IROS 2017 论文列表

This paper addresses the problem of human-robot collaboration in scenarios where a robot assists a human by executing a complex motion involving the manipulation of an object. We focus on tasks in which success in the task depends on reaching a target pose that is controlled by the human. We contribute a reinforcement learning-based approach that allows the robot to reason about its own ability to successfully complete the task given the current target pose and indirectly adjust that pose by prompting the human user. Our approach allows the robot both to trade-off the benefits of adjusting the target position against the cost of bothering the human user while, at the same time, adapting to each user's responses. Our approach was tested in a real-world human-robot collaboration scenario involving the Baxter robot.

Six-degree of freedom (DOF) inertial measurement units (IMUs) are widely used for attitude estimation. However, such systems' accuracy is limited by the accuracy of calibration of bias, scale factors, and non-orthogonality of sensor measurements. This paper reports a stable adaptive estimator of measurement bias in six-DOF IMUs and preliminary simulation results employing a commercially available IMU comprising a 3-axis fiber optic gyroscope (FOG) with a 3-axis micro-electro-mechanical systems (MEMS) accelerometer. A stability proof of the adaptive estimator and preliminary numerical simulation results are reported. The simulation results for a rotating IMU configuration are promising, and further experimental evaluation and extension of the algorithm for the case of a translating IMU configuration typically found on moving robotic vehicles are needed.

This paper presents a method to achieve online gait adaptation of a dynamically walking biped when collaborating with an external agent-either a human or a robot-acting as a leader. Adaptation occurs without any explicit information on the leader's intended motion; only implicit information is used through the interaction force developed between the leader and the biped. An adaptive supervisory control scheme is proposed and guarantees for boundedness of the state despite switching under external force are provided. The supervisory controller leverages the availability of a library of exponentially stable limit-cycle gaits, and orchestrates switching among them in an online fashion to achieve adaptation. As a result, the range of leader speeds that the biped can adapt to is drastically enlarged while the leader's effort is reduced.

We present an optimization framework for the design and analysis of underactuated biped walkers, characterized by passive or actuated joints with rigid or non-negligible elastic actuation/transmission elements. The framework is based on optimal control, dealing with geometric constraints and various dynamic objective functions, as well as boundary conditions, which helps in selecting optimal values both for the actuation and the transmission parameters. Solutions of the formulated problems are shown for different kinds of bipedal architectures, and comparisons drawn between traditional rigid robots and compliant ones show the energy-efficiency of compliant actuators in the context of locomotion.

In active vision, the camera motion is controlled in order to improve a certain visual sensing strategy. In this paper, we formulate an active vision task function to improve pose estimation. This is done by defining an optimality metric on the Fisher Information Matrix. This task is then incorporated into a weighted multi-objective optimization framework. To test this approach, we apply it on the three image point visual servoing problem which has a degenerate configuration - a singularity cylinder. The simulation results show that the singular configurations of pose estimation are avoided during visual servoing. We then discuss the potential of active vision to be integrated into more complex multi-task frameworks.

We consider an agricultural automation scenario where a robot, equipped with a camera mounted on a manipulator, is charged with counting the number of apples in an orchard. We focus on the subtask of planning views so as to accurately estimate the number of apples in an apple cluster. We present a method to efficiently enumerate combinatorially distinct world models and to compute the most likely model from one or more views. These are incorporated into single and multi-step planners. We evaluate these planners in simulation as well as with experiments on a real robot.

Designing and manufacturing actuators using soft materials are among the most important subjects for future robotics. In nature, animals made by soft tissues such as the octopus have attracted the attention of the robotics community in the last years. Suckers (or suction cups) are one of the most important and peculiar organs of the octopus body, giving it the ability to apply high forces on the external environment. The integration of suction cups in soft robots can enhance their ability to manipulate objects and interact with the environment similarly to what the octopus does. However, artificial suction cups are currently actuated using fluid pressure so most of them require external compressors, which will greatly increase the size of the soft robot. In this work, we proposed the use of the ElectroHydroDynamics (EHD) principle to actuate a suction cup. EHD is a fluidic phenomenon coupled with electrochemical reaction that can induce pressure through the application of a high-intensity electric field. We succeeded in developing a suction cup driven by EHD keeping the whole structure extremely simple, fabricated by using a 3D printer and a cutting plotter. We can control the adhesion of the suction cup by controlling the direction of the fluidic flow in our EHD pump. Thanks to a symmetrical arrangement of the electrodes, composed by plates parallel to the direction of the channel, we can change the direction of the flow by changing the sign of the applied voltage. We obtained the pressure of 643 Pa in one unit of EHD pump and pressure of 1428 Pa in five units of EHD pump applying 6 kV. The suction cup actuator was able to hold and release a 2.86 g piece of paper. We propose the soft actuator driven by the EHD pump, and expand the possibility to miniaturize the size of soft robots.

Animal locomotion exhibits all the features of complex non linear systems such as multi-stability, critical fluctuation, limit cycle behavior and chaos. Studying these aspects on real robots has been proved difficult and therefore results mostly rely on the use of computer simulation. Simple control approaches - based on phase oscillators - have been proposed and exhibit several of these features. In this work, we compare two types of controllers: (a) an open loop control approach based on phase oscillators and (b) the Tegotae-based closed loop extension of this controller. The first controller has been shown to exhibit synchronization features between the body and the controller when applied to a quadruped robot with compliant leg structures. In this contribution, we apply both controllers to the locomotion of a stiff quadruped structure. We show that the Tegotae-controller exhibits self-organizing behavior, such as spontaneous gait transition and critical fluctuation. Moreover, it exhibits features such as the ability to stabilize both asymmetric and symmetric morphological changes, despite the lack of compliance in the leg.

High accuracy navigation in GPS-deprived environments is of prime importance to various robotics applications and has been extensively investigated in the last two decades. Recent approaches have shown that incorporating sensor's calibration states in addition to the 6DOF pose states may cause better performance of the system. However, these approaches typically consider a passive setting, where robot actions are externally defined. On the other hand, belief space planning (BSP) approaches account for different sources of uncertainty, thus identifying actions that improve certain aspects in inference, such as accuracy. Yet, existing BSP approaches typically do not consider sensor calibration, nor a visual-inertial SLAM setup. In this paper we contribute a BSP approach for active sensor calibration of a visual-inertial SLAM setup. For this purpose we incorporate within the belief both robot's pose and sensor calibration states while considering operation in partially unknown and uncertain environment. In particular, we leverage the recently developed concept of IMU pre-integration and develop appropriate factor graph formulation for future beliefs to facilitate computationally efficient inference within BSP. Our approach is valid for general cost functions, and can be used to identify best robot actions from a given set of candidate actions or to calculate locally-optimal actions using direct trajectory optimization techniques. We demonstrate our approach in high-fidelity synthetic simulation and show that incorporate sensors calibration state into the BSP significantly improved estimation accuracy.

In the present work, we propose an active learning framework based on optimal query paths to efficiently address the problem of tactile object shape exploration. Most previous approaches perform active touch probing at discrete query points, which leads to inefficient touch-and-retract motions. In contrast, in this paper we propose to query information efficient sliding paths instead of only touch locations. This is realized by three components: A Gaussian process implicit surface model represents the shape and uncertainty of the object. A compliant task/force controller framework fuses the information of this GP model into the parameterization of its tasks, which enables the robot to slide over the unknown object safely and robustly. Thirdly, we develop two strategies to solve the proposed active path querying learning problem. Sliding along those query paths not only creates more dense data than touch probing, but additionally greatly reduces the uncertainty of the object. We demonstrate the effectiveness of our proposed framework both in simulation and on the PR2 robot platform. Furthermore, it is shown that our methodology can be extended to other learning tasks, such as finding a desired surface normal on an unknown object, e.g. for pushing.

Acquiring High Dynamic Range (HDR) photos from several images, with an active shutter providing different exposures (sensor integration periods), has been widely commercialised in photography for static camera positions. In the case of a mobile video sensor (as is the case in robotics), this problem is more difficult due to real-time motion of the sensor which transforms the perspective between the acquired images. HDR approaches for a set of images from different perspectives have therefore been significantly overlooked since this would require sophisticated dense mapping approaches to eliminate the motion component. Recent dense visual SLAM (Simultaneous Localization And Mapping) approaches provide this framework, however, few works have attempted to perform HDR visual SLAM. Current approaches are thus highly dependant on illumination conditions and camera shutter settings. In this paper a new approach is proposed that enables 3D HDR environment maps to be acquired actively from a dynamic set of images in real-time. The 6 DOF pose, the dense scene structure and the HDR texture map will be estimated simultaneously with the objective of maximising the dynamic range. This will allow to obtain a radiance map of the scene by fusing a real-time stream of low dynamic range images (LDR) into a graph of HDR key-frame images. In particular, a method is proposed to actively control the shutter based on information theory to optimise the information content of the 3D HDR environment map for RGB-D sensors. As will be shown in the results, a 3D HDR environment map allows mapping (visual SLAM) to take actively advantage of varying luminosity in different parts of the scene.

This paper considers the problem of active object recognition using touch only. The focus is on adaptively selecting a sequence of wrist poses that achieves accurate recognition by enclosure grasps. It seeks to minimize the number of touches and maximize recognition confidence. The actions are formulated as wrist poses relative to each other, making the algorithm independent of absolute workspace coordinates. The optimal sequence is approximated by Monte Carlo tree search. We demonstrate results in a physics engine and on a real robot. In the physics engine, most object instances were recognized in at most 16 grasps. On a real robot, our method recognized objects in 2-9 grasps and outperformed a greedy baseline.

As robots begin to inhabit public and social spaces, it is increasingly important to ensure that they behave in a socially appropriate way. However, manually coding social behaviours is prohibitively difficult since social norms are hard to quantify. Therefore, learning from demonstration (LfD), wherein control policies are inferred from demonstrations of correct behaviour, is a powerful tool for helping robots acquire social intelligence. In this paper, we propose a deep learning approach to learning social behaviours from demonstration. We apply this method to two challenging social tasks for a semi-autonomous telepresence robot. Our results show that our approach outperforms gradient boosting regression and performs well against a hard-coded controller. Furthermore, ablation experiments confirm that each element of our method is essential to its success.

In this paper, we proposed a model for a wire-driven continuum manipulator without assuming constancy in curvatures of the manipulator's shape as a natural extension of the famous and useful constant curvature model. We show that the tip orientation of the continuum manipulator can be decided completely by pulling-wire lengths, regardless of its whole shape, which is a useful property same as the constant curvature model. Moreover, we can derive the tip position, which depends on the whole shape, by assuming distributed forces along wires mainly due to the friction between the wires and their surrounding materials. Experimental validation results show that, based on the model, we can estimate the tip orientation from pulling-wire lengths with good accuracy and precision. We also provide experimental results for discussion about estimating the tip position with the realistic assumption on the force distribution along wires.

In this paper, a wearable hand system for virtual reality is proposed. The proposed system can deliver accurate force to three fingertips (thumb, index, and middle fingers), which play dominant roles in manipulating and interacting with the external environment, and also measure fingertip positions of the three fingers. The proposed system has five degrees of freedom (DOFs) for each finger structure to permit natural finger motion, can be easily worn by clicking the silicon-coated fingertip structures and buckling the wearable structure around the palm. The forces delivered to the fingertips are controlled by the motor current with the robust control algorithm, and the fingertip position is calculated by forward kinematics. The performance of the proposed system was verified experimentally; the calculated fingertip position was compared to a camera-based motion capture system and interactions with objects of different stiffness were tested.

This paper presents a vision-based learning-by-demonstration approach for multi-robot manipulation. With this method, a vision system is involved in both the task demonstration and reproduction stages, and the speed and accuracy of the task reproduction are adapted according to the context of the demonstration. An expert first demonstrates how to use tools to perform a task, while the tool motion is observed using a vision system. The demonstrations are then encoded using a statistical model to generate a reference motion trajectory. Equipped with the same tools and the learned model, the robot is guided by vision to reproduce the task. The task performance was evaluated in terms of both accuracy and speed. However, simply increasing the robot's speed could decrease the reproduction accuracy. To this end, a dual-rate Kalman filter is employed to compensate for latency between the robot and vision system. More importantly, the robot speed is adapted according to the learned motion model. We demonstrate the effectiveness of our approach by performing two tasks: a trajectory reproduction task and a bimanual sewing task. We show that using our vision-based approach, the robots can conduct effective learning by demonstrations and perform accurate and fast task reproduction. The proposed approach is generalisable to other manipulation tasks, where bimanual or multi-robot cooperation is required.

Re-configurable modular robotic (RMR) systems are advantageous for their reconfigurability and versatility. A new modular robot can be built for a specific task by using modules as building blocks. However, constructing a kinematic model for a newly conceived robot requires significant work. Due to the finite size of module-types, models of all module-types can be built individually and stored in a database beforehand. With this priori knowledge, the model construction process can be automated by detecting the modules and their corresponding interconnections. Previous literature proposed theoretical frameworks for constructing kinematic models of modular robots, assuming that such information was known a priori. While well-devised mechanisms and built-in sensors can be employed to detect these parameters, they significantly complicate the module design and thus are expensive. In this paper, we propose a vision-based method to identify kinematic chains and automatically construct robot models for modular robots. Each module is affixed with augmented reality (AR) tags that are encoded with unique IDs. An image of a modular robot is taken and the detected modules are recognized by querying a database that maintains all module information. The poses of detected module-links are used to compute: (i) the connection between modules and (ii) joint angles of joint-modules. Finally, the robot serial-link chain is identified and the kinematic model is constructed and visualized. Our experimental results validate the effectiveness of our approach. While implementation with only our RMR is shown, our method can be applied to other RMRs where self-identification is not possible.

In this paper, we propose a classification method for single views of 3D objects with missing data retrieval. A mobile robot equipped with range sensors basically obtains only single view information of a 3D scene. Therefore, large amount of information is missing by self-occlusion, which leads to severe restriction on the object classification exploiting the whole shapes of 3D objects. Humans can precisely identify the objects from single view, since they already have concepts of the entire shape of 3D objects by learning process. Based on these concepts, humans can infer the entire shape and category of the object from a single view. Inspired from this, the proposed algorithm learns concepts in abbreviated form for the shapes of 3D objects, then infers the entire shape and object category from these concepts simultaneously. We apply a generative model based on variational auto-encoder (VAE) to learn the concepts for complex shapes of 3D objects. Our method is evaluated on 3D CAD model dataset, and also compared with other state-of-the-art methods.

Remote magnetic navigation of catheters is a technique used to perform radiofrequency ablation of heart tissue in order to treat cardiac arrhythmias. The flexible magnetic catheters used in this context are in some cases not sufficiently dexterous to navigate the complex and patient-specific anatomy of the heart. To overcome such limitations, this paper proposes a new approach that relies on the integration of variable stiffness segments into a magnetic catheter. The magnetic variable stiffness (VS) catheter presented here is based on silicone and a low melting point alloy (LMPA) that transforms from a solid to liquid phase upon joule heating. This dramatically changes the bending stiffness of the segment in which it is integrated, improving dexterity. Compared to standard catheters, a VS catheter can partially (just one segment) or completely lock its shape (shape fixity) in order to explore a larger 3D volume inside a magnetic navigation system, thus extending regions of the heart that can be reached for performing ablation procedures.