IROS 2017 论文列表

The Non-Anthropomorphic Biped (NABi) has demonstrated agile motion without complex control due to its unique leg alignment in the sagittal plane. This paper presents NABi-S, a small version of NABi. The “S” designation refers to the new spring or “soleus” mechanism used to store energy and stabilize walking. An important feature of NABi-S is its ability to self-stabilize - it can recover from any tipping in the sagittal plane. Using this feature, NABi-S demonstrates stable motion using CPG control, a difficult task to achieve on a biped. In this paper, the soleus prototype is designed and analyzed. An open-loop, CPG algorithm is then implemented, and a parameter space search is performed to optimize robot locomotion.

Unmanned aerial vehicles (UAVs) can offer timely and cost-effective delivery of high-quality sensing data. However, deciding when and where to take measurements in complex environments remains an open challenge. To address this issue, we introduce a new multiresolution mapping approach for informative path planning in terrain monitoring using UAVs. Our strategy exploits the spatial correlation encoded in a Gaussian Process model as a prior for Bayesian data fusion with probabilistic sensors. This allows us to incorporate altitude-dependent sensor models for aerial imaging and perform constant-time measurement updates. The resulting maps are used to plan information-rich trajectories in continuous 3-D space through a combination of grid search and evolutionary optimization. We evaluate our framework on the application of agricultural biomass monitoring. Extensive simulations show that our planner performs better than existing methods, with mean error reductions of up to 45% compared to traditional “lawnmower” coverage. We demonstrate proof of concept using a multirotor to map color in different environments.

In this paper, we show the achievement of a transformable aerial robot with internal communication system for multiple objects transportation. As it is not easy to make the flight endurance of an aerial robot longer, we study the problem to transport multiple objects at the same time to improve the efficiency of transportation. However, for conventional aerial robots, multiple objects transportation is difficult because the CoG position changes when the number of grasped objects changes, resulting in the instability of the flight. Therefore, to solve this problem, we focus on the multirotor with two-dimensional multilinks proposed in our previous work, which possesses the ability to modify the CoG position actively and can keep the flight stable. First, we introduce the hardware platform including the structure of link module and internal communication system to achieve the extensibility in terms of the link number. We then propose a method to find the optimal form for the multilinks based on the flight stability. Finally, we present experimental results which include aerial transformation and multiple objects transportation.

Visual scene understanding is an important capability that enables robots to purposefully act in their environment. In this paper, we propose a novel deep neural network approach to predict semantic segmentation from RGB-D sequences. The key innovation is to train our network to predict multi-view consistent semantics in a self-supervised way. At test time, its semantics predictions can be fused more consistently in semantic keyframe maps than predictions of a network trained on individual views. We base our network architecture on a recent single-view deep learning approach to RGB and depth fusion for semantic object-class segmentation and enhance it with multi-scale loss minimization. We obtain the camera trajectory using RGB-D SLAM and warp the predictions of RGB-D images into ground-truth annotated frames in order to enforce multi-view consistency during training. At test time, predictions from multiple views are fused into keyframes. We propose and analyze several methods for enforcing multi-view consistency during training and testing. We evaluate the benefit of multi-view consistency training and demonstrate that pooling of deep features and fusion over multiple views outperforms single-view baselines on the NYUDv2 benchmark for semantic segmentation. Our end-to-end trained network achieves state-of-the-art performance on the NYUDv2 dataset in single-view segmentation as well as multi-view semantic fusion.

This paper considers the problem of finding pose matches between trajectories of multiple robots in their respective coordinate frames or equivalent matches between trajectories obtained during different sessions. Pose correspondences between trajectories are mediated by common landmarks represented in a topological map lacking distinct metric coordinates. Despite such lack of explicit metric level associations, we mine preliminary pose level correspondences between trajectories through a novel multi-scale heat-kernel descriptor and correspondence graph framework. These serve as an improved initialization for ICP (Iterative Closest Point) to yield dense pose correspondences. We perform extensive analysis of the proposed method under varying levels of pose and landmark noise and showcase its superiority in obtaining pose matches in comparison with standard ICP like methods. To the best of our knowledge, this is the first work of the kind that brings in elements from spectral graph theory to solve the problem of pose correspondences in a multi-robotic setting and differentiates itself from other works.

Binaural sound localization is known to be improved by incorporating the movement of the sensor. “Active” schemes based on this paradigm can overcome conventional limitations such as front-back ambiguity and source range recovery. Starting from a Gaussian prior on the relative position of a source, this paper determines the motion of a binaural sensor which leads to the most effective path for localization. To this aim, a reward function is defined as the conditional expectation, over the yet unknown N next observations, of the entropy of the N-step-ahead posterior pdf of the relative source position. The optimal motion of the binaural sensor is obtained from a constrained optimization problem involving the automatic differentiation of the reward function. The method is validated in simulation, and is being implemented on a real-life robotic test bed.

This paper presents a generalized multi-stage Bayesian approach for an unmanned aerial vehicle to estimate the location of a mobile target. The major hardware components of the proposed approach are a camera with a fisheye lens and another camera with a normal lens and a pan/tilt unit. With wide angle of view (AOV), the fisheye lens camera first detects the bearing of the target, and the PT camera next captures the target in its AOV. The recursive Bayesian estimation steadily locates the target in a globally defined space. The paper also proposes a multi-stage detection method for the fisheye lens camera. The level of confidence is defined in association with the probability of detection (POD) for each detection technique, and the fisheye lens enables continuous detection by gradually increasing the POD. The observation likelihood is finally derived from the POD in a generalized manner. The proposed approach was applied to the detection of a mobile target by a multi-rotor helicopter, and results have demonstrated the effectiveness of both the proposed multi-stage Bayesian approach and multi-stage fisheye lens detection method.

Multi-robot transfer learning allows a robot to use data generated by a second, similar robot to improve its own behavior. The potential advantages are reducing the time of training and the unavoidable risks that exist during the training phase. Transfer learning algorithms aim to find an optimal transfer map between different robots. In this paper, we investigate, through a theoretical study of single-input single-output (SISO) systems, the properties of such optimal transfer maps. We first show that the optimal transfer learning map is, in general, a dynamic system. The main contribution of the paper is to provide an algorithm for determining the properties of this optimal dynamic map including its order and regressors (i.e., the variables it depends on). The proposed algorithm does not require detailed knowledge of the robots' dynamics, but relies on basic system properties easily obtainable through simple experimental tests. We validate the proposed algorithm experimentally through an example of transfer learning between two different quadrotor platforms. Experimental results show that an optimal dynamic map, with correct properties obtained from our proposed algorithm, achieves 60-70% reduction of transfer learning error compared to the cases when the data is directly transferred or transferred using an optimal static map.

This paper presents a methodology for mapping and localization of Unmanned Aerial Vehicles (UAVs) based on the integration of sensors from different modalities. Particularly, we integrate distance estimations to Ultra-Wideband (UWB) sensors and 3D point-clouds from RGB-D sensors. First, a novel approach for environment mapping is introduced, exploiting the synergies between UWB sensors and point-clouds to produce a multi-modal 3D map that integrates the estimated UWB sensors position. This map is further integrated into a Monte Carlo Localization method to robustly estimate the UAV pose. Finally, the full approach is tested with real indoor flights and validated with a motion tracking system.

We present an interactive and stable multi-contact dynamic simulation algorithm for rigid bodies. Our approach is based on fast frictional dynamics (FFD) [14], which is designed for large sets of non-convex rigid bodies. We use a new friction model that performs velocity-level multi-contact simulation using impulse decomposition. Moreover, we accurately handle friction at each contact point using contact distribution and frictional impulse solvers, which also account for relative motion. We evaluate our algorithm's performance on many complex multi-body benchmarks with thousands of contacts. In practice, our dynamics simulation algorithm takes a few milliseconds per timestep and exhibits more stable behaviors.

Introducing humanoid robots in areas where space is limited, for example in search-and-rescue scenarios or industrial manufacturing, represents a huge challenge, especially when the environment is cluttered and unknown. The robot should be capable of utilizing multiple contact points distributed across the entire body and not just its feet and hands. Extra contacts on the whole body, for instance including the knees and elbows, enable the robot to increase its agility and robustness by enhancing the support polygon. This paper applies our passivity-based approach for hierarchical whole-body control including balancing to scenarios involving contacts distributed all over the body of the robot as required in confined spaces. The approach is experimentally validated on the torque-controlled humanoid robot TORO to demonstrate the general applicability of the presented framework.

We introduce a variant of the multi-vehicle routing problem which accounts for nonholonomic constraints and dense, dynamic obstacles, called MVRP-DDO. The problem is strongly motivated by an industrial mining application. This paper illustrates how MVRP-DDO relates to other extensions of the vehicle routing problem. We provide an application-independent formulation of MVRP-DDO, as well as a concrete instantiation in a surface mining application. We propose a multi-abstraction search approach to compute an executable plan for the drilling operations of several machines in a very constrained environment. The approach is evaluated in terms of makespan and computation time, both of which are hard industrial requirements.

Continuum Reconfigurable Incisionless Surgical Parallel (CRISP) robots consist of multiple needle-diameter flexible instruments that are assembled into a parallel structure inside the human body. With a camera placed at the tip of one of the instruments, the CRISP robot can be used to inspect anatomical sites in constrained body cavities in a minimally invasive manner. We introduce a motion planner for CRISP robots that computes manipulations of the flexible instruments outside the body such that the camera can visually inspect a user-specified site of clinical interest inside the body. Our sampling-based motion planner ensures avoidance of collisions with anatomical obstacles inside the body, enforces remote-center-of-motion constraints on the instrument's entry points into the body, and efficiently handles the expensive computation of CRISP robot kinematics. We also extend the motion planner to estimate the set of points inside a body cavity that can be visually inspected by the camera of a CRISP robot for a given setup. We demonstrate our method in a simulated endoscopic medical procedure in the pleural space around a lung.

The interaction test between a robotic rat and living rat is considered as a possible way to quantitatively characterize the rat sociality. In such robot-rat interactions, the robot should be designed to fully replicate a real rat in terms of morphological and behavioral characteristics. To address this problem, a multi-jointed robot prototype has been modified based on our previous work. We optimally update the forelimb of the robot and redesign the control board to make it more dexterous and increase its behavioral capability. Then, we systematically and kinematically analyze the rotational range of joint variables and the workspace of the robot by using traversal method. To evaluate the motion capability of the modified robot, we propose two quantitative parameters: maximum reachable height (MRH) and minimum bendable distance (MBD). Additionally, we achieve to quantitatively evaluate the behavioral similarity between the robot and rat with the calculated accumulative distance (AD) by using dynamic time warping (DTW). These evaluated methods show high promise to improve the robot-rat interaction to be more similar to rat-rat interaction.

The increasing complexity of soft and hybrid-soft robots highlights the need for more efficient methods of minimizing machine learning solution spaces, and creative ways to ease the process of rapid prototyping. In this paper, we present an initial exploration of this process, using hand-chosen morphologies. Four different choices of muscle groups will be actuated on a tensegrity quadruped called MountainGoat: three for a primarily spine-driven morphology, and one for a primarily leg-driven morphology, and the locomotion speed will be compared. Each iteration of design seeks to reduce the total number of active muscles, and consequently reduce the dimensionality of the problem for machine learning, while still producing effective locomotion. The reduction in active muscles seeks to simplify future rapid prototyping of the robot.

In many robotic applications, softness leads to improved performance, robustness, and safety, while lowering manufacturing cost, increasing versatility, and simplifying control. The advantages of soft robots derive from the fact that their behavior partially results from interactions of the robot's morphology with its environment, which is commonly referred to as morphological computation (MC). But not all MC is good in the sense that it supports the desired behavior. One of the challenges in soft robotics is to build systems that exploit the morphology (good MC) while avoiding body-environment interactions that are harmful with respect to the desired functionality (bad MC). Up to this point, constructing a competent soft robot design requires experience and intuition from the designer. This work is the first to propose a systematic approach that can be used in an automated design process. It is based on calculating a low-dimensional representation of an observed behavior, which can be used to distinguish between good and bad MC. We evaluate our method based on a set of grasping experiments, with variations in hand design, controller, and objects. Finally, we show that the information contained in the low-dimensional representation is comprehensive in the sense that it can be used to guide an automated design process.

Fast and accurate depth estimation is essential for autonomous driving vehicles and autonomous robots. Monocular cameras have been expected to be widely applied due to their lower cost, but since they can't provide depth directly, fast and accurate monocular depth estimation methods are required. This paper proposes a novel method of monocular depth estimation based on flat surface using feature triangulation and two frames per estimation. Experiments under ideal conditions and with the KITTI Vision Benchmark Suite showed that the method achieved smaller errors comparing to existing methods that rely on inertial sensors and many frames.

This paper presents a novel approach to metric scale reconstruction of a three-dimensional (3D) scene using a monocular camera. Using a sequence of images from a monocular camera with a fixed focus lens, metric distance to a set of features in the environment is estimated from image blur due to defocus. The blur texture ambiguity which causes scale errors in depth from defocus is corrected in an EKF framework that exploits image velocity measurements. We show in real experiments that our method converges to a metric scale, accurate, sparse depth map and 3D camera poses with images from a monocular camera. Therefore, the proposed approach has the potential to enhance robot navigation algorithms that rely on monocular cameras.

This paper presents a mono-camera based simultaneous obstacle recognition and distance estimation method for power transmission lines(PTLs) inspection robot to avoid obstacles on or around them. The proposed robot inspects the PTLs while moving along them between power transmission towers. For autonomous navigation, our robot recognizes obstacles and avoids a collision with them. In addition, it can stop at the ends of the PTLs by recognizing its structures, such as insulators, installed at the ends of them. In order to recognize obstacles or insulators efficiently, initially, a robust PTLs detection method based on vanishing point estimation is proposed since they are installed on or around the PTLs. Then, obstacle models with various scales are built, and multiple regions-of-interest(ROIs) according to the scales of the model are constructed along the detected PTLs. Finally, obstacles are recognized within the multiple ROIs. Because each ROI represents a corresponding scale of the obstacle model to be matched, the proposed approach can efficiently deal with scale changes and also estimate distance between the robot and the obstacle simultaneously with obstacle recognition. Experimental results not only show that the proposed method correctly recognizes obstacles but also that the distance between the robot and the obstacles is estimated efficiently.