IROS 2017 论文列表

In this paper, we propose a trajectory generation framework for quadrotor autonomous navigation in unknown 3-D complex environments using gradient information. We decouple the trajectory generation problem as front-end path searching and back-end trajectory refinement. Based on the map that is incrementally built onboard, we adopt a sampling-based informed path searching method to find a safe path passing through obstacles. We convert the path consists of line segments to an initial safe trajectory. An optimization-based method which minimizes the penalty of collision cost, smoothness and dynamical feasibility is used to refine the trajectory. Our method shows the ability to online generate smooth and dynamical feasible trajectories with safety guarantee. We integrate the state estimation, dense mapping and motion planning module into a customized light-weight quadrotor platform. We validate our proposed method by presenting fully autonomous navigation in unknown cluttered indoor and outdoor environments.

This paper discusses the possibilities of extending and adapting the CHOMP motion planner [1] to work with a non-holonomic vehicle such as an autonomous truck with a single trailer. A detailed study has been done to find out the different ways of implementing these constraints on the motion planner. CHOMP, which is a successful motion planner for articulated robots produces very fast and collision-free trajectories. This nature is important for a local path adaptor in a multi-vehicle path planning for resolving path-conflicts in a very fast manner and hence, CHOMP was adapted. Secondly, this paper also details the experimental integration of the modified CHOMP with the sensor fusion and control system of an autonomous Volvo FH-16 truck and a set of experiments conducted on a real-time environment. Finally, additional simulations were also conducted to compare the performance of the different approaches developed to study the feasibility of employing CHOMP to autonomous vehicles.

Defining a state representation on which optimal control can perform well is a tedious but crucial process. It typically requires expert knowledge, does not generalize straightforwardly over different tasks and strongly influences the quality of the learned controller. In this paper, we present an autonomous feature construction method for learning low-dimensional manifolds of goal-relevant features jointly with an optimal controller using reinforcement learning. Our method combines information-theoretic algorithms with principal component analysis to performs a return-weighted reduction of the state representation. The method does not require any preprocessing of the data, does not assume strong restrictions on the state representation, and substantially improves the performance of learning by reducing the number of samples required. We show that our method can learn high quality controller in redundant spaces, even from pixels, and outperforms both classical and state-of-the-art deep learning approaches.

In this paper, we present a multi-sensory terrain classification algorithm with a generalized terrain representation using semantic and geometric features. We compute geometric features from lidar point clouds and extract pixel-wise semantic labels from a fully convolutional network that is trained using a dataset with a strong focus on urban navigation. We use data augmentation to overcome the biases of the original dataset and apply transfer learning to adapt the model to new semantic labels in off-road environments. Finally, we fuse the visual and geometric features using a random forest to classify the terrain traversability into three classes: safe, risky and obstacle. We implement the algorithm on our four-wheeled robot and test it in novel environments including both urban and off-road scenes which are distinct from the training environments and under summer and winter conditions. We provide experimental result to show that our algorithm can perform accurate and fast prediction of terrain traversability in a mixture of environments with a small set of training data.

Depth estimation from a single image is very challenging due to the inherent ambiguity of mapping a color image to a depth map. Previous work tackles this problem by exploiting various levels of features with multi-scale deep convolutional neural networks. However, most of the local geometric structure related monocular depth cues are lost when being propagated through convolutional neural network. Moreover, the error of depth cues related to local geometric structures is not considered in the loss function. In this work, we propose the GeoCueDepth convolutional neural network to exploit local geometric structure cues and propose a training loss that takes the geometric error into consideration, which significantly improve the performance of depth prediction in both accuracy and sharpness. Experiments show that the proposed method achieves 0.122 average relative error and 0.078 square relative error on the NYU Depth v2 data set, which outperforms state-of-the-art monocular depth estimation approaches.

Series-elastic and viscoelastic robots can provide performance gains in applications with high dynamics. Harnessing these, requires an understanding of the dynamics of the system, which can be gained using optimization-based methods. The result are motions which make optimal use of the intrinsic behavior, possibly exceeding the performance of an equivalent rigid-body robot. We present a collocation framework which enables both automatic computation of contact-switching patterns and allows the full utilization of the dynamics of the compliant system. The formulation also addresses the problem of redundant torque generation in viscoelastic actuators. The effectiveness of this method was demonstrated in simulations as well as experiments with a compliant bipedal robot. The approach is capable of providing gait primitives, longer gait sequences containing multiple steps as well as generating extremely dynamic motions, e.g. somersaults.

Modular robots can be rapidly reconfigured into customized articulated legged morphologies capable of mobile manipulation and inspection. However, current gait generation methods do not keep pace with the speed of physical reconfiguration. This work focuses on quickly creating gaits for modular legged robots. We build on a recent method that uses trajectory optimization to design quasi-static gaits given only robot geometry and foot contact patterns. We develop methods to automatically generate contact patterns for new gaits and transitions between them. We show the utility of these methods applied to robots with many limbs, such that limbs can be fluidly reassigned to locomotion, manipulation, or inspection tasks, or to adapt gaits to hardware failures online. We demonstrate gait and transition generation with our modular hexapod and dodecapod robots. The robots switch between gaits that use all limbs for locomotion and those that leave some limbs free to pick up objects or position a camera.

The aim of this paper is to establish a generalized parameter estimation scheme to online estimate the Center of Gravity (COG) for multirotors, while using a geometric controller to perform position tracking for applications in aerial manipulation. The proposed scheme is developed so the controller uses the estimated COG to compensate and remove constant offset in the position tracking. The efficiency and validity of the proposed parameter estimation and compensation scheme is proved through two experimental evaluations, one when step changes to the COG are applied and one tracking experiment where a compact aerial manipulator is attached to the multirotor and performs sweeping motions.

Inertial measurement units (IMUs) enable human motion measurement in any environment, which can be useful for human robot interaction, exoskeletons, and active prosthetics. This paper proposes an approach for estimating the orientation between a wearable IMU sensor and the body frame of the wearer using a simple and fast calibration procedure. The proposed approach uses the generalized Hebbian algorithm to incrementally estimate the axis aligned with gravity using acceleration measurements obtained during a static pose, and the axis perpendicular to the saggital plane using gyro measurements obtained during sagittal plane movements. An automated convergence criterion based on the sensor measurement variance is used. The proposed approach is tested in simulation and with human movement and demonstrates excellent and fast calibration performance.

In many sensing and control applications, data are represented in the form of symmetric positive definite (SPD) matrices. Considering the underlying geometry of this data space can be beneficial in many robotics applications. In this paper, we present an extension of Gaussian mixture regression (GMR) with input and/or output data on SPD manifolds. As the covariance of SPD datapoints is a 4th-order tensor, we develop a method for parallel transport of high order covariances on SPD manifolds. The proposed approach is experimented in the context of prosthetic hands, with the estimation of wrist movements based on spatial covariance features computed from surface electromyography (sEMG) signals.

Point set registration is a long addressed problem in lots of pattern recognition tasks. This paper presents a robust point set registration algorithm based on optimization of distance between two probability distributions. A major problem encountered in the point to point algorithms is the definition of correspondence between two point sets. This paper follows the idea of some probability based point set registration methods and the point set is represented as Gaussian Mixture Models (GMMs). Through optimizing distance between the two GMMs, the rigid transformation (rotation and translation) between two point sets will be obtained while averting the trouble of finding correspondence. Previous studies used L2 distance, KL distance, etc. to measure similarity between two GMMs, the problem therein is the robustness to noise and outliers, especially when the covariance matrix is large or there exists local minimum. So in this work, the signature quadratic form distance is derived for the distribution similarity measurement. The contribution of this paper is as follows. First, we derive the signature quadratic form distance for GMMs similarity measurement. Second, the signature quadratic form distance is adopted to the point set registration algorithm. And performance of the proposed method compared with some existing widely used point set registration algorithms is also presented. Experimental results show precision and robustness of this algorithm. The results also demonstrate this algorithm outperforms some state-of-the-art point set registration algorithms in terms of noise, outliers, partial structures and misalignment initialization, etc.

Today's cutting-edge in dynamic legged robots use leg topologies which enable agile behaviors such as running and jumping but most of these dynamic behaviors occur primarily in the robot's sagittal plane. Navigating complex environments, however, requires the use of omni-directional dynamic maneuvers for traversing obstacles in confined spaces. This paper introduces GOAT, the Gearless Omnidirectional Acceleration-vectoring Topology - a novel 3-DoF, 3-RSR leg topology combined with a direct-drive or quasi-direct-drive (QDD) actuation scheme to enable high fidelity proprioceptive force control and multi-modal mobility via high-energy, omni-directional jumping, running, and walking. GOAT's kinematics, mechanical design, actuation, drive electronics, and control algorithms are presented. Experimental results successfully demonstrate a mechanically robust GOAT leg performing a) explosive vertical jumping up to 82 cm b) landing using high fidelity virtual compliance and c) high-speed omnidirectional running/jumping trajectories while mounted on a test rig.

Estimation of 3D object position is a crucial step for a variety of robotics and computer vision applications including 3D reconstruction and object manipulation. When working in microscale, new types of visual sensors are used such as Scanning Electron Microscope (SEM). Nowadays, micro- and nanomanipulation tasks, namely components assembly, are performed in teleoperated mode in most of the cases. Measuring object position and orientation is a crucial step towards automatic object handling. Current methods of pose estimation in SEM allow recovering full object movement using its computer-aided design (CAD) model. If the model is not known, most methods allow to estimate only in-plane translations and rotation around camera optical axis. In the literature, SEM is considered as a camera with parallel projection or an affine camera, which means image invariance to z-movement and bas-relief ambiguity. In this paper, authors address the problem of measuring full 3D rotation of the unknown scene for uncalibrated SEM without additional sensors. Rotations are estimated from image triplets by solving a spherical triangle from fundamental matrices only, without need of intrinsic calibration, allowing to avoid parallel projection ambiguities. The presented results, obtained in simulation and on real data, allow validating the proposed scheme.

The traction force of any moving object on a floor is the frictional force between the object and the floor. Therefore, investigation of the frictional constraint is very important for such moving object on a floor as well as a biped walking robot. Conventionally, the study of frictional constraint of a biped robot was only constrained in translational slip, without considering rotational slip, which is important in actual biped walking. This is one of the causes the walking speed of biped robots lower than the one of human beings. In this paper, an evaluation method of frictional constraints for biped robots is proposed. The translational friction force and the twist frictional torque acting on the biped robot's sole are both deduced and verified for single support phase. First, based on a biped robot model, an approach to calculate the frictional force and torque is deduced as the quantitative expression of frictional constraint. Then, the experimental verification method is designed and the proposed calculation approach is confirmed. Finally, the consideration of the result of experiments is discussed. This approach is expected to be used for helping the realization of fast biped walkings.

Compliant end effectors in robotics are an important prerequisite for the field of object manipulation and environment interactions. However, current manipulators are usually stiff position-controlled systems, generating the need to add compliance. In this work, we present Forward Dynamics Compliance Control (FDCC), a new threefold control concept that realizes Cartesian compliance through combining Admittance, Impedance and Force Control into one control strategy. We close the control loop only through a force-torque sensor, allowing a system independent and decoupled configuration of the end effector compliance. As a key component in FDCC, we leverage forward dynamics simulations of a virtual model to directly map Cartesian inputs to joint control commands, leading to excellent stability in singularities. Experiments on three different robotic manipulators verify the key advantages of this approach.

Accurate specification and rigorous analysis of Jacobian matrix are indispensable to guarantee correct evaluation on the manipulator kinematics performance. In this paper, a formal analysis method of the Jacobian matrix in the screw theory is presented by using the higher-order logic theorem prover HOL4. Formalizations of twists and the forward kinematics are characterized with the product of exponential formula and the theory of functional matrices. To the best of our knowledge, this work is the first to formally reason about the spatial Jacobian using theorem proving. The formal modeling and analysis of a 3-DOF planar manipulator substantiate the effectiveness and applicability of the proposed approach to formally verify the kinematics properties of manipulator.

The presence of force feedback in medical instruments has been proven to reduce tissue damage. In order to provide force feedback, information about the interaction forces between the instrument and the environment must be known. Direct measurement of these forces by commercial sensors is not feasible due to space limitations. Thus, in this study we propose to estimate the interaction forces using strain measurements from Fiber Bragg Grating (FBG) sensors. These measurements can also be used for shape sensing and as a result both force and shape can be sensed simultaneously. For force sensing two models are proposed and compared. The first is based on a Rigid Link approximation, while the second uses the Cosserat rod theory. The models are validated experimentally using a tendon-driven continuum manipulator that is subjected to forces at the tip. The force estimates from the models are compared to the measurements from a commercial force sensor. Mean absolute errors of 11.2 mN (6.9%) and 15.9 mN (8.3%) are observed for the Rigid Link model and Cosserat model, respectively.

In this work, we concern the problem of motion planning for a hexapod walking robot crawling in a semi-structured environment where a precise foot-tip positioning is necessary. We propose pipelined approach utilizing an RGB-D camera to perceive and map the forthcoming terrain in 2.5 D which is then processed for available foot-tip positions. The robot motion control is based on sampling-based planning to determine the most suitable leg supporting configurations for the individual body positions in the created terrain map. The individual body positions are connected into a roadmap with taking into account a feasibility of the robot transition between the individual configurations. The resulting trajectory is then planned in the created roadmap using a standard A* planner. The proposed method has been experimentally evaluated in the on-line and onboard setup with a real hexapod crawling robot. The herein reported results support feasibility of the proposed approach for a precise motion planning of small hexapod crawling robot in a semi-structured environment.

A novel biped control is proposed. It enables a robust foot-guided maneuver and suits to the locomotion in complex environments. It treats ZMP as an indirect manipulation variable rather than the control variable, and does not need to observe the acceleration. ZMP behavior is only constrained to approach the pivot foot during a step, so that the robot can flexibly accelerate itself. The most distinctive point to the previous methods is that the motion stability is guaranteed based on the capture point only at landing, which improves agility. The optimal input, i.e., the desired position of ZMP, is analytically obtained immediately from the current state of COM, time to land, and the desired landing position, so that it can be implemented with even less computation cost. Though the limitation of the supporting region is not explicitly taken into account, it is guaranteed that the desired ZMP is confined within the region based on its convexity.