ICRA 2019 论文列表

Perception of the local environment is a precondition for mobile robots to navigate safely in dynamic environments. Most robots, i.e., humanoids and smaller wheeled robots rely on planar regions. For humanoids, a simple 2D occupancy map as environment representation on which a path is planned is hereby not sufficient since they can step over and onto objects and therefore need height information. Considering dynamic obstacles introduces another level of complexity, since they can lead to necessary replanning or collisions at later stages. In this paper, we present a framework that first extracts planar regions in height maps and detects dynamic obstacles. Our system then uses this information to create a set of prediction maps, in which paths can be efficiently planned in real time at low CPU cost. We show in simulation and real-world experiments that our framework keeps run times well under 10ms for one computation cycle and allows for foresighted real-time 3D footstep planning.

In several applications, autonomous mobile robots benefit from knowing the structure of the indoor environments where they operate. This knowledge can be extracted from the metric maps built (e.g., using SLAM algorithms) from the data perceived by the robots' sensors. The layout is a way to represent the structure of an indoor environment with geometrical primitives. Most of the current methods for reconstructing the layout from a metric map represent the parts of the environment that have been fully observed. In this paper, we propose an approach that predicts the layout of rooms which are only partially known in a 2D metric grid map. The prediction is made according to the global structure of the environment, as identified from its known parts. Experiments show that our approach is able to effectively predict the layout of several indoor environments that have been observed to different degrees.

Anticipating possible behaviors of traffic participants is an essential capability of autonomous vehicles. Many behavior detection and maneuver recognition methods only have a very limited prediction horizon that leaves inadequate time and space for planning. To avoid unsatisfactory reactive decisions, it is essential to count long-term future rewards in planning, which requires extending the prediction horizon. In this paper, we uncover that clues to vehicle behaviors over an extended horizon can be found in vehicle interaction, which makes it possible to anticipate the likelihood of a certain behavior, even in the absence of any clear maneuver pattern. We adopt a recurrent neural network (RNN) for observation encoding, and based on that, we propose a novel vehicle behavior interaction network (VBIN) to capture the vehicle interaction from the hidden states and connection feature of each interaction pair. The output of our method is a probabilistic likelihood of multiple behavior classes, which matches the multimodal and uncertain nature of the distant future. A systematic comparison of our method against two state-of-the-art methods and another two baseline methods on a publicly available real highway dataset is provided, showing that our method has superior accuracy and advanced capability for interaction modeling.

This paper describes recent developments of a robotic hummingbird project, aimed at achieving precision stationary hovering. To this end, the early version of our flapping mechanism is modified which, besides being more efficient, reduces significantly the asymmetry of the wing trajectory of the previous version. A cascade control strategy is used to compensate for the residual parasitic torques and the misalignment of the lift vector and the autopilot.

This paper focuses on important implementation issues of numerical optimal control that are often overlooked. In particular, transcription methods should be carefully implemented for obtaining a discrete representation of the problem. For this purpose, we explain the algorithms to solve the minimum-effort problem by applying a direct collocation method based on B-Splines. In addition, we describe how to compute the gradient of the objective function as well as the Jacobian of the constraints without the use of finite differences and automatic differentiation. Geometric algorithms based on Lie groups and Lie algebra are examined to efficiently compute the analytical derivatives of the equations of motion of articulated robots. These ingredients allow the fast computation of dynamically feasible robot motions. We provide numerical comparisons with a biped robot to validate our recipe against classical direct collocation methods.

In response to an abundance of applications, Unmanned Aerial Vehicles are being called upon to perform missions of high difficulty for increasingly long periods of time. Traditional paradigms of propeller design and actuation are reaching a design ceiling, motivating creative approaches to the design of propeller-based propulsion mechanisms. Within the last decade, one particular kind of mechanism, the variable-pitch propeller, has been studied by researchers for its applications to the class of small UAVs. This paper pushes for new results in this area by exploring the use of Variable Pitch Propulsion (VPP) to minimize power consumption for small, versatile UAVs. A control algorithm is presented to minimize the consumed electrical power during a quasi-steady propulsive state. In particular, the algorithm is not confined to operation in limited regions of the state space, but it seeks to minimize power at whatever point in the state space a steady state is reached. Several experimental results are presented to validate the approach.

Key design and operation parameters of the system for magnetically guided actuation of miniature robots on planar surfaces are analyzed and discussed. The study is carried out on the numerical analysis and also on the experimental measurement on the prototype of the system. Special attention is paid to robot actuation under uncertainty, which can be caused by both external and internal effects. A technique based on a superposition of actuation and lock-up field is proposed.

Non-linearities in cable transmissions are important limitations for the accurate control of flexible instruments used in medical endoscopic systems. Hysteresis effects greatly impact the accuracy of conventional kinematic models. This is especially critical for implementing automatic motions in flexible medical robotic systems. In this paper, we propose a method for improving open-loop accuracy of flexible instruments by implementing a Position Inverse Kinematic Model which is able to take into account hysteresis effects. In order to avoid complex physical modeling, the method relies on the off-line learning of the behavior of the instruments. Basic knowledge of the kinematic is also incorporated in the learning process in order to make it fast. The validity of the approach is demonstrated by the execution of 2D and 3D trajectories with the instruments of the STRAS medical robot. The accuracy is shown to be significantly improved with respect to other learning-based methods.

We present a novel pose and posture estimation framework of aerial skeleton system for outdoor flying. To exploit redundant/independent sensing while rendering the system “modular”, we attach an IMU (inertial measurement unit) sensor and a GNSS (global navigation satellite system) module on each link and perform SE(3)-motion EKF (extended Kalman filtering). We then apply the kinematic constraints of the aerial skeleton system to these EKF estimates of all the links through SCKF (smoothly constrained Kalman filtering), thereby, enforcing the kinematic coherency of the skeleton system and, consequently, significantly enhancing the estimation accuracy and the control performance/stability of the aerial skeleton system. A semi-distributed version of the obtained estimation framework is also presented to address the issue of scalability. The theory is then verified/demonstrated with real outdoor flying experiments and simulation studies of a three-link aerial skeleton system.

Unsupervised Learning based monocular visual odometry (VO) has lately drawn significant attention for its potential in label-free leaning ability and robustness to camera parameters and environmental variations. However, partially due to the lack of drift correction technique, these methods are still by far less accurate than geometric approaches for large-scale odometry estimation. In this paper, we propose to leverage graph optimization and loop closure detection to overcome limitations of unsupervised learning based monocular visual odometry. To this end, we propose a hybrid VO system which combines an unsupervised monocular VO called NeuralBundler with a pose graph optimization back-end. NeuralBundler is a neural network architecture that uses temporal and spatial photometric loss as main supervision and generates a windowed pose graph consists of multi-view 6DoF constraints. We propose a novel pose cycle consistency loss to relieve the tensions in the windowed pose graph, leading to improved performance and robustness. In the back-end, a global pose graph is built from local and loop 6DoF constraints estimated by NeuralBundler, and is optimized over SE(3). Empirical evaluation on the KITTI odometry dataset demonstrates that 1) NeuralBundler achieves state-of-the-art performance on unsupervised monocular VO estimation, and 2) our whole approach can achieve efficient loop closing and show favorable overall translational accuracy compared to established monocular SLAM systems.

In this paper, we propose an end-to-end grasp evaluation model to address the challenging problem of localizing robot grasp configurations directly from the point cloud. Compared to recent grasp evaluation metrics that are based on handcrafted depth features and a convolutional neural network (CNN), our proposed PointNetGPD is lightweight and can directly process the 3D point cloud that locates within the gripper for grasp evaluation. Taking the raw point cloud as input, our proposed grasp evaluation network can capture the complex geometric structure of the contact area between the gripper and the object even if the point cloud is very sparse. To further improve our proposed model, we generate a large-scale grasp dataset with 350k real point cloud and grasps with the YCB object set for training. The performance of the proposed model is quantitatively measured both in simulation and on robotic hardware. Experiments on object grasping and clutter removal show that our proposed model generalizes well to novel objects and outperforms state-of-the-art methods. Code and video are available at https://lianghongzhuo.github.io/PointNetGPD.

The use of 3D LiDAR, which has proven its capabilities in autonomous driving systems, is now expanding into many other fields. The sharing and transmission of point cloud data from 3D LiDAR sensors has broad application prospects in robotics. However, due to the sparseness and disorderly nature of this data, it is difficult to compress it directly into a very low volume. A potential solution is utilizing raw LiDAR data. We can rearrange the raw data from each frame losslessly in a 2D matrix, making the data compact and orderly. Due to the special structure of 3D LiDAR data, the texture of the 2D matrix is irregular, in contrast to 2D matrices of camera images. In order to compress this raw, 2D formatted LiDAR data efficiently, in this paper we propose a method which uses a recurrent neural network and residual blocks to progressively compress one frame's information from 3D LiDAR. Compared to our previous image compression based method and generic octree point cloud compression method, the proposed approach needs much less volume while giving the same decompression accuracy. Potential application scenarios for point cloud compression are also considered in this paper. We describe how decompressed point cloud data can be used with SLAM (simultaneous localization and mapping) as well as for localization using a given map, illustrating potential uses of the proposed method in real robotics applications.

When performing some surgical tasks inside colon with NOTES technology, colon tissue could block the task space and occlude the endoscopic vision. In order to solve this problem, we developed a pneumatically driven deployable and undeployable structure which can distend collapsing tissue and can be delivered through a 4.5 mm endoscopic channel. The structure is designed to be flexible enough to pass through colon's tortuous pathway. Also, it is designed to hold its shape without continually applying air pressure after deployment. This allows to make use of an endoscopic channel for the other surgical instruments. Besides, due to the compliant nature of the device, it is safe to deploy inside a smaller space than its maximum deployable size. The functionality of the device was verified with an in-vitro experiment. The structure was successfully deployed inside a pig's colon with an inner diameter of 60 mm by applying 3.5 bars of air pressure and created a sufficient task space for surgical operations.

We propose a novel plug-and-play (PnP) module for improving depth prediction with taking arbitrary patterns of sparse depths as input. Given any pre-trained depth prediction model, our PnP module updates the intermediate feature map such that the model outputs new depths consistent with the given sparse depths. Our method requires no additional training and can be applied to practical applications such as leveraging both RGB and sparse LiDAR points to robustly estimate dense depth map. Our approach achieves consistent improvements on various state-of-the-art methods on indoor (i.e., NYU-v2) and outdoor (i.e., KITTI) datasets. Various types of LiDARs are also synthesized in our experiments to verify the general applicability of our PnP module in practice.

This paper addresses the problem of using autonomous robots to record events that obey narrative structure. The work is motivated by a vision of robot teams that can, for example, produce individualized highlight videos for each runner in a large-scale road race such as a marathon. We introduce a method for specifying the desired structure as a function that describes how well the captured events can be used to produce an output that meets the specification. This function is specified in a compact, legible form similar to a weighted finite automaton. Then we describe a planner that uses simple predictions of future events to coordinate the robots' efforts to capture the most important events, as determined by the specification. We describe an implementation of this approach, and demonstrate its effectiveness in a simulated race scenario both in simulation and in a hardware testbed.

Online action classification is an important field of research, enabling the particularly interesting application scenario of controlling wearable devices which actively support the user's motions. The majority of machine learning applications of real-world systems are based on pre-trained average-user models without any personalization. Our long-term goal is to provide a system that adapts to its user's personal behavior patterns on the fly and in real-time. Ideally, we want to initiate a continuous collaboration between the system and the user where both alternatively adjust to each other to maximize the system's utility. Such tasks are not feasible with static models. In this paper, we investigate the potential and benefits of personalized online learning in the task of online action classification. We record motion sequences of different subjects wearing the Xsens bodysuit, which incorporates multiple inertial measuring units, enabling a fine-grained discrimination of motions. On this basis, we first perform a feature selection, showing that only a few sensors are necessary to achieve a high classification performance. Subsequently, we compare the recognition capabilities of offline average user models against personalized models trained in an online way. Our experiments conclude that personalized models require only few data to outperform average user systems and are particularly valuable for applications with limited computational hardware which rely on the raw sensor inputs only.

This paper proposes a method to deploy teams of robots with constrained energy capacities to persistently maintain a map of an uncertain environment. Typical occupancy map approaches assume a static world; however, we introduce a decay in confidence that degrades the occupancy probability of grid cells and promotes revisitation. Further, sections of the map whose occupancy differs between observations are visited more frequently, while unchanging areas are scheduled less frequently. While naive planning is intractable through the entire space of multi-agent spatio-temporal states, the proposed algorithm decouples planning such that constraints are resolved separately by solving tracTable subproblems. We evaluate this approach in simulation and show how the uncertainty of our world model is maintained below an acceptable threshold while the algorithm retains a tractable computation time.

Wireless magnetic microrobots can perform complex tasks for small-scale minimally invasive surgery (MIS) that requires high precision and dexterity. The choice of the configuration of the electromagnetic actuation (EMA) system is a key issue for reliable medical applications. This paper addresses the study of a robotic EMA platform firstly devoted to ophthalmic MIS, aiming at improving the manipulability and dexterity of the procedure. To this end, a robotic EMA system comprising four static and four mobile electromagnets is investigated. Evaluation of the magnetic force and torque, the manipulability and the dexterity indexes of EMA platforms are studied. The results demonstrate that a robotic EMA platform increases the versatility of the EMA system, and becomes resourceful to perform various tasks.

We present a Pedestrian Dominance Model (PDM) to identify the dominance characteristics of pedestrians for robot navigation. Through a perception study on a simulated dataset of pedestrians, PDM models the perceived dominance levels of pedestrians with varying motion behaviors corresponding to trajectory, speed, and personal space. At runtime, we use PDM to identify the dominance levels of pedestrians to facilitate socially-aware navigation for the robots. PDM can predict dominance levels from trajectories with ~85% accuracy. Prior studies in psychology literature indicate that when interacting with humans, people are more comfortable around people that exhibit complementary movement behaviors. Our algorithm leverages this by enabling the robots to exhibit complementing responses to pedestrian dominance. We also present an application of PDM for generating dominance-based collision-avoidance behaviors in the navigation of autonomous vehicles among pedestrians. We demonstrate the benefits of our algorithm for robots navigating among tens of pedestrians in simulated environments.