ICRA 2018 论文列表

This paper presents the development of a fish-like magnetically propelled microswimmer fabricated by 3D laser lithography. The microswimmer consists of a head and a caudal fin, just like a natural fish. There is a joint between the head and the fin so that the caudal fin can oscillate around the head to generate thrust, and the oscillation of the fin hardly transfers to the head, which benefits the stable motion of the microswimmer. The caudal fin of the microswimmer is deposited with a layer of 50 nm nickel (Ni) for magnetic actuation. Through applying an oscillating uniform magnetic field, the microswimmer can move along with the direction guided by the external magnetic field. A magnetic control system with permanent magnets is designed to provide such an oscillating uniform magnetic field, where the oscillating frequency and amplitude are controllable. A micro probe operation platform is used to detach the fabricated microswimmers from glass substrate in manufacturing. The proposed magnetically propelled microswimmer can be potentially used as powerful detoxification and biosensing tools for medical diagnosis and treatment in precision medicine.

Robot malfunction is inevitable in practical applications of the robot formation control due to uncontrolled crashing, system malfunction or communication loss. In this paper, we study the synchronous formation control problem in the presence of robot malfunctions. Our main idea is to improve the network connectivity and motion synchronism of the robot formation through a series of topology switchings and robot replacements. Firstly, the synchronous formation control method is introduced which enables the robots to tracking their desired trajectories while keeping predefined formation shapes. Secondly, a recursive switched topology control strategy is proposed to restore the formation shape as well as to improve the network connectivity and motion synchronism in the presence of robot malfunctions. Thirdly, the convergence analysis of the proposed control system is presented and a sufficient condition is obtained under an average dwell time scheme. What's more, the proposed approach is fully distributed and the communication delays between neighboring robots also have been taken into consideration. Simulation results demonstrate the effectiveness of the proposed approach.

Humans' fingertips can perceive not only the magnitude and the direction of force but also the texture of object. When we grasp an object, the surface texture sensing of the fingertip helps us recognize the object and the force feeling that is parallel to the skin helps us grasp stably. Focusing on these points, we have developed a dual-modal vision-based tactile sensor that can measure the texture of object and a distribution of force vectors. The tactile sensor consists of a transparent elastomer, a camera, a piece of transparent acrylic board, LEDs and supporting structures. A reflective membrane and markers array are on the surface of the elastomer. An applied force on the elastic body results in movements of the markers, which are acquired by the CCD camera. In addition, the shape and texture of the object's contact surface can be reflected by the membrane deformations. The distribution of force vectors is determined by the BP neural network. The local binary pattern algorithm using captured images calculates the texture information. This paper reports experimental evaluation results concerning accuracy of determination of magnitude, direction of force, and texture recognition rate.

The acceptance of teleoperation applications like tele-driving, tele-surgery, tele-maintenance, etc., is challenged by the quality-reducing effect of end-to-end latency. Particularly, when users wear Head-Mounted Displays to enhance the immersive experience, the lag between head motion and display response leads to unbearable motion sickness, indisposition, and, in the worst case, abortion of the teleoperation session. In this paper, we propose a delay compensation approach with head motion prediction that can be applied to pan-tilt-unit-based stereoscopic telepresence systems. We provide the user with the impression of a 3D 360° video that represents the remote scene without noticing the present delay, even when rotating the head. To this end, we propose a novel prediction paradigm for head motion estimation to substantially mitigate the negative impact of the latency on the quality of experience. We re-implemented state-of-the-art head movement predictors and compare them to our proposed approach by means of qualitative measures. In our experiments, we used two real and independent head motion datasets for validation and tested communication delays between 100-1000ms. Our results show that mean compensation rates of more than 99% are able with our approach.

The need for fast and precise measurements of plant attributes makes robotic solutions an ideal replacement for labor-intensive phenotyping processes. In this work we present a deep learning-based high throughput, online pipeline for in-situ sorghum stalk detection and grasping. We use a variation of Generative Adversarial Network (GAN) for stalk segmentation trained on a relatively small number of images followed by a grasp point generation pipeline. The presented pipeline is robust to field challenges such as occlusions, high stalk density and lighting variation, and was deployed on a custom-built ground robot. We tested our end-to-end system in a field of Sorghum bicolor in South Carolina, USA, achieving an average grasping accuracy of 74.13% and a stalk detection F1 score of 0.90. Grasp point detection for plant manipulation takes an average of 0.98 seconds, and pixel-wise stalk detection takes 0.2 seconds per image.

We present a semantically rich graph representation for indoor robotic navigation. Our graph representation encodes: semantic locations such as offices or corridors as nodes, and navigational behaviors such as enter office or cross a corridor as edges. In particular, our navigational behaviors operate directly from visual inputs to produce motor controls and are implemented with deep learning architectures. This enables the robot to avoid explicit computation of its precise location or the geometry of the environment, and enables navigation at a higher level of semantic abstraction. We evaluate the effectiveness of our representation by simulating navigation tasks in a large number of virtual environments. Our results show that using a simple sets of perceptual and navigational behaviors, the proposed approach can successfully guide the way of the robot as it completes navigational missions such as going to a specific office. Furthermore, our implementation shows to be effective to control the selection and switching of behaviors.

Context is an essential capability for robots that are to be as adaptive as possible in challenging environments. Although there are many context modeling efforts, they assume a fixed structure and number of contexts. In this paper, we propose an incremental deep model that extends Restricted Boltzmann Machines. Our model gets one scene at a time, and gradually extends the contextual model when necessary, either by adding a new context or a new context layer to form a hierarchy. We show on a scene classification benchmark that our method converges to a good estimate of the contexts of the scenes, and performs better or on-par on several tasks compared to other incremental models or non-incremental models.

This paper reports on a data-driven, interaction-aware motion prediction approach for pedestrians in environments cluttered with static obstacles. When navigating in such workspaces shared with humans, robots need accurate motion predictions of the surrounding pedestrians. Human navigation behavior is mostly influenced by their surrounding pedestrians and by the static obstacles in their vicinity. In this paper we introduce a new model based on Long-Short Term Memory (LSTM) neural networks, which is able to learn human motion behavior from demonstrated data. To the best of our knowledge, this is the first approach using LSTMs, that incorporates both static obstacles and surrounding pedestrians for trajectory forecasting. As part of the model, we introduce a new way of encoding surrounding pedestrians based on a 1d-grid in polar angle space. We evaluate the benefit of interaction-aware motion prediction and the added value of incorporating static obstacles on both simulation and real-world datasets by comparing with state-of-the-art approaches. The results show, that our new approach outperforms the other approaches while being very computationally efficient and that taking into account static obstacles for motion predictions significantly improves the prediction accuracy, especially in cluttered environments.

Event cameras offer many advantages for dynamic robotics due to their low latency response to motion, high dynamic range, and inherent compression of the visual signal. Many algorithms easily achieve real-time performance when testing on off-line datasets, however with an increase in camera resolution and applications on fast-moving robots, latency-free operation is not guaranteed. The event-rate is not constant, but is proportional to the amount of movement in the scene, or the velocity of the camera itself. Recently, algorithms have instead reported a maximum event-rate that can be achieved in real-time. In this paper we present the event-driven framework used on the iCub robot, which closes the loop between algorithm processing rate and the actual event-rate of the camera in order to smoothly control and limit the latency, while allowing the algorithm to degrade gracefully when large bursts of events occur. We show two algorithms that process events differently from each other and demonstrate the trade-off between latency and algorithm performance that the framework provides.

Flying robots require a combination of accuracy and low latency in their state estimation in order to achieve stable and robust flight. However, due to the power and payload constraints of aerial platforms, state estimation algorithms must provide these qualities under the computational constraints of embedded hardware. Cameras and inertial measurement units (IMUs) satisfy these power and payload constraints, so visual-inertial odometry (VIO) algorithms are popular choices for state estimation in these scenarios, in addition to their ability to operate without external localization from motion capture or global positioning systems. It is not clear from existing results in the literature, however, which VIO algorithms perform well under the accuracy, latency, and computational constraints of a flying robot with onboard state estimation. This paper evaluates an array of publicly-available VIO pipelines (MSCKF, OKVIS, ROVIO, VINS-Mono, SVO+MSF, and SVO+GTSAM) on different hardware configurations, including several single-board computer systems that are typically found on flying robots. The evaluation considers the pose estimation accuracy, per-frame processing time, and CPU and memory load while processing the EuRoC datasets, which contain six degree of freedom (6DoF) trajectories typical of flying robots. We present our complete results as a benchmark for the research community.

This paper presents a novel 3DOF pedestrian trajectory prediction approach for autonomous mobile service robots. While most previously reported methods are based on learning of 2D positions in monocular camera images, our approach uses range-finder sensors to learn and predict 3DOF pose trajectories (i.e. 2D position plus 1D rotation within the world coordinate system). Our approach, T-Pose-LSTM (Temporal 3DOF-Pose Long-Short-Term Memory), is trained using long-term data from real-world robot deployments and aims to learn context-dependent (environment- and time-specific) human activities. Our approach incorporates long-term temporal information (i.e. date and time) with short-term pose observations as input. A sequence-to-sequence LSTM encoder-decoder is trained, which encodes observations into LSTM and then decodes the resulting predictions. On deployment, the approach can perform on-the-fly prediction in real-time. Instead of using manually annotated data, we rely on a robust human detection, tracking and SLAM system, providing us with examples in a global coordinate system. We validate the approach using more than 15 km of pedestrian trajectories recorded in a care home environment over a period of three months. The experiments show that the proposed T-Pose-LSTM model outperforms the state-of-the-art 2D-based method for human trajectory prediction in long-term mobile robot deployments.

In this paper, we present a probabilistic framework to recover the extrinsic calibration parameters of a lidar-IMU sensing system. Unlike global-shutter cameras, lidars do not take single snapshots of the environment. Instead, lidars collect a succession of 3D-points generally grouped in scans. If these points are assumed to be expressed in a common frame, this becomes an issue when the sensor moves rapidly in the environment causing motion distortion. The fundamental idea of our proposed framework is to use preintegration over interpolated inertial measurements to characterise the motion distortion in each lidar scan. Moreover, by using a set of planes as a calibration target, the proposed method makes use of lidar point-to-plane distances to jointly calibrate and localise the system using on-manifold optimisation. The calibration does not rely on a predefined target as arbitrary planes are detected and modelled in the first lidar scan. Simulated and real data are used to show the effectiveness of the proposed method.

We propose an approach to estimate 3D human pose in real world units from a single RGBD image and show that it exceeds performance of monocular 3D pose estimation approaches from color as well as pose estimation exclusively from depth. Our approach builds on robust human keypoint detectors for color images and incorporates depth for lifting into 3D. We combine the system with our learning from demonstration framework to instruct a service robot without the need of markers. Experiments in real world settings demonstrate that our approach enables a PR2 robot to imitate manipulation actions observed from a human teacher.

Simultaneous Localization And Mapping (SLAM) is one of the major bricks needed to build truly autonomous mobile robots. The probabilistic formulation of SLAM is based on two models: the motion model and the observation model. In practice, these models, together with the SLAM map representation, do not model perfectly the robot's real dynamics, the sensor measurement errors and the environment. Consequently, systematic errors affect SLAM estimations. In this paper, we propose two approaches to predict corrections to be applied to SLAM estimations. Both are based on the Ensemble Multilayer Perceptron model. The first approach uses successive estimated poses to predict the errors, with no assumptions on the underlying SLAM process or sensor used. The second method is specific to 2D likelihood SLAM approaches, thus, the likelihood distributions are used to predict the corrections, making this second approach independent of the sensor used. We also build a hybrid correction module based on successive estimated poses and the likelihood distributions. The validity of both approaches is evaluated through two experiments using different evaluation metrics and sensor configurations.

This paper describes nonprehensile manipulation realized using the vibration of a plate. A novel manipulation strategy is proposed wherein the three degrees-of-freedom (DoF) of a part are controlled by only one actuator. First, a manipulator driven by a single actuator is introduced. The end effector of this manipulator is a flat plate. The manipulator employs an active-passive hybrid joint mechanism with nonparallel axes. Based on the sinusoidal displacement input to the actuator, the manipulator can generate the velocity of a part omnidirectionally on the plate. Next, simulation results are presented to show that the velocity map of the part varies depending upon the surface friction property of the plate. Further, the control of the rotational behavior of the part on the boundary of two areas with different friction properties by means of the input frequency is shown. Based on this control, a 3- DoF manipulation strategy using a virtual turntable is developed to realize the desired position and orientation of the part. Finally, the proposed method is demonstrated via experiments.

This paper revisits computed torque method for robot manipulators and aims at developing its new framework based on the L_{1}L_{1} robustness, in which the L_{\infty}L_{\infty} norm together with its induced norm is employed to characterize model uncertainties and a performance measure. More precisely, we consider the L_{1}L_{1} robust stability and performance for a given robot manipulator with a computed torque controller. We first show that the modelling errors in the computed torque method can be divided into an exogenous disturbance and a multiplicative model uncertainty, which are bounded in terms of the L_{\infty}L_{\infty} norm and its induced norm, respectively. It is next shown that the robot manipulator with the computed torque controller can be equivalently represented by an interconnection of a continuous-time linear time-invariant (LTI) nominal plant and a stabilizing controller together with the L_{\infty}L_{\infty}-induced norm bounded model uncertainty. Based on the interconnected representation, the L_{1}L_{1} robust stability condition and an upper bound of the L_{1}L_{1} performance against the exogenous disturbance with respect to all model uncertainties in a class of a bounded L_{\infty}L_{\infty}-induced norm are dealt with by using the small-gain theorem. Finally, the effectiveness of the theoretical results is demonstrated through some experiment results.