ICRA 2018 论文列表

We consider a virtual manipulator in grasping scenarios which allows us to capture the effect of the object dynamics. This modeling approach turns a multi-arm robot into an underactuated system. We observe that controlling floating-base multi-leg robots is fundamentally similar. The Projected Inverse Dynamics Control approach is employed for decoupling contact consistent motion generation and controlling contact wrenches. The proposed framework for underactuated robots has been evaluated on an enormous robot hand composed of four KUKA LWR IV+ representing fingers cooperatively manipulating a 9kg box with total 28 actuated DOF and six virtual DOF representing the object as additional free-floating robot link. Finally, we validate the same approach on ANYmal, a floating-base quadruped with 12 actuated DOF. Experiments are performed both in simulation and real world.

This paper describes the dynamic modeling and locomotion control of a two-link underactuated brachiating robot traversing a flexible cable. A multi-body system comprised of a two-link robot, a flexible cable, and coupling soft junctions is modeled dynamically. This model is used to formulate an energy-minimizing optimal control strategy that includes the effects of cable vibration induced by robot locomotion. Optimized trajectories and control torque profiles are obtained via multiple-shooting and parametric trajectory approaches. Simulation results show that these optimal torque profiles result in energy-efficient continuous brachiation over a flexible cable. Additional studies examine how the optimal torque profiles change depending on the robot's initial position along a catenary cable.

The exploration of an “intelligent” orthotic shoe sole to negate, or minimize, longitudinal slip by momentarily increasing friction force is presented. The conceptual device takes the form of a rubberized shoe sole containing pockets of air that can be released via valves controlled by a microprocessor. During a slip event, the valves would be opened and the bladders would be collapsed by the weight of the user, which modulates contact and friction forces. The goal is to increase friction forces in this process, by creating an impact force between the user and ground surface, with the potential to increase friction and mitigate slip. Simulations of bladder walls, air flow through valves, contact forces, and friction forces are modeled and combined into a lumped parameter model to predict device behavior. Prototypes of the device are created and evaluated to validate models and slip-mitigating potential.

Strain wave gears are frequently used in small and medium size industrial robots. In order to describe and quantify friction effects in gearboxes of such type, a structurally simple, yet powerful model is proposed taking into account both speed-and load-dependent friction effects. Moreover, position-dependent disturbances in a robotic joint are considered. An identification procedure is presented that allows to separate the individual components of the model and identify them subsequently. The effectiveness of the model and identification procedure is validated using experimental data gathered from four different robotic joints of varying size. Furthermore, the benefits of improved friction modeling are shown by means of different applications, including smooth lead-through programming and sensorless force control.

One of the most challenging tasks in the development of path planners for intelligent vehicles is the design of the cost function that models the desired behavior of the vehicle. While this task has been traditionally accomplished by hand-tuning the model parameters, recent approaches propose to learn the model automatically from demonstrated driving data using Inverse Reinforcement Learning (IRL). To determine if the model has correctly captured the demonstrated behavior, most IRL methods require obtaining a policy by solving the forward control problem repetitively. Calculating the full policy is a costly task in continuous or large domains and thus often approximated by finding a single trajectory using traditional path-planning techniques. In this work, we propose to find such a trajectory using a conformal spatiotemporal state lattice, which offers two main advantages. First, by conforming the lattice to the environment, the search is focused only on feasible motions for the robot, saving computational power. And second, by considering time as part of the state, the trajectory is optimized with respect to the motion of the dynamic obstacles in the scene. As a consequence, the resulting trajectory can be used for the model assessment. We show how the proposed IRL framework can successfully handle highly dynamic environments by modeling the highway tactical driving task from demonstrated driving data gathered with an instrumented vehicle.

We address the integrated prediction, planning, and control problem that enables a single follower robot (the photographer) to quickly re-establish visual contact with a moving target (the subject) that has escaped the follower's field of view. We deal with this scenario, which reactive controllers are typically ill-equipped to handle, by making plausible predictions about the long- and short-term behavior of the target, and planning pursuit paths that will maximize the chance of seeing the target again. At the core of our pursuit method is the use of predictive models of target behavior, which help narrow down the set of possible future locations of the target to a few discrete hypotheses, as well as the use of combinatorial search in physical space to check those hypotheses efficiently. We model target behavior in terms of a learned navigation reward function, using Inverse Reinforcement Learning, based on semantic terrain features of satellite maps. Our pursuit algorithm continuously predicts the latent destination of the target and its position in the future, and relies on efficient graph representation and search methods in order to navigate to locations at which the target is most likely to be seen at an anticipated time. We perform extensive evaluation of our predictive pursuit algorithm over multiple satellite maps, thousands of simulation scenarios, against state-of-the art MDP and POMDP solvers. We show that our method significantly outperforms them by exploiting domain-specific knowledge, while being able to run in real-time.

The dynamics of a humanoid robot cannot be correctly described independently from the external forces acting on it. These forces have to be reconstructed to enable the robot to control them or to compensate for them. Force sensors are usually used to measure these forces, but because of their cost, they are often put only on the ankle/feet and possibly the wrists. This paper addresses the issue of the estimation of external forces and moments that apply at any part of a robot without direct force measurements and without torque measurements. The sensors used are the regular force sensors and the IMUs of the robot. The method relies on a model-based estimator able to make the fusion between these sensors and the whole body dynamics. The estimator reconstructs a single state vector containing the floating-base kinematics, a filtered measurement of contact force and an additional estimation external force that we evaluate in this paper. Validation is performed on HRP-2 in a multi-contact motion.

This paper investigates a multi-rotor with a suspended load in perspectives of 1) real-time path planning, 2) obstacle avoidance, and 3) transportation of a suspended object. A suspended load cannot be controlled with conventional controllers designed for nominal multi-rotors due to the dynamic coupling between the multi-rotor and load. Although several control and planning algorithms have been proposed based on elaborately derived dynamic equations, most existing studies separate control and path planning problems by following predefined trajectories after trajectory generation. Moreover, many state-of-the-art trajectory generation algorithms cannot work real-time for a system with high degrees of freedom, which makes it not suitable to operate the system in dynamic environments where obstacles appear abruptly or move unexpectedly. With this in mind, we apply Model Predictive Control (MPC) with Sequential Linear Quadratic (SLQ) solver to compute feasible and optimal trajectory real-time and to operate a multi-rotor with a suspended load in dynamic environments. We design an obstacle-avoidance algorithm suitable for the current platform flying in cluttered environments. Flight experiments shows that the proposed algorithm successfully controls the multi-rotor and allows to avoid obstacles simultaneously.

We introduce ModQuad, a novel flying modular robotic structure that is able to self-assemble in midair and cooperatively fly. The structure is composed by agile flying modules that can easily move in a three dimensional environment. The module is based on a quadrotor platform within a cuboid frame which allows it to attach to other modules by matching vertical faces. Using this mechanism, a ModQuad swarm is able to rapidly assemble flying structures in midair using the robot bodies as building units. In this paper, we focus on two important tasks for modular flying structures. First, we propose a decentralized modular attitude controller to allow a team of physically connected modules to fly cooperatively. Second, we develop a docking method that drives pairs of structures to be attached in midair. Our method precisely aligns, and corrects motion errors during the docking process. In our experiments, we tested and analyzed the performance of the cooperative flying method for multiple configurations. We also tested the docking method with successful results.

We present the new HippoCampus micro underwater vehicle, first introduced in [1]. It is designed for monitoring confined fluid volumes. These tightly constrained settings demand agile vehicle dynamics. Moreover, we adapt a robust attitude control scheme for aerial drones to the underwater domain. We demonstrate the performance of the controller with a challenging maneuver. A submerged Furuta pendulum is stabilized by HippoCampus after a swing-up. The experimental results reveal the robustness of the control method, as the system quickly recovers from strong physical disturbances, which are applied to the system.

Dense reconstructions often contain errors that prior work has so far minimised using high quality sensors and regularising the output. Nevertheless, errors still persist. This paper proposes a machine learning technique to identify errors in three dimensional (3D) meshes. Beyond simply identifying errors, our method quantifies both the magnitude and the direction of depth estimate errors when viewing the scene. This enables us to Improve the reconstruction accuracy. We train a suitably deep network architecture with two 3D meshes: a high-quality laser reconstruction, and a lower quality stereo image reconstruction. The network predicts the amount of error in the lower quality reconstruction with respect to the high-quality one, having only view the former through its input. We evaluate our approach by correcting two dimensional (2D) inverse-depth images extracted from the 3D model, and show that our method improves the quality of these depth reconstructions by up to a relative 10% RMSE.

We present here, a novel network architecture called MergeNet for discovering small obstacles for on-road scenes in the context of autonomous driving. The basis of the architecture rests on the central consideration of training with less amount of data since the physical setup and the annotation process for small obstacles is hard to scale. For making effective use of the limited data, we propose a multi-stage training procedure involving weight-sharing, separate learning of low and high level features from the RGBD input and a refining stage which learns to fuse the obtained complementary features. The model is trained and evaluated on the Lost and Found dataset and is able to achieve state-of-art results with just 135 images in comparison to the 1000 images used by the previous benchmark. Additionally, we also compare our results with recent methods trained on 6000 images and show that our method achieves comparable performance with only 1000 training samples.

To precisely reach for an object with a humanoid robot, it is of central importance to have good knowledge of both end-effector, object pose and shape. In this work we propose a framework for markerless visual servoing on unknown objects, which is divided in four main parts: I) a leastsquares minimization problem is formulated to find the volume of the object graspable by the robot's hand using its stereo vision; II) a recursive Bayesian filtering technique, based on Sequential Monte Carlo (SMC) filtering, estimates the 6D pose (position and orientation) of the robot's end-effector without the use of markers; III) a nonlinear constrained optimization problem is formulated to compute the desired graspable pose about the object; IV) an image-based visual servo control commands the robot's end-effector toward the desired pose. We demonstrate effectiveness and robustness of our approach with extensive experiments on the iCub humanoid robot platform, achieving real-time computation, smooth trajectories and subpixel precisions.

This paper introduces an Augmented Reality (AR) system for open liver surgery. Although open surgery remains the gold-standard for the treatment of complex tumors and central lesions, technological issues actually prevent using AR with sufficient accuracy for clinical use. We propose a markers-based method allowing for the tracking and the deformation of a preoperative model in real-time during the surgery. Markers are manually placed on the surface of the organ after opening the abdominal cavity, and tracked in real-time by a set of infrared cameras. Our framework is composed of both a nonrigid initial registration method, providing an estimation of the location of the markers in the preoperative model, and a realtime tracking algorithm to deform the model during the surgery (even for large deformation or partial occlusion of the organ). The method is validated on both synthetic and ex-vivo samples; in addition, we demonstrate its applicability in the operating room during a liver resection surgery on a human patient. Preliminary studies provided promising results to improve the location of tumors, and to help surgeons into planning the ideal resection intraoperatively.

This paper presents a weakly-supervised learning system for real-time road marking detection using images of complex urban environments obtained from a monocular camera. We avoid expensive manual labelling by exploiting additional sensor modalities to generate large quantities of annotated images in a weakly-supervised way, which are then used to train a deep semantic segmentation network. At run time, the road markings in the scene are detected in real time in a variety of traffic situations and under different lighting and weather conditions without relying on any preprocessing steps or predefined models. We achieve reliable qualitative performance on the Oxford RobotCar dataset, and demonstrate quantitatively on the CamVid dataset that exploiting these annotations significantly reduces the required labelling effort and improves performance.

Various autonomous robotic systems require maps for robust and safe navigation. Particularly when robots are employed in dynamic environments, accurate knowledge about which components of the robot perceptions belong to dynamic and static aspects in the environment can greatly improve navigation functions. In this paper we propose a novel method for building 3D grid maps using laser range data in dynamic environments. Our approach uses a neural network to estimate the pointwise probability of a point belonging to a dynamic object. The output from our network is fed to the mapping module for building a 3D grid map containing only static parts of the environment. We present experimental results obtained by training our neural network using the KITTI dataset and evaluating it in a mapping process using our own dataset. In extensive experiments, we show that maps generated using the proposed probability about dynamic objects increases the accuracy of the resulting maps.

This work presents a method to improve vehicle localization by using the information from a prior occupancy grid to bound the possible poses. The method, named Map-Aware Particle Filter, uses a nonlinear approach to map-matching that can be integrated into a particle filter framework for localization. Each particle is re-weighted based on the validity of its current position in the map. In addition, we buffer the trajectory followed by the vehicle and then append it to each particle's pose. We then quantify the overlap between the trajectory and the map's free space. This serves as a measure of each particle's validity given the trajectory and the shape of the map. We evaluated the method by performing experiments with different types of localization sensors: First, (i) we significantly reduced the drift inherent to dead reckoning. By only using wheel odometry and map information we achieved loop closure over a distance of approximately 3 km. We also (ii) increased the accuracy of GPS localization. Finally, (iii) we fused a fragile 2D LiDAR localization with the map information. The resulting system had a higher robustness and managed to close the loop in an outdated map where it had failed before.

The complex physical properties of highly deformable materials such as clothes pose significant challenges for autonomous robotic manipulation systems. We present a novel visual feedback dictionary-based method for manipulating deformable objects towards a desired configuration. Our approach is based on visual servoing and we use an efficient technique to extract key features from the RGB sensor stream in the form of a histogram of deformable model features. These histogram features serve as high-level representations of the state of the deformable material. Next, we collect manipulation data and use a visual feedback dictionary that maps the velocity in the high-dimensional feature space to the velocity of the robotic end-effectors for manipulation. We have evaluated our approach on a set of complex manipulation tasks and human-robot manipulation tasks on different cloth pieces with varying material characteristics.

Teleoperation systems could benefit from force sensory substitution when kinesthetic force feedback systems are too bulky or expensive, and when they cause instability by magnifying force feedback. We aim to magnify force feedback using sensory substitution via force-controlled tactile skin deformation, using a device with the ability to provide tangential and normal force directly to the fingerpads. The sensory substitution device is able to provide skin deformation force feedback over ten times the maximum stable kinesthetic force feedback on a da Vinci Research Kit teleoperation system. We evaluated the effect of this force magnification in two experimental tasks where the goal was to minimize interaction force with the environment. In a peg transfer task, magnified force feedback using sensory substitution improved participants' performance for force magnifications up to ten times, but decreased performance for higher force magnifications. In a tube connection task, sensory substitution that doubled the force feedback maximized performance; there was no improvement at the larger magnifications. These experiments demonstrate that magnified force feedback using sensory substitution via force-controlled skin deformation feedback can decrease applied forces similarly to magnified kinesthetic force feedback during teleoperation.