IROS 2018 论文列表

The remote center of motion (RCM) mechanism is a prominent candidate to aid bone drilling. The surgeon can simply place a drill with the RCM mechanism near the entry point to provide drill alignment with the target. Using this assistive mechanism for bone drilling improves drilling accuracy and reduces the complexity of bone drilling robotic systems. However, because most RCM mechanisms have been developed for laparoscopic surgery or needle insertion into soft tissue, they lack rigidity and are unsuitable for bone drilling. One of the most difficult and important surgical procedures in bone drilling is maintaining as well as guiding the orientation of the drill with respect to the target. This paper proposes an improved RCM mechanism in which a pair of linear actuators and a gearless arc-guide are employed to achieve high rigidity and resolution, which enable bone drilling. A vision-guided navigation system is also integrated into the proposed system to automatically guide the orientation. To verify that the proposed RCM mechanism has sufficient rigidity and targeting accuracy, a series of experiments was performed. The results obtained confirm that the proposed mechanism can maintain its tilting angle under up to 50 N, with a targeting error of approximately 0.28mm.

In this paper, we present the Trifo Visual Inertial Odometry (Trifo-VIO), a tightly-coupled filtering-based stereo VIO system using both points and lines. Line features help improve system robustness in challenging scenarios when point features cannot be reliably detected or tracked, e.g. low-texture environment or lighting change. In addition, we propose a novel lightweight filtering-based loop closing technique to reduce accumulated drift without global bundle adjustment or pose graph optimization. We formulate loop closure as EKF updates to optimally relocate the current sliding window maintained by the filter to past keyframes. We also present the Trifo Ironsides dataset, a new visual-inertial dataset, featuring high-quality synchronized stereo camera and IMU data from the Ironsides sensor [3] with various motion types and textures and millimeter-accuracy groundtruth. To validate the performance of the proposed system, we conduct extensive comparison with state-of-the-art approaches (OKVIS, VINS-MONO and S-MSCKF) using both the public EuRoC dataset and the Trifo Ironsides dataset.

Tree species identification using bark images is a challenging problem that could prove useful for many forestry related tasks. However, while the recent progress in deep learning showed impressive results on standard vision problems, a lack of datasets prevented its use on tree bark species classification. In this work, we present, and make publicly available, a novel dataset called BarkNet 1.0 containing more than 23,000 high-resolution bark images from 23 different tree species over a wide range of tree diameters. With it, we demonstrate the feasibility of species recognition through bark images, using deep learning. More specifically, we obtain an accuracy of 93.88% on single crop, and an accuracy of 97.81% using a majority voting approach on all of the images of a tree. We also empirically demonstrate that, for a fixed number of images, it is better to maximize the number of tree individuals in the training database, thus directing future data collection efforts.

This paper presents a novel optimization-based passivity control algorithm for haptic-enabled bilateral teleoperation systems involving multiple degrees of freedom. In particular, in the context of energy-bounding control, the contribution focuses on the implementation of a passivity layer for an existing time-domain scheme, ensuring optimal transparency of the interaction along subsets of the environment space which are preponderant for the given task, while preserving the energy bounds required for passivity. The involved optimization problem is convex and amenable to real-time implementation. The effectiveness of the proposed design is validated via an experiment performed on a virtual teleoperated environment.

Hand-eye coordination is a requirement for many manipulation tasks including grasping and reaching. However, accurate hand-eye coordination has shown to be especially difficult to achieve in complex robots like the iCub humanoid. In this work, we solve the hand-eye coordination task using a visuomotor deep neural network predictor that estimates the arm's joint configuration given a stereo image pair of the arm and the underlying head configuration. As there are various unavoidable sources of sensing error on the physical robot, we train the predictor on images obtained from simulation. The images from simulation were modified to look realistic using an image-to-image translation approach. In various experiments, we first show that the visuomotor predictor provides accurate joint estimates of the iCub's hand in simulation. We then show that the predictor can be used to obtain the systematic error of the robot's joint measurements on the physical iCub robot. We demonstrate that a calibrator can be designed to automatically compensate this error. Finally, we validate that this enables accurate reaching of objects while circumventing manual fine-calibration of the robot.

One desirable capability of autonomous cars is to accurately predict the pedestrian motion near intersections for safe and efficient trajectory planning. We are interested in developing transfer learning algorithms that can be trained on the pedestrian trajectories collected at one intersection and yet still provide accurate predictions of the trajectories at another, previously unseen intersection. We first discussed the feature selection for transferable pedestrian motion models in general. Following this discussion, we developed one transferable pedestrian motion prediction algorithm based on Inverse Reinforcement Learning (IRL) that infers pedestrian intentions and predicts future trajectories based on observed trajectory. We evaluated our algorithm at three intersections. We used the accuracy of augmented semi-nonnegative sparse coding (ASNSC), trained and tested at the same intersection as a baseline. The result shows that the proposed algorithm improves the baseline accuracy by a statistically significant percentage in both non-transfer task and transfer task.

We describe a trajectory planning method for heterogeneous mobile robot teams in known environments. We consider two core problems that arise with heterogeneous robot teams: asymmetric inter-robot collision constraints and varying dynamic limits. Asymmetric collision constraints are important for close-proximity flight of rotorcraft due to the downwash effect, which complicates spatial coordination. Varying dynamic limits complicate temporal coordination between robots and must be taken into account during planning. Our method builds upon a hybrid planner that combines graph-planning techniques with trajectory optimization and scales well to large homogeneous robot teams. We extend the hybrid planning approach to include the additional spatial and temporal coordination to support heterogeneous teams. Our method scales well with the number of robots and robot types and we demonstrate our approach on a team of 15 physical robots of 4 different types, including quadrotors and differential drive robots.

Emerging robot-assisted endovascular intervention has the potential to reduce X-ray radiations to the operator while enhancing the stability and dexterity of catheter manipulation. Supervised and shared autonomy of endovascular procedures could add further improvements in reduced fatigue and cognitive workloads of the operator, higher success rates of cannulation and improved surgical outcomes. However, robotic path planning for endovascular procedure is challenging due to complex and non-linear flow dynamics inside the vasculature. This paper presents a learning-based robotic catheterization platform addressing those challenges, this approach incorporates path integral reinforcement learning (RL) framework based on dynamic movement primitives (DMP) to enhance catheterization tasks by a customized robotic manipulator. The robotic trajectories were optimized through RL in order to avoid unwanted contacts between the catheter tip and the vessel wall. The proposed methods can adapt to different flow simulations, vascular models, and catheterization tasks. The quality of the catheterization was evaluated with performance metrics. The results show significant refinement of catheter paths by the proposed approach, resulting in shorter overall lengths and fewer contact forces, which can potentially reduce risks in endothelial wall damages, embolization, and stroke. The results support the development of robotic path planning for endovascular procedures as well as designing intelligent, hands-on robotic navigation platforms.

Artificial agents might not understand human interests and actions if these agents cannot anticipate how a person understands a situation and, based on this, what could be his/her expectations. In many cases, understanding, expectations and behaviors are constrained, if not driven, by culture. Can we provide human culture to an artificial agent? Can we provide formal representations of different cultures? In this paper we discuss the (elusive) notion of culture and propose an approach based on the notion of trait which, we argue, allows building formal modules suitable to represent culture (broadly understood). We distinguish the trait types (knowledge, rule, behavior, interpretation) that such modules should contain and briefly discuss how they could be organized. Finally, we exemplify the role of a trait module in the flow of information internal to an agent highlighting surprising potentialities.

In order to perform safe robotic interventions in harsh environments it is necessary to help the robotic operator with a Human-Robot Interface that provides multimodal interactions, from low level interaction methods to bilateral teleoperation with force feedback. These interaction modalities, though, rely purely on the operator's skills. With the objective of providing a safer system, higher-level applications can be integrated in the interface in order to provide some help to the operator, without relying uniquely on his/her capacities. This paper presents a novel object recognition and tracking system which runs in real-time on the robot while the operator is operating it. The tracking system enters in the teleoperation loop and helps the operator to achieve the requested goals. The system is optimized to track featureless objects such as metallic plates, metallic connectors and monochromatic objects. Moreover, the algorithm provides improvements with respect to previous tracking experiments, including depth estimation in order to better interact with the velocity control of the robotic arm when approaching the target, as well as high reliability with partial occlusions. This vision-based control system is used in real interventions in hazardous environments, in order to track and manipulate metallic parts of scientific and engineering machines, giving a performance success over 95%, and reaching the 100% under the remote human supervision.

We propose a method to track from a multi-rotor drone a moving source, such as a human speaker or an emergency whistle, whose sound is mixed with the strong ego-noise generated by rotating motors and propellers. The proposed method is independent of the specific drone and does not need pre-training nor reference signals. We first employ a time-frequency spatial filter to estimate, on short audio segments, the direction of arrival of the moving source and then we track these noisy estimations with a particle filter. We quantitatively evaluate the results using a ground-truth trajectory of the sound source obtained with an on-board camera and compare the performance of the proposed method with baseline solutions.

This paper represents a step towards vision-based manipulation of plastic materials. Manipulating deformable objects is made challenging by: 1) the absence of a model for the object deformation, 2) the inherent difficulty of visual tracking of deformable objects, 3) the difficulty in defining a visual error and 4) the difficulty in generating control inputs to minimise the visual error. We propose a novel representation of the task of manipulating deformable objects. In this preliminary case study, the shaping of kinetic sand, we assume a finite set of actions: pushing, tapping and incising. We consider that these action types affect only a subset of the state, i.e., their effect does not affect the entire state of the system (specialized actions). We report the results of a user study to validate these hypotheses and release the recorded dataset. The actions (pushing, tapping and incising) are clearly adopted during the task, although it is clear that 1) participants use also mixed actions and 2) actions' effects can marginally affect the entire state, requesting a relaxation of our specialized actions hypothesis. Moreover, we compute task errors and corresponding control inputs (in the image space) using image processing. Finally, we show how machine learning can be applied to infer the mapping from error to action on the data extracted from the user study.

Percutaneous Nephrolithotomy is a recommended treatment method for large kidney stone removal. However, the first and most important step, i.e., getting the percutaneous access to create the tract between the targeted calyx and the flank skin, is challenging as the surgeon is often occupied by several tasks at a given time. Therefore, in this paper, we propose a robotic assisted system that collaborates with the surgeon and provides assistance in order for the surgeons to focus on more critical jobs resulting in better surgical performance. A procedure for this robot including three working stages is described. This procedure allows the surgeon to choose a suitable percutaneous target using an ultrasound probe based on his or her experience and the robot will track the respiratory motion of the target kidney stone and insert the needle automatically after the surgeon releases the probe. Experiments are conducted to demonstrate the procedure with the proposed assisted robot for PCNL.

We report on lift experiments conducted at scale for an artificial platform mimicking the dragonfly species: Sympetrum sanguineum. The platform, as well as the lift sensor, was custom designed and built. The flapping mechanism consisted of a piezoelectric bending-beam actuator, transmission using carbon-fiber elements and polymide-film joints, and wings constructed of polyester film with carbon-fiber support structure. The flapping kinematics of the Sympetrum san-guineum was replicated as closely as possible although only a pair of forewings were used in these experiments. The lift generated, when accounting for the addition of a pair of hindwings, is sufficient to allow for the hovering of a real-life dragonfly. The results, the first at-scale fully transient measurements of artificial dragonfly forewings, show that the lift curves quantitatively as well as qualitatively validate existing 2D and 3D computer simulations of dragonfly forewings.

Robotics can play a significant role in the rehabilitation of patients with spasticity by improving their early diagnosis and reducing the costs associated with care. Spasticity is a muscle control disorder characterized by an increase in muscle tone with exaggerated stretch reflexes, as one component of the upper motor neuron syndrome. Furthermore, spasticity is present in other pathologies, such as cerebral palsy, spina bifida, brain stroke among others. This video shows the ongoing research on developing a platform for the modelling and the assessment of spasticity using collaborative robots as clinical tool. Our aim is to develop methods for non-invasive biomechanical modelling of upper limbs joints using 7-DOF Rosen Kinematics [1], mixed with a non-linear state of Hills force-velocity relation [2], improved by introducing new parameters such as rigidity, viscoelasticity, extensibility and thixotropy. After a learning phase performed by the therapist, the robot replicates the trajectories required to perform the assessment. The video also describes the detailed analysis of passive movement response (force/torque and position/velocity)of the limb. These parameters will be used to determine the degree of spasticity of patients in a fast and objective manner, while simultaneously developing new clinical scales, such as a modified version of Ashworth [3].

As roles for unmanned aerial vehicles (UAVs)continue to diversify, the ability to sense and interact closely with the environment becomes increasingly important. Within this paper we report on the initial flight tests of a novel adaptively compliant actuator which will allow a UAV to carry out such tasks as the “pick and placement” of remote sensors, structural testing and contact-based inspection. Three key results are discussed and presented; the ability to physically apply forces with the UAV through the use of an active compliant manipulator; the ability to tailor these forces through tuning of the manipulator controller gains; and the ability to apply a rapid series of physical pulses in order to excite remotely placed sensors, e.g. vibration sensors. A series of over sixty flight tests have been used to generate initial results which clearly demonstrate the potential of this new type of compliant aerial actuator.

Soft material robots are attractive for safe interaction with humans and unstructured environments due to their compliance and low intrinsic stiffness and mass. These properties enable new capabilities such as the ability to conform to environmental geometry for tactile sensing and to undergo large shape changes for actuation. Due to the complex coupling between sensing and actuation in high-dimensional nonlinear soft systems, prior work in soft robotics has primarily focused on either sensing or actuation. This paper presents SOFTcell, a novel controllable stiffness tactile device that incorporates both optical sensing and pneumatic actuation. We report details on the device's design and implementation and analyze results from characterization experiments on sensitivity and performance, which show that SOFTcell can controllably increase its effective modulus from 4.4kPa to 46.1kPa. Additionally, we demonstrate the utility of SOFTcell for grasping in a reactive control task in which tactile data is used to detect fingertip shear as a grasped object slips, and cell pressurization is used to prevent the slip without the need to adjust fingertip position.

Over the next few years, more and more autonomous mobile robot systems will find their way into modern shop floors. However, it will be necessary to provide human-machine interfaces for interventions in unexpected situations like system-deadlocks, algorithm failures or inabilities. Using virtual or mixed reality-technologies, multi-modal teleoperation offers potential for being a suitable human-machine interface. Essential challenges in this field are, among others, a real-time remote control, a time-efficient and holistic environment detection using multiple sensors, a noise-reduced visualization of sensor-data, and capabilities of object recognition. This paper summarizes research results regarding an architecture capable of a near realtime, interoperable, and operator-supporting teleoperation. The focus of this paper is on a method to efficiently process and visualize point-clouds to meet high frame rate demands of virtual reality applications. To provide near real-time feedback of the robot and its environment over large distances, the presented method is capable to segment known objects from unknown objects to reduce bandwidth requirements. The results of this paper were evaluated using a industrial articulated robotic arm for teleoperation via a long distance UDP/IP communication.

Adapting to the ground enables stable footholds in legged locomotion by exploiting the structure of the terrain. On that account, we present a passive adaptive planar foot with three rotational degrees of freedom that is lightweight and thus suited for highly dynamic legged robots. Its low laying pivot joint provides high stability towards kinking. Information about the relative foot sole pose, and accordingly, the ground orientation is gathered by inertial measurement units (IMUs) placed on the foot sole and the shank. A complementary filter is presented that fuses these orientation estimates with an angular encoder to obtain a drift-free relative foot sole pose. The passive adaptive planar foot has been tested and compared to the classical point foot design on a variety of terrains and shows superior traction performance, especially on compressible soils. Being mounted on the quadrupedal robot ANYmal, the foot provides a reliable contact detection based on the fusion of the built-in 6-axis force/torque transducer and the IMUs. This allows to walk and trot on uneven terrain, loose soils, as well as climbing up a ramp and stairs while keeping the entire foot sole in ground contact all the time.