IROS 2017 论文列表

The knowledge of the static scene parts and the moving objects in a dynamic scene plays a vital role for scene modelling, understanding, and landmark-based robot navigation. The key information for these tasks lies on semantic labels of the scene parts and the motion trajectories of the dynamic objects. In this work, we propose a method that segments the 3D feature trajectories based on their motion behaviours, and assigns them semantic labels using 2D-to-3D label transfer. These feature trajectories are constructed by using the proposed trajectory recovery algorithm which takes the loss of feature tracking into account. We introduce a complete framework for static-map and dynamic objects' reconstruction, as well as semantic scene understanding for a calibrated and moving 2D-3D camera setup. Our motion segmentation approach is faster by two orders of magnitude, while performing better than the state-of-the-art 3D motion segmentation methods, and successfully handles the previously discarded incomplete trajectory scenarios.

This paper presents a new teleoperated spherical tensegrity robot capable of performing locomotion on steep inclined surfaces. With a novel control scheme centered around the simultaneous actuation of multiple cables, the robot demonstrates robust climbing on inclined surfaces in hardware experiments and speeds significantly faster than previous spherical tensegrity models. This robot is an improvement over other iterations in the TT-series and the first tensegrity to achieve reliable locomotion on inclined surfaces of up to 24°. We analyze locomotion in simulation and hardware under single and multi-cable actuation, and introduce two novel multi-cable actuation policies, suited for steep incline climbing and speed, respectively. We propose compelling justifications for the increased dynamic ability of the robot and motivate development of optimization algorithms able to take advantage of the robot's increased control authority.

In this paper we propose a method for pivoting an object held by a parallel gripper, without requiring accurate dynamical models or advanced hardware. Our solution uses the motion of the robot arm for generating inertial forces to move the object. It also controls the rotational friction at the pivoting point by commanding a desired distance to the gripper's fingers. This method relies neither on fast and precise tracking systems to obtain the position of the tool, nor on real-time and high-frequency controllable robotic grippers to quickly adjust the finger distance. We demonstrate the efficacy of our method by applying it on a Baxter robot.

Automatically correlating plant observable characteristics to their underlying genetics will streamline selection methods in plant breeding. Measurement of plant observable characteristics is called phenotyping, and knowing plant phenotypes accurately and throughout a plant's growth is central to making breeding decisions. In-field plant phenotyping in an automated and noninvasive manner is hence crucial to accelerating plant breeding methods. However, most of the existing methods on plant phenotyping using visual imaging are confined to controlled greenhouse environments. This paper presents an automated method of mapping 2D images collected in an outdoor sorghum field to segmented 3D plant units that are of interest for phenotyping. This method leverages multiple horizontal and vertical viewpoints while capturing 2D images from a robotic platform so as to generate in-field 3D reconstructions of the sorghum plant. We develop and quantitatively evaluate segmentation methods on these 3D reconstructions and also compare against reconstructions obtained from a controlled greenhouse environment. We present analysis that contrasts the role of purely local geometric features and the effect of addition of global context in both datasets. This work furthers capabilities of in-field phenotyping which paves the way forward for plant biologists to study the coupled effect of genetics and environment on improving crop yields.

In the physical human-robot interaction system, improving transparency performance which allows humans to move as if there is no robot is a challenging topic. In general physical human-robot interaction, the multiaxial force sensor is used to control the robot. However the signal measured from the force sensor involves not only the force applied to the robot by the human motion intention but also the influence of the natural force feedback between the robot and the human hand due to the motion of the robot. Therefore, in order to improve the transparency performance, it is necessary to consider the dynamics of the human hand as well as the dynamics of the robot. In this paper, an novel algorithm to improve transparency performance is proposed. The proposed algorithm uses only the multiaxial force sensor and compensates the natural force feedback by using the human hand impedance and further improves the transparency performance which has a limitation adjusting admittance parameters. The validity of the proposed algorithm is described through the system model analysis and the experiment using the hydraulic upper limb exoskeleton.

Ensuring safety is a primary goal in physical human-robot interaction. In various collision experiments it was found that the robot's effective mass, velocity, and geometry are the key parameters which influence the human injury severity during an impact. Recently, a velocity controller was proposed that limits the robot speed to a biomechanically safe value, taking into account the mass and the curvature in the direction of movement for a given point of interest. The mass and the geometry depend on the mechanical design, however, the effective mass also depends on the robot configuration. In this paper, we exploit the redundant degree(s) of freedom of a joint torque controlled seven- and eight-DOF robot to minimize the effective mass without affecting the desired Cartesian end-effector trajectory and with the goal to improve the performance of the safe velocity controller at the same time. Given recent results in robotics injury analysis, we analyze when such a redundancy resolution scheme actually improves safety. For the considered robots, we find reflected mass extrema that can be obtained by null space motions, and propose a real-time, torque-based redundancy resolution scheme, which is finally verified in experiments.

Surgical robot research is driven by the desire of improving surgical outcomes. This paper proposed a Recurrent Neural Network based controller to address two problems: 1) improving control precision, 2) increasing adaptiveness for robot motion (explained in Section I). RNN was adopted in this work mainly because 1) the problem formulation naturally matches RNN structure, 2) RNN has advantages as an biologically inspired method. The proposed method was explained in detail and analysis shows that the proposed method is able to dynamically regulate outputs to increase the adaptiveness and the control precision. This paper uses Raven II surgical robot as an example to show the application of the proposed method, and the numeral simulation results from the proposed method and three other controllers show that the proposed method has improved precision, improved high robustness against noise and increased movement smoothness, and it keeps the manipulator links as far away as possible from physical boundaries, which potentially increases surgical safety and leads to improved surgical outcomes.

The place recognition problem comprises two distinct subproblems; recognizing a specific location in the world (“specific” or “ordinary” place recognition) and recognizing the type of place (place categorization). Both are important competencies for mobile robots and have each received significant attention in the robotics and computer vision community, but usually as separate areas of investigation. In this paper, we leverage the powerful complementary nature of place recognition and place categorization processes to create a new hybrid place recognition system that uses place context to inform place recognition. We show that semantic place categorization creates an informative natural segmenting of physical space that in turn enables significantly better place recognition performance in comparison to existing techniques. In particular, this new semantically-informed approach adds robustness to significant local changes within the environment, such as transitioning between indoor and outdoor environments or between dark and light rooms in a house, complementing the capabilities of current condition-invariant techniques that are robust to globally consistent change (such as day to night cycles). We perform experiments using 4 novel benchmark datasets and show that semantically-informed place recognition outperforms the previous state-of-the-art systems. Like it does for object recognition [1], we believe that semantics can play a key role in boosting conventional place recognition and navigation performance for robotic systems.

For certain manipulation tasks, object pose estimation from head-mounted cameras may not be sufficiently accurate. This is at least in part due to our inability to perfectly calibrate the coordinate frames of today's high degree of freedom robot arms that link the head to the end-effectors. We present a novel framework combining pre-touch sensing and deep learning to more accurately estimate pose in an efficient manner. The use of pre-touch sensing allows our method to localize the object directly with respect to the robot's end effector, thereby avoiding error caused by miscalibration of the arms. Instead of requiring the robot to scan the entire object with its pre-touch sensor, we use a deep neural network to detect object regions that contain distinctive geometric features. By focusing pre-touch sensing on these regions, the robot can more efficiently gather the information necessary to adjust its original pose estimate. Our region detection network was trained using a new dataset containing objects of widely varying geometries and has been labeled in a scalable fashion that is free from human bias. This dataset is applicable to any task that involves a pre-touch sensor gathering geometric information, and has been made publicly available. We evaluate our framework by having the robot re-estimate the pose of a number of objects of varying geometries. Compared to two simpler region proposal methods, we find that our deep neural network performs significantly better. In addition, we find that after a sequence of scans, objects can typically be localized to within 0.5 cm of their true position. We also observe that the original pose estimate can often be significantly improved after collecting a single quick scan.

In this paper, we present a design and control approach of a modular off-board actuation system with mono-articular hip exosuit, enabling the instantaneous assistive profile modification of magnitude, shape, and timing. The off-board system consists an actuation unit with two degrees of freedom that can transmit forces to our soft exosuit via Bowden cables. To perform accurate force tracking and cable slack management, we implemented a switching admittance-position control approach that utilizes the advantages of both admittance and position control. In particular, feedforward models (FFMs) are added to the admittance control to compensate variability of individual kinematics, nonlinear hip exosuit stiffness and actuator transmission loss to enable a more accurate force tracking. This system allows us to track force profiles accurately and robustly under normal walking or jogging speed. The performance of force tracking was evaluated on three subjects walking on a treadmill at 1.25 m/s with various magnitude, shape, and timings of assistive profiles and jogging on a treadmill at 2.3 m/s with varied magnitude of a representative assistance profile. Results showed that assistance could be delivered reliably across different profiles. In case of walking experiments, the errors of the magnitude, RMS-E, and timing were within 2.0 N, 8.4 N, and 1.8% when varying the magnitude, shape, and timing respectively. In case of jogging experiments, the RMS-E was within 3.0 N. This system will be used for future research examining how to individualize assistive profiles for different wearers during walking and jogging.

A GelSight sensor uses an elastomeric slab covered with a reflective membrane to measure tactile signals. It measures the 3D geometry and contact force information with high spacial resolution, and successfully helped many challenging robot tasks. A previous sensor [1], based on a semi-specular membrane, produces high resolution but with limited geometry accuracy. In this paper, we describe a new design of GelSight for robot gripper, using a Lambertian membrane and new illumination system, which gives greatly improved geometric accuracy while retaining the compact size. We demonstrate its use in measuring surface normals and reconstructing height maps using photometric stereo. We also use it for the task of slip detection, using a combination of information about relative motions on the membrane surface and the shear distortions. Using a robotic arm and a set of 37 everyday objects with varied properties, we find that the sensor can detect translational and rotational slip in general cases, and can be used to improve the stability of the grasp.

The use of concentric tube robots has recognized advantages for accessing target lesions while conforming to certain anatomical constraints. However, their complex kinematics makes their safe telemanipulation in convoluted anatomy a challenging task. Collaborative control schemes, which guide the operator through haptic and visual feedback, can simplify this task and reduce the cognitive burden of the operator. Guaranteeing stable, collision-free robot configurations during manipulation, however, is computationally demanding and, until now, either required long periods of pre-computation time or distributed computing clusters. Furthermore, the operator is often presented with guidance paths which have to be followed approximately. This paper presents a heterogeneous (CPU/GPU) computing approach to enable rapid workspace analysis on a single computer. The method is used in a new navigation scheme that guides the robot operator towards locations of high dexterity or manipulability of the robot. Under this guidance scheme, the user can make informed decisions and maintain full control of the path planning and manipulation processes, with intuitive visual feedback on when the robot's limitations are being reached.

This study proposes an actuator whose design relies on a coordinated approach to control and hardware design: impedance control is supplemented with the introduction of a spring-damper coupler between the actuator and reference (ground). The coupler between the actuator and reference has the effect of further reducing the impedance apparent to the environment. Unlike traditional series elastic actuation, where the deformation of a physical spring coupler introduced between an actuator and link is fed back for control, our coupler is located between actuator and reference and its deformation is not fed back for control. The proposed control further employs time-delay estimation to account for the system dynamics that include both the physical and virtual couplers and to realize accurate and robust control over end-effector position and contact forces. We present a simple method for estimating interaction forces and regulating the contact force without a priori knowledge of the environment. A numerical simulation demonstrates the efficacy of the proposed approach.

Magnetically actuated microswimmers have shown great potentials in multiple application scenarios, attracting researchers to investigate their swimming characteristics and controlling methods. However, among those studies, only a small number of closed-loop control schemes have been applied, which is crucial for the accuracy and repeatability in applications. In this paper, we proposed an Image-Based Visual Servoing(IBVS) method for arbitrary planar path following using features directly presented in image space as feedbacks. During experiments we found that IBVS guarantees convergence while not requiring camera calibrations. Furthermore, the proposed path following method is intuitive and flexible, and provides great potentials in various applications.

This paper proposes an image-based localization method that enables to estimate a bounded domain of the pose of an unmanned aerial vehicle (UAV) from uncertain measurements of known landmarks in the image. The approach computes a domain that should contain the actual robot pose, assuming bounded image measurement errors and landmark position uncertainty. It relies on interval analysis and constraint propagation techniques to rigorously back-propagate the errors through the non-linear observation model. Attitude information from onboard sensors is merged with image observations to reduce the pose uncertainty domain, along with prediction based on velocity measurements. As tracking landmarks in the image is prone to errors, the proposed method also enable fault detection from measurement inconsistencies. This method is tested using a quadcopter UAV with an onboard camera.

In current approaches to combined task and motion planning, usually symbolic planning and sampling based motion-planning are integrated. One problem is here to come up with good samples. We address the problem of identifying useful poses for a robot close to working surfaces such as tables or shelves. Our approach is based on reachability inversion which answers the question: where should the robot be located in order to reach a certain object? We extend the concept from point-based objects to flat polygonal surfaces in order to enable the robot to have a a good grasping position for many objects. Our approach allows to quickly sample multiple distinct poses for the robot from an prior computed distribution. Further we show how sampling from an inverse reachability distribution can be integrated into a CTAMP system.

Exoskeletons are wearable robotic systems to assist the human body concerning power and accuracy. In this work, an upper limb exoskeleton with seven degrees of freedom provides haptic guidance to the user enhancing accuracy of the target position and constant thrust force during drilling tasks. Therefore, an inverse kinematics algorithm is introduced using selective damping, depending on joint velocities to minimize the end effector position error for redundant systems. Additionally, the method recognizes user intention with structurally integrated torque sensors for null space optimization. An implicit, hybrid force-position controller is implemented to perform position controlled drilling tasks with constant thrust force. The performance evaluation with the exoskeleton shows a maximum position error of 1.27 mm and steady-state thrust force response with a maximum overshoot of 1N.

Trajectory optimization is a fundamental problem in robotics. While optimization of continuous control trajectories is well developed, many applications require both discrete and continuous, i.e. hybrid controls. Finding an optimal sequence of hybrid controls is challenging due to the exponential explosion of discrete control combinations. Our method, based on Differential Dynamic Programming (DDP), circumvents this problem by incorporating discrete actions inside DDP: we first optimize continuous mixtures of discrete actions, and, subsequently force the mixtures into fully discrete actions. Moreover, we show how our approach can be extended to partially observable Markov decision processes (POMDPs) for trajectory planning under uncertainty. We validate the approach in a car driving problem where the robot has to switch discrete gears and in a box pushing application where the robot can switch the side of the box to push. The pose and the friction parameters of the pushed box are initially unknown and only indirectly observable.

Modern autonomous systems such as driverless vehicles need to safely operate in a wide range of conditions. A potential solution is to employ a hybrid systems approach, where safety is guaranteed in each individual mode within the system. This offsets complexity and responsibility from the individual controllers onto the complexity of determining discrete mode transitions. In this work we propose an efficient framework based on recursive neural networks and coreset data summarization to learn the transitions between an arbitrary number of controller modes that can have arbitrary complexity. Our approach allows us to efficiently gather annotation data from the large-scale datasets that are required to train such hybrid nonlinear systems to be safe under all operating conditions, favoring underexplored parts of the data. We demonstrate the construction of the embedding, and efficient detection of switching points for autonomous and non-autonomous car data. We further show how our approach enables efficient sampling of training data, to further improve either our embedding or the controllers.