IROS 2019 论文列表

Machining applications using robots are still not common in industrial settings. Reasons are the unintuitive programming concepts which typically require expert knowledge and the inflexibility regarding small alterations of the workpieces. We present a prototypical solution for an intuitive and flexible robotic machining concept for unmodeled work pieces. For this we use a high resolution laser scanner to record very dense point clouds. Algorithms to detect linear edges with obtuse angled corners, linear edges with acute angled corners, linear inner edges and circular edges were developed, demonstrated and validated. To accurately execute generated trajectories in practice, an algorithm to directly calibrate the transformation between the sensor and the milling tool was developed. For the algorithms and the calibration process a repeatability tolerance of 0.2 mm is achieved.

We consider the problem of accurate and high-rate self-localization for a mobile robot. We adaptively combine the speed information acquired by proprioceptive sensors with intermittent positioning samples acquired via ultra-wideband (UWB) radios. These are triggered only if and when needed to reduce the positioning uncertainty, itself modeled by a probabilistic cost function. Our formulation is agnostic w.r.t. the source of uncertainty and enables an intuitive specification of user navigation requirements along with stochastic guarantees on the system operation. Experimental results in simulation and with a real platform show that our approach i) meets these guarantees in practice ii) achieves the same accuracy of a fixed periodic sampling but with significantly higher scalability and lower energy consumption iii) is resilient to errors in UWB estimates, enabling the use of low-accuracy ranging schemes which further improve these two performance metrics.

Indoor localization is one of the crucial enablers for deployment of service robots. Although several successful techniques for indoor localization have been proposed, the majority of them relies on maps generated from data gathered with the same sensor modality used for localization. Typically, tedious labor by experts is needed to acquire this data, thus limiting the readiness of the system as well as its ease of installation for inexperienced operators. In this paper, we propose a memory and computationally efficient monocular camera-based localization system that allows a robot to estimate its pose given an architectural floor plan. Our method employs a convolutional neural network to predict room layout edges from a single camera image and estimates the robot pose using a particle filter that matches the extracted edges to the given floor plan. We evaluate our localization system using multiple real-world experiments and demonstrate that it has the robustness and accuracy required for reliable indoor navigation.

Much work in robotics has focused on “humanin-the-loop” learning techniques that improve the efficiency of the learning process. However, these algorithms have made the strong assumption of a cooperating human supervisor that assists the robot. In reality, human observers tend to also act in an adversarial manner towards deployed robotic systems. We show that this can in fact improve the robustness of the learned models by proposing a physical framework that leverages perturbations applied by a human adversary, guiding the robot towards more robust models. In a manipulation task, we show that grasping success improves significantly when the robot trains with a human adversary as compared to training in a self-supervised manner.

Robotic grasping in cluttered environments is often infeasible due to obstacles preventing possible grasps. Then, pre-grasping manipulation like shifting or pushing an object becomes necessary. We developed an algorithm that can learn, in addition to grasping, to shift objects in such a way that their grasp probability increases. Our research contribution is threefold: First, we present an algorithm for learning the optimal pose of manipulation primitives like clamping or shifting. Second, we learn non-prehensible actions that explicitly increase the grasping probability. Making one skill (shifting) directly dependent on another (grasping) removes the need of sparse rewards, leading to more data-efficient learning. Third, we apply a real-world solution to the industrial task of bin picking, resulting in the ability to empty bins completely. The system is trained in a self-supervised manner with around 25 000 grasp and 2500 shift actions. Our robot is able to grasp and file objects with 274±3 picks per hour. Furthermore, we demonstrate the system's ability to generalize to novel objects.

An end-effector such as a gripper or multi-fingered hand is essential to enable robots to grasp and manipulate objects of various size and shape. Soft skin increases the grasp stability and can provide space for tactile sensors. However, covering the joints with skin is challenging, typically causing a considerable surface area of multi-segment robot fingers not to be covered by skin. This also creates the risk that objects get pinched in the joints when flexing the fingers. The current paper suggests using a remote center motion (RCM) mechanism to move the center of joint rotation to the surface of a thick skin layer. In particular, a 6-bar mechanism is used. Thereby, a thick soft skin layer with a continuous surface can be realized. Furthermore, adaptive joint coupling with linkages is implemented. In the current paper a 2-fingered gripper is realized, and objects of various size and shape are grasped (from thin paper to objects of 135 mm diameter). The current gripper was manufactured with 3D-printed material to enable rapid prototyping, therefore the payload was limited to only 1 kg for this version. Overall, this paper shows the feasibility of an RCM for a robot finger and discusses the benefits and limitations of such a mechanism.

Unmanned surface vehicles (USVs) are typically designed for open area marine applications. In this paper, we present a new autonomy system (Roboat) for urban waterways which requires robust localization, perception, planning, and control. A novel localization system, based on the extended Kalman filter (EKF), is proposed for USVs, which utilizes LiDAR, camera, and IMU to provide a decimeter-level precision in dynamic GPS-attenuated urban waterways. Area and shape filters are proposed to crop water reflections and street obstacles from a pointcloud. Euclidean clustering and multi-object contour tracking are then introduced to detect and track the static and moving objects reliably in urban waters. An efficient path planner is tailored to calculate optimal trajectories to avoid these static and dynamic obstacles. Lastly, a nonlinear model predictive control (NMPC) scheme with full state integration is formulated for the four-control-input robot to accurately track the trajectory from the planner in rough water. Extensive experiments show that the robot is able to autonomously navigate in both the indoor waterway and the cluttered outdoor waterway in the presence of static and dynamic obstacles, implying that Roboat could have a great impact on the future of transportation in many coastal and riverside cities.

We present RoFICoM, a new retractable connection device that allows for mechanical, electric, and data communication connection between separable robotic modules. The device is intentionally designed to be used in lattice-type metamorphic robots, however, its applicability is much wider. The main novelty of our solution lies primarily in a new unique flat design and spatial compactness of the connector. With a flat connector, much more space is left for the body of a robotic module in the structure. Moreover, the connector is also fully self-contained device with well defined mechanical, electrical and data interfaces, hence it can be easily embedded in various robotic solutions. Our RoFICoM connector is easy to produce, it is open-hardware and free for non-commercial use. In the paper, we give construction details and report on a couple of experiments we performed to demonstrate key features of the connection achieved with two RoFICoM devices.

Environmental monitoring and surveying operations on rivers currently are performed primarily with manually-operated boats. In this domain, autonomous coverage of areas is of vital importance, for improving both the quality and the efficiency of coverage. This paper leverages human expertise in river exploration and data collection strategies to automate and optimize these processes using autonomous surface vehicles (ASVs). In particular, three deterministic algorithms for both partial and complete coverage of a river segment are proposed, providing varying path length, coverage density, and turning patterns. These strategies resulted in increases in accuracy and efficiency compared to human performance. The proposed methods were extensively tested in simulation using maps of real rivers of different shapes and sizes. In addition, to verify their performance in real world operations, the algorithms were deployed successfully on several parts of the Congaree River in South Carolina, USA, resulting in total of more than 35km of coverage trajectories in the field.

We expand on previous work for negotiating human-robot navigation contention around doorways to produce a more socially-compliant autonomous robot behaviour. Our goal is to improve the integration of robots navigating in human environments by eliciting human recognition of the robot's right of way. This is achieved by incorporating feedback from a user study of our previous system to create a more communicative, reciprocal, and assertive behaviour. Our contribution includes both the updated behaviour and a new user study that evaluates and compares the system to its predecessor. Results show that participants are more likely to respect the robot's right of way given the new robot behaviour, but their responses also highlight the challenges of socially integrating robots into human spaces.

Visual localization is a crucial problem in mobile robotics and autonomous driving. One solution is to retrieve images with known pose from a database for the localization of query images. However, in environments with drastically varying conditions (e.g. illumination changes, seasons, occlusion, dynamic objects), retrieval-based localization is severely hampered and becomes a challenging problem. In this paper, a novel domain-invariant feature learning method (DIFL) is proposed based on ComboGAN, a multi-domain image translation network architecture. By introducing a feature consistency loss (FCL) between the encoded features of the original image and translated image in another domain, we are able to train the encoders to generate domain-invariant features in a self-supervised manner. To retrieve a target image from the database, the query image is first encoded using the encoder belonging to the query domain to obtain a domain-invariant feature vector. We then preform retrieval by selecting the database image with the most similar domain-invariant feature vector. We validate the proposed approach on the CMU-Seasons dataset, where we outperform state-of-the-art learning-based descriptors in retrieval-based localization for high and medium precision scenarios.

Joint attention has been shown to be not only crucial for human-human interaction but also human-robot interaction. Joint attention can help to make cooperation more efficient, support disambiguation in instances of uncertainty and make interactions appear more natural and familiar. In this paper, we present an autonomous gaze system that uses multimodal perception capabilities to model responsive joint attention mechanisms. We investigate the effects of our system on people’s perception of a robot within a problem-solving task. Results from a user study suggest that responsive joint attention mechanisms evoke higher perceived feelings of social presence on scales that regard the direction of the robot’s perception.

Motivated by requirements for future automotive radar, we study the problem of resolving target elevation from measurements by a one-dimensional horizontal radar antenna array. This is a challenging and ill-posed problem, since such measurements contain only indirect and highly ambiguous elevation cues. As a consequence, traditional model-based approaches fail. We instead propose to use a machine-learning-based approach that learns to exploit the subtle elevation cues and prior knowledge of the scene from the data. We design an encoder-decoder structured deep convolutional neural network that takes a radar return intensity image in the range-azimuth plane as input and produces a depth image in the elevation-azimuth plane as output. We train the network with over 200 000 radar frames collected in highway environments. Through experimental evaluations, we demonstrate the feasibility of resolving the highly ambiguous elevation information in such environments.

We propose a method to maintain high resource availability in a networked heterogeneous multi-robot system subject to resource failures. In our model, resources such as sensing and computation are available on robots. The robots are engaged in a joint task using these pooled resources. When a resource on a particular robot becomes unavailable (e.g., a sensor ceases to function), the system automatically reconfigures so that the robot continues to have access to this resource by communicating with other robots. Specifically, we consider the problem of selecting edges to be modified in the system's communication graph after a resource failure has occurred. We define a metric that allows us to characterize the quality of the resource distribution in the network represented by the communication graph. Upon a resource becoming unavailable due to failure, we reconFigure the network so that the resource distribution is brought as close to the maximal resource distribution as possible without a large change in the number of active inter-robot communication links. Our approach uses mixed integer semi-definite programming to achieve this goal. We employ a simulated annealing method to compute a spatial formation that satisfies the inter-robot distances imposed by the topology, along with other constraints. Our method can compute a communication topology, spatial formation, and formation change motion planning in a few seconds. We validate our method in simulation and real-robot experiments with a team of seven quadrotors.

The enrichment of the application scenarios of rotorcrafts presents new challenges for the study of their aerodynamic characteristics, such as operating above a building surface of finite size. In this paper, the ground effect is divided into infinite ground effect and finite ground effect, and three types of finite ground effects with different blocked area are studied. Through numerical simulations, the rotor thrust data and flow field figures in ground effect are obtained. Based on the rotor thrust data, mathematical models are established to describe the rotor thrust alteration caused by infinite and finite ground effect. The analysis of the flow field reveals the mechanism of the finite ground effect.

Since deep-learning-based method has been widely-used and is capable of generating generic model, most existing methods about action recognition use either two-stream structure, considering spatial and temporal features separately, or C3D, costing lots of prices in memory and time. We aim to design a robust system to extract spatiotemporal features with aggregation mechanism to integrate local features in temporal order. In light of this, we propose ResFlow to estimate optical flow and predict action recognition simultaneously. Leveraging the characteristic of optical flow estimation, we extract spatiotemporal feature via an autoencoder. Via a novel Sequentially Pooling Mechanism which literally pool global spatiotemporal feature sequentially, we extract spatiotemporal feature at each time and aggregate these local features into global feature. This design use only RGB images as input with temporal information encoded, pre-trained by optical flow, and sequentially aggregate spatiotemporal features in high efficiency. We evaluate our ability of estimating optical flow on FlyingChairs dataset and show the promising results of action recognition on UCF-101 dataset through a series of experiments.

It is difficult to create robust, reusable, and reactive behaviors for robots that can be easily extended and combined. Frameworks such as Behavior Trees are flexible but difficult to characterize, especially when designing reactions and recovery behaviors to consistently converge to a desired goal condition. We propose a framework which we call Robust Logical-Dynamical Systems (RLDS), which combines the advantages of task representations like behavior trees with theoretical guarantees on performance. RLDS can also be constructed automatically from simple sequential task plans and will still achieve robust, reactive behavior in dynamic real-world environments. In this work, we describe both our proposed framework and a case study on a simple household manipulation task, with examples for how specific pieces can be implemented to achieve robust behavior. Finally, we show how in the context of these manipulation tasks, a combination of an RLDS with planning can achieve better results under adversarial conditions.

Three-dimensional (3D) point cloud processing has attracted a great deal of attention in computer vision, robotics, and the machine learning community because of significant progress in deep neural networks on 3D data. Another trend in the community is learning of generative models based on generative adversarial networks. In this paper, we propose a framework for 3D point cloud generation based on a combination of auto-encoders and generative adversarial networks. The framework first trains auto-encoders to learn latent representations, and then trains generative adversarial networks in the learned latent space. We focus on improving the training method for auto-encoders in order to generate 3D point clouds with higher fidelity and coverage. We add parallel sub-decoders that reconstruct subsets of the input point cloud. In order to construct these subsets, we introduce a point sampling algorithm that imposes a method to sample spatially localized point sets. These local subsets are utilized to measure local reconstruction losses. We train auto-encoders to learn an effective latent representation for both global and local shape reconstruction based on the multi-task learning approach. Furthermore, we add global and local adversarial losses to generate more plausible point clouds. Quantitative and qualitative evaluations demonstrate that the proposed method outperforms state-of-the-art method on the task of 3D point cloud generation.

This paper proposes a practical and less conservative passivity analysis for series elastic actuators (SEAs) by introducing load port definition and shows that the achievable stiffness by the impedance control of SEA can be set higher than the inherent stiffness of SEA depending on the condition of the load dynamics. Since SEA can inherently measure or estimate a transmitted force thanks to its embedded spring element, impedance control is often exploited to render compliant behaviors related between the motion and the force. Although the stability of the SEA control system is of great importance, the conventional passivity analysis gives conservative criteria, and indeed limits the actual actuator performance. To tackle the conservatism of the conventional passivity in SEAs, we first explore the dynamic characteristics of SEA including load dynamics, which has been ignored for the sake of simplicity of the passivity analysis by excluding uncertain load dynamics. The inclusion of the load dynamics into the passivity analysis allows us to properly derive the less conservative limit of achievable stiffness by impedance control and the factors that determine the limit. The proposed analysis is verified by numerical simulations and applied to a passivity observer design for experimental validation on an actual SEA setup.