ICRA 2018 论文列表

When working alongside human collaborators in dynamic environments such as a disaster recovery, an unmanned ground vehicle (UGV) may require fast field adaptation to perform its duties or learn novel tasks. In disaster recovery situations, personnel and equipment are constrained, so training must be accomplished with minimal human supervision. In this paper, we introduce a novel framework which uses learned visual perception and inverse optimal control trained with minimal human supervisory examples. This approach is used to learn to mimic navigation behavior and is demonstrated through extensive evaluation in a real-world environment. Finally, we demonstrate the ability to learn an additional behavior with minimal human demonstration in the field.

Advanced motor skills are essential for robots to physically coexist with humans. Much research on robot dynamics and control has achieved success on hyper robot motor capabilities, but mostly through heavily case-specific engineering. Meanwhile, in terms of robot acquiring skills in a ubiquitous manner, robot learning from human demonstration (LfD) has achieved great progress, but still has limitations handling dynamic skills and compound actions. We present a composite learning scheme which goes beyond LfD and integrates robot learning from human definition, demonstration, and evaluation. The method tackles advanced motor skills that require dynamic time-critical maneuver, complex contact control, and handling partly soft partly rigid objects. We also introduce the “nunchaku flipping challenge”, an extreme test that puts hard requirements to all these three aspects. Continued from our previous presentations, this paper introduces the latest update of the composite learning scheme and the physical success of the nunchaku flipping challenge.

In order to conduct many desirable functions, service robots will need to actuate buttons and switches that are designed for humans. This paper presents the design of a robot named SwitchIt that is small, relatively inexpensive, easily mounted on a mobile robot, and actuates buttons reliably. Its operating characteristics were developed after conducting a systematic study of buttons and switches in human environments. From this study, we develop a categorization of buttons based on a set of physical properties relevant for robots to operate them. After a human calibrates and annotates buttons in the robot's environment using a hand-held tablet, the system automatically recognizes, pushes, and detects the state of a variety of buttons. Empirical tests demonstrate that the system succeeds in operating 95.7% of 234 total buttons/switches in an office building and a household environment.

Despite the ubiquity of carpentered items, the customization of carpentered items remains labor intensive. The generation of laymen editable templates for carpentry is difficult. Current design tools rely heavily on CNC fabrication, limiting applicability. We develop a template based system for carpentry and a robotic fabrication system using mobile robots and standard carpentry tools. Our end-to-end design and fabrication tool democratizes design and fabrication of carpentered items. Our method combines expert knowledge for template design, allows laymen users to customize and verify specific designs, and uses robotics system to fabricate parts. We validate our system using multiple designs to make customizable, verifiable templates and fabrication plans and show an end-to-end example that was designed, manufactured, and assembled using our tools.

In many industrial robotics applications, such as spot-welding, spray-painting or drilling, the robot is required to visit successively multiple targets. The robot travel time among the targets is a significant component of the overall execution time. This travel time is in turn greatly affected by the order of visit of the targets, and by the robot configurations used to reach each target. Therefore, it is crucial to optimize these two elements, a problem known in the literature as the Robotic Task Sequencing Problem (RTSP). Our contribution in this paper is two-fold. First, we propose a fast, near-optimal, algorithm to solve RTSP. The key to our approach is to exploit the classical distinction between task space and configuration space, which, surprisingly, has been so far overlooked in the RTSP literature. Second, we provide an open-source implementation of the above algorithm, which has been carefully benchmarked to yield an efficient, ready-to-use, software solution. We discuss the relationship between RTSP and other Traveling Salesman Problem (TSP) variants, such as the Generalized Traveling Salesman Problem (GTSP), and show experimentally that our method finds motion sequences of the same quality but using several orders of magnitude less computation time than existing approaches.

The need for temporal and spatial coordination of two robot arms moving independently in a shared workspace frequently arises in industrial and service-oriented robotics alike. Today, this problem is often solved manually, leading to a negative impact on user experience as well as on execution performance. In this paper, we present an algorithm that is able to automatically coordinate independently planned motions of a dual-arm manipulator during execution. In addition, the algorithm is capable of refining the plan upon receiving new motion commands during the robot motion. We demonstrate the effectiveness and efficiency of the proposed approach on an ABB YuMi robot working on an industrial palletizing task.

Despite several robots having been proposed for vitreoretinal surgery, there is limited information on their dynamic modeling. This gap leads to sub-optimal motor selection and hinders the application of advanced control schemes that would fulfill the goal of micro-precise surgery. This paper presents the design process and a dynamics study of a multi-Degree of Freedom (DoF) robotic system, which is inspired by established co-manipulation architectures. A rigorous kinematics and dynamics analysis of the robot's part that is responsible for manipulating the surgical tool during the retinal surgery phase is provided. In particular, the Euler-Lagrange equations of motion, which describe the dynamics of the 3-link surgical manipulator, are combined with novel analytical models of each link's corresponding transmission mechanism, including an anti-backlash lead screw assembly and a worm drive. The resulting models, transferable to existing manipulators, provide a meticulous analysis of the robot's performance that can be used both for mechanical design and control purposes.

The monocular visual-inertial system (VINS), which consists one camera and one low-cost inertial measurement unit (IMU), is a popular approach to achieve accurate 6-DOF state estimation. However, such locally accurate visual-inertial odometry is prone to drift and cannot provide absolute pose estimation. Leveraging history information to relocalize and correct drift has become a hot topic. In this paper, we propose a monocular visual-inertial SLAM system, which can relocalize camera and get the absolute pose in a previous-built map. Then 4-DOF pose graph optimization is performed to correct drifts and achieve global consistent. The 4-DOF contains x, y, z, and yaw angle, which is the actual drifted direction in the visual-inertial system. Furthermore, the proposed system can reuse a map by saving and loading it in an efficient way. Current map and previous map can be merged together by the global pose graph optimization. We validate the accuracy of our system on public datasets and compare against other state-of-the-art algorithms. We also evaluate the map merging ability of our system in the large-scale outdoor environment. The source code of map reuse is integrated into our public code, VINS-Monol11https://github.com/HKUST-Aerial-Robotics/VINS-Mono.

A crucial challenge in robotics is achieving reliable results in spite of sensing and control uncertainty. In this work, we explore the conformant planning approach to robot manipulation. In particular, we tackle the problem of pushing multiple planar objects simultaneously to achieve a specified arrangement without external sensing. Conformant planning is a belief-state planning problem. A belief state is the set of all possible states of the world, and the goal is to find a sequence of actions that will bring an initial belief state to a goal belief state. To do forward belief-state planning, we created a deterministic belief-state transition model from supervised learning based on off-line physics simulations. We compare our method with an on-line physics-based manipulation approach and show significantly reduced planning times and increased robustness in simulated experiments. Finally, we demonstrate the success of this approach in simulations and physical robot experiments.

Reinforcement learning (RL) allows robots to solve control tasks through interaction with their environment. In this paper we study a model-based value-function RL approach, which is suitable for computationally limited robots and light embedded systems. We develop a diving agent, which uses the RL algorithm for underwater depth stabilization. Simulations and experiments with the micro diving agent demonstrate its ability to learn the depth stabilization task.

In the framework of robotics, Reinforcement Learning (RL) deals with the learning of a task by the robot itself. This paper presents a hierarchical-planning approach in which the robot learns the optimal behavior for different levels in a decoupled way. For high-level discrete actions, Q-learning was chosen, whereas for the low level we utilize Policy Improvement with Path Integrals (PI2) algorithm to learn the parameters of policies, represented by rhythmic Dynamic Movement Primitives (DMPs). The paper studies the case of a 4-finger-gripper manipulator, which performs the task of continuously spinning a ball around a desired axis. The results demonstrate the efficacy of the hierarchical planning and the increased performance of the task when PI2 is used in conjunction with rhythmic DMPs in a real environment.

We present the Regression-based Linear Quadratic Regulator (R-LQR), a new approach for determining locally-optimal control feedback policies for robots with non-linear dynamics and non-quadratic cost functions. Our proposal uses a free-derivative algorithm based on local quadratic regressions to obtain the robot motion policy. In addition, our methodology allows to define a notion of scale that translates into the definition of neighborhoods of valid policy and into the exploration of larger areas of the search space to find the optimal policies. The results show that our formulation allows to reach policies with lower costs than existing algorithms and to avoid problems when the behavior of the cost function makes the optimization difficult.

We present a novel, reflection-aware method for 3D sound localization in indoor environments. Unlike prior approaches, which are mainly based on continuous sound signals from a stationary source, our formulation is designed to localize the position instantaneously from signals within a single frame. We consider direct sound and indirect sound signals that reach the microphones after reflecting off surfaces such as ceilings or walls. We then generate and trace direct and reflected acoustic paths using inverse acoustic ray tracing and utilize these paths with Monte Carlo localization to estimate a 3D sound source position. We have implemented our method on a robot with a cube-shaped microphone array and tested it against different settings with continuous and intermittent sound signals with a stationary or a mobile source. Across different settings, our approach can localize the sound with an average distance error of 0.8 m tested in a room of 7 m by 7 m area with 3 m height, including a mobile and non-line-of-sight sound source. We also reveal that the modeling of indirect rays increases the localization accuracy by 40% compared to only using direct acoustic rays.

A novel 2-RER reconfigurable parallel mechanism (ReConBot) considered as the flexible torso of the mobile manipulator is proposed. This paper deals with the analysis of reconfiguration, kinematics, and motion planning. The ReConBot is composed of straight bar-shape base and moving platforms and two metamorphic kinematic chains (MKC) consisted of a revolute (R) joint, a planar (E) joint, and an R joint in sequence. Firstly, mobility and reconfiguration analysis discuss the conditions and mutual mode transition rules of 12 possible configuration states. And then, the kinematics model covers all states with Cartesian coordinate and axis/angle representations. What's more, the motion planning following the rules of the mode transition is explained and illustrated together with a case study. Furthermore, the method of handling the transition at singularity position is discussed. Finally, the robotic system and its experiments verify the correctness of the theoretical analysis and the validation of reconfiguration rules.

The ability to recognize objects is an essential skill for a robotic system acting in human-populated environments. Despite decades of effort from the robotic and vision research communities, robots are still missing good visual perceptual systems, preventing the use of autonomous agents for realworld applications. The progress is slowed down by the lack of a testbed able to accurately represent the world perceived by the robot in-the-wild. In order to fill this gap, we introduce a large-scale, multi-view object dataset collected with an RGB-D camera mounted on a mobile robot. The dataset embeds the challenges faced by a robot in a real-life application and provides a useful tool for validating object recognition algorithms. Besides describing the characteristics of the dataset, the paper evaluates the performance of a collection of well-established deep convolutional networks on the new dataset and analyzes the transferability of deep representations from Web images to robotic data. Despite the promising results obtained with such representations, the experiments demonstrate that object classification with real-life robotic data is far from being solved. Finally, we provide a comparative study to analyze and highlight the open challenges in robot vision, explaining the discrepancies in the performance.

Rearranging objects on a tabletop surface by means of nonprehensile manipulation is a task which requires skillful interaction with the physical world. Usually, this is achieved by precisely modeling physical properties of the objects, robot, and the environment for explicit planning. In contrast, as explicitly modeling the physical environment is not always feasible and involves various uncertainties, we learn a nonprehensile rearrangement strategy with deep reinforcement learning based on only visual feedback. For this, we model the task with rewards and train a deep Q-network. Our potential field-based heuristic exploration strategy reduces the amount of collisions which lead to suboptimal outcomes and we actively balance the training set to avoid bias towards poor examples. Our training process leads to quicker learning and better performance on the task as compared to uniform exploration and standard experience replay. We demonstrate empirical evidence from simulation that our method leads to a success rate of 85%, show that our system can cope with sudden changes of the environment, and compare our performance with human level performance.

The efficiency and execution time of a road scene evaluation subsystem directly influences a self-driving car's control effectiveness. This article presents a novel approach to fuse data from various sensors (camera, LIDAR, radar, IMU) for pedestrian and vehicle tracking. Our approach, thanks to modern methods of image processing and power of GPU for LIDAR processing, achieves 25Hz update frequency for tracking. The system has been tested on the DiDi-Udacity Self-Driving Challenge datasets.

Accurate and robust object state estimation enables successful object manipulation. Visual sensing is widely used to estimate object poses. However, in a cluttered scene or in a tight workspace, the robot's end-effector often occludes the object from the visual sensor. The robot then loses visual feedback and must fall back on open-loop execution. In this paper, we integrate both tactile and visual input using a framework for solving the SLAM problem, incremental smoothing and mapping (iSAM), to provide a fast and flexible solution. Visual sensing provides global pose information but is noisy in general, whereas contact sensing is local, but its measurements are more accurate relative to the end-effector. By combining them, we aim to exploit their advantages and overcome their limitations. We explore the technique in the context of a pusher-slider system. We adapt iSAM's measurement cost and motion cost to the pushing scenario, and use an instrumented setup to evaluate the estimation quality with different object shapes, on different surface materials, and under different contact modes.

We present a method for planning the motion of arbitrarily-shaped volumetric deformable bodies or robots through complex environments. Such robots have very high-dimensional configuration spaces and we compute trajectories that satisfy the dynamics constraints using a two-stage learning method. First, we train a multitask controller parameterized using dynamic movement primitives (DMP), which encodes various locomotion or movement skills. Next, we train a neural-network controller to select the DMP task to navigate the robot through environments while avoiding obstacles. By combining the finite element method (FEM), model reduction, and contact invariant optimization (CIO), the DMP controller's parameters can be optimized efficiently using a gradient-based method, while the neural-network's parameters are optimized using Deep Q-Learning (DQL). This two-stage learning algorithm also allows us to reuse the trained DMP controller for different navigation tasks, such as moving through different environmental types and to different goal positions. Our results show that the learned motion planner can navigate swimming and walking deformable robots with thousands of DOFs at realtime.