ICRA 2021 论文列表

Autonomous fetal ultrasonography with a robotic ultrasound (US) can potentially solve the issue of the shortage of ob-gyn physicians in prenatal care. In fetal cardiac diagnosis, acoustic shading derived from the fetal skeleton makes the scanning procedure using a robotic US cumbersome. We hypothesize that fetal bone distribution can be used for determining the fetal position and for subsequently estimating the heart position. This paper serves a method for estimating fetal heart position for autonomous cardiac diagnosis. Our proposed system is comprised of following three processes. First, the fetal bone distribution is generated by following steps: detect shadow areas derived from fetal bones; calculate the three-dimensional (3D) position of the bones; and identify bone types (skull, spine, and others) based on its shadow features. Next, the fetal head position and body axis are estimated based on the generated 3D bone distribution. Finally, the fetal chest position and thoracic sagittal axis of the fetus are estimated based on anatomical structures, and the position of the heart is determined. We validated the proposed method by evaluating the accuracies of bone detection and heart position estimation with three pregnant volunteers. The false positive rate of bone detection is less than 10%, which is sufficient for the proposed method; further, the estimation error for the heart position was within an acceptable error of 15 mm. These results demonstrate the sufficient accuracy of the proposed estimation method and suggest the potential of autonomous fetal cardiac diagnosis using robotic US scanning.

Shared autonomy teleoperation can guarantee safety, but does so by reducing the human operator’s control authority, which can lead to reduced levels of human-robot agreement and user satisfaction. This paper presents a novel haptic shared autonomy teleoperation paradigm that uses haptic feedback to inform the user about the inner state of a shared autonomy paradigm, while still guaranteeing safety. This differs from haptic shared control, which uses haptic feedback to inform the user’s actions, but gives the human operator full control over the robot’s actions. We conducted a user study in which twelve users flew a simulated UAV in a search-and-rescue task with no assistance or assistance provided by haptic shared control, shared autonomy, or haptic shared autonomy. All assistive teleoperation methods use control barrier functions to find a control command that is both safe and as close as possible to the human-generated control command. For assistive teleoperation conditions with haptic feedback, we apply a force to the user that is proportional to the difference between the human-generated control and the safe control. We find that haptic shared autonomy improves the user’s task performance and satisfaction. We also find that haptic feedback in assistive teleoperation can improve the user’s situational awareness. Finally, results show that adding haptic feedback to shared-autonomy teleoperation can improve human-robot agreement.

In this paper, we investigate the problems that Convolutional Neural Networks (CNN)-based pose estimators have with symmetric objects. We considered the value of the CNN’s output representation when continuously rotating the object and found that it has to form a closed loop after each step of symmetry. Otherwise, the CNN (which is itself a continuous function) has to replicate an uncontinuous function. On a 1-DOF toy example we show that commonly used representations do not fulfill this demand and analyze the problems caused thereby. In particular, we find that the popular min-over-symmetries approach for creating a symmetry-aware loss tends not to work well with gradient-based optimization, i.e. deep learning.We propose a representation called "closed symmetry loop" (csl) from these insights, where the angle of relevant vectors is multiplied by the symmetry order and then generalize it to 6-DOF. The representation extends our algorithm from [1] including a method to disambiguate symmetric equivalents during the final pose estimation. The algorithm handles continuous rotational symmetry (e.g. a bottle) and discrete rotational symmetry (e.g. a 4-fold symmetric box). It is evaluated on the T-LESS dataset, where it reaches state-of-the-art for unrefining RGB-based methods.

Traditional aerial vehicles are usually custom-designed for specific tasks. Although they offer an efficient solution, they are not always able to adapt to changes in the task specification, e.g., increasing the payload. This applies to quadrotors, having a maximum payload and only four controllable degrees of freedom, limiting their adaptability to the task’s variations. We propose a versatile modular robotic system that can increase its payload and degrees of freedom by assembling heterogeneous modules; we call it H-ModQuad. It consists of cuboid modules propelled by quadrotors with tilted propellers that can generate forces in different directions. By connecting different types of modules, an H-ModQuad can increase its controllable degrees of freedom from 4 to 5 and 6. We model the general structure and propose three controllers, one for each number of controllable degrees of freedom. We extend the concept of the actuation ellipsoid to find the best reference orientation that can maximize the performance of the structure. Our approach is validated with experiments using actual robots, showing the independence of the translation and orientation of a structure.

Demands on leg degrees of freedom and control precision for bipedal robotics are steadily increasing, especially for the tasks involving walking on a rough terrain. In this paper we present an alternative, as well as a working proof-of-concept. Meet gyrubot: a 5-link almost planar bipedal robot with a torso complemented by a nonanthropomorphic stabilization system, capable of blindly walking through uneven areas. Despite being almost planar, the robot does not need any support in the frontal plane! This paper describes the mechanical design and the architecture of the controllers. We also provide the experimental evidence of the ability of gyrubot to navigate across non-flat terrains.

Medical diagnostic robot systems have been paid more and more attention due to its objectivity and accuracy. The diagnosis of mild cognitive impairment (MCI) is considered an effective means to prevent Alzheimer's disease (AD). Doctors diagnose MCI based on various clinical examinations, which are expensive and the diagnosis results rely on the knowledge of doctors. Therefore, it is necessary to develop a robot diagnostic system to eliminate the influence of human factors and obtain a higher accuracy rate. In this paper, we propose a novel Group Feature Domain Adversarial Neural Network (GF- DANN) for amnestic MCI (aMCI) diagnosis, which involves two important modules. A Group Feature Extraction (GFE) module is proposed to reduce individual differences by learning group- level features through adversarial learning. A Dual Branch Domain Adaptation (DBDA) module is carefully designed to reduce the distribution difference between the source and target domain in a domain adaption way. On three types of data set, GF-DANN achieves the best accuracy compared with classic machine learning and deep learning methods. On the DMS data set, GF-DANN has obtained an accuracy rate of 89.47%, and the sensitivity and specificity are 90% and 89%. In addition, by comparing three EEG data collection paradigms, our results demonstrate that the DMS paradigm has the potential to build an aMCI diagnose robot system.

Modern LiDAR-SLAM (L-SLAM) systems have shown excellent results in large-scale, real-world scenarios. However, they commonly have a high latency due to the expensive data association and nonlinear optimization. This paper demonstrates that actively selecting a subset of features significantly improves both the accuracy and efficiency of an L-SLAM system. We formulate the feature selection as a combinatorial optimization problem under a cardinality constraint to preserve the information matrix's spectral attributes. The stochastic-greedy algorithm is applied to approximate the optimal results in real-time. To avoid ill-conditioned estimation, we also propose a general strategy to evaluate the environment's degeneracy and modify the feature number online. The proposed feature selector is integrated into a multi-LiDAR SLAM system. We validate this enhanced system with extensive experiments covering various scenarios on two sensor setups and computation platforms. We show that our approach exhibits low localization error and speedup compared to the state-of-the-art L-SLAM systems. To benefit the community, we have released the source code: https://ram-lab.com/file/site/m-loam.

Billions of people use chopsticks, a simple yet versatile tool, for fine manipulation of everyday objects. The small, curved, and slippery tips of chopsticks pose a challenge for picking up small objects, making them a suitably complex test case. This paper leverages human demonstrations to develop an autonomous chopsticks-equipped robotic manipulator. Due to the lack of accurate models for fine manipulation, we explore model-free imitation learning, which traditionally suffers from the covariate shift phenomenon that causes poor generalization. We propose two approaches to reduce covariate shift, neither of which requires access to an interactive expert or a model, unlike previous approaches. First, we alleviate singlestep prediction errors by applying an invariant operator to increase the data support at critical steps for grasping. Second, we generate synthetic corrective labels by adding bounded noise and combining parametric and non-parametric methods to prevent error accumulation. We demonstrate our methods on a real chopstick-equipped robot that we built, and observe the agent’s success rate increase from 37.3% to 80%, which is comparable to the human expert performance of 82.6%.

As it happens, despite yet unmatched by robots perception and motor skills humans drop objects during handover because of false grasp detection and early release. Accordingly, the fluent robot-human handover is still an open challenge. This paper presents an approach to a natural robot to human handover using Capacitive Proximity Sensor (CPS) for robust grasp detection and release trigger. We propose an experimental setup for the evaluation using a collaborative robot, an eye-in-hand depth camera, and CPS integrated into the gripper. Three grasp detection methods were implemented and an object release was triggered based on torque-sensing, capacitive sensing, and the combination of both. Finally, a user study was designed and conducted, indicating that the capacitive method is the most preferred type with the shortest human idle time and the highest fluency ratings.

Soft isoperimetric truss robots have demonstrated an ability to grasp and manipulate objects using the members of their structure. The compliance of the members affords large contact areas with even force distribution, allowing for successful grasping even with imprecise open-loop control. In this work we present methods of analyzing and controlling isoperimetric truss robots in the context of grasping and manipulating objects. We use a direct stiffness model to characterize the structural properties of the robot and its interactions with external objects. With this approach we can estimate grasp forces and stiffnesses with limited computation compared to higher fidelity finite elements methods, which, given the many degrees-of-freedom of truss robots, are prohibitively expensive to run on-board. In conjunction with the structural model, we build upon a literature of differential kinematics for truss robots and apply it to the task of manipulating an object within the robot’s workspace.

Currently, state-of-the-art exploration methods maintain high-resolution map representations in order to optimize exploration goals in each step that maximizes information gain. However, during exploring, those "optimal" selections could quickly become obsolete due to the influx of new information, especially in large-scale environments, and result in high-frequency re-planning that hinders the overall exploration efficiency. In this paper, we propose a graph-based topological planning framework, building a sparse topological map in three-dimensional (3D) space to guide exploration steps with high-level intents so as to render consistent exploration maneuvers. Specifically, this work presents a novel method to estimate 3D space’s geometry with convex polyhedrons. Then, the geometry information is utilized to group space into distinctive regions. And those regions are added as nodes into the topological map, directing the exploration process. We compared our method with the state-of-the-art in simulated environments. The proposed method achieves higher space coverage and outperforms exploration efficiency by more than 40% during experiments. Finally, a field experiment was conducted to further evaluate the applicability of our method to empower efficient and robust exploration in real-world environments.

One of the most crucial yet challenging tasks for autonomous vehicles in urban environments is predicting the future behaviour of nearby pedestrians, especially at points of crossing. Predicting behaviour depends on many social and environmental factors, particularly interactions between road users. Capturing such interactions requires a global view of the scene and dynamics of the road users in three-dimensional space. This information, however, is missing from the current pedestrian behaviour benchmark datasets. Motivated by these challenges, we propose 1) a novel graph-based model for predicting pedestrian crossing action. Our method models pedestrians’ interactions with nearby road users through clustering and relative importance weighting of interactions using features obtained from the bird’s-eye-view. 2) We introduce a new dataset that provides 3D bounding box and pedestrian behavioural annotations for the existing nuScenes dataset. On the new data, our approach achieves state-of-the-art performance by improving on various metrics by more than 15% in comparison to existing methods. The dataset is available at https://github.com/huawei-noah/datasets/PePScenes.

Modular self-reconfigurable robotic (MSRR) systems are potentially more robust and more adaptive than conventional systems. Following our previous work where we proposed a freeform MSRR module called FreeBOT, this paper presents a novel configuration detection system for FreeBOT using a magnetic sensor array. A FreeBOT module can be connected by up to 11 modules, and the proposed configuration detection system can locate a variable number of connection points accurately in real-time. By equipping FreeBOT with 24 magnetic sensors, the magnetic field density produced by magnets and steel spherical shells can be monitored. The connectable area is split into 199 non-uniform regions, including 84 uniform regions. Using a Graph Convolutional Network (GCN) based algorithm, the connection points can be located accurately under ferromagnetic environments. The system can locate a variable number of connection points for such a region division with only single connection point training data. Finally, the localization algorithm can run faster than 40 Hz on FreeBOT. With the real-time configuration detection system, the FreeBOT system has the potential to reconfigure automatically and accurately.

This paper proposes an online optimal perception-aware strategy meant to maximize the information collected along the trajectory via the available measurements while simultaneously minimizing the negative effects of actuation/process noise. Indeed, in several robotic applications, the actuation/process noise is far from negligible and its negative effects are particularly relevant especially with intermittent measurements (e.g. collected by a vision system with limited Field-Of-View). New metrics are proposed as combinations of the Constructability Gramian, for measuring the amount of information collected via the available sensors, and the Reachability Gramian, for measuring the degrading effects of actuation/process noise. Control inputs that optimize those cost functions are provided. To show the effectiveness of our method, we consider one case study involving a unicycle-like vehicle, subject to Gaussian measurement noise and Gaussian or Brownian actuation noise, that estimates its state using intermittent distances from known environmental markers.

Visuomotor control (VMC) is an effective means of achieving basic manipulation tasks such as pushing or pick- and-place from raw images. Conditioning VMC on desired goal states is a promising way of achieving versatile skill primitives. However, common conditioning schemes either rely on task-specific fine tuning - e.g. using one-shot imitation learning (IL) - or on sampling approaches using a forward model of scene dynamics i.e. model-predictive control (MPC), leaving deployability and planning horizon severely limited. In this paper we propose a conditioning scheme which avoids these pitfalls by learning the controller and its conditioning in an end-to-end manner. Our model predicts complex action sequences based directly on a dynamic image representation of the robot motion and the distance to a given target observation. In contrast to related works, this enables our approach to efficiently perform complex manipulation tasks from raw image observations without predefined control primitives or test time demonstrations. We report significant improvements in task success over representative MPC and IL baselines. We also demonstrate our model's generalisation capabilities in challenging, unseen tasks featuring visual noise, cluttered scenes and unseen object geometries.

When learning policies for robotic systems from data, safety is a major concern, as violation of safety constraints may cause hardware damage. SafeOpt is an efficient Bayesian optimization (BO) algorithm that can learn policies while guaranteeing safety with high probability. However, its search space is limited to an initially given safe region. We extend this method by exploring outside the initial safe area while still guaranteeing safety with high probability. This is achieved by learning a set of initial conditions from which we can recover safely using a learned backup controller in case of a potential failure. We derive conditions for guaranteed convergence to the global optimum and validate GoSafe in hardware experiments.

We combine Boston Dynamics Spot® with a light-weight, external robot arm to perform dynamic grasping maneuvers. While Spot is a reliable, robust and easy-to-control mobile robot, these highly desirable qualities come with the price that the control access granted to the user is restricted. Consequently Spot’s behavior must largely be treated as a black box, which causes difficulties when combined with a moving payload such as a robotic arm. We overcome the arising challenges by building a model of the combined platform, fitting the corresponding model parameters using experimental data and a straight-forward optimization framework. We use this model to generate control commands for the physical platform using trajectory optimization. We demonstrate that even with a simple model, and control trajectories deployed in a feed-forward manner, the combined platform is capable of executing grasping tasks in a dynamic fashion. Furthermore, we show how the platform can use the additional degrees of freedom of the legs to extend the reachability of the arm.

Redundant robots offer the possibility of improving agility, compared to their non-redundant counterparts, by exploiting the additional kinematic DOFs to increase a measure called manipulability. While it is common to maximize the manipulability measure during redundancy resolution locally, global optimization of a full trajectory is usually computationally too expensive and thus only considered for offline procedures in current literature. However, local maximization is prone to be sub-optimal and at times even fails at preserving agility of a robot that ought to be reactive. In this work we build upon our previous contributions on online trajectory generation on SE(3) and closed-form task space manipulability of a 7-DOF serial robot, and combine it with graph search techniques for global optimization. This enables, for the first time, online trajectory generation with globally optimal redundancy resolution regarding manipulability, to maintain agility in reactive robot behavior.

Global position control for underactuated bipedal walking is a challenging problem due to the lack of actuation on the feet of the robots. In this paper, we apply the Hybrid-Linear Inverted Pendulum (H-LIP) based stepping on 3D underactuated bipedal robots for global position control. The step-to-step (S2S) dynamics of the H-LIP walking approximates the actual S2S dynamics of the walking of the robot, where the step size is considered as the input. Thus the feedback controller based on the H-LIP approximately controls the robot to behave like the H-LIP, the differences between which stay in an error invariant set. Model Predictive Control (MPC) is applied to the H-LIP for global position control in 3D. The H-LIP stepping then generates desired step sizes for the robot to track. Moreover, turning behavior is integrated with the step planning. The proposed framework is verified on the 3D underactuated bipedal robot Cassie in simulation together with a proof-of-concept experiment.