IROS 2015 论文列表

We address the problem of robust instance detection in RGB-D image in the presence of noisy data, cluttering, partial occlusion and large pose variation. We extract contour points from the depth image, construct a Surface Oriented Traverse (SOT) feature for each contour point and further classify it as either belonging or not belonging to the instance of interest. Starting from each contour point, its SOT feature is constructed by traversing and uniformly sampling along an oriented geodesic path on the object surface. After classification, all contour points vote for an instance-specific saliency map, from which the instance of interest is finally localized. Compared with the holistic template-based and learning-based methods, our method inherits advantages of the feature-based methods in dealing with cluttering, partial occlusion, and large pose variation. Furthermore, our method does not require accurate 3D models or high quality laser scan data as input and takes noisy data from commodity 3D sensors. Experimental results on the public RGB-D Object Dataset and our FindMe RGB-D Dataset demonstrate the effectiveness and robustness of our proposed instance detection algorithm.

The aim of this research is to develop a musculoskeletal robot arm, which has a similar mechanical structure to that of a human, to simply make the robot move similarly to a human based on his/her electromyographic signals. In recent years, many musculoskeletal robots have been developed to show advantages of their bio-inspired designs. In this research, we propose a new perspective of their advantages that the control using biological signals can be simplified thanks to the similarity of mechanical structure. To this end, the shoulder complex, which consists of several bones and joints, is focused because it is the most difficult part to develop in humanlike musculoskeletal robot arms. In particular, we develop a linkage mechanism, which can realize similar function to the shoulder complex, to make a surface electromyography(sEMG) based posture control applicable for this complex system. The advantage, that a simple mapping is still available to control the posture to be the same to that of a human, has been successfully shown in an experiment.

We present a light, actuated smart staff with 5DOF tip force sensing which can be used by a humanoid robot operating in challenging terrain. The staff has an extension mechanism that employs mechanical multiplexing to achieve a high extension ratio in a stiff and compact package. The tip force sensor uses two metal diaphragms to achieve decoupling of axial and radial forces and the ability to tune the maximum range of forces in each direction independently. With the force sensor, the robot can characterize the coefficient of friction and orientation of a surface with simple motion primitives. Two sets of experiments were conducted with the smart staff manipulated by a 7 DOF robot arm. Using the sensor, the robot was able to determine the orientations of sloped surfaces within two degrees in two orthogonal directions.

Although superresolution has been studied to huge impact in visual imaging, it is relatively unexplored in tactile robotics. Here we demonstrate a novel optical sensor design (the TacTip) capable of achieving 40-fold localization superresolution to 0.1mm accuracy compared with a 4mm resolution between tactile elements. This superresolution is reached for localizing a 40mm diameter hemicylinder with a tactile finger pad also of 40mm diameter. Deformations of the sensor surface are measured as displacements of molded internal pins, with pin separation thus defining sensor resolution. Active Bayesian perception for classifying object location was used to ensure robust localization and hence the magnitude of the superresolution. These results are comparable with those for capacitive tactile sensors, which we interpret as originating from a convergence in the taxel-based design of the optical sensor and capacitive tactile sensors. The attained superresolution is comparable to the best perceptual hyperacuity in humans.

Pinching is an important capability for mobile robots handling small items or tools. Successful pinching requires force-closure and, in underwater applications, gentle suction flow at the fingertips can dramatically improve the handling of light objects by counteracting the negative effects of water lubrication and enhancing friction. In addition, monitoring the flow gives a measure of suction-engagement and can act as a binary tactile sensor. Although a suction system adds complexity, elastic tubes can double as passive spring elements for desired finger kinematics.

Autonomous robots operating in semi- or unstructured environments, e.g. during search and rescue missions, require methods for online on-board creation of maps to support path planning and obstacle avoidance. Perception based on stereo cameras is well suited for mixed indoor/outdoor environments. The creation of full 3D maps in GPS-denied areas however is still a challenging task for current robot systems, in particular due to depth errors resulting from stereo reconstruction. State-of-the-art 6D SLAM approaches employ graph-based optimization on the relative transformations between keyframes or local submaps. To achieve loop closures, correct data association is crucial, in particular for sensor input received at different points in time. In order to approach this challenge, we propose a novel method for submap matching. It is based on robust keypoints, which we derive from local obstacle classification. By describing geometrical 3D features, we achieve invariance to changing viewpoints and varying light conditions. We performed experiments in indoor, outdoor and mixed environments. In all three scenarios we achieved a final 3D position error of less than 0.23% of the full trajectory. In addition, we compared our approach with a 3D RBPF SLAM from previous work, achieving an improvement of at least 27% in mean 2D localization accuracy in different scenarios.

Snakes and worms are capable of travelling through tightly confined spaces on a resistance medium using rectilinear locomotion. To apply this kind of gait into snake-inspired robots, some fundamental properties remain elusive. In this paper, we present a n-segment kinematic model of the snake robot based on an ideal anchor assumption. Then, the relationships between the motion performance and the gait patterns are identified based on two force constraints hypothesises. To validate theoretical analysis of kinematic model and prediction of average velocity, the experiments are conducted on a novel snake-like robot prototype. Each module of the robot contains unique one-direction wheels, which generate sufficient backward friction force severing as anchoring force.

Generally, wheel mechanisms are inferior to a tracked or walking mechanism in terms of step climbability or traversability in rough terrain; however, they are superior in terms of energy efficiency, structural simplicity, and carrying capacity. This paper proposes a new wheel mechanism, the swing-grouser wheel, which can climb high steps (especially in low friction environments) and has high energy efficiency. In addition, the swing-grouser wheel can climb regardless of the body inclination. Its merits are compared to the results of prior studies. Furthermore, the performance of the swing-grouser wheel was confirmed using a real device experiment and a 2D physics simulation, and improved using a full search of the parameters of the swing-grouser wheel. As a result, one improved parameter resulted in climbing at over 68% of the wheel diameter in a low friction environment; additionally, the energy efficiency was better than that of the previous model.

In this paper, structure-borne vibrations in a one-dimensional structure are examined as a means of communication and sensing for networks of robots. The concept is inspired by the observation that insects may use structural vibrations to communicate and to detect features of their environment. A 12 × 5 × 6 cm mobile robot capable of traversing an acoustically favorable structure is presented. A technique for measuring distance between robots and communicating commands to robots using vibrations generated on a common one-dimensional substrate are presented, and demonstrated through preliminary experimental results. Design considerations for this type of system are also discussed.

Micro- and nanorobots can be built in large numbers, but generating independent control inputs for each robot is prohibitively difficult. Instead, micro- and nanorobots are often controlled by a global field. In previous work we conducted large-scale human-user experiments where humans played games that steered large swarms of simple robots to complete tasks such as manipulating blocks. One surprising result was that humans completed a block-pushing task faster when provided with only the mean and variance of the robot swarm than with full-state feedback. Inspired by human operators, this paper investigates controllers that use only the mean and variance of a robot swarm. We prove that the mean position is controllable, and show how an obstacle can make the variance controllable. We next derive automatic controllers for these and a hybrid, hysteresis-based switching control to regulate the first two moments of the robot distribution. Finally, we employ these controllers as primitives for a block-pushing task.

This paper presents a stochastic approach for automatic collision avoidance for tele-operated unmanned aerial vehicles (UAVs). Collision detection and mitigation in the presence of uncertainty is an important problem to address because on-board sensing and state estimation uncertainties are inherent in real-world systems. A feedforward-based algorithm is described that continually extrapolates the future trajectory of the vehicle given the current operator control input for collision avoidance. If the predicted probability of a collision is greater than a user-defined confidence bound, the algorithm overrides the operator control input with the nearest, safe command signal to steer the robot away from obstacles, while maintaining user intent. The algorithm is implemented on a simulated quadrotor helicopter (quadcopter) with varying amounts of artificial uncertainty. Simulation results show that for a given confidence bound, the aerial robot is able to avoid collisions, even in a situation where the operator is deliberately attempting to crash the vehicle.

Motion planning in stochastic dynamic environments is difficult due to the need for constant plan adjustment caused by the uncertainty of the environment. There are many motion planning problems, including flight coordination and autonomous vehicles, that require an algorithm to predict obstacle motion and plan safely. In this paper, we propose Stochastic Ensemble Simulation (SES)-based planning, a novel framework to efficiently predict and produce safe trajectories in the presence of stochastic obstacles. The stochastic obstacles can be introduced in several ways including stochastic motion or position/speed uncertainty. SES-based planning works by first predicting an obstacle's future position offline through an ensemble of Monte Carlo simulations. These runs simulate the stochastic obstacle dynamics and store the simulation results in temporal snapshots of predicted positions. An online planner then uses this output to identify a predicted collision-free direct path to the goal. If the direct path is not expected to be collision-free, a more expensive tree-based planner is used. Our experiments show SES-based planning outperforms other methods that have high planning success in environments with 900 stochastically moving obstacles. Furthermore, our method plans trajectories with an 80% success rate for a 7 DOF robot in an environment with 250 stochastic moving obstacles and 50 obstacles with speed/position uncertainty. This complex problem is currently beyond the capability of several comparison methods.

This paper presents a new antagonistic tendon-driven joint (TDJ) which has an advantage of enhanced stiffness characteristics. Detailed mathematical analyses on the stiffness of typical TDJs are worked out to compare the stiffness among them. Prototypes of the proposed TDJ design are introduced in the form of packaged modular structure that merges two TDJs generating a combined roll-pitch motion or similar ones. Numerical examples are given to confirm the superior stiffness of the proposed TDJ design.

Microassembly represents a very promising solution to microproducts and complex Micro-Electro-Mechanical Systems (MEMS) fabrication. Since, in the case of teleoperated assembly, an operator is the main source of errors, there is a great interest in microassembly automation. Its main issue consists in precise estimation of object position. Previous studies demonstrate the possibility of application of modelbased visual tracking algorithms from ViSP (Visual Servoing Platform) library. However, the methods of macroassembly cannot be directly applied when working with microobjects. The characterization of single-view visual tracking notably revealed the complexity of depth estimation in microscale, which is due to small depth variation in the seen images compared with the distance from camera. So, an algorithm of Z coordinate reconstruction using a second camera was developed and analyzed for visual servoing task. It was then used to automate microassembly. Experiments demonstrate the possibility of complex microcomponent automatic microassembly with precision better than 10 micrometers.

This paper describes a visual SLAM system based on stereo cameras and focused on real-time localization for mobile robots. To achieve this, it heavily exploits the parallel nature of the SLAM problem, separating the time-constrained pose estimation from less pressing matters such as map building and refinement tasks. On the other hand, the stereo setting allows to reconstruct a metric 3D map for each frame of stereo images, improving the accuracy of the mapping process with respect to monocular SLAM and avoiding the well-known bootstrapping problem. Also, the real scale of the environment is an essential feature for robots which have to interact with their surrounding workspace. A series of experiments, on-line on a robot as well as off-line with public datasets, are performed to validate the accuracy and real-time performance of the developed method.

There are a few types of step-climbing wheelchairs in the world, but most of them are large and heavy because they are power-assisted. Therefore, they require large space to maneuver, which is not always feasible with existing house architectures. This study proposes a novel step-climbing wheelchair based on lever propulsion control using human upper limbs. The developed step-climbing wheelchair device consists of manual wheels with casters for moving around and a rotary-legs mechanism that is capable of climbing steps. The wheelchair also has a passive mechanism for posture transition to shift the center of gravity of the person between the desired positions for planar locomotion and step-climbing. The proposed design consists of passive parts, and this leads the wheelchair being compact and lightweight. In this paper, we present the design of this step-climbing wheelchair and some preliminary experiments to test its usability.

This paper introduces a new state estimator for biped robots fusing encoder, inertial and force torque measurements. The estimator is implemented as a Kalman filter that uses the dynamical model of the linear inverted pendulum with the center of mass (CoM) state as output. In order to compensate for disturbances and model errors we extend the model by a state for the external force and an additional input which is calculated from the dynamics error in pattern generation. Several simulation results underline the effectiveness of the proposed filter and show its robustness against disturbances. Experimental results and an application example validate the method under real world conditions.

This paper introduces a state estimation methodology for a hexapod robot that makes use of proprioceptive sensors and a kinematic model of the robot. The methodology focuses on providing reliable full pose state estimation for a commercially-available hexapod robot platform with the use of only commonly-available sensors. The presented methodology provides the derivation of the kinematic model and implements an Extended Kalman Filter (EKF) state estimation framework similar to that recently validated on a quadruped. The EKF fuses the kinematic model with on-board IMU measurements to estimate the pose of the robot. The methodology was tested with experiments using a physical hexapod robot and validated with independent ground truth measurements.

Omnidirectional wheels used on Omniclimber inspection robot and in other robots enable a holonomic drive and a good maneuverability. On the other hand, they have a poor wheel traction and suffer from vertical and horizontal vibration, decreasing the trajectory following accuracy of the robot. In this study, we address this problem by integrating an orientation estimation and correction algorithm in the Omniclimber control by integration of an accelerometer. Moreover, since the Omniclimber chassis adapts to curved structures, the kinematics of the robot change when moving on a curved surface. We integrated an additional algorithm which corrects the robot's kinematics based on the curvature diameter and the current robot orientation. By integrating these two algorithms we could make remarkable improvements on the path following accuracy of the Omniclimber on flat and curved structures.