IROS 2020 论文列表

In this work, we focus on developing algorithms to maintain and enhance the connectivity of a multi-robot system with minimal disruption to the primary tasks that the robots are performing. Such algorithms are useful for collaborating robots to be resilient to reduction in connectivity of the communication graph of the robot team when robots can arrive or leave. These algorithms are also useful in a supervisory control setting when an operator wants to enhance the connectivity of the robot team. In contrast to many existing works that can only maintain the current connectivity of the multi-robot graph, we propose a generalized connectivity control framework that allows for reconfiguration of the multi-robot system to provably satisfy any connectivity demand, while minimally disrupting the execution of their original tasks. In particular, we propose a novel k-Connected Minimum Resilient Graph (k-CMRG) algorithm to compute an optimal k-connectivity graph that minimally constrains the robots' original task-related motion, and employ the Finite-Time Convergence Control Barrier Function (FCBF) to enforce the pairwise robot motion constraints defined by the edges of the graph. The original controllers are minimally modified to drive the robots and form the k-CMRG. We demonstrate the effectiveness of our approach via simulations in the presence of multiple tasks and robot failures.

Mobile robots of various types have been proposed for infrastructure inspection and disaster investigation. For such mobile robot applications, accessing the areas is of primary importance for missions. Therefore, various locomotive mechanisms have been studied. We introduce a novel mobile robot system, named DIR-3, combining a crawler robot and a microdrone. By rotating its arm back and forth, DIR-3, a very simple, lightweight crawler robot with a single 360-degree rotatable U-shaped arm, can climb up/down an 18 cm high step, 1.5 times its height.Furthermore, to inspect high places, which is considered difficult for conventional mobile robots, a drone mooring system for mobile robots is presented. The tethered microdrone of DIR-3 can be controlled freely as a flying camera by switching operating modes on the graphic user interface. The drone mooring system has a unique tension-controlled winding mechanism that enables stable landing on DIR-3 from any location in the air, in addition to measurement and estimation of relative positions of the drone. We evaluated the landing capability, position estimation accuracy, and following control of the drone using the winding mechanism. Results show the feasibility of the proposed system for inspection of cracks in a 5 m high concrete wall.

Robotic insertion tasks are characterized by contact and friction mechanics, making them challenging for conventional feedback control methods due to unmodeled physical effects. Reinforcement learning (RL) is a promising approach for learning control policies in such settings. However, RL can be unsafe during exploration and might require a large amount of real-world training data, which is expensive to collect. In this paper, we study how to use meta-reinforcement learning to solve the bulk of the problem in simulation by solving a family of simulated industrial insertion tasks and then adapt policies quickly in the real world. We demonstrate our approach by training an agent to successfully perform challenging real-world insertion tasks using less than 20 trials of real-world experience.

Personal robots and driverless cars need to be able to operate in novel environments and thus quickly and efficiently learn to recognise new object classes. We address this problem by considering the task of video object segmentation. Previous accurate methods for this task finetune a model using the first annotated frame, and/or use additional inputs such as optical flow and complex post-processing. In contrast, we develop a fast, causal algorithm that requires no finetuning, auxiliary inputs or post-processing, and segments a variable number of objects in a single forward-pass. We represent an object with clusters, or "visual words", in the embedding space, which correspond to object parts in the image space. This allows us to robustly match to the reference objects throughout the video, because although the global appearance of an object changes as it undergoes occlusions and deformations, the appearance of more local parts may stay consistent. We learn these visual words in an unsupervised manner, using meta-learning to ensure that our training objective matches our inference procedure. We achieve comparable accuracy to finetuning based methods (whilst being 1 to 2 orders of magnitude faster), and state-of-the-art in terms of speed/accuracy trade-offs on four video segmentation datasets. Code is available at https://github.com/harkiratbehl/MetaVOS.

Video object segmentation plays a vital role to many robotic tasks, beyond the satisfied accuracy, quickly adapt to the new scenario with very limited annotations and conduct a quick inference are also important. In this paper, we are specifically concerned with the task of fast segmenting all pixels of a target object in all frames, given the annotation mask in the first frame. Even when such annotation is available, this remains a challenging problem because of the changing appearance and shape of the object over time. In this paper, we tackle this task by formulating it as a meta-learning problem, where the base learner grasping the semantic scene understanding for a general type of objects, and the meta learner quickly adapting the appearance of the target object with a few examples. Our proposed meta-learning method uses a closed form optimizer, the so-called "ridge regression", which has been shown to be conducive for fast and better training convergence. Moreover, we propose a mechanism, named "block splitting", to further speed up the training process as well as to reduce the number of learning parameters. In comparison with the state-of-the art methods, our proposed framework achieves significant boost up in processing speed, while having highly comparable performance compared to the best performing methods on the widely used datasets. Video demo can be found here 1.

In this study, a passive arm-support exoskeleton was designed to provide assistive aid for manufacturing workers. The exoskeleton has two operating states which can be altered using an unique ratchet bar mechanism with two blocks fixed on the ratchet bar. When the upper arm is elevated to the highest poiont, the pawl module will touch the lower block to allow the pawl separated, so that the arm can move freely without any resistance. When the upper arm is depressed to the lowest point, the pawl module will touch the upper block to make the pawl re-engaged, so that the upper arm can be locked at any vertical position. For purpose to improve the ergonomical property, the structural parameters of the exoskeleton were determined by particle swarm optimization. The designed exoskeleton was simulated in the Adams model to investigate its actual performance. A preliminary experimental study was conducted to evaluate the effectiveness of the designed exoskeleton on alleviating users' physical loads in holding heavy tools; the muscular activities on the shoulder muscle groups involved in the weights bearing, elicited by the surface electromyography (EMG) over the shoulder, were significantly reduced from three healthy subjects who carried hand-held tools. The simulation and experiment results show that the designed exoskeleton could effectively relieve the shoulder burden by transferring the bearing load to the waist, where the motion of the arm was not obstructed.

Recent advancements in Learning from Human Feedback present an effective way to train robot agents via inputs from non-expert humans, without a need for a specially designed reward function. However, this approach needs a human to be present and attentive during robot learning to provide evaluative feedback. In addition, the amount of feedback needed grows with the level of task difficulty and the quality of human feedback might decrease over time because of fatigue. To overcome these limitations and enable learning more robot tasks with higher complexities, there is a need to maximize the quality of expensive feedback received and reduce the amount of human cognitive involvement required. In this work, we present an approach that uses active learning to smartly choose queries for the human supervisor based on the uncertainty of the robot and effectively reduces the amount of feedback needed to learn a given task. We also use a novel multiple buffer system to improve robustness to feedback noise and guard against catastrophic forgetting as the robot learning evolves. This makes it possible to learn tasks with more complexity using lesser amounts of human feedback compared to previous methods. We demonstrate the utility of our proposed method on a robot arm reaching task where the robot learns to reach a location in 3D without colliding with obstacles. Our approach is able to learn this task faster, with less human feedback and cognitive involvement, compared to previous methods that do not use active learning.

Most path planning algorithms for covering a complex 3D object ignore physical limitations or constraints on a robot's motion. Adhering to such constraints for a given path can slow down the time to cover the path because the motion may need to be adjusted. This work considers a scenario in computer numerical control (CNC) milling applications, where the robot is a cutting tool that needs to cover the surface of a complex 3D object under the following constraint: for every point on the generated path, the robot must be assigned an accessible orientation to avoid collisions between it and other parts of the object. Our proposed approach, which we call max orientation coverage, employs a two-step optimization scheme. It can improve path efficiency with respect to both the length of the path and the cost of dealing with the collision-avoiding constraints. We evaluate our approach through extensive simulation studies on four CAD benchmarks against a state-of-the-art baseline. We show that our proposed approach can improve the efficiency of the path by 29.7% on average compared with the baseline and the improvement goes up to 46.5% for certain complex objects.

In this paper, we propose an autonomous exploration and a tapping mechanism-based material mapping system for a mobile robot in unknown environments. The goal of the proposed system is to integrate simultaneous localization and mapping (SLAM) modules and sound-based material classification to enable a mobile robot to explore an unknown environment autonomously and at the same time identify the various objects and materials in the environment. This creates a material map that localizes the various materials in the environment which has potential applications for search and rescue scenarios. A tapping mechanism and tapping audio signal processing based on machine learning techniques are exploited for a robot to identify the objects and materials. We demonstrate the proposed system through experiments using a mobile robot platform installed with Velodyne LiDAR, a linear solenoid, and microphones in an exploration-like scenario with various materials. Experiment results demonstrate that the proposed system can create useful material maps in unknown environments.

When exploring an unknown environment, a mobile robot must decide where to observe next. It must do this whilst minimising the risk of failure, by only exploring areas that it expects to be safe. In this context, safety refers to the robot remaining in regions where critical environment features (e.g. terrain steepness, radiation levels) are within ranges the robot is able to tolerate. More specifically, we consider a setting where a robot explores an environment modelled with a Markov decision process, subject to bounds on the values of one or more environment features which can only be sensed at runtime. We use a Gaussian process to predict the value of the environment feature in unvisited regions, and propose an estimated Markov decision process, a model that integrates the Gaussian process predictions with the environment model transition probabilities. Building on this model, we propose an exploration algorithm that, contrary to previous approaches, considers probabilistic transitions and explicitly reasons about the uncertainty over the Gaussian process predictions. Furthermore, our approach increases the speed of exploration by selecting locations to visit further away from the currently explored area. We evaluate our approach on a real-world gamma radiation dataset, tackling the challenge of a nuclear material inspection robot exploring an a priori unknown area.

One of the main challenges of the active assistive devices is how to estimate the motion of the missing/impaired limbs and joints in line with the remaining limbs. To do so, a motion planner is required. This study proposes a motion planner that can be used for active prosthetic/orthotic knees. The aim is to continuously estimate the knee joint positions based on the thigh motion, using as few inputs as possible. Data from thigh-mounted IMU (thigh acceleration and angle) are used as inputs to estimate knee joint positions as outputs. It is aimed to continuously estimate the outputs as opposed to the state-machine approaches which divide the gait cycles into different sections and require switching rules. The performance of the motion planner is investigated for five walking speeds (0.6, 0.9, 1.2, 1.4 and 1.6 m/s). The strengths and limitations of the motion planer are investigated at different scenarios.

Significant progress has been made with artistic robots. However, existing robots fail to produce high-quality portraits in a short time. In this work, we present a drawing robot, which can automatically transfer a facial picture to a vivid portrait, and then draw it on paper within two minutes averagely. At the heart of our system is a novel portrait synthesis algorithm based on deep learning. Innovatively, we employ a self-consistency loss, which makes the algorithm capable of generating continuous and smooth brush-strokes. Besides, we propose a componential sparsity constraint to reduce the number of brush-strokes over insignificant areas. We also implement a local sketch synthesis algorithm, and several pre- and post-processing techniques to deal with the background and details. The portrait produced by our algorithm successfully captures individual characteristics by using a sparse set of continuous brush-strokes. Finally, the portrait is converted to a sequence of trajectories and reproduced by a 3-degree-of-freedom robotic arm. The whole portrait drawing robotic system is named AiSketcher. Extensive experiments show that AiSketcher can produce considerably high-quality sketches for a wide range of pictures, including faces in-the-wild and universal images of arbitrary content. To our best knowledge, AiSketcher is the first portrait drawing robot that uses neural style transfer techniques. AiSketcher has attended a quite number of exhibitions and shown remarkable performance under diverse circumstances.

In this paper, we consider the problem of distributed pose graph optimization (PGO) that has extensive applications in multi-robot simultaneous localization and mapping (SLAM). We propose majorization minimization methods for distributed PGO and show that our methods are guaranteed to converge to first-order critical points under mild conditions. Furthermore, since our methods rely a proximal operator of distributed PGO, the convergence rate can be significantly accelerated with Nesterov's method, and more importantly, the acceleration induces no compromise of convergence guarantees. In addition, we also present accelerated majorization minimization methods for the distributed chordal initialization that have a quadratic convergence, which can be used to compute an initial guess for distributed PGO. The efficacy of this work is validated through applications on a number of 2D and 3D SLAM datasets and comparisons with existing state-of-the- art methods, which indicates that our methods have faster convergence and result in better solutions to distributed PGO.

One of the key advantages of robots is the high speeds at which they can operate. In industrial settings, increased velocities can lead to higher throughputs and improved efficiency. Some manipulation tasks might require the robot to perform highly dynamic operations such as shaking, or swinging while grasping an object. These fast movements may produce high accelerations and thus give rise to inertial forces that can cause a grasped object to slip. In this paper a method is proposed to determine the inertial forces that arise on a grasped object during a trajectory, find the instances at which the object might slip, and avoid these slippages by changing the trajectory, namely the orientation of the object. To exemplify the usage of this approach, two grasping tasks are realised: a prehensile and a non-prehensile grasp, and strategies to successfully perform these tasks without changing the overall duration of the trajectory are defined and evaluated.

In this paper, we present a magnetized cell-robot using macrophages as templates, which can be controlled under a strong gradient magnetic field, to approach and kill cancer cells in both vitro and vivo environment. Firstly, we establish a magnetic control system using only four coils which can generate gradient field up to 4.14 T/m utilizing the coupled field contributed by multiple electromagnets acting in concert. Most importantly, the cell-robot which is based on the macrophage is proposed, and can be transported to the vicinity of cancer cells precisely using strong gradient magnetic field. Then the cell-robot will actively phagocytose the cancer cells and eventually kill them, achieving the cancer treatment at the cellular level. It has important significance for guiding accurate targeted therapy in vivo for the future, under the premise of zero harm to the human body.

The modular assembly and actuation of 3D printed milliscale cuboid robots using a globally applied magnetic field is presented. Cuboids are composed of a rectangular resin shell embedded with two spherical permanent magnets that can independently align with any applied magnetic field. Placing cuboids within short distances of each other allows for modular assembly and disassembly by changing magnetic field direction. Assembled cuboids are demonstrated to stably self-propel under sequential field inputs allowing for both rolling and pivot walking motion modes. Swarms of cuboids could be actuated within the working space and exhibit near identical behavior. Specialized `trap robots' were developed to capture objects, transport them within the working space, and subsequently release the payload in a new location. Cuboids with male and female connectors were developed to exhibit the selective mating between cuboids. The results show that cuboids are a diverse and adaptable platform that has the potential to be scaled down to the sub-millimeter regime for use in medical or small-scale assembly applications.

Tissue engineering is trying to use modular tissue micro-rings to construct artificial biological microtubes as substitute of autologous tissue tubes to alleviate the shortage of donor sources. However, because of the lack of effective assembly strategies, it is still challenging to achieve high-speed fabrication of biological microtubes with high cell density. In this paper, we proposed a robotic-based magnetic assembly strategy to handle this challenge. We first encapsulated magnetic alginate microfibers into micro-rings formed by cell self-assembly to enhance the controllability. Afterwards, a 3D long-stroke manipulator with visual servoing system was designed to achieve magnetic pick-and-place operations of micro-rings for 3D assembly. Moreover, we developed a mathematical model of the motion of micro-ring in solution environments. Based on visual feedback, we analyzed the feasibility of automatic assembly and following response of micro-rings with the moving magnets, which shows our proposed method has great potential to achieve high-speed bio-assembly. Finally, we successfully assembled multi-micro-rings into a biological microtube with high cell density.

Autonomous driving requires the inference of actionable information such as detecting and classifying objects, and determining the drivable space. To this end, we present Multi-View LidarNet (MVLidarNet), a two-stage deep neural network for multi-class object detection and drivable space segmentation using multiple views of a single LiDAR point cloud. The first stage processes the point cloud projected onto a perspective view in order to semantically segment the scene. The second stage then processes the point cloud (along with semantic labels from the first stage) projected onto a bird's eye view, to detect and classify objects. Both stages use an encoder-decoder architecture. We show that our multi-view, multi-stage, multi-class approach is able to detect and classify objects while simultaneously determining the drivable space using a single LiDAR scan as input, in challenging scenes with more than one hundred vehicles and pedestrians at a time. The system operates efficiently at 150 fps on an embedded GPU designed for a self-driving car, including a postprocessing step to maintain identities over time. We show results on both KITTI and a much larger internal dataset, thus demonstrating the method's ability to scale by an order of magnitude.

In this study, a deep-learning-based multi-stage network architecture called Multi-Stage Depth Prediction Network (MSDPN) is proposed to predict a dense depth map using a 2D LiDAR and a monocular camera. Our proposed network consists of a multi-stage encoder-decoder architecture and Cross Stage Feature Aggregation (CSFA). The proposed multi-stage encoder-decoder architecture alleviates the partial observation problem caused by the characteristics of a 2D LiDAR, and CSFA prevents the multi-stage network from diluting the features and allows the network to learn the interspatial relationship between features better. Previous works use sub-sampled data from the ground truth as an input rather than actual 2D LiDAR data. In contrast, our approach trains the model and conducts experiments with a physically-collected 2D LiDAR dataset. To this end, we acquired our own dataset called KAIST RGBD-scan dataset and validated the effectiveness and the robustness of MSDPN under realistic conditions. As verified experimentally, our network yields promising performance against state-of-the-art methods. Additionally, we analyzed the performance of different input methods and confirmed that the reference depth map is robust in untrained scenarios.