ICRA 2020 论文列表

We present a novel analysis of AO-RRT: a tree-based planner for motion planning with kinodynamic constraints, originally described by Hauser and Zhou (AO-X, 2016). AO-RRT explores the state-cost space and has been shown to efficiently obtain high-quality solutions in practice without relying on the availability of a computationally-intensive two-point boundary-value solver. Our main contribution is an optimality proof for the single-tree version of the algorithm-a variant that was not analyzed before. Our proof only requires a mild and easily-verifiable set of assumptions on the problem and system: Lipschitz-continuity of the cost function and the dynamics. In particular, we prove that for any system satisfying these assumptions, any trajectory having a piecewise-constant control function and positive clearance from the obstacles can be approximated arbitrarily well by a trajectory found by AORRT. We also discuss practical aspects of AORRT and present experimental comparisons of variants of the algorithm.

This paper introduces a Model Predictive Control (MPC) strategy for fully-constrained Cable-Driven Parallel Robots. The main advantage of the proposed scheme lies in its ability to explicitly handle cable tension limits. Indeed, the cable tension distribution is performed as an integral part of the main control architecture. This characteristic significantly improves the safety of the system. Experimental results demonstrate this advantage addressing a typical pick-and-place task with two different scenarios: nominal cable tension limits and reduced maximum tension. Satisfactory tracking errors were obtained in the first scenario. In the second scenario, the desired trajectory escapes from the workspace defined by the new set of tension limits. The proposed MPC scheme is able to minimize the tracking errors without violating the tension limits. Satisfying results were also obtained regarding robustness against uncertainties on the payload mass.

How to reduce uncertainty in object pose estimation under complex contacts is crucial to autonomous robotic manipulation and assembly. In this paper, we introduce an approach through forecasting contact force from simulated complex contacts with calibration based on real force sensing. A constraint-based haptic simulation algorithm is used with sphere-tree representation of contacting objects to compute contact poses and forces, and through matching the computed forces to measured real force data via a regression model, the least-uncertain estimate of the relative contact pose is obtained. Our approach can handle multi-region complex contacts and does not make any assumption about contact types or contact locations. It also does not have restriction on object shapes. We have applied the force forecast approach to reducing uncertainty in estimating object poses in challenging peg-in-hole robotic assembly tasks and demonstrate the effectiveness of the approach by successful completion of contact-rich two-pin and three-pin real peg-in-hole assembly tasks with complex shapes of pins and holes.

Adding more cameras to SLAM systems improves robustness and accuracy but complicates the design of the visual front-end significantly. Thus, most systems in the literature are tailored for specific camera configurations. In this work, we aim at an adaptive SLAM system that works for arbitrary multi-camera setups. To this end, we revisit several common building blocks in visual SLAM. In particular, we propose an adaptive initialization scheme, a sensor-agnostic, information- theoretic keyframe selection algorithm, and a scalable voxel- based map. These techniques make little assumption about the actual camera setups and prefer theoretically grounded methods over heuristics. We adapt a state-of-the-art visual- inertial odometry with these modifications, and experimental results show that the modified pipeline can adapt to a wide range of camera setups (e.g., 2 to 6 cameras in one experiment) without the need of sensor-specific modifications or tuning.

Lower leg prostheses could improve the life quality of amputees by increasing comfort and reducing energy to locomote, but currently control methods are limited in modulating behaviors based upon the human's experience. This paper describes the first steps toward learning complex controllers for dynamical robotic assistive devices. We provide the first example of behavioral cloning to control a powered transfemoral prostheses using a Gated Recurrent Unit (GRU) based recurrent neural network (RNN) running on a custom hardware accelerator that exploits temporal sparsity. The RNN is trained on data collected from the original prosthesis controller. The RNN inference is realized by a novel EdgeDRNN accelerator in real-time. Experimental results show that the RNN can replace the nominal PD controller to realize end-to-end control of the AMPRO3 prosthetic leg walking on flat ground and unforeseen slopes with comparable tracking accuracy. EdgeDRNN computes the RNN about 240 times faster than real time, opening the possibility of running larger networks for more complex tasks in the future. Implementing an RNN on this real-time dynamical system with impacts sets the ground work to incorporate other learned elements of the human-prosthesis system into prosthesis control.

We propose thrombolysis using a magnetic nanoparticle microswarm with tissue plasminogen activator (tPA) under ultrasound imaging. The microswarm is generated in blood using an oscillating magnetic field and can be navigated with locomotion along both the long and short axis. By modulating the input field, the aspect ratio of the microswarm can be reversibly tuned, showing the ability to adapt to different confined environments. Simulation results indicate that both in-plane and out-of-plane fluid convection are induced around the microswarm, which can be further enhanced by tuning the aspect ratio of the microswarm. Under ultrasound imaging, the microswarm is navigated in a microchannel towards a blood clot and deformed to obtain optimal lysis. Experimental results show that the lysis rate reaches -0.1725 ± 0.0612 mm3/min in the 37°C blood environment under the influence of the microswarm-induced fluid convection and tPA. The lysis rate is enhanced 2.5-fold compared to that without the microswarm (-0.0681 ± 0.0263 mm3/min). Our method provides a new strategy to increase the efficiency of thrombolysis by applying microswarm-induced fluid convection, indicating that swarming micro/nanorobots have the potential to act as effective tools towards targeted therapy.

We consider recognition and reconfiguration of lattice-based cellular structures by very simple robots with only basic functionality. The underlying motivation is the construction and modification of space facilities of enormous dimensions, where the combination of new materials with extremely simple robots promises structures of previously unthinkable size and flexibility; this is also closely related to the newly emerging field of programmable matter. Aiming for large-scale scalability, both in terms of the number of the cellular components of a structure, as well as the number of robots that are being deployed for construction requires simple yet robust robots and mechanisms, while also dealing with various basic constraints, such as connectivity of a structure during reconfiguration. To this end, we propose an approach that combines ultra-light, cellular building materials with extremely simple robots. We develop basic algorithmic methods that are able to detect and reconfigure arbitrary cellular structures, based on robots that have only constant-sized memory. As a proof of concept, we demonstrate the feasibility of this approach for specific cellular materials and robots that have been developed at NASA.

Autonomous underwater robots working with teams of human divers may need to distinguish between different divers, e.g., to recognize a lead diver or to follow a specific team member. This paper describes a technique that enables autonomous underwater robots to track divers in real time as well as to reidentify them. The approach is an extension of Simple Online Realtime Tracking (SORT) with an appearance metric (deep SORT). Initial diver detection is performed with a custom CNN designed for realtime diver detection, and appearance features are subsequently extracted for each detected diver. Next, realtime tracking-by-detection is performed with an extension of the deep SORT algorithm. We evaluate this technique on a series of videos of divers performing human-robot collaborative tasks and show that our methods result in more divers being accurately identified during tracking. We also discuss the practical considerations of applying multi-person tracking to on-board autonomous robot operations, and we consider how failure cases can be addressed during on-board tracking.

We consider problems in which robots conspire to present a view of the world that differs from reality. The inquiry is motivated by the problem of validating robot behavior physically despite there being a discrepancy between the robots we have at hand and those we wish to study, or the environment for testing that is available versus that which is desired, or other potential mismatches in this vein. After formulating the concept of a convincing illusion, essentially a notion of system simulation that takes place in the real world, we examine the implications of this type of simulability in terms of infrastructure requirements. Time is one important resource: some robots may be able to simulate some others but, perhaps, only at a rate that is slower than real-time. This difference gives a way of relating the simulating and the simulated systems in a form that is relative. We establish some theorems, including one with the flavor of an impossibility result, and providing several examples throughout. Finally, we present data from a simple multi-robot experiment based on this theory, with a robot navigating amid an unbounded field of obstacles."Truth is beautiful, without doubt; but so are lies."-Ralph Waldo Emerson.

The paper presents a working pipeline which integrates hardware and software in an automated robotic rose cutter. To the best of our knowledge, this is the first robot able to prune rose bushes in a natural environment. Unlike similar approaches like tree stem cutting, the proposed method does not require to scan the full plant, have multiple cameras around the bush, or assume that a stem does not move. It relies on a single stereo camera mounted on the end-effector of the robot and real-time visual servoing to navigate to the desired cutting location on the stem. The evaluation of the whole pipeline shows a good performance in a garden with unconstrained conditions, where finding and approaching a specific location on a stem is challenging due to occlusions caused by other stems and dynamic changes caused by the wind.

In this study, we developed a real-time simulation method for non-deformable continuous tracks having grousers for rough terrain by explicitly considering the collision and friction between the tracks and the ground. In the proposed simulation method, an arbitrary trajectory of a track is represented with multiple linear and circular segments, each of which is a link connected to a robot body. The proposed method sets velocity constraints between each segment link and the robot body, to simulate the track rotation around the body. To maintain the shape of a track, it also restores the positions of the segment links when required. Experimental comparisons with other existing real-time simulation methods demonstrated that while the proposed method considered the grousers and the friction with the ground, it was comparable to them in terms of the computational speed. Experimental comparison of the simulations based on the proposed method and a physical robot exhibited that the former was comparable to the precise motion of the robot on rough or uneven terrain.

Surgical robots have been introduced to operating rooms over the past few decades due to their high sensitivity, small size, and remote controllability. The cable-driven nature of many surgical robots allows the systems to be dexterous and lightweight, with diameters as low as 5mm. However, due to the slack and stretch of the cables and the backlash of the gears, inevitable uncertainties are brought into the kinematics calcu-lation [1]. Since the reported end effector position of surgical robots like RAVEN-II [2] is directly calculated using the motor encoder measurements and forward kinematics, it may contain relatively large error up to 10mm, whereas semi-autonomous functions being introduced into abdominal surgeries require position inaccuracy of at most 1mm. To resolve the problem, a cost-effective, real-time and data-driven pipeline for robot end effector position precision estimation is proposed and tested on RAVEN-II. Analysis shows an improved end effector position error of around 1mm RMS traversing through the entire robot workspace without high-resolution motion tracker. The open source code, data sets, videos, and user guide can be found at //github.com/HaonanPeng/RAVEN Neural Network Estimator.

In this paper an automated data labeling (ADL) neural network is proposed to streamline dataset collecting for real-time predicting the continuous motion of hand and wrist, these gestures are only decoded from a surface electromyography (sEMG) array of eight channels. Unlike collecting both the bio-signals and hand motion signals as samples and labels in supervised learning, this algorithm only collects unlabeled sEMG into an unsupervised neural network, in which the hand motion labels are auto-generated. The coefficient of determination (R2) for three DOFs, i.e. wrist flex/extension, wrist pro/supination, hand open/close, was 0.86, 0.89 and 0.87 respectively. The comparison between real motion labels and auto-generated labels shows that the latter has earlier response than former. The results of Fitts’ law test indicate that ADL has capability of controlling multi-DOFs simultaneously even though the training set only contains sEMG data from single DOF gesture. Moreover, no more hand motion measurement needed which greatly helps upper limb amputee imagine the gesture of residual limb to control a dexterous prosthesis.

Rotation invariance has been an important topic in computer vision tasks. Ideally, robot grasp detection should be rotation-invariant. However, rotation-invariance in robotic grasp detection has been only recently studied by using rotation anchor box that are often time-consuming and unreliable for multiple objects. In this paper, we propose a rotation ensemble module (REM) for robotic grasp detection using convolutions that rotates network weights. Our proposed REM was able to outperform current state-of-the-art methods by achieving up to 99.2% (image-wise), 98.6% (object-wise) accuracies on the Cornell dataset with real-time computation (50 frames per second). Our proposed method was also able to yield reliable grasps for multiple objects and up to 93.8% success rate for the real-time robotic grasping task with a 4-axis robot arm for small novel objects that was significantly higher than the baseline methods by 11-56%.

Unmanned aerial vehicles (UAVs) are frequently used for large-scale scene mapping and reconstruction. However, in most cases, drones are operated manually, which should be more effective and intelligent. In this article, we present a method of real-time UAV path planning for autonomous urban scene reconstruction. Considering the obstacles and time costs, we utilize the top view to generate the initial path. Then we estimate the building heights and take close-up pictures that reveal building details through a SLAM framework. To predict the coverage of the scene, we propose a novel method which combines information on reconstructed point clouds and possible coverage areas. The experimental results reveal that the reconstruction quality of our method is good enough. Our method is also more time-saving than the state-of-the-arts.

Scene understanding is paramount in robotics, self-navigation, augmented reality, and many other fields. To fully accomplish this task, an autonomous agent has to infer the 3D structure of the sensed scene (to know where it looks at) and its content (to know what it sees). To tackle the two tasks, deep neural networks trained to infer semantic segmentation and depth from stereo images are often the preferred choices. Specifically, Semantic Stereo Matching can be tackled by either standalone models trained for the two tasks independently or joint end-to-end architectures. Nonetheless, as proposed so far, both solutions are inefficient because requiring two forward passes in the former case or due to the complexity of a single network in the latter, although jointly tackling both tasks is usually beneficial in terms of accuracy. In this paper, we propose a single compact and lightweight architecture for real-time semantic stereo matching. Our framework relies on coarse-to-fine estimations in a multi-stage fashion, allowing: i) very fast inference even on embedded devices, with marginal drops in accuracy, compared to state-of-the-art networks, ii) trade accuracy for speed, according to the specific application requirements. Experimental results on high-end GPUs as well as on an embedded Jetson TX2 confirm the superiority of semantic stereo matching compared to standalone tasks and highlight the versatility of our framework on any hardware and for any application.

In this paper, we propose a real-time human-robot interface (HRI) system, where Electrooculography (EOG) and Electromyography (EMG) signals were decoded to perform reach-to-grasp movements. For that, five different eye movements (up, down, left, right and rest) were classified in real-time and translated into commands to steer an industrial robot (UR-10) to one of the four approximate target directions. Thereafter, EMG signals were decoded to perform the grasping task using an attached gripper to the UR-10 robot arm. The proposed system was tested offline on three different healthy subjects, and mean validation accuracy of 93.62% and 99.50% were obtained across the three subjects for EOG and EMG decoding, respectively. Furthermore, the system was successfully tested in real-time with one subject, and mean online accuracy of 91.66% and 100% were achieved for EOG and EMG decoding, respectively. Our results obtained by combining real-time decoding of EOG and EMG signals for robot control show overall the potential of this approach to develop powerful and less complex HRI systems. Overall, this work provides a proof-of-concept for successful real-time control of robot arms using EMG and EOG signals, paving the way for the development of more dexterous and human-controlled assistive devices.

Simultaneous Localization and Mapping (SLAM) is one of the basic problems in mobile robotics. While most approaches are based on occupancy grid maps, Normal Distributions Transforms (NDT) and mixtures like Occupancy Normal Distribution Transforms (ONDT) have been shown to represent sensor measurements more accurately. In this work, we slightly re-formulate the (O)NDT matching function such that it becomes a least squares problem that can be solved with various robust numerical and analytical non-linear optimizers. Further, we propose a novel global (O)NDT scan matcher for loop closure. In our evaluation, our NDT and ONDT methods are able to outperform the occupancy grid map based ones we adopted from Google's Cartographer implementation.

Human-multi-robot collaboration is becoming more and more common in intelligent manufacturing. Optimal assembly scheduling of such systems plays a critical role in their production efficiency. Existing approaches mostly consider humans as agents with assumed or known capabilities, which leads to suboptimal performance in realistic applications where human capabilities usually change. In addition, most robot adaptation focuses on human-single-robot interaction and the adaptation in human-multi-robot interaction with changing human capability still remains challenging due to the complexity of the heterogeneous multi-agent interactions. This paper proposes a real-time adaptive assembly scheduling approach for human-multi-robot collaboration by modeling and incorporating changing human capability. A genetic algorithm is also designed to derive implementable solutions for the formulated adaptive assembly scheduling problem. The proposed approaches are validated through different simulated human-multi-robot assembly tasks and the results demonstrate the effectiveness and advantages of the proposed approaches.