IROS 2017 论文列表

Progress has been achieved recently in object detection given advancements in deep learning. Nevertheless, such tools typically require a large amount of training data and significant manual effort to label objects. This limits their applicability in robotics, where solutions must scale to a large number of objects and variety of conditions. This work proposes an autonomous process for training a Convolutional Neural Network (CNN) for object detection and pose estimation in robotic setups. The focus is on detecting objects placed in cluttered, tight environments, such as a shelf with multiple objects. In particular, given access to 3D object models, several aspects of the environment are physically simulated. The models are placed in physically realistic poses with respect to their environment to generate a labeled synthetic dataset. To further improve object detection, the network self-trains over real images that are labeled using a robust multi-view pose estimation process. The proposed training process is evaluated on several existing datasets and on a dataset collected for this paper with a Motoman robotic arm. Results show that the proposed approach outperforms popular training processes relying on synthetic - but not physically realistic - data and manual annotation. The key contributions are the incorporation of physical reasoning in the synthetic data generation process and the automation of the annotation process over real images.

Self-folding is capable of forming complex three-dimensional structures from planar sheets. However, existing self-folding machines cannot generate large forces or bear high loads. In contrast, traditional robots are expected to operate under large loads but these robots are generally dense, resulting in heavy machines with relatively low strength-to-weight ratios. In this paper, we present a new self-folding technique that is meter-scale and load-bearing. We demonstrate the design and fabrication of both structural beams and pneumatically actuated joints. We model and measure the stiffness and strength of self-folded structural beams. Finally, we integrate these results into a robotic arm that weighs 0.3 kg and can lift up to 1.0 kg. These results indicate that functional, load-bearing, self-folding robots are possible.

A Series Elastic Actuator (SEA) is designed by placing a passive compliant element between a conventional stiff actuator and link. The intrinsically compliant mechanical structure provides several superiorities, e.g., safety, energy efficiency, high force fidelity, low cost force measurement, high transparency, etc., in advanced robot applications, such as humanoids, quadrupeds and exoskeletons. However, the motion control problem of an SEA is more complicated than that of a conventional stiff actuator due to its higher order dynamics. This paper proposes a novel Active Disturbance Rejection (ADR) based robust force controller for SEAs by combining Differential Flatness (DF) and Disturbance Observer (DOb) in state space. The robust state and control input references are systematically generated in terms of a fictitious variable, namely differentially flat output, estimated disturbances and their successive derivatives. A second order DOb is designed in state space so that disturbances and their first and second order derivatives are estimated. It is experimentally shown that high performance force control applications can be performed without requiring the precise dynamic models of the actuator and environment when the proposed robust force controller is implemented.

The autonomous measurement of tree traits, such as branching structure, branch diameters, branch lengths, and branch angles, is required for tasks such as robotic pruning of trees as well as structural phenotyping. We propose a robotic vision system called the Robotic System for Tree Shape Estimation (RoTSE) to determine tree traits in field settings. The process is composed of the following stages: image acquisition with a mobile robot unit, segmentation, reconstruction, curve skeletonization, conversion to a graph representation, and then computation of traits. Quantitative and qualitative results on apple trees are shown in terms of accuracy, computation time, and robustness. Compared to ground truth measurements, the RoTSE produced the following estimates: branch diameter (mean-squared error 0.99 mm), branch length (mean-squared error 45.64 mm), and branch angle (mean-squared error 10.36 degrees). The average run time was 8.47 minutes when the voxel resolution was 3 mm3.

Piezoelectric ceramics(PZT)actuator has been widely used in flexure-guided nanopositioning stage because of their high resolution. However, it is quite hard to achieve high-rate precision positioning control because of the complex hysteresis nonlinearity effect of PZT actuator. Thus, an online RELM algorithm with forgetting property(FReOS-ELM) is proposed to handle this issue. Firstly, we adopt regularized extreme learning machine(RELM)to build an intelligent hysteresis model. The training of the algorithm is completed only in one step, which avoids the shortcomings of the traditional hysteresis model based on artificial neural network(ANN) that slow training speed and easy to fall into the local minimum. Then, based on the regularized on-line sequential extreme learning machine(ReOS-ELM), an on-line RELM algorithm with forgetting property(FReOS-ELM) is designed, which can avoid the computational load of ReOS-ELM in the process of adding new data for learning on-line. In the experiment, a real-time voltage signal with varying frequencies and amplitudes is adopted, and the output displacement data of the nanopositioning stage is also acquired and analyzed. The results powerfully verify that the performance of the established hysteresis model based on the proposed FReOS-ELM is satisfactory, which can be used to improve the practical positioning performance for flexure nanopositioning stage.

Lighting and shading can have a great impact on a robot's ability to recognize, match, and classify objects in indoor scenes. Additionally, the realism of augmented reality applications benefit greatly from understanding the lighting and shading in a scene. To aid in extracting this information from a mobile platform we introduce a novel framework to solve the intrinsic image decomposition problem for RGB-D streams. In our pipeline, the task is formulated as a Bayesian estimation problem. Compared to frame-based methods that must solve a full conditional random field (CRF) optimization problem at each time step, our framework can utilize the knowledge of past frames to predict the intrinsic images at a given frame. Our approach produces more reliable and consistent predictions over time, and our filtering-based framework achieves significant performance gains. Furthermore, our framework can be easily integrated into standard perception loops in many robotic systems that use a similar recursive filtering structure. We show qualitative results on real data and generate quantitative results using ground truth from a photorealistic synthetic dataset produced using a state-of-the-art ray tracer and high fidelity 3D model.

Despite the availability of increasingly more robust machine perception and control algorithms, human operators are still needed to directly control robotic systems in extreme environments where there is a high cost of failure. This paper reports a novel architecture for semi-autonomous teleoperation over high-latency telemetry and the results of a 15-subject user study used to evaluate that architecture. The intent-recognition-based traded control (IRTC) architecture recognizes operator intent based on remote scene and task modeling, and automatically trades control to an autonomous reactive subsystem. Each subject participated in one two-hour session during which they controlled a single simulated redundant robotic manipulator with 4000ms of closed-loop latency. Subjects continuously assemble and disassemble a 3D lattice structure with or without IRTC-based assistance. The user study showed that the IRTC system improved operator performance in the simulated multi-object assembly task. All source code for conducting the experiment including the simulation and data processing is available online under a permissive open-source license.

We present a 3D mesh re-sampling algorithm, carefully tailored for 3D object detection using point pair features (PPF). Computing a sparse representation of objects is critical for the success of state-of-the-art object detection, recognition and pose estimation methods. Yet, sparsity needs to preserve fidelity. To this end, we develop a simple, yet very effective point sampling strategy for detection of any CAD model through geometric hashing. Our approach relies on rendering the object coordinates from a set of views evenly distributed on a sphere. Actual sampling takes place on 2D domain over these renderings; the resulting samples are efficiently merged in 3D with the aid of a special voxel structure and relaxed with Lloyd iterations. The generated vertices are not concentrated only on critical points, as in many keypoint extraction algorithms, and there is even spacing between selected vertices. This is valuable for quantization based detection methods, such as geometric hashing of point pair features. The algorithm is fast and can easily handle the elongated/acute triangles and sharp edges typically existent in industrial CAD models, while automatically pruning the invisible structures. We do not introduce structural changes such as smoothing or interpolation and sample the normals on the original CAD model, achieving the maximum fidelity. We demonstrate the strength of this approach on 3D object detection in comparison to similar sampling algorithms.

A range of pneumatically-driven devices have been developed. However, these require a large air compressor, making them heavy, bulky, and therefore inappropriate for use in portable devices. In this study, we proposed and tested a portable pneumatic power source using the chemical reaction between sodium bicarbonate and citric acid. The carbon dioxide generated in the reaction was used as a pneumatic power source. The chemicals used are inexpensive, nontoxic, and readily available. From the result of basic experiment demonstrated in this paper, we conclude that the method can provide enough high pressure (more than 1 MPa) for activating pneumatic actuators and the chemical reaction is safe since it is endothermic reaction. We also demonstrated a novel system that provides continuous mixing of the reagents, using a pressure booster. The system keeps a pressure inside of a tank moderate, and tank can become light weight because it doesn't need to endure high pressure.

This paper proposes a passive and explicit integrator for simulating a rotational rigid-body dynamics rendered by a robot. Considering the Euler integration method, active energy terms are identified. These sources of energy are due to the external torque and the coupled dynamics which can lead to a non-physical behavior of the simulated dynamics. The proposed method dissipates this energy using a variable damper regulated by an energy observer. The new algorithm guarantees not only passivity but also a consistent energetic integration. The integration method is sustained by simulations and tested on a real-time hardware-in-the-loop simulator.

The torso of a humanoid robot is a fundamental part of its kinematic structure because it defines the reachable workspace, supports the entire upper-body and can be used to control the position of the center of mass. The majority of the torso joints are designed exploiting serial or differential mechanisms, while parallel kinematic structures are less used mainly because of their greater design complexity. This paper describes the design and construction of a 4 degrees of freedom (DoF) torso for our new humanoid robot. Three degrees of freedom, namely roll, pitch and heave, have been implemented using a 3 DoF parallel kinematic structure, while the fourth DoF, namely yaw, has been implemented with a rotational joint on top of the parallel structure. The design has been optimized to reduce the cost and the volume of the system. A first prototype of the torso has been constructed and validated with respect to our design requirements. Eventually, experimental tests have been conducted to assess the functionality of the proposed system.

This article presents the development of a fully soft, exoskeleton robot for power augmentation and rehabilitation of a human wrist joint. The system is powered by novel bending pneumatic muscles which have been combined with contractor pneumatic muscles to provide actuation for the glove. This research has assessed the behaviour of the new bending actuators and demonstrated them in a prototype wrist exoskeleton. An accompanying mathematical model of the force generated by the proposed extensor bending artificial muscles has been developed. The proposed wearable robot is capable of producing flexion-extension and abduction-adduction motions of the human wrist to generate rehabilitation motions. The exoskeleton is designed to fit any adult hand size without the need for any mechanical changes, meaning it can easily be swapped between users.

Despite significant advancements in the field of wearable robots (WRs), commercial WRs still use traditional direct-drive actuation units to power their joints. On the other hand, in research prototypes compliant actuators are increasingly being used to more adequately address the issues of safety, robustness, control and overall system efficiency. The advantages of mechanical compliance are exploited in a novel modular actuator prototype designed for the knee joint. Due to its modularity, the actuator can be implemented in a knee joint of a standalone or a multi-joint lower-limbs orthosis, for use in gait rehabilitation and/or walking assistance. Differently from any other actuator used in orthotic research prototypes, it combines a Variable Stiffness Actuator (VSA) and a Parallel Elasticity Actuation (PEA) unit in a single modular system. Although independent, the units are designed to work together in order to fully mimic dynamic behavior of the human knee joint. In this paper, design aspects and functional evaluation of the new actuator are presented and a rationale for such a design in biomechanics of the human knee joint is given. The VSA subsystem is characterized in a quasi-static benchmarking environment and the results showing main performance indicators are presented.

Robotic systems have recently drawn attention in minimally invasive surgeries due to their increased dexterity feature. A major drawback of these systems is image blurring due to lens contamination which cause imaging impairment during up to 40% of surgery time. This paper demonstrates a novel laparoscopic magnetic driven camera system with implemented in-vivo lens cleaning and debris prevention systems. This camera robot can cover 150 degrees field of view inside the abdominal cavity and provide adjustable illumination system to improve the video quality. Design details for different modules, such as anchoring, actuation, video capturing, illumination, debris prevention and lens cleaning of this robot have been provided and discussed. This camera robot can decrease the possibility of lens contamination by creating CO2 gas barrier in front of lens. In case of contamination it can clean the lens in-vivo without removing the camera from abdominal cavity. The lens cleaning module has been tested for water vapor and water droplets. The robot is manufactured and each module has been validated by designed experiments.

In an outdoor autonomous navigational context, classic compass sensors such as magnetometers have to deal with unpredictable magnetic disturbances. In this paper, we propose to get inspiration from the insect navigational abilities to design a celestial compass based on linear polarization of ultraviolet (UV) skylight. To compute the solar meridian relative orientation, our 3D-printed celestial compass uses only two pixels created by two UV-light photo-sensors topped with linear polarizers arranged orthogonally to each other, in the same manner that was observed in insects' Dorsal Rim Area ommatidia. The compass was then embedded on our hexapod walking robot called Hexabot. We first tested the UV-polarized light compass to compensate for yaw random disturbances. We then used the compass to maintain Hexabot's heading direction constant in a straight-forward task, knowing the robot has important yaw drifts. Experiments under various meteorological conditions provided steady state heading direction errors from 0.3° under clear sky conditions to 1.9° under overcast sky, which suggests interesting precision and reliability to make this optical compass suitable for robotics.

In this paper, we demonstrate that it is feasible to conduct reliable multi-axis force control motor neuroimaging experiments in humans using a novel five degree-of-freedom Haptic fMRI interface (HFI-5). We demonstrate that HFI-5 supports accurate force and torque control for all its axes. In addition, it minimizes the force-to-torque coupling induced by its closed chain kinematic structure. HFI-5 has low backdrive friction (<; 0.5N; <; 0.05Nm), which improves its haptic transparency. While the device is large and consequently flexible, a linear model can correct kinematic accuracies due to device deflection. Resulting position error measurements are similar in scale to human hand jitter, and are thus satisfactory. To test HFI-5's efficacy, we performed force perception and force discrimination tests with it on four human subjects. The test revealed that humans could readily perceive forces above 1N and could discern forces that were about 33% apart. As such, HFI-5's force range (x:±16.36N; y:±4.62N; z:±6.9N; passively gravity compensated) can support a wide variety of forces that are perceptible and distinguishable. Finally, we performed ten fMRI scans with a human subject who performed a visually guided force-control experiment. Temporal noise patterns in fMRI measurements matched the fMRI baseline (0.8%), which demonstrates that HFI-5's electromagnetic motors do not introduce substantial noise. HFI-5 is thus ready to be used in motor neuroimaging experiments.

Minimally invasive robotic surgery holds a fundamental role in modern surgery. However, one of its major limitations compared to classic laparoscopy is that the surgeon can only rely on visual perception, for the lack of haptic force feedback. A new solution for a force sensor placed at the end-tip of the trocar is presented here. This solution allows measuring the interaction forces between the surgical instrument and the environment without any changes to the instrument structure and with full adaptability to different robot platforms and surgical tools. A prototype of the sensor has been realized with 3D printed technology for a proof of concept. The static and dynamic characterization of the sensor is provided together with experimental validation.

This paper describes the design and development of an automated construction quality assessment robot system (QuicaBot) for hollowness, crack, evenness, alignments and inclination problems. To the best of our knowledge, this work is the first attempt to pave the way towards a fully autonomous robotic system for post construction quality assessment of buildings. The main goal of the novel robot is twofold: to systematize the manual inspection work through automation resulting in more reliable and objective inspection reports, and to speed up the inspection process resulting in a cost reduction. Based-on our initial on-site tests, the developed robot increases the overall efficiency in all the aforementioned five problems.

This paper presents a hexapod crawling robot which has bioinspired design and locomotion posture from the insects. The robot, called S-Hex II, is an upgraded version of the different printable monolithic hexapod robot which is named as S-Hex I. The S-Hex II is studied for the project of developing mesoscale walking robots actuated by the soft dielectric elastomer actuators. In comparison with the S-Hex I, the S-Hex II owns smaller dimension, lighter weight, and significantly faster walking speed. We improve and increase the total number of degree-of-freedom (DOF) of the soft dielectric elastomer actuators (DEAs) from three, used in the S-Hex I, up to five, employed in the S-Hex II, and that provides the promising versatile locomotion ability to the S-Hex II robot. This robot has successfully demonstrated the back and forth ambulation on flat surfaces using the alternating tripod gait with the walking speed of 52 mm/s (approximately 0.35 body-lengths per second) at 7 Hz of actuation frequency.