to‐end trainable fully convolutional network that simultaneously estimates plant stem positions as well as the spatial extent of crop plants and weeds. It jointly learns how to detect the stems and the pixel‐wise semantic segmentation and incorporates spatial information by considering image sequences of local field strips. The jointly learned feature representation for both tasks furthermore exploits the crop arrangement information that is often present in crop fields. This information is considered even if it is only observable from the image sequences and not a single image. Such image sequences, as typically provided by robots navigating over the field along crop rows, enable our approach to robustly estimate the semantic segmentation and stem positions despite the large variations encountered in different fields. We implemented and thoroughly tested our approach on images from multiple farms in different countries. The experiments show that our system generalizes well to previously unseen fields under varying environmental conditions—a key capability to deploy such systems in the real world. Compared to state‐of‐the‐art approaches, our approach generalizes well to unseen fields and not only substantially improves the stem detection accuracy, that is, distinguishing crop and weed stems, but also improves the semantic segmentation performance.

Truly autonomous driving without the need for human intervention can only be attained when self-driving cars fully understand their surroundings. Most of these vehicles rely on a suite of active and passive sensors. LiDAR sensors are a cornerstone in most of these hardware stacks, and leveraging them as a complement to other passive sensors such as RGB cameras is an enticing goal. Understanding the semantic class of each point in a LiDAR sweep is important, as well as knowing to which instance of that class it belongs to. To this end, we present a novel, single-stage, and real-time capable panoptic segmentation approach using a shared encoder with a semantic and instance decoder. We leverage the geometric information of the LiDAR scan to perform a novel, distance-aware tri-linear upsampling, which allows our approach to use larger output strides than using transpose convolutions leading to substantial savings in computation time. Our experimental evaluation and ablation studies for each module show that combining our geometric and semantic embeddings with our learned, variable instance thresholds, a category-specific loss, and the novel trilinear upsampling module leads to higher panoptic quality. We will release the code of our approach in our LiDAR processing library LiDAR-Bonnetal [27].

Inferring semantic information towards an understanding of the surrounding environment is crucial for autonomous vehicles to drive safely. Deep learning-based segmentation methods can infer semantic information directly from laser range data, even in the absence of other sensor modalities such as cameras. In this paper, we address improving the generalization capabilities of such deep learning models to range data that was captured using a different sensor and in situations where no labeled data is available for the new sensor setup. Our approach assists the domain transfer of a LiDAR-only semantic segmentation model to a different sensor and environment exploiting existing geometric mapping systems. To this end, we fuse sequential scans in the source dataset into a dense mesh and render semi-synthetic scans that match those of the target sensor setup. Unlike simulation, this approach provides a real-to-real transfer of geometric information and delivers additionally more accurate remission information. We implemented and thoroughly tested our approach by transferring semantic scans between two different real-world datasets with different sensor setups. Our experiments show that we can improve the segmentation performance substantially with zero manual re-labeling. This approach solves the number one feature request since we released our semantic segmentation library LiDAR-bonnetal [18].

Simultaneous localization and mapping (SLAM) is a fundamental capability required by most autonomous systems. In this paper, we address the problem of loop closing for SLAM based on 3D laser scans recorded by autonomous cars. Our approach utilizes a deep neural network exploiting different cues generated from LiDAR data for finding loop closures. It estimates an image overlap generalized to range images and provides a relative yaw angle estimate between pairs of scans. Based on such predictions, we tackle loop closure detection and integrate our approach into an existing SLAM system to improve its mapping results. We evaluate our approach on sequences of the KITTI odometry benchmark and the Ford campus dataset. We show that our method can effectively detect loop closures surpassing the detection performance of state-of-the-art methods. To highlight the generalization capabilities of our approach, we evaluate our model on the Ford campus dataset while using only KITTI for training. The experiments show that the learned representation is able to provide reliable loop closure candidates, also in unseen environments.

Humanoid robots are often supposed to share their workspace with humans and thus have to deal with objects used by humans in their everyday life. In this article, we present our novel approach to humanoid navigation through cluttered environments, which exploits knowledge about different obstacle classes to decide how to deal with obstacles and select appropriate robot actions. To classify objects from RGB images and decide whether an obstacle can be overcome by the robot with a corresponding action, e.g., by pushing or carrying it aside or stepping over or onto it, we train and exploit a convolutional neural network (CNN). Based on associated action costs, we compute a cost grid containing newly observed objects in addition to static obstacles on which a 2D path can be efficiently planned. This path encodes the necessary actions that need to be carried out by the robot to reach the goal. We implemented our framework in the Robot Operating System (ROS) and tested it in various scenarios with a Nao robot as well as in simulation with the REEM-C robot. As the experiments demonstrate, using our CNN, the robot can robustly classify the observed obstacles into the different classes and decide on suitable actions to find efficient solution paths. Our system finds paths also through regions where traditional motion planning methods are not able to calculate a solution or require substantially more time.

Annotated datasets are essential for supervised learning. However, annotating large datasets is a tedious and time-intensive task. This paper addresses active learning in the context of semantic segmentation with the goal of reducing the human labeling effort. Our application is agricultural robotics and we focus on the task of distinguishing between crop and weed plants from image data. A key challenge in this application is the transfer of an existing semantic segmentation CNN to a new field, in which growth stage, weeds, soil, and weather conditions differ. We propose a novel approach that, given a trained model on one field together with rough foreground segmentation, refines the network on a substantially different field providing an effective method of selecting samples to annotate for supporting the transfer. We evaluated our approach on two challenging datasets from the agricultural robotics domain and show that we achieve a higher accuracy with a smaller number of samples compared to random sampling as well as entropy based sampling, which consequently reduces the required human labeling effort.

Semantic scene understanding is important for various applications. In particular, self-driving cars need a fine-grained understanding of the surfaces and objects in their vicinity. Light detection and ranging (LiDAR) provides precise geometric information about the environment and is thus a part of the sensor suites of almost all self-driving cars. Despite the relevance of semantic scene understanding for this application, there is a lack of a large dataset for this task which is based on an automotive LiDAR. In this paper, we introduce a large dataset to propel research on laser-based semantic segmentation. We annotated all sequences of the KITTI Vision Odometry Benchmark and provide dense point-wise annotations for the complete $360^{o}$ field-of-view of the employed automotive LiDAR. We propose three benchmark tasks based on this dataset: (i) semantic segmentation of point clouds using a single scan, (ii) semantic segmentation using multiple past scans, and (iii) semantic scene completion, which requires to anticipate the semantic scene in the future. We provide baseline experiments and show that there is a need for more sophisticated models to efficiently tackle these tasks. Our dataset opens the door for the development of more advanced methods, but also provides plentiful data to investigate new research directions.

Semantic scene understanding is important for autonomous robots that aim to navigate dynamic environments, manipulate objects, or interact with humans in a natural way. In this paper, we address the problem of jointly performing semantic segmentation as well as instance segmentation in an online fashion, so that autonomous robots can use this information on-the-go and without sacrificing accuracy. We achieve this by exploiting a local connectivity prior of objects in the real world and a multi-task convolutional neural network architecture. The network identifies the individual object instances and their classes without region proposals or pre-segmentation of the images into individual classes. We implemented and thoroughly evaluated our approach, and our experiments suggest that our method can be used to accurately segment instance masks of objects and identify their class in an online fashion.

Reliable and accurate localization and mapping are key components of most autonomous systems. Besides geometric information about the mapped environment, the semantics plays an important role to enable intelligent navigation behaviors. In most realistic environments, this task is particularly complicated due to dynamics caused by moving objects, which can corrupt the mapping step or derail localization. In this paper, we propose an extension of a recently published surfel-based mapping approach exploiting three-dimensional laser range scans by integrating semantic information to facilitate the mapping process. The semantic information is efficiently extracted by a fully convolutional neural network and rendered on a spherical projection of the laser range data. This computed semantic segmentation results in point-wise labels for the whole scan, allowing us to build a semantically-enriched map with labeled surfels. This semantic map enables us to reliably filter moving objects, but also improve the projective scan matching via semantic constraints. Our experimental evaluation on challenging highways sequences from KITTI dataset with very few static structures and a large amount of moving cars shows the advantage of our semantic SLAM approach in comparison to a purely geometric, state-of-the-art approach.

Semantic scene understanding is important for various applications. In particular, self-driving cars need a fine-grained understanding of the surfaces and objects in their vicinity. Light detection and ranging (LiDAR) provides precise geometric information about the environment and is thus a part of the sensor suites of almost all self-driving cars. Despite the relevance of semantic scene understanding for this application, there is a lack of a large dataset for this task which is based on an automotive LiDAR. In this paper, we introduce a large dataset to propel research on laser-based semantic segmentation. We annotated all sequences of the KITTI Vision Odometry Benchmark and provide dense point-wise annotations for the complete $360^{o}$ field-of-view of the employed automotive LiDAR. We propose three benchmark tasks based on this dataset: (i) semantic segmentation of point clouds using a single scan, (ii) semantic segmentation using sequences comprised of multiple past scans, and (iii) semantic scene completion, which requires to anticipate the semantic scene in the future. We provide baseline experiments and show that there is a need for more sophisticated models to efficiently tackle these tasks. Our dataset opens the door for the development of more advanced methods, but also provides plentiful data to investigate new research directions.

Perception in autonomous vehicles is often carried out through a suite of different sensing modalities. Given the massive amount of openly available labeled RGB data and the advent of high-quality deep learning algorithms for image-based recognition, high-level semantic perception tasks are pre-dominantly solved using high-resolution cameras. As a result of that, other sensor modalities potentially useful for this task are often ignored. In this paper, we push the state of the art in LiDAR-only semantic segmentation forward in order to provide another independent source of semantic information to the vehicle. Our approach can accurately perform full semantic segmentation of LiDAR point clouds at sensor frame rate. We exploit range images as an intermediate representation in combination with a Convolutional Neural Network (CNN) exploiting the rotating LiDAR sensor model. To obtain accurate results, we propose a novel post-processing algorithm that deals with problems arising from this intermediate representation such as discretization errors and blurry CNN outputs. We implemented and thoroughly evaluated our approach including several comparisons to the state of the art. Our experiments show that our approach outperforms state-of-the-art approaches, while still running online on a single embedded GPU. The code can be accessed at https://github.com/PRBonn/lidar-bonnetal.

Reliable and accurate localization and mapping are key components of most autonomous systems. Besides geometric information about the mapped environment, the semantics plays an important role to enable intelligent navigation behaviors. In most realistic environments, this task is particularly complicated due to dynamics caused by moving objects, which can corrupt the mapping step or derail localization. In this paper, we propose an extension of a recently published surfel-based mapping approach exploiting three-dimensional laser range scans by integrating semantic information to facilitate the mapping process. The semantic information is efficiently extracted by a fully convolutional neural network and rendered on a spherical projection of the laser range data. This computed semantic segmentation results in point-wise labels for the whole scan, allowing us to build a semantically-enriched map with labeled surfels. This semantic map enables us to reliably filter moving objects, but also improve the projective scan matching via semantic constraints. Our experimental evaluation on challenging highways sequences from KITTI dataset with very few static structures and a large amount of moving cars shows the advantage of our semantic SLAM approach in comparison to a purely geometric, state-of-the-art approach.