ICRA 2020 论文列表

The paper presents an approach for interactive programming of the robotic manipulator using mixed reality. The developed system is based on the HoloLens glasses connected through Robotic Operation System to Unity engine and robotic manipulators. The system gives a possibility to recognize the real robot location by the point cloud analysis, to use virtual markers and menus for the task creation, to generate a trajectory for execution in the simulator or on the real manipulator. It also provides the possibility of scaling virtual and real worlds for more accurate planning. The proposed framework has been tested on pick-and-place and contact operations execution by UR10e and KUKA iiwa robots.

Humans, in comparison to robots, are remarkably adept at reaching for objects in cluttered environments. The best existing robot planners are based on random sampling of configuration space- which becomes excessively high-dimensional with large number of objects. Consequently, most planners often fail to efficiently find object manipulation plans in such environments. We addressed this problem by identifying high-level manipulation plans in humans, and transferring these skills to robot planners. We used virtual reality to capture human participants reaching for a target object on a tabletop cluttered with obstacles. From this, we devised a qualitative representation of the task space to abstract the decision making, irrespective of the number of obstacles. Based on this representation, human demonstrations were segmented and used to train decision classifiers. Using these classifiers, our planner produced a list of waypoints in task space. These waypoints provided a high-level plan, which could be transferred to an arbitrary robot model and used to initialise a local trajectory optimiser. We evaluated this approach through testing on unseen human VR data, a physics-based robot simulation, and a real robot (dataset and code are publicly available1). We found that the human-like planner outperformed a state-of-the-art standard trajectory optimisation algorithm, and was able to generate effective strategies for rapid planning- irrespective of the number of obstacles in the environment.

As robot autonomy improves, robots are increasingly being considered in the role of autonomous observation systems - free-flying cameras capable of actively tracking human activity within some predefined area of interest. In this work, we formulate the autonomous observation problem through multi-objective optimization, presenting a novel Semi-MDP formulation of the autonomous human observation problem that maximizes observation rewards while accounting for both human- and robot-centric costs. We demonstrate that the problem can be solved with both scalarization-based Multi-Objective MDP methods and Constrained MDP methods, and discuss the relative benefits of each approach. We validate our work on activity tracking using a NASA Astrobee robot operating within a simulated International Space Station environment.

When a robot manipulator approaches a kinematic singular configuration, control strategies need to be employed to ensure safe and robust operation. If this manipulator is being controlled by a human through physical human-robot collaboration, the choice of strategy for handling singularities can have a significant effect on the feelings and impressions of the user. To date the preferences of humans during physical human-robot collaboration regarding strategies for managing kinematic singularities have yet to be thoroughly explored.This work presents an empirical study of a damping-based strategy for handling singularities with regard to the preferences of the human operator. Two different parameters, damping rate and damping asymmetry, are tested using a double-blind A/B pairwise comparison testing protocol. Participants included two cohorts made up of the general public (n=51) and people working within a robotic research centre (n=18). In total 105 individual trials were performed. Results indicate a preference for a faster, asymmetric damping behavior that slows motions towards singularities whilst allowing for faster motions away.

Soft growing robots are proposed for use in applications such as complex manipulation tasks or navigation in disaster scenarios. Safe interaction and ease of production promote the usage of this technology, but soft robots can be challenging to teleoperate due to their unique degrees of freedom. In this paper, we propose a human-centered interface that allows users to teleoperate a soft growing robot for manipulation tasks using arm movements. A study was conducted to assess the intuitiveness of the interface and the performance of our soft robot, involving a pick-and-place manipulation task. The results show that users were able to complete the task 97% of the time and achieve placement errors below 2 cm on average. These results demonstrate that our body-movement-based interface is an effective method for control of a soft growing robot manipulator.

How autonomous vehicles and human drivers share public transportation systems is an important problem, as fully automatic transportation environments are still a long way off. Understanding human drivers’ behavior can be beneficial for autonomous vehicle decision making and planning, especially when the autonomous vehicle is surrounded by human drivers who have various driving behaviors and patterns of interaction with other vehicles. In this paper, we propose an LSTM-based trajectory prediction method for human drivers which can help the autonomous vehicle make better decisions, especially in urban intersection scenarios. Meanwhile, in order to collect human drivers’ driving behavior data in the urban scenario, we describe a system called UrbanFlow which includes the whole procedure from raw bird’s-eye view data collection via drone to the final processed trajectories. The system is mainly intended for urban scenarios but can be extended to be used for any traffic scenarios.

The current state of the art in automotive high definition digital (HD) map generation based on dedicated mapping vehicles cannot reliably keep these maps up to date because of the low traversal frequencies. Anonymized data collected from the fleet of vehicles that is already on the road provides a huge potential to outperform such state of the art solutions in robustness, safety and up-to-dateness of the map while achieving comparable quality. We thus present a solution based on crowdsourced data to (i) detect changes in the map independent of the type of change, (ii) automatically trigger map update jobs for parts of the map, and (iii) create and integrate map patches to keep the map always up to date. The developed solution provides a crowdsourced up to date HD map to make reliable prior information on lane markings and road edges available to automated driving functions.

There is a long history demonstrating humans' tendency to create artificial copies of living creatures. For moving machines called robots, actuators play a key role in developing human-like movements. Among different types of actuation, PAMs (pneumatic artificial muscles) are known as the most similar ones to biological muscles. In addition to similarities in force generation mechanism (tension based), the well-accepted argumentation from Klute et al., states that the PAM force-length (fl) behavior is close to biological muscles, while the force-velocity (fv) pattern is different. Using the multiplicative formulation of the pressure (as an activation term), fl and fv beside an additive passive parallel elastic element, we present a new model of PAM. This muscle-based model can predict PAM dynamic behaviors with high precision. With a second experiment on a two-segmented leg, the proposed model is verified to predict the generated forces of PAMs in an antagonistic arrangement. Such a dynamic muscle-like model of artificial muscles can be used for the design and control of legged robots to generate robust, efficient and versatile gaits.

Texture sensing is one of the types of information sensed by humans through touch, and is thus of interest to robotics that this type of information can be acquired and processed. In this work we present a texture topography sensor based on a ciliary structure, a biological structure found in many organisms. The device consists of up to 9 elastic cilia with permanent magnetization assembled on top of a highly sensitive tunneling magnetoresistance (TMR) sensor, within a compact footprint of 6×6 mm2. When these cilia brush against some textured surface, their movement and vibrations give rise to a signal that can be correlated to the characteristics of the texture being measured. We also present an electronic signal acquisition board, used in this work. Various configurations of cilia sizes are tested, with the most precise being capable of differentiating different types of sandpaper from 9.2 μm to 213 μm average surface roughness with a 7 μm resolution. As a topography scanner the sensor was able to scan a 20 μm high step in a flat surface.

We revise, improve and extend the system previously introduced by us and named SSM-VPR (Semantic and Spatial Matching Visual Place Recognition), largely boosting its performance above the current state of the art. The system encodes images of places by employing the activations of different layers of a pre-trained, off-the-shelf, VGG16 Convolutional Neural Network (CNN) architecture. It consists of two stages: given a query image of a place, (1) a list of candidates is selected from a database of places and (2) the candidates are geometrically compared with the query. The comparison is made by matching CNN features and, equally important, their spatial locations, selecting the best candidate as the recognized place. The performance of the system is maximized by finding optimal image resolutions during the second stage and by exploiting temporal correlation between consecutive frames in the employed datasets.

We present a novel region growing algorithm for plane extraction of organized point clouds using the spherical convex hull. Instead of explicit plane parameterization, our approach interprets potential underlying planes as a series of geometric constraints on the sphere that are refined during region growing. Unlike existing schemes relying on downsampling for sequential execution in real time, our approach enables pixelwise plane extraction that is highly parallelizable. We further test the proposed approach with a fully parallel implementation on a GPU. Evaluation based on public data sets has shown state-of-the-art extraction accuracy and superior speed compared to existing approaches, while guaranteeing real-time processing at full input resolution of a typical RGB-D camera.

We consider the problem of planning views for a robot to acquire images of an object for visual inspection and reconstruction. In contrast to offline methods which require a 3D model of the object as input or online methods which rely on only local measurements, our method uses a neural network which encodes shape information for a large number of objects. We build on recent deep learning methods capable of generating a complete 3D reconstruction of an object from a single image. Specifically, in this work, we extend a recent method which uses Higher Order Functions (HOF) to represent the shape of the object. We present a new generalization of this method to incorporate multiple images as input and establish a connection between visibility and reconstruction quality. This relationship forms the foundation of our view planning method where we compute viewpoints to visually cover the output of the multiview HOF network with as few images as possible. Experiments indicate that our method provides a good compromise between online and offline methods: Similar to online methods, our method does not require the true object model as input. In terms of number of views, it is much more efficient. In most cases, its performance is comparable to the optimal offline case even on object classes the network has not been trained on.

Flexible medical instruments, such as Continuum Dexterous Manipulators (CDM), constitute an important class of tools for minimally invasive surgery. Accurate CDM shape reconstruction during surgery is of great importance, yet a challenging task. Fiber Bragg grating (FBG) sensors have demonstrated great potential in shape sensing and consequently tip position estimation of CDMs. However, due to the limited number of sensing locations, these sensors can only accurately recover basic shapes, and become unreliable in the presence of obstacles or many inflection points such as s-bends. Optical Frequency Domain Reflectometry (OFDR), on the other hand, can achieve much higher spatial resolution, and can therefore accurately reconstruct more complex shapes. Additionally, Random Optical Gratings by Ultraviolet laser Exposure (ROGUEs) can be written in the fibers to increase signal to noise ratio of the sensors. In this comparison study, the tip position error is used as a metric to compare both FBG and OFDR shape reconstructions for a 35 mm long CDM developed for orthopedic surgeries, using a pair of stereo cameras as ground truth. Three sets of experiments were conducted to measure the accuracy of each technique in various surgical scenarios. The tip position error for the OFDR (and FBG) technique was found to be 0.32 (0.83) mm in free-bending environment, 0.41 (0.80) mm when interacting with obstacles, and 0.45 (2.27) mm in s-bending. Moreover, the maximum tip position error remains sub-millimeter for the OFDR reconstruction, while it reaches 3.40 mm for FBG reconstruction. These results propose a cost-effective, robust and more accurate alternative to FBG sensors for reconstructing complex CDM shapes.

In this paper, we propose a parallel six-wheel-legged robot that can traverse irregular terrain while carrying objectives to do heavy-duty work. This robot is equipped with six Stewart platforms as legs and tightly integrates the additional degrees of freedom introduced by the wheels. The presented control strategy with physical system used to adapt the diverse degrees of each leg to irregular terrain such that robot increases the traversability, and simultaneously to maintain the horizontal whole-torso pose. This strategy makes use of Contact Scheduler (CS) and Whole-Torso Control (WTC) to control the multiple degrees of freedom (DOF) leg for performing high-flexibility locomotion and adapting the rough terrain like actively parallel suspension system. We conducted experiments on flat, slope, soft and sandgravel surface, which validate the proposed control method and physical system. Especially, we attempt to traverse over sand-gravel terrain with 3 people about 240kg payload.

Electromyography (EMG) based interfaces have been used in various robotics studies ranging from teleoperation and telemanipulation applications to the EMG based control of prosthetic, assistive, or robotic rehabilitation devices. But most of these studies have focused on the decoding of user's motion or on the control of the robotic devices in the execution of simple tasks (e.g., grasping tasks). In this work, we present a learning scheme that employs High Density Electromyography (HD-EMG) sensors to decode a set of dexterous, in-hand manipulation motions (in the object space) based on the myoelectric activations of human forearm and hand muscles. To do that, the subjects were asked to perform roll, pitch, and yaw motions manipulating two different cubes. The first cube was designed to have a center of mass coinciding with the geometric center of the cube, while for the second cube the center of mass was shifted 14 mm to the right (off-centered design). Regarding the acquisition of the myoelectric data, custom HD-EMG electrode arrays were designed and fabricated. Using these arrays, a total of 89 EMG signals were extracted. The object motion decoding was formulated as a regression problem using the Random Forests (RF) technique and the muscle importances were studied using the inherent feature variables importance calculation procedure of the RF. The muscle importance results show that different subjects use different strategies to execute the same motions on same object when the weight is off-centered. Finally, the decoded motions were used to control a five fingered robotic hand in a proof-of-concept application.

If robots and humans are to coexist and cooperate in society, it would be useful for robots to be able to engage in tactile interactions. Touch is an intuitive communication tool as well as a fundamental method by which we assist each other physically. Tactile abilities are challenging to engineer in robots, since both mechanical safety and sensory intelligence are imperative. Existing work reveals a trade-off between these principles- tactile interfaces that are high in resolution are not easily adapted to human-sized geometries, nor are they generally compliant enough to guarantee safety. On the other hand, soft tactile interfaces deliver intrinsically safe mechanical properties, but their non-linear characteristics render them difficult for use in timely sensing and control. We propose a robotic system that is equipped with a completely soft and therefore safe tactile interface that is large enough to interact with human upper limbs, while producing high resolution tactile sensory readings via depth camera imaging of the soft interface. We present and validate a data-driven model that maps point cloud data to contact forces, and verify its efficacy by demonstrating two real-world applications. In particular, the robot is able to react to a human finger's pokes and change its pose based on the tactile input. In addition, we also demonstrate that the robot can act as an assistive device that dynamically supports and follows a human forearm from underneath.

This paper presents a novel physical gripping framework intended for controlled, high force density attachment on a range of surfaces. Our framework utilizes a light-activated chemical adhesive to attach to surfaces. The cured adhesive is part of a "sacrificial layer," which is shed when the gripper separates from the surface. In order to control adhesive behavior we utilize ultraviolet (UV) light sensitive acrylics which are capable of rapid curing when activated with 380nm light. Once cured, zero input power is needed to hold load. Thin plastic parts can be used as the sacrificial layers, and these can be released using an electric motor. This new gripping framework including the curing load capacity, adhesive deposition, and sacrificial methods are described in detail. Two proof-of concept prototypes are designed, built, and tested. The experimental results illustrate the response time (15-75s depending on load), high holding force-to-weight ratio (10-30), and robustness to material type. Additionally, two drawbacks of this design are discussed: corruption of the gripped surface and a limited number of layers.

Learning-based grasp pose detection algorithms have boosted the performance of robot grasping, but they usually need manually fine-tuning steps to find the balance between detection accuracy and efficient. In this paper, we discard these intermediate procedures, like sampling grasps and generating grasp proposals, and propose an end-to-end grasp pose detection model. Our model uses the RGB image as the input and predicts the single grasp pose in each small grid of the image. Furthermore, the best grasps are found by non-maximum suppression (NMS) strategy. The clustering and ranking procedures are left for NMS while the network only generates dense grasp predictions, which keeps the network simple and efficient. To achieve dense predictions, the predicted grasps of our detection model are represented by the 6 channels images with each pixel location representing a rated grasp. To the best of our knowledge, our model is the first neural network that attaches a grasp pose in pixel level. The model achieves 96.5% accuracy which costs 14ms for prediction of a 480×360 resolution RGB image in Cornell Grasp Dataset, and 90.4% robot grasping success rate for unknown objects with a parallel plate gripper in the real environment.

Accurate, robust and real-time localization under constrained-resources is a critical problem to be solved. In this paper, we present a new sparse pose-graph visual-inertial SLAM (SPVIS). Unlike the existing methods that are costly to deal with a large number of redundant features and 3D map points, which are inefficient for improving positioning accuracy, we focus on the concise visual cues for high-precision pose estimating. We propose a novel hierarchical quadtree based optical flow tracking algorithm, it achieves high accuracy and robustness within very few concise features, which is only about one fifth features of the state-of-the-art visual-inertial SLAM algorithms. Benefiting from the efficient optical flow tracking, our sparse pose-graph optimization time cost achieves bounded complexity. By selecting and optimizing the informative features in sliding window and local VIO, the computational complexity is bounded, it achieves low time cost in long-term operation. We compare with the state-of-the-art VIO/VI-SLAM systems on the challenging public datasets by the embedded platform without GPUs, the results effectively verify that the proposed method has better real-time performance and localization accuracy.