ICRA 2017 论文列表

High-performance collision handling, which is divided into the five phases detection, isolation, estimation, classification and reaction, is a fundamental robot capability for safe and sensitive operation/interaction in unknown environments. For complex humanoid robots collision handling is obviously significantly more complex than for classical static manipulators. In particular, the robot stability during the collision reaction phase has to be carefully designed and relies on high fidelity contact information that is generated during the first three phases. In this paper, a unified realtime algorithm is presented for determining unknown contact forces and contact locations for humanoid robots based on proprioceptive sensing only, i.e. joint position, velocity and torque, as well as force/torque sensing along the structure. The proposed scheme is based on nonlinear model-based momentum observers that are able to recover the unknown contact forces and the respective locations. The dynamic loads acting on internal force/torque sensors are also corrected based on a novel nonlinear compensator. The theoretical capabilities of the presented methods are evaluated in simulation with the Atlas robot. In summary, we propose a full solution to the problem of collision detection, collision isolation and collision identification for the general class of humanoid robots.

In sampling-based 3D rigid body motion planning one of the major subroutines is collision detection. Especially for problems with narrow passages many samples have to be checked by a collision detection algorithm. In this application, the runtime of the motion planning algorithm is dominated by collision detection and the samples have the very specific characteristic that many of them are in collision and have small penetration volumes. In our work, we introduce a data structure and an algorithm that makes use of this characteristic by combining well-known data structures like a distance field and an octree with the swap algorithm by Llanas et al. For 3D rigid body motion planning with narrow passages, our approach achieves a speedup of up to 5.0 compared to well-established collision detection libraries like the Proximity Query Package (PQP) and the Flexible Collision Library (FCL).

Collision avoidance, in particular between robots, is an important component for autonomous robots. It is a necessary component in numerous applications such as humanrobot interaction, automotive or unmanned aerial vehicles. While many collision avoidance algorithms take into account actuation constraints, only a few consider sensing limitations. In this paper, we present a reciprocal collision avoidance algorithm based on the velocity obstacle approach that guarantees collision-free maneuvers even when the robots are only capable to sense their environment within a limited Field Of View (FOV). We also present the challenges associated to sensors with limited FOV, show the conditions under which maneuvering can be safely done, and the modifications that a velocity obstacle approach requires to satisfy such conditions. We provide simulations and real robot experiments to validate our approach.

This paper shows a strategy based on passive force control for collaborative object transportation using Micro Aerial Vehicles (MAVs), focusing on the transportation of a bulky object by two hexacopters. The goal is to develop a robust approach which does not rely on: (a) communication links between the MAVs, (b) the knowledge of the payload shape and (c) the position of grasping point. The proposed approach is based on the master-slave paradigm, in which the slave agent guarantees compliance to the external force applied by the master to the payload via an admittance controller. The external force acting on the slave is estimated using a non-linear estimator based on the Unscented Kalman Filter (UKF) from the information provided by a Visual-Inertial (VI) navigation system. Experimental results (online video [1]) demonstrate the performance of the force estimator and show the collaborative transportation of a 1.2 m long object.

For collaborative robots to become useful, end users who are not robotics experts must be able to instruct them to perform a variety of tasks. With this goal in mind, we developed a system for end-user creation of robust task plans with a broad range of capabilities. CoSTAR: the Collaborative System for Task Automation and Recognition is our winning entry in the 2016 KUKA Innovation Award competition at the Hannover Messe trade show, which this year focused on Flexible Manufacturing. CoSTAR is unique in how it creates natural abstractions that use perception to represent the world in a way users can both understand and utilize to author capable and robust task plans. Our Behavior Tree-based task editor integrates high-level information from known object segmentation and pose estimation with spatial reasoning and robot actions to create robust task plans. We describe the cross-platform design and implementation of this system on multiple industrial robots and evaluate its suitability for a wide variety of use cases.

In this paper we introduce Co-Fusion, a dense SLAM system that takes a live stream of RGB-D images as input and segments the scene into different objects (using either motion or semantic cues) while simultaneously tracking and reconstructing their 3D shape in real time. We use a multiple model fitting approach where each object can move independently from the background and still be effectively tracked and its shape fused over time using only the information from pixels associated with that object label. Previous attempts to deal with dynamic scenes have typically considered moving regions as outliers, and consequently do not model their shape or track their motion over time. In contrast, we enable the robot to maintain 3D models for each of the segmented objects and to improve them over time through fusion. As a result, our system can enable a robot to maintain a scene description at the object level which has the potential to allow interactions with its working environment; even in the case of dynamic scenes.

High-level, closed-loop traversal through obstacle-strewn environments remains an open and challenging endeavor for snake-like robotic platforms. Rectilinear forms of locomotion, despite their unique mobility advantages compared to other gait shapes, have seen relatively little progress toward this objective. A dynamical exposition of traveling wave rectilinear motion is reviewed and applied to generate a functional mapping from gait parameter space to corresponding averaged steady-behavior body velocities. We demonstrate the system dynamics, in this case, resemble that of a fixed forward-velocity unicycle where average body curvature presents itself as a versatile control input, modulating angular body velocity in a linear manner. Target body velocities computed to track non-trivial planned paths, then, are mapped to average body curvature commands, enabling autonomous path following, by a physical, multi-link robotic snake, through a series of distinct obstacle arrangements.

Semiconductor manufacturing industry has adopted multi-cluster tools as wafer fabrication equipment that is extraordinarily pricey but highly attractive owing to their higher productivity than single cluster tools can achieve. A challenging issue is how to schedule these tools. It is especially difficult to schedule their frequently occurring close-down processes subject to wafer residency constraints. Such processes appear frequently as caused by wafer lot switches and preventive and emergency maintenances. They are dynamical and non-cyclic. We analyze the synchronization conditions for multiple robots to perform concurrent activities. Upon these conditions, for the situations that an optimal schedule can be found in the steady state, a linear program model is proposed to find a feasible and optimal schedule for close-down processes. An example shows the application and efficiency of our proposed method.

This paper develops a method to use RGB-D cameras to track the motions of a human spinal cord injury patient undergoing spinal stimulation and physical rehabilitation. Because clinicians must remain close to the patient during training sessions, the patient is usually under permanent and transient occlusions due to the training equipment and the movements of the attending clinicians. These occlusions can significantly degrade the accuracy of existing human tracking methods. To improve the data association problem in these circumstances, we present a new global feature based on the geodesic distances of surface mesh points to a set of anchor points. Transient occlusions are handled via a multi-hypothesis tracking framework. To evaluate the method, we simulated different occlusion sizes on a data set captured from a human in varying movement patterns, and compared the proposed feature with other tracking methods. The results show that the proposed method achieves robustness to both surface deformations and transient occlusions.

This paper presents a bionic soft robot with chromatic surface micro-structure (CSM), as a new approach for the measurement of body deformation of the soft robots. Firstly, the CSM films are fabricated by diffraction gratings mold using material polydimethylsiloxane (PDMS). Then, the CSM films are attached to the surface of the soft actuators based on shape memory alloy (SMA) wires. By recording and matching the pattern through two cameras, the continuous deformation curvature degrees of SMA soft actuators are attained. Finally, by integrating three actuators mentioned previously, a bio-inspired prototype of three-legged soft robot is manufactured to carry out motions of both freely and grasp obstacles. By recording and matching the pattern of the robot, we demonstrated a non-contact deformation measurement for flexible devices. As a new non-contact measurement method for body deformation of soft robots, it may shed some light on the issues of the soft robots' sensing and controlling in the future.

Increasing the size of operationnal workspace is one of the main problems parallel robots are faced with. Among all the proposed solutions to that, crossing Type 2 singularities using dedicated trajectory generation and multi-model controller has a great potential. Yet, this approach is not sufficient for the robot to operate autonomously, as assembly mode detection during the motion currently requires additional redundant information. To tackle this problem, we propose an algorithm based on Interval Analysis (IA) that is able to track the end-effector of the robot even under assembly mode change. IA-based solvers for the forward kinematic problem of parallel robots are well known, but they cannot be used under assembly mode change. Compared to those classical approaches, the major modification introduced is the tracking of end-effector velocity in addition to its pose. Using this new information of velocity, the algorithm is capable to monitor the assembly mode change of the robot happening when the singularities are crossed. The behavior and the reliability of this algorithm are analyzed experimentally on a five-bar planar parallel mechanism.

Tethers are used to supply power and transfer data for teleoperated robots. They are known to limit the robot's workspace and could have the effect of hampering its motion. What if we could take advantage of the tether? In this paper a new visual servoing scheme for catenary shaped deformable objects is introduced in order to control the tether parametric shape by properly moving its fixation point. In most of the visual servoing approaches the target object is rigid and distant from the controlled robot. On the contrary, in this paper, the object is deformable and attached to the robot, thus its 3D shape changes while the robot is moving. The experimental system is composed of two terrestrial mobile robots of the same motion capabilities linked with a slack rope. Simulation and real experiments validate the proposed control scheme for proper tether handling.

This paper discusses robust capture of a free-floating target using a robotic arm. The position error resulting from sensor errors and time delay can cause undesired contact and unstable control. In this paper, we propose a caging-based rigid gripper and impedance control, which enables the robot to capture the target robustly without precise motion tracking and large force interaction. The performance of the proposed method is verified experimentally using an air-floating system that emulates planar microgravity motion.

Path planning on a 2D-grid is a well-studied problem in robotics. It usually involves searching for a shortest path between two vertices on a grid. Single-source path planning is a modified problem which asks to find distances from a given point to all other points on the map. A high-performance algorithm for single-source any-angle path planning on a grid that we named CWave is proposed in this work. “Any-angle” attribute of a path planning algorithm implies that such algorithm can find paths which may include any angle segments, as opposed to standard A* on an 8-connected graph, the path can turn with 45°-increments only. The key idea of the presented algorithm is that it does not represent the grid as a graph and uses discrete geometric primitives to define the wave front. In its purest form, CWave requires for computation only integer arithmetics and multiplication by two, but can accumulate the distance error at turning points. A modified version of CWave with minimal usage of floating-point calculations is also developed. It allows to eliminate any accumulative errors which is proven mathematically and experimentally on several maps. The performance of the algorithm on three maps is demonstrated to be significantly faster than that of Theta*, Lazy Theta* and Field A* adapted for single-source planning. The limitations of the current implementations of the algorithm as well as potential improvements are discussed.

In this paper, a lightweight central pattern generator(CPG) model is designed for a snake-like robot, to achieve smooth transition of body shape and locomotion speed. First, based on the convergence behavior of the gradient system, a lightweight CPG model with fast computing time is designed and compared with other widely adopted CPG models. Then, the body shape and locomotion speed transitions in rolling gait are simulated based on the proposed CPG model. Compared with the sinusoid-based method, a smooth transition process can be achieved, without generating undesired movement or abnormal torque. Finally, extensive prototype experiments are conducted to demonstrate that the CPG-based control can effectively ensure smooth transition process and avoid abnormal torque, when the body shape and locomotion speed are changed.

This paper demonstrates the use of a single forward facing camera for obstacle avoidance on a quadrotor. We train a CNN for estimating depth from a single image. The depth map is then fed to a behaviour arbitration based control algorithm that steers the quadrotor away from obstacles. We conduct experiments with simulated and real drones in a variety of environments.

Learning from demonstrations has been shown to be a successful method for non-experts to teach manipulation tasks to robots. These methods typically build generative models from demonstrations and then use regression to reproduce skills. However, this approach has limitations to capture hard geometric constraints imposed by the task. On the other hand, while sampling and optimization-based motion planners exist that reason about geometric constraints, these are typically carefully hand-crafted by an expert. To address this technical gap, we contribute with C-LEARN, a method that learns multi-step manipulation tasks from demonstrations as a sequence of keyframes and a set of geometric constraints. The system builds a knowledge base for reaching and grasping objects, which is then leveraged to learn multi-step tasks from a single demonstration. C-LEARN supports multi-step tasks with multiple end effectors; reasons about SE(3) volumetric and CAD constraints, such as the need for two axes to be parallel; and offers a principled way to transfer skills between robots with different kinematics. We embed the execution of the learned tasks within a shared autonomy framework, and evaluate our approach by analyzing the success rate when performing physical tasks with a dual-arm Optimas robot, comparing the contribution of different constraints models, and demonstrating the ability of C-LEARN to transfer learned tasks by performing them with a legged dual-arm Atlas robot in simulation.

This paper introduces a novel robotic platform for aerial remote gas sensing. Spectroscopic measurement methods for remote sensing of selected gases lend themselves for use on mini-copters, which offer a number of advantages for inspection and surveillance. No direct contact with the target gas is needed and thus the influence of the aerial platform on the measured gas plume can be kept to a minimum. This allows to overcome one of the major issues with gas-sensitive mini-copters. On the other hand, remote gas sensors, most prominently Tunable Diode Laser Absorption Spectroscopy (TDLAS) sensors have been too bulky given the payload and energy restrictions of mini-copters. Here, we introduce and present the Unmanned Aerial Vehicle for Remote Gas Sensing (UAV-REGAS), which combines a novel lightweight TDLAS sensor with a 3-axis aerial stabilization gimbal for aiming on a versatile hexacopter. The proposed system can be deployed in scenarios that cannot be addressed by currently available robots and thus constitutes a significant step forward for the field of Mobile Robot Olfaction (MRO). It enables tomographic reconstruction of gas plumes and a localization of gas sources. We also present first results showing the gas sensing and aiming capabilities under realistic conditions.

This paper examines the capability to incorporate spatial force feedback to the human toe when teleoperating a robotic arm in a force task. Due to the growing complexity of teleoperated systems new means of feedback get increasingly important. To investigate the viability of spatial toe-feedback, experiments with 12 subjects were conducted. The participants had to teleoperate a DLR Light-Weight Robot (LWR) via optical tracking of one finger in order to push a toy train. The orientation of the rail was unknown to the subject and had to be explored using the haptic feedback - a three-dimensional spatial force to the toe, reflecting the contact forces at the robotic end-effector - in absence of visual feedback. The rail was mounted in one of four possible orientations (differences of 45°). The main task of the experiment was to identify the present orientation. In our study subjects could successfully identify the orientation of the rail in more than two thirds of all trials (68%). In almost half of the trials (44%) the subjects were able to move the train along the rails long enough to reach the bumpers at the end and identify them as such. Assuming no feedback would be provided at all, the first metric has a chance level of 25%, and reaching the bumper can be considered impossible. Thus, we can conclude that humans can incorporate spatial force feedback to the toe into their sensorimotor loop.