IROS 2016 论文列表

To properly perform tasks based on abstract instructions, autonomous robots need refined reasoning skills in order to bridge the gap between the ambiguous descriptions and the comprehensive information needed to execute the implied actions. In this article, we present an automated knowledge acquisition system from human executed tasks in virtual environments, and extend the knowledge processing system KNOWROB[1] to be capable to reason on the acquired data. We have set up two scenarios in a physics based simulator: creating a pancake, and garnishing a pizza dough. Users where asked to execute these tasks using the provided tools and ingredients. Using a data processing module we then collect the low-level data and the relevant abstract events from the performed episodes. The recorded data is then made available in a format that robots can understand, by using a symbolic layer to interconnect the two data types in a seamless way.

We study the problem of objects search in clutter. In cluttered environments, partial occlusion among objects prevents vision systems from correctly recognizing objects. Hence, the agent needs to move objects around to gather information, which helps reduce uncertainty in perception. At the same time, the agent needs to minimize the efforts of moving objects to reduce the time required to complete the task. We model the problem as a Partially Observable Markov Decision Process (POMDP), formulating it as a problem of optimal decision making under uncertainty. By exploiting spatial constraints, we are able to adapt online POMDP planners to handle objects search problems with large state space and action space. Experiments show that the POMDP solution outperforms greedy approaches, especially in cases where multi-step manipulation is required.

To achieve tasks in unknown environments with high reliability, highly accurate localization during task execution is necessary for humanoid robots. In this paper, we discuss a localization system which can be applied to a humanoid robot when executing tasks in the real world. During such tasks, humanoid robots typically do not possess a referential to a constant horizontal plane which can in turn be used as part of fast and cost efficient localization methods. We solve this problem by first computing an improved odometry estimate through fusing visual odometry, feedforward commands from gait generator and orientation from inertia sensors. This estimate is used to generate a 3D point cloud from the accumulation of successive laser scans and such point cloud is then properly sliced to create a constant height horizontal virtual scan. Finally, this slice is used as an observation base and fed to a 2D SLAM method. The fusion process uses a velocity error model to achieve greater accuracy, which parameters are measured on the real robot. We evaluate our localization system in a real world task execution experiment using the JAXON robot and show how our system can be used as a practical solution for humanoid robots localization during complex tasks execution processes.

The torque control of finger joints is important for effective hand rehabilitation after neural disorders such as stroke. This paper presents an approach for accurate torque control of finger joints with UT hand exoskeleton. We present (1) how we obtained an accurate kinematics model, (2) how we built the torque actuation model with Bowden cable SEA, and (3) how we controlled the torque of finger joints with the kinematic model and the Bowden cable SEA. We have validated our approach with a testbed finger and results show that the UT hand exoskeleton accurately controls the torque of finger joints.

The reliable manipulation of micro-objects has been a still difficult work in scientific and technical field due to scale effects. This paper presents two types of release methods, using local stream and inertia force generated by 3D high speed motion of an end effector, for releasing and accurate positioning of biological cells. Two-fingered microhand driven by DC motors for both end effectors and PZT actuators for right end effector is employed. A parallel mechanism controlled by three PZT actuators generates 3D high speed motions to release cells adhered to one of the end effector. The local stream and inertia force created by high speed motion of the right end effector detach the cells adhered to the left end effector and right end effector, respectively. To generate the necessary external forces for separation of the attached cells, the vibration having high frequency and suitable amplitude is applied. For accurate positioning of the object, circular motions are proposed. To verify the advantage of the proposed motion, we compare five motions, three 1D motions and two circular motions. Experiments were conducted employing 16μm NIH3T3 cells. From these analyses of experiments, we conclude that the proposed motions can detach micro objects (100%) with high position accuracy (3±0.7μm) on desired position after release.

To render a wearable robot imperceptible to a user is a very challenging control task. The constant and intrinsic interaction between robot and human, and person-dependent behaviours are the main difficulties when designing such cooperative control. In this contribution we introduce and discuss a novel and promising transparency control framework. The foundation of the framework is to measure the acceleration of the human limbs and to exploit this measurement to generate feedforward control commands by using a rigid body model of the robot. The framework includes also an acceleration feedback controller and a state estimator to enhance the overall performance. We present a simplified stability analysis with different feedback controllers and preliminary experimental data that demonstrate the potential of the proposed method in reducing interaction forces and mimicking human motions.

The present paper introduces a manipulator that is developed to combine safe and dynamic interaction tasks. The system is built from lightweight carbon fiber links and novel high-performance series elastic actuator units that provide dynamic movement capability, low-impedance joint torque control, and inherent interaction safety. This enabled the implementation of a model-based direct force control method purely based on joint torque regulation. Using unified force and motion control, the end-effector position can be accurately and dynamically tracked in task space while acting safely upon (unexpected) contacts with the environment. The force control component is implemented in a novel way that shows reduced forces in comparison to existing methods when navigating across a surface of unpredictable orientation and friction. ANYpulator is tested using a haptic feedback method that renders the system dynamics and contact forces back to the user.

This paper introduces ANYmal, a quadrupedal robot that features outstanding mobility and dynamic motion capability. Thanks to novel, compliant joint modules with integrated electronics, the 30 kg, 0.5 m tall robotic dog is torque controllable and very robust against impulsive loads during running or jumping. The presented machine was designed with a focus on outdoor suitability, simple maintenance, and user-friendly handling to enable future operation in real world scenarios. Performance tests with the joint actuators indicated a torque control bandwidth of more than 70 Hz, high disturbance rejection capability, as well as impact robustness when moving with maximal velocity. It is demonstrated in a series of experiments that ANYmal can execute walking gaits, dynamically trot at moderate speed, and is able to perform special maneuvers to stand up or crawl very steep stairs. Detailed measurements unveil that even full-speed running requires less than 280 W, resulting in an autonomy of more than 2 h.

This paper presents a novel design of a compact and light-weight robotic hand for a social humanoid robot. The proposed system is able to perform common hand gestures and self-adaptable grasps by mixing under-actuated and self-adaptable hand kinematics in a unique design. The hand answers the need for precise finger postures and sensor-less force feedback during gestures and for finger adaptation and autonomous force distribution during grasps. These are provided by a dual actuation system embodied within the palm and the fingers. Coexistence is ensured by compliant transmissions based on elastomer bars rather than classical tension springs, thanks to their high elastic coefficient at reduced sizes and strains. The proposed solution significantly reduces the weight and the size of the hand by using a reduced number of small actuators for gesturing and a single motor for grasping. The hand prototype (ALPHA) is realized to confirm the design feasibility and functional capabilities. It is controlled to provide safe human-robot interaction and preserve mechanical integrity in order to be embodied on a humanoid robot.

Cellular mechanical properties as the main physical performance characteristics have been actively studied in the past years for the study of cytobiology and the development of medicine. In this study, by combining Hertz model, a novel strategy is proposed to simultaneously measure the cellular mechanical properties including cellular mass, elasticity and viscosity, based on the principle of forced vibration stimulated by simple harmonic force, with piezoelectric transducer (PZT) as vibrator and Atomic Force Microscope (AFM) as detector. The corresponding theoretical model was derived and the simulation was realized based on the proposed model. The experiments of indentations and vibrations with myoblasts and myotubes were implemented to calculate the three mechanical parameters of cells according to the proposed strategy. The results validated the proposed approach. This work would be useful for the development of cytology, medicine, previously diagnose, specific therapy and so on.

Cleaning up the past half century of nuclear waste represents the largest environmental remediation project in the whole Europe. Nuclear waste must be sorted, segregated and stored according to its radiation level in order to optimize maintenance costs. The objective of this work is to develop a shared control framework for remote manipulation of objects using visual information. In the presented scenario, the human operator must control a system composed of two robotic arms, one equipped with a gripper and the other one with a camera. In order to facilitate the operator's task, a subset of the gripper motion are assumed to be regulated by an autonomous algorithm exploiting the camera view of the scene. At the same time, the operator has control over the remaining null-space motions w.r.t. the primary (autonomous) task by acting on a force feedback device. A novel force feedback algorithm is also proposed with the aim of informing the user about possible constraints of the robotic system such as, for instance, joint limits. Human/hardware-in-the-loop experiments with simulated slave robots and a real master device are finally reported for demonstrating the feasibility and effectiveness of the approach.

This paper presents an intelligent sewing system for personalized stent graft manufacturing, a challenging sewing task that is currently performed manually. Inspired by medical suturing robots, we have adopted a single-sided sewing technique using a curved needle to perform the task of sewing stents onto fabric. A motorized surgical needle driver was attached to a 7 d.o.f robot arm to manipulate the needle with a second robot controlling the position of the mandrel. A learning-from-demonstration approach was used to program the robot to sew stents onto fabric. The demonstrated sewing skill was segmented to several phases, each of which was encoded with a Gaussian Mixture Model. Generalized sewing movements were then generated from these models and were used for task execution. During execution, a stereo vision system was adopted to guide the robots and adjust the learnt movements according to the needle pose. Two experiments are presented here with this system and the results show that our system can robustly perform the sewing task as well as adapt to various needle poses. The accuracy of the sewing system was within 2mm.

We present a new descriptor for tactile 3D object classification. It is invariant to object movement and simple to construct, using only the relative geometry of points on the object surface. We demonstrate successful classification of 185 objects in 10 categories, at sparse to dense surface sampling rate in point cloud simulation, with an accuracy of 77.5% at the sparsest and 90.1% at the densest. In a physics-based simulation, we show that contact clouds resembling the object shape can be obtained by a series of gripper closures using a robotic hand equipped with sparse tactile arrays. Despite sparser sampling of the object's surface, classification still performs well, at 74.7%. On a real robot, we show the ability of the descriptor to discriminate among different object instances, using data collected by a tactile hand.

This paper focuses on a real system implementation, analysis, and evaluation of a cooperative sensor fusion algorithm based on a Gaussian Mixture Probability Hypothesis Density (GM-PHD) filter, using simulated and real vehicles endowed with automotive-grade sensors. We have extended our previously presented cooperative sensor fusion algorithm with a fusion weight optimization method and implemented it on a vehicle that we denote as the ego vehicle. The algorithm fuses information obtained from one or more vehicles located within a certain range (that we call cooperative), which are running a multi-object tracking PHD filter, and which are sharing their object estimates. The algorithm is evaluated on two Citroën C-ZERO prototype vehicles equipped with Mobileye cameras for object tracking and lidar sensors from which the ground truth positions of the tracked objects are extracted. Moreover, the algorithm is evaluated in simulation using simulated C-ZERO vehicles and simulated Mobileye cameras. The ground truth positions of tracked objects are in this case provided by the simulator. Multiple experimental runs are conducted in both simulated and real-world conditions in which a few legacy vehicles were tracked. Results show that the cooperative fusion algorithm allows for extending the sensing field of view, while keeping the tracking accuracy and errors similar to the case in which the vehicles act alone.

In this article we present a symbolic closed-form matrix formulation to obtain the dynamic equations of branched articulated multibody systems (AMS)s. The proposed approach uses geometric mechanics based on Screw Theory and Lie groups. Both Lagrange's and Newton-Euler's equation of motion are derived. Furthermore, the structure of the proposed set of geometric equations holds the intrinsic robot parameters explicitly arranged like symbolic matrices. The formulation is valid for any branched AMS without closed kinematic chains and whose joints have one degree of freedom (DoF) (revolute and/or prismatic). All these properties allow the use of these equations in different algorithms such as identification, simulation and control of branched AMSs like hands or humanoids. Finally, the proposed equations have been validated and verified with the multi-body simulation software package MSC=ADAMS© by computing the inverse dynamics of a two arm torso of 16 DoF.

In the last decade, consumer electronic devices such as smartphones, are packaged with small cameras, gyroscopes, and accelerometers, all sensors allowing autonomous deployment of aerial robots in GPS-denied environments. Our previous work [1], demonstrated the feasibility of using smartphones for autonomous flight. In many applications, there is a large interest to the use multiple autonomous aerial vehicles in a cooperative manner to speed up the operation of the mission. In this work, we present the first fully autonomous smartphone-based swarm of quadrotors. Multiple vehicles are able to plan safe trajectories avoiding inter-robot collisions, optimizing at the same time a given task and concurrently building in a cooperative manner a 3-D map of the environment. The sensing, sensor fusion, control, and planning are all done on an offthe- shelf Samsung Galaxy S5 smartphone using just the single camera and IMU available on the phone. The work allows any consumer with multiple smartphones to autonomously drive a swarm of multiple vehicles without GPS, by downloading an app, and have the swarm cooperatively map a 3-D environment.

A novel three axis force sensor, based on magnetic flux measurements, was used in the fingers of a gripper. The force sensor uses three Hall Effect sensors orthogonally placed at the base of a hemisphere made of silicon rubber. A neodymium permanent magnet was inside the hemisphere. When a force was applied to the perimeter of hemisphere, it compressed the hemisphere displacing the magnet. This displacement caused change in the magnetic field around the Hall-effect sensors. By analysing these changes, we calculated the force in three directions using a lookup table. This sensor can be used in robot grippers to manipulate objects dexterously with tactile feedback. The cheap construction, robustness and reliability are few advantages of this sensor for it to be used in industrial applications. The sensor design, simulation and its characterization are presented in this work. Furthermore, as an application, a peg in a hole experiment was carried out to present the ability of the sensors to be used in robot grippers for manipulation tasks.

This work demonstrates an all-elastomer MEMS capacitive strain sensor with high dynamic range (5000:1), and features an inexpensive molding microfabrication process. The sensor is comprised of conductive elastometric comb capacitors embedded in a dielectric. Two different sensor designs, lateral combs (LC) and transverse combs (TC), were developed to evaluate sensor sensitivity as a function of load orientation. Both sensors have combs with a gap of 30 μm, length of 4 mm, and depth of 65 μm. A linear elastic analytical model was developed to predict change in capacitance as a function of strain, and experimental results show a reasonable agreement with the theoretical predictions. The observed strain responses have high linearity and dynamic range, and negligible hysteresis. The strain resolution of the LC and TC sensors is 100 μstrain and 500 μstrain, respectively, tested up to 50% strain.

We present the novel kinematics, workspace characterization, functional prototype and impedance control of a six degrees of freedom haptic interface designed to train surgeons for laparoscopic procedures, through virtual reality simulations. The parallel kinematics of the device is constructed by connecting a 3RRP planar parallel mechanism to a linearly actuated modified delta mechanism with a connecting link. The configuration level forward and inverse kinematics of the device assume analytic solutions, while its workspace can be shaped to enable large end-effector translations and rotations, making it well-suited for laparoscopy operations. Furthermore, the haptic interface features a low apparent inertia with high structural stiffness, thanks to its parallel kinematics with grounded actuators. A model-based open-loop impedance controller with feed-forward gravity compensation has been implemented for the device and various virtual tissue/organ stiffness levels have been rendered.