ICRA 2018 论文列表

In this paper, we present a real-time falling robot stabilization system for a humanoid robot in which the robot can prevent falling using hand contact with walls and other surfaces in the environment. Instead of ignoring or avoiding interaction with environmental obstacles, our system uses obstacle geometry to determine a contact point that reduces impact and necessary friction. It uses a planar dynamic model that is appropriate for falling stabilization in the robot's sagittal plane and frontal plane. The hand contact is determined with an optimal control approach, and to make the algorithm run in realtime, a simplified three-link robot model and a pre-computed database of subproblems for the hand contact optimization are adopted. Moreover, if the robot is not leaning too far after stabilization, we employ a heuristic push-up strategy to recover the robot to a standing posture. System integration is performed on the Darwin-Mini robot and validation is conducted in several environments and falling scenarios.

As human gaze provides information on our cognitive states, actions, and intentions, gaze-based interaction has the potential to enable a fluent and natural human-robot collaboration. In this work, we focus on reliable gaze estimation in remote eye tracking based on calibration-free methods. Although these methods work well in controlled settings, they fail when illumination conditions change or other objects induce noise. We propose a novel, adaptive method based on a probabilistic model, which reliably detects glints from stereo images and evaluate our method using a data set that contains different challenges with regarding to light and reflections.

In this work we develop a novel framework that enables the real-time 3D reconstruction of underwater environments using features from 2D sonar images. Due to noisy and low-resolution imagery as compared with standard cameras, automatic feature extractors for sonar images are not reliable in many scenarios. Thus, a human often needs to hand-select features in sonar imagery for environment reconstructions. Given the high data capture rates of standard imaging sonars (on the order of 20Hz), hand-annotating the features in every frame cannot be done in real-time. To address this we use a Convolutional Neural Network (CNN) that analyzes incoming imagery in real-time and proposes only a small subset of high-quality frames to the user for feature annotation. We demonstrate that our approach provides real-time reconstruction capability without loss in classification performance on datasets captured onboard our underwater vehicle while operating in a variety of environments.

Precision farming robots, which target to reduce the amount of herbicides that need to be brought out in the fields, must have the ability to identify crops and weeds in real time to trigger weeding actions. In this paper, we address the problem of CNN-based semantic segmentation of crop fields separating sugar beet plants, weeds, and background solely based on RGB data. We propose a CNN that exploits existing vegetation indexes and provides a classification in real time. Furthermore, it can be effectively re-trained to so far unseen fields with a comparably small amount of training data. We implemented and thoroughly evaluated our system on a real agricultural robot operating in different fields in Germany and Switzerland. The results show that our system generalizes well, can operate at around 20 Hz, and is suitable for online operation in the fields.

While object tracking for 3D point clouds has been widely researched in recent years, most trackers employ a direct point-to-point matching method under the assumption that target object clouds are dense, although the method is not suitable for sparse point clouds. In this paper, we introduce a novel object-tracking strategy that enables even sparse point clouds to be tracked properly. The strategy involves estimating distributions, called as Estimation of Vertical Distributions (EVD), by the proposed interpolation method to augment data and by a point-to-distribution matching technique. The EVD step generates vertical distributions of unoccupied areas on a target object using the distributions of the occupied areas and then seeks the optimal solution through a coarse-to-fine grid search to guarantee real-time performance. In order to verify the proposed tracking algorithm, we have tested our tracker on real world data collected by our own platform, and the results have demonstrated that the tracker outperforms other trackers.

In this work, we present a successful application of a policy search algorithm to a real-time robotic learning problem, where the goal is to maximize the efficiency of lift generation on a dynamically scaled flapping robotic wing. The robotic wing has two degrees-of-freedom, i.e., stroke and pitch, and operates in a tank filled with mineral oil. For all experiments, the Reynolds number is maintained constant at 1000, where learning is performed for different prescribed stroke amplitudes to find the optimal wing pitching amplitude and the stroke-pitch phase difference that maximize the power loading (PL) of lift generation, a measure of aerodynamic efficiency. For the investigated stroke amplitude range (30°-90°), the efficiency is observed to increase with the stroke amplitude and the lift is mainly generated through the delayed stall, a quasi-steady aerodynamic mechanism. Furthermore, the wing rotation becomes more asymmetric with respect to stroke reversal as the stroke amplitude decreases, indicating an increased use of unsteady lift generation mechanisms at lower stroke amplitudes.

Deep anterior lamellar keratoplasty (DALK) is a promising technique for corneal transplantation that avoids the chronic immunosuppression comorbidities and graft rejection risk associated with penetrating keratoplasty (PKP), the standard procedure. In DALK, surgeons must insert a needle 90% through the 500 μm cornea without penetrating its underlying membrane. This pushes surgeons to their manipulation and visualization limits such that 59% of DALK attempts fail due to corneal perforation or inadequate needle depth. We propose a robot-assisted solution to jointly solve the manipulation and visualization challenges using a cooperatively-controlled, precise robot arm and live optical coherence tomography (OCT) imaging, respectively. Our system features an interface handle, with which the surgeon and robot cooperatively hold the tool, and a posterior corneal boundary virtual fixture driven by real-time OCT segmentation. A study in which three operators performed DALK needle insertions manually and cooperatively in ex vivo human corneas demonstrated an 84% improvement in perforation-free needle depth without an increased perforation rate.

The paper describes an estimation and identification procedure that allows to reconstruct the inertial parameters of a rigid load attached to the end-effector of an industrial manipulator. In particular, the proposed method adopts a multirate quaternion-based Kalman filter, fusing measurements obtained from robot kinematics and inertial sensors at possibly different sampling frequencies, to estimate linear accelerations and angular velocities/accelerations of the load. Then, a recursive total least-squares (RTLS) process is executed to identify the load parameters. Both steps of the estimation and identification procedure are performed in real-time, without the need for offline post-processing of measured data.

This paper presents a reactive controller for planar manipulation tasks that leverages machine learning to achieve real-time performance. The approach is based on a Model Predictive Control (MPC) formulation, where the goal is to find an optimal sequence of robot motions to achieve a desired object motion. Due to the multiple contact modes associated with frictional interactions, the resulting optimization program suffers from combinatorial complexity when tasked with determining the optimal sequence of modes. To overcome this difficulty, we formulate the search for the optimal mode sequences offline, separately from the search for optimal control inputs online. Using tools from machine learning, this leads to a convex hybrid MPC program that can be solved in real-time. We validate our algorithm on a planar manipulation experimental setup where results show that the convex hybrid MPC formulation with learned modes achieves good closed-loop performance on a trajectory tracking problem.

In this paper, we present a reactive robot navigation method for a non-holonomic mobile robot taking inspiration from the phenomena observed in magnetic fields. The algorithm is shown to be able to guide mobile robots in arbitrary-shaped convex environment without being trapped in local minima by exploiting the local sensory information without priori knowledge about the environment. A preliminary validation study involving simulation of and experiments with a TurtleBot mobile robot platform show the advantage of the proposed method over existing ones.

Reactivity to unexpected situations is one of the most important characteristics of walking for real world applications. In this study, we introduce a reactive biped robot walking method that reflects only the current state of the robot. Therefore, time plan and trajectory tracking control are not required for robot walking, and this enables reactive behavior to unexpected contact or disturbance. The walking algorithm is realized through a whole-body control algorithm based on the operational space control framework, that possesses the capability to command the required force for tasks and also implement compliant task behavior by adjusting corresponding task gains. The performance of the proposed method is verified by experiments with a 12-DoF torque controlled biped robot. Robust walking is demonstrated when the foot is stopped by an unexpected obstacle or when the lateral motion is unexpectedly blocked and released by a human.

This paper focuses on the problem of deploying a set of autonomous robots to efficiently service tasks that arrive sequentially in an environment over time. Each task is serviced when the robot visits the corresponding task location. Robots can then redeploy while waiting for the next task to arrive. The objective is to redeploy the robots taking into account the next N task arrivals. We seek to minimize a linear combination of the expected cost to service tasks and the redeployment cost between task arrivals. In the single robot case, we propose a one-stage greedy algorithm and prove its optimality. For multiple robots, the problem is NP-hard, and we propose two constant-factor approximation algorithm, one for the problem with a horizon of two task arrivals and the other for the infinite horizon when redeployment cost is weighted more heavily than service cost. Finally, we present extensive benchmarking results to characterize both solution quality and runtime.

Classical RRT algorithm is blind to efficiently explore configuration space for expanding the tree through a narrow passage when solving a motion planning (MP) problem. Although there have been several attempts to deal with narrow passages which are based on a wide spectrum of assumptions and configuration setups, we solve this problem in rather general way. We use dominant eigenvectors of the configuration sets formed by properly sampling the space around the nearest node, to efficiently expand the tree around the obstacles and through narrow passages. Unlike classical RRT, our algorithm is aware of having the tree nodes in front of a narrow passage and in a narrow passage, which enables a proper tree expansion in a vine-like manner. A thorough comparison with RRT, RRT-connect, and DDRRT algorithm is provided by solving three different difficult MP problems. The results suggest a significant superiority the proposed Rapidly-exploring Random Vines (RRV) algorithm might have in configuration spaces with narrow passages.

Kinodynamic RRT planners are considered to be general tools for effectively finding feasible trajectories for high-dimensional dynamical systems. However, they struggle when holonomic constraints are present in the system, such as those arising in parallel manipulators, in robots that cooperate to fulfill a given task, or in situations involving contacts with the environment. In such cases, the state space becomes an implicitly-defined manifold, which makes the diffusion heuristic inefficient and leads to inaccurate dynamical simulations. To address these issues, this paper presents an extension of the kinodynamic RRT planner that constructs an atlas of the state-space manifold incrementally, and uses this atlas both to generate random states and to dynamically steer the system towards such states. To the best of our knowledge, this is the first randomized kinodynamic planner that explicitly takes holonomic constraints into account. We validate the approach in significantly-complex systems.

This paper details the system and methods developed to enable autonomous nuclear radiation source localization and mapping using aerial robots in GPS-denied environments. A Thallium-doped Cesium Iodide (CsI(Tl)) scintillator and a Silicon Photomultiplier are combined with custom-built electronics for counting and spectroscopy, and the provided radiation measurements are pose-annotated using visual-inertial localization enabling autonomous operation in GPS-denied environments. Provided this capability, a strategy for radioactive source localization, as well as active source search path planning was developed. The proposed method is motivated and accounts for the limited endurance of the vehicle, which entails a very small amount of dwell points, and the fact that GPS-denied localization implies varying uncertainty of the robot's position estimate. The complete system is evaluated in multiple experimental studies using a small aerial robot and a Cesium-137 radiation source. As shown, accurate radioactive source localization is achieved, enabling efficient radiation mapping of indoor GPS-denied environments.

Research on multi-robot systems has demonstrated promising results in manifold applications and domains. Still, efficiently learning an effective robot behaviors is very difficult, due to unstructured scenarios, high uncertainties, and large state dimensionality (e.g, hyper-redundant and groups of robot). To alleviate this problem, we present Q-CP a cooperative model-based reinforcement learning algorithm, which exploits action values to both (1) guide the exploration of the state space and (2) generate effective policies. Specifically, we exploit Q-learning to attack the curse-of-dimensionality in the iterations of a Monte-Carlo Tree Search. We implement and evaluate Q-CP on different stochastic cooperative (general-sum) games: (1) a simple cooperative navigation problem among 3 robots, (2) a cooperation scenario between a pair of KUKA YouBots performing hand-overs, and (3) a coordination task between two mobile robots entering a door. The obtained results show the effectiveness of Q- CP in the chosen applications, where action values drive the exploration and reduce the computational demand of the planning process while achieving good performance.

For robots to have a wide range of applications, they must be able to execute numerous tasks. However, recent studies into robot manipulation using deep neural networks (DNN) have primarily focused on single tasks. Therefore, we investigate a robot manipulation model that uses DNNs and can execute long sequential dynamic tasks by performing multiple short sequential tasks at appropriate times. To generate compound tasks, we propose a model comprising two DNNs: a convolutional autoencoder that extracts image features and a multiple timescale recurrent neural network (MTRNN) to generate motion. The internal state of the MTRNN is constrained to have similar values at the initial and final motion steps; thus, motions can be differentiated based on the initial image input. As an example compound task, we demonstrate that the robot can generate a “Put-In-Box” task that is divided into three subtasks: open the box, grasp the object and put it into the box, and close the box. The subtasks were trained as discrete tasks, and the connections between each subtask were not trained. With the proposed model, the robot could perform the Put-In-Box task by switching among subtasks and could skip or repeat subtasks depending on the situation.

Localizing radio frequency (RF) sources is an important application for unmanned aerial vehicles (UAVs), Localization is often carried out by estimating bearing to an RF source, which can be achieved by rotating a directional antenna in place. Multirotor UAVs are well-suited for this sensing modality because they can efficiently rotate in place. However, a full rotation from a single location is needed to account for scale factors affecting the directional antenna's measurements. Although easy to perform, these rotations tend to be slow and delay localization. In this paper, we equip a multirotor UAV with a directional antenna and an omnidirectional antenna. The omnidirectional antenna serves to normalize measurements made by the directional antenna, yielding “pseudo-bearing” measurements. These bearing-like measurements are less informative than bearing measurements but do not require a full rotation, leading to more measurements and faster localization. We validate the normalization with antenna theory and ground tests. Claims of improved localization are validated with simulations and flight tests on a multirotor UAV. Our setup significantly reduces localization time compared to a multirotor UAV equipped with only a directional antenna.

Inspired by the ability of animals to rely on proprioception and vestibular feedback to adapt their gait, we propose a modular framework for autonomous locomotion that relies on force sensing and inertial information. A first controller exploits anti-compliance, a new application of positive force feedback, to quickly react against obstacles upon impact. We hypothesize that, in situations where a robot experiences occasional impacts with the environment, anti-compliance can help negotiate unknown obstacles, similar to biological systems where positive feedback enables fast responses to external stimuli. A novel parallel controller, based on a bi-stable dynamical system, continuously adjusts the robot's direction of locomotion, and reverts it in reaction to major swerves. We present experimental results, demonstrating how our framework allows a snake robot to autonomously locomote through a row of unevenly-spaced obstacles. Finally, we extend our proprioceptive controller to legged locomotion, showing how a hexaprint robot can adapt its motion to climb over obstacles.