ICRA 2017 论文列表

Many robotics and Augmented Reality (AR) systems that use sparse keypoint-based visual maps operate in large and highly repetitive environments, where pose tracking and localization are challenging tasks. Additionally, these systems usually face further challenges, such as limited computational power, or insufficient memory for storing large maps of the entire environment. Thus, developing compact map representations and improving retrieval is of considerable interest for enabling large-scale visual place recognition and loop-closure. In this paper, we propose a novel approach to compress descriptors while increasing their discriminability and match-ability, based on recent advances in neural networks. At the same time, we target resource-constrained robotics applications in our design choices. The main contributions of this work are twofold. First, we propose a linear projection from descriptor space to a lower-dimensional Euclidean space, based on a novel supervised learning strategy employing a triplet loss. Second, we show the importance of including contextual appearance information to the visual feature in order to improve matching under strong viewpoint, illumination and scene changes. Through detailed experiments on three challenging datasets, we demonstrate significant gains in performance over state-of-the-art methods.

Various leg exoskeletons have been designed for gait rehabilitation. The transparency of these exoskeletons is crucial to their effectiveness in gait training. The weight and inertia of an exoskeleton are two important factors affecting its transparency. In this study, using a light-weight leg exoskeleton C-ALEX, we conducted a series of experiments to explore the effect of exoskeleton weight and inertia on the natural walking of twelve healthy subjects. They walked in C-ALEX under three levels of mass added to the leg: (i) no added mass, (ii) 1.8 kg, and (iii) 3.6 kg, and three different setups of C-ALEX: (i) freewalking without C-ALEX, (ii) with C-ALEX, and (iii) with C-ALEX compensating for the weight of the added mass. The result shows that increasing exoskeleton mass increases step length, decreases step height, and reduces maximum knee flexion. After weight compensation, the step height, and the maximum knee flexion partially restored, but the step length did not, implying that the inertia is responsible for the change in step length. The study demonstrates that compensating for weight alone cannot eliminate the changes due to exoskeleton mass. On the other hand, reducing the overall mass of the exoskeleton can better preserve the natural gait of the subjects.

Rigid haptic devices enable humans to physically interact with virtual environments, and the range of impedances that can be safely rendered using these rigid devices is quantified by the Z-Width metric. Series elastic actuators (SEAs) similarly modulate the impedance felt by the human operator when interacting with a robotic device, and, in particular, the robot's perceived stiffness can be controlled by changing the elastic element's equilibrium position. In this paper, we explore the K-Width of SEAs, while specifically focusing on how discretization inherent in the computer-control architecture affects the system's passivity. We first propose a hybrid model for a single degree-of-freedom (DoF) SEA based on prior hybrid models for rigid haptic systems. Next, we derive a closed-form bound on the K-Width of SEAs that is a generalization of known constraints for both rigid haptic systems and continuous time SEA models. This bound is first derived under a continuous time approximation, and is then numerically supported with discrete time analysis. Finally, experimental results validate our finding that large pure masses are the most destabilizing operator in human-SEA interactions, and demonstrate the accuracy of our theoretical K-Width bound.

A Meal Assistance Robot is an assistive device that is used to aid individuals who cannot independently direct food to their mouths for consuming. For individuals who undergo loss of upper limb functions due to amputations, spinal cord injuries or cerebral palsy, self-feeding can be impossible, and to assist such individuals in regaining their independence meal assistance robots have been introduced. In this paper we propose a meal assistance robot that is controlled using user intentions based on Electroencephalography (EEG) signals while incorporating camera-based automatic mouth position tracking and mouth open detection systems. In the proposed system, users select any solid food item that they desire to consume from three different containers by looking at corresponding flickering LED matrices. User intentions are identified through EEG signals using a Steady State Visual Evoked Potentials (SSVEP) based intention detection method. Initial motion commands for scooping food from the containers are generated and sent to the meal assistance robot from this first stage. At the second stage, a camera-based mouth position tracking method is proposed for automatically detecting the user's mouth position and thereby moving the spoon or end-effector of the meal assistance robot towards the mouth of the user. This method is capable of automatically tracking the mouth position of users irrespective of their individual body differences and seating positions. A mouth open/closed recognition method is implemented at the final stage in order to feed food to the users when they desire consumption, indicated by the opening/closing of their mouth. A set of experiments were carried out with healthy subjects to validate the proposed system and results are here presented.

Identifying collision-free paths over long time windows in environments with stochastically moving obstacles is difficult, in part because long-term predictions of obstacle positions typically have low fidelity, and the region of possible obstacle occupancy is typically large. As a result, planning methods that are restricted to identifying paths with a low probability of collision may not be able to find a valid path. However, allowing paths with a higher probability of collision may limit detection of imminent collisions. In this paper, we present Dynamic Risk Tolerance (DRT), a framework that dynamically evaluates risk tolerance, a function which is formulated as a time-varying upper bound on the acceptable likelihood of collision for a given path. DRT is implemented with forward stochastic reachable sets to predict the exact distribution of obstacles in a scalable manner over an arbitrarily long time window. In effect, DRT identifies actions that balance risks posed by both near and far obstacles. We empirically compare DRT to other state of the art methods that are capable of generating real-time solutions in highly crowded environments, and demonstrate the success rates for DRT that is 46% higher than the best performing comparison method, in the most difficult problem tested.

This paper presents a new method for planning and controlling dynamic multi-contact motions for humanoid robots. Our motion planner takes a sequence of multi-contact stances and generates closed-form reference trajectories for the robot center of mass (CoM) position, velocity, and acceleration, based on the concept of Divergent Component of Motion (DCM). The timing of the contact transitions and the end-effector trajectories are automatically computed such that the motion is feasible with respect to kinematic and dynamic constraints. We verify the constraints using a simplified model of the robot to achieve a very fast planner that finds a feasible solution within a few seconds. The reference trajectories serve as inputs to a passivity-based whole-body controller which includes a DCM controller for tracking the CoM trajectory. We demonstrate the robustness of our approach in simulation and experiments with the humanoid robot TORO.

This paper introduces dynamic manipulability of the center of mass (CoM) as a metric to measure robots' physical abilities to accelerate their CoMs in different directions. By decomposing the effects of velocity dependent constraints, such as unilateral contacts and friction cones, CoM dynamic manipulability is defined as a velocity independent metric which depends only on robot's configuration and inertial parameters. Thus, this metric is independent of any choice of controller and expresses only physical abilities of robots. This important property makes the proposed metric a proper tool to study, analyse and design current and future robots. The outcome of the CoM dynamic manipulability analysis in this paper is an ellipsoid in the CoM acceleration space which graphically shows accessible points due to the unit weighted norm of joint torques. Physical meanings and concepts of two reasonable choices for the weighting matrix, which is used in the weighted norm of joint torques, are discussed and illustrative examples are presented. Since the proposed metric measures physical ability to accelerate the CoM, it is claimed to be a suitable tool to study balance ability of legged robots.

We present a control methodology for underactuated aerial manipulators that is both easy to implement on real systems and able to achieve highly dynamic behaviors. The method is composed by two parts: i) a nominal input/state trajectory generator that takes into account the full-body dynamics of the system exploiting its differential flatness property; ii) a decentralized feedback controller acting on the actuated degrees of freedom that confers the needed robustness to the closed-loop system. We demonstrate that the proposed controller is able to precisely track dynamic trajectories when implemented on a standard hardware. Comparative experiments clearly show the benefit of using the nominal input/state generator.

The coverage control problem for multiple mobile robot system with limited and dissimilar sensing abilities is addressed in this article. There exist missing areas while taking the previously developed r-limited Voronoi partitions due to distinct sensing ranges. Thus, a modified partitioning approach is proposed to enhance coverage performance. Furthermore, the dynamic coverage controllers are developed with time-varying density function so that the system is more applicable to various tasks such as target tracking or information decays. In addition to theoretical analysis, numerical examples and experiments with four mobile robots are conducted to verify stability and coverage performance of mobile robot network utilizing the proposed coverage partition algorithm.

Transport of objects is a major application in robotics nowadays. While ground robots can carry heavy payloads for long distances, they are limited in rugged terrains. Aerial robots can deliver objects in arbitrary terrains; however they tend to be limited in payload. It has been previously shown that, for heavy payloads, it can be beneficial to carry them using multiple flying robots. In this paper, we propose a novel collaborative transport scheme, in which two quadrotors transport a cable-suspended payload at accelerations that exceed the capabilities of previous collaborative approaches, which make quasi-static assumptions. Furthermore, this is achieved completely without explicit communication between the collaborating robots, making our system robust to communication failures and making consensus on a common reference frame unnecessary. Instead, they only rely on visual and inertial cues obtained from on-board sensors. We implement and validate the proposed method on a real system.

Duckietown is an open, inexpensive and flexible platform for autonomy education and research. The platform comprises small autonomous vehicles (“Duckiebots”) built from off-the-shelf components, and cities (“Duckietowns”) complete with roads, signage, traffic lights, obstacles, and citizens (duckies) in need of transportation. The Duckietown platform offers a wide range of functionalities at a low cost. Duckiebots sense the world with only one monocular camera and perform all processing onboard with a Raspberry Pi 2, yet are able to: follow lanes while avoiding obstacles, pedestrians (duckies) and other Duckiebots, localize within a global map, navigate a city, and coordinate with other Duckiebots to avoid collisions. Duckietown is a useful tool since educators and researchers can save money and time by not having to develop all of the necessary supporting infrastructure and capabilities. All materials are available as open source, and the hope is that others in the community will adopt the platform for education and research.

Deep learning has rapidly transformed the state of the art algorithms used to address a variety of problems in computer vision and robotics. These breakthroughs have relied upon massive amounts of human annotated training data. This time consuming process has begun impeding the progress of these deep learning efforts. This paper describes a method to incorporate photo-realistic computer images from a simulation engine to rapidly generate annotated data that can be used for the training of machine learning algorithms. We demonstrate that a state of the art architecture, which is trained only using these synthetic annotations, performs better than the identical architecture trained on human annotated real-world data, when tested on the KITTI data set for vehicle detection. By training machine learning algorithms on a rich virtual world, real objects in real scenes can be learned and classified using synthetic data. This approach offers the possibility of accelerating deep learning's application to sensor-based classification problems like those that appear in self-driving cars. The source code and data to train and validate the networks described in this paper are made available for researchers.

We present a solution for online simultaneous localization and mapping (SLAM) self-calibration in the presence of drift in calibration parameters in order to support accurate long-term operation. Calibration parameters such as the camera focal length or camera-to-IMU extrinsics are frequently subject to drift over long periods of operation, inducing cumulative error in the reconstruction. The key contributions are modeling calibration parameters as a spatiotemporal quantity: sensor-to-sensor spatial calibration and sensor intrinsic parameters are continuously time-varying, with statistical tests for change detection and regression. An analysis of the long term effects of inappropriately modeling time-varying sensor calibration is also provided. Constant-time operation is achieved by selecting only a fixed number of informative segments of the trajectory for calibration parameter estimation, giving the added benefit of avoiding early linearization errors by not rolling past measurements into a prior distribution. Our approach is validated with simulated and real-world data.

In this work, multiple robots circulate around the boundary of a desired region in order to create a virtual fence. The aim of the this fence is to avoid internal or external agents crossing through the delimited area. In this paper, we propose a distributed technique that allows a team of robots to plan the deformation of the boundary shape in order to escort the safe region from one place to a goal. Our proposal is composed of two parts. First, we present a distributed planning method for the dynamic boundary. We model the resulting plan as a twice differentiable function. Second, we use the obtained function to guide the robot team, where every member uses only local information for the controller. The robots distribute themselves along the time-varying perimeter and patrol around it. We show in simulation how the robots behave in partially/totally unknown environments with static obstacles.

In this paper, we study the fixed-time cooperative tracking control problem for multi-robot systems with doubleintegrator dynamics. First, a novel distributed observer is proposed for each follower to estimate the leader state in a fixed time, then a local tracking controller based on sliding mode technique is proposed such that the estimated leader state is tracked in a fixed time. Both cases of a stationary leader and a dynamic leader are investigated. Since nonholonomic dynamics can better describe the mobile robots in reality, we further extend the results to achieve fixed-time cooperative tracking for multi-robot systems with nonholonomic dynamics. Different from the conventional finite-time cooperative tracking strategies, the fixed-time approach in this work guarantees that an upper bound of settling time can be prescribed without dependence on initial states of robots, which provides additional system information in advance. Finally, numerical simulations are given to demonstrate the effectiveness of the theoretical results.

We consider a team of multiple dynamical and heterogeneous robots which are deployed for gathering different types of data within a common workspace. The robots have different roles due to different capabilities: some gather data from the workspace (Type-A robots) and others receive data from Type-A robots and upload them to a data center (Type-B robots). The data-gathering tasks are specified locally to each Type-A robot as high-level Linear Temporal Logic (LTL) formulas. All robots have a limited buffer to store the data. Thus the data gathered by Type-A robots should be transferred to Type-B robots before the buffers overflow, respecting at the same time limited communication range for all robots. The main contribution of this work is a distributed task coordination and intermittent meeting scheme that guarantees the satisfaction of all local tasks while obeying the above constraints. We present numerical simulations to demonstrate the advantages of the proposed method over most existing approaches that require all-time network connectivity.

The paper presents a two stages distributed algorithm for cooperative manipulating an unknown object rigidly grasped by mobile manipulators, in the absence of both a central unit and any explicit information exchange among robots. In the first stage, robots cooperatively estimate the object kinematic and dynamic parameters by properly moving the object or applying specific contact wrenches. In the second stage, the estimated parameters are used in a distributed cooperative algorithm aimed at controlling the object pose while limiting both the squeezing wrenches exerted by the manipulators and the wrench exerted by the environment on the object. Numerical simulations demonstrate the feasibility of the approach.

Data association is one of the fundamental problems in multi-sensor systems. Most current techniques rely on pairwise data associations which can be spurious even after the employment of outlier rejection schemes. Considering multiple pairwise associations at once significantly increases accuracy and leads to consistency. In this work, we propose a fully decentralized method for globally consistent data association from pairwise data associations based on a distributed averaging scheme on the set of doubly stochastic matrices. We demonstrate the effectiveness of the proposed method using theoretical analysis and experimental evaluation.

We discuss selected mechanical aspects of self-reconfiguration of densely-packed modular robots. The change of connection topology and transport of modules are fundamental mechanisms for these systems, which determine their desired emergent behavior, e.g., movement, shape change or interaction with their surroundings. At the same time, reconfiguration affects the forces between modules. We present a distributed procedure by which a robot can predict if the next planned reconfiguration step will overstress intermodular connections. We use a Finite Element model of a modular robot, with one-node-per-module discretization and beam elements representing intermodular connections. The analysis is restricted to static loads and linear elasticity. We present a distributed procedure of aggregation of the stiffness matrix and iterative solution of the resulting equations of elasticity. The procedure is illustrated with numerical examples and analyzed in terms of its efficiency.