IROS 2017 论文列表

This paper presents the mechatronic design of a robotic arm that is mounted on a ground rover and used to deploy and recover small-scale unmanned aerial vehicles. The arm and rover are part of a network of collaborative robotic agents aiming to enhance current rescue operations by supporting human operators without burdening them with servicing tasks. The robust autonomy of the system, guaranteed in part through the addition of this robotic arm, is a main contributing factor. Design requirements are derived from the context of the rescue mission and a kinematic analysis is provided that leads to a customized design, including variable stiffness joints for compliant interaction with the environment. Experiments demonstrate the system's ability to perform the required Cartesian trajectory control and manipulation tasks, and to achieve a desired variable end-effector compliance.

The use of magnetic resonance imaging (MRI) for guiding robotic surgical devices has shown great potential for performing precisely targeted and controlled interventions. To fully realize these benefits, devices must work safely within the tight confines of the MRI bore without negatively impacting image quality. Here we expand on previous work exploring MRI guided robots for neural interventions by presenting the mechanical design and assessment of a device for positioning, orienting, and inserting an interstitial ultrasound-based ablation probe. From our previous work we have added a 2 degree of freedom (DOF) needle driver for use with the aforementioned probe, revised the mechanical design to improve strength and function, and performed an evaluation of the mechanism's accuracy and effect on MR image quality. The result of this work is a 7-DOF MRI robot capable of positioning a needle tip and orienting it's axis with accuracy of 1.37 ± 0.06mm and 0.79° ± 0.41°, inserting it along it's axis with an accuracy of 0.06 ± 0.07mm, and rotating it about it's axis to an accuracy of 0.77° ± 1.31°. This was accomplished with no significant reduction in SNR caused by the robot's presence in the MRI bore, <; 10.3% reduction in SNR from running the robot's motors during a scan, and no visible paramagnetic artifacts.

Human attention processes play a major role in the optimization of human-robot interaction (HRI) systems. This work describes a novel methodology to measure and predict situation awareness and from this overall performance from gaze features in real-time. The awareness about scene objects of interest is described by 3D gaze analysis using data from wearable eye tracking glasses and a precise optical tracking system. A probabilistic framework of uncertainty considers coping with measurement errors in eye and position estimation. Comprehensive experiments on HRI were conducted with typical tasks including handover in a lab based prototypical manufacturing environment. The methodology is proven to predict standard measures of situation awareness (SAGAT, SART) as well as performance in the HRI task in real-time and will open new opportunities for human factors based performance optimization in HRI applications.

For intelligent robots to interact in meaningful ways with their environment, they must understand both the geometric and semantic properties of the scene surrounding them. The majority of research to date has addressed these mapping challenges separately, focusing on either geometric or semantic mapping. In this paper we address the problem of building environmental maps that include both semantically meaningful, object-level entities and point- or mesh-based geometrical representations. We simultaneously build geometric point cloud models of previously unseen instances of known object classes and create a map that contains these object models as central entities. Our system leverages sparse, feature-based RGB-D SLAM, image-based deep-learning object detection and 3D unsupervised segmentation.

The performance of a master-slave robotic system depends significantly on the ergonomics and the capability of its master device to correctly interface the user with the slave robot. Master manipulators generating commands in task space represent a commonly adopted solution for controlling a range of slave robots while retaining an ergonomic design. However, these devices present several drawbacks, such as requiring the use of clutching mechanics to compensate for the mismatch between slave and master workspaces, and the lack of capability to intuitively transmit important information such as specific joint limits to the user. In this paper, a novel joint-space master manipulator is presented. This manipulator emulates the kinematic structure of highly flexible surgical instruments which it is designed to control. This system uses 6 active degrees of freedom to compensate for its own weight, as well as to provide force feedback corresponding to the slave robot's joint limits. A force/torque sensor integrated at the end effector is used to relay user-generated forces and torques directly to specific joints. This is performed to counteract the friction stemming from structural constraints imposed by the kinematic design of the instruments. Finally, a usability study is carried out to test the validity of the system, proving that the instruments can be intuitively controlled even at the extremities of the workspace.

Visual-Inertial SLAM methods have become a very important technology for several applications in robotics. This kind of approach usually is composed by sensors as rate gyros, accelerometers and monocular cameras. Magnetometers and GPS modules generally used for outdoors are absent in the SLAM system observation, since the magnetometer measurements deteriorate in the presence of ferromagnetic materials and the GPS module signals are unavailable indoors or in urban environments. In order to make use of all these sensors, we propose Markovian jump linear systems (MJLS) to model the modes of operation of the navigation system based on available sensors and their reliability. An extended Kalman filter for MJLS fuses the sensor data and estimates the motion using the best mode of operation for each particular time instant. Experimental results are presented to show the effectiveness of the proposed method, in situations that would pose a challenge for standard data fusion techniques.

This work focuses on multi-robot coordination based on the Hoplites framework for solving the Multi-Robot Task Allocation (MRTA) problem. In particular, we investigate three variations of increasing complexity for the MRTA problem: spatial task allocation based on distance, spatial task allocation based on time and distance, and persistent coverage. The Fast Marching Method (FMM) has been used for robot path planning and providing estimates of the plans that robots bid on, in the context of the market. The use of this framework for solving the persistent coverage problem provides interesting insights by taking a high-level approach that is different from the commonly used solutions to this problem such as computing robot trajectories to keep the desired coverage level. A high fidelity simulation tool, Webots, along with the Robotic Operating System (ROS) have been utilized to provide our simulations with similar complexity to the real-world tests. Results confirm that this pipeline is a very effective tool for our evaluations given that our simulations closely follow the results in reality. By modifying the replanning to prevent having costly or invalid plans by means of priority planning and turn taking, and basing the coordination on maximum plan length as opposed to time, we have been able to make improvements and adapt the Hoplites framework to our applications. The proposed approach is able to solve the spatial task allocation and persistent coverage problems in general. However, there exist some limitations. Particularly, in the case of persistent coverage, this method is suitable for applications where moderate spatial resolutions are sufficient such as patrolling.

Occupancy mapping is an important component of a robot navigation system since it enables intelligent planning of actions to accomplish a task. In some cases, occupancy information is mapped as part of a SLAM system. In other cases, such as visual SLAM, an occupancy map must be separately created, harnessing localization information. We present a novel occupancy mapping method that adaptively and efficiently takes advantage of improved trajectory estimates from the localization system, without requiring storage of all poses of the trajectory or the sensor information from all poses. Occupancy information is stored in local maps anchored to poses in the trajectory. Local maps are created adaptively depending on the uncertainty of localization. The local maps maintain consistency of occupancy information independently of changes in the localization map due to loop closures or convergence of the estimation algorithm. At any time, a global occupancy map (snapshot map) can be rendered from the local occupancy maps. We evaluate the performance of the system using real data and compare it to two baseline methods.

We propose an approach to mapping tissue and vascular systems without the use of contrast agents, based on moving and measuring magnetic particles. To this end, we consider a swarm of particles in a 1D or 2D grid that can be tracked and controlled by an external agent. Control inputs are applied uniformly so that each particle experiences the same applied forces. We present algorithms for three tasks: (1) Mapping, i.e., building a representation of the free and obstacle regions of the workspace; (2) Subset Coverage, i.e., ensuring that at least one particle reaches each of a set of desired locations; and (3) Coverage, i.e., ensuring that every free region on the map is visited by at least one particle. These tasks relate to a large body of previous work from robot navigation, both from theory and practice, which is based on individual control. We provide theoretical insights that have potential relevance for fast MRI scans with magnetically controlled contrast media. In particular, we develop a fundamentally new approach for searching for an object at an unknown distance D, where the search is subject to two different and independent cost parameters for moving and for measuring. We show that regardless of the relative cost of these two operations, there is a simple O(log D/log log D)-competitive strategy, which is the best possible. Also, we provide practically useful and computationally efficient strategies for higher-dimensional settings. These algorithms extend to any number of particles and show that additional particles tend to reduce the mean and the standard deviation of the time required for each task.

Motion planning for manipulators with rotational joints is challenging because the actuation range for each link is constrained by the placement and orientation of other links. Thus, finding paths that avoid self-collision is non-trivial. However, rotational joints are often used in industrial robots. We develop a reparameterization of the planning problem called directed reachable volumes that provides an explicit representation of the workspace regions that the joints and end effectors can reach given the placement and orientation of other links. This formulation, while similar in spirit to prior reachable volume work, does not rely on the same restrictive assumptions that preclude prior work from handling rotational joints. We provide primitive planning operations that can be used in the context of state-of-the-art motion planning methods. We present experimental validation of directed reachable volumes by demonstrating a simulated pick-and-place scenario using realistic robots with rotational joints.

This paper aims to characterise the magnetic interaction in neighbouring sets of local electromagnetic actuation (LEMA) actuators in a robotic platform for abdominal surgery. The analysis looks into the affect of the magnetic fields contributed by a stator-rotor set (the actuation unit) located adjacent to the rotor of interest. Each rotor drives one of the degree-of-freedoms (DOFs) on a surgical robotic device. In this study, a two-DOF setup is used for the magnetic interaction analysis, which can be expanded to general case n-DOF setup with the Principle of Superposition of magnetic fields from multiple sources. The magnetic model is then used to compute the dynamics of the system, which involves the equation of motion of the rotors and associated robotic mechanism it drives, and the actuator (electrical) model that takes into account the back EMF generated by the permanent magnet rotors. The magnetic field effect of the neighbouring set onto the rotor is observed by obtaining the speed response of the rotor through simulation so that the dynamic model can be validated against the experimental results. The outcomes are useful for the design specification of the LEMA system configuration, involving the feasible / pragmatic distance between the stator sets such that the interference is minimised, and for the design of the necessary control strategy.

This paper proposes a flexible descriptor for representing 6-D rigid body motion trajectories, which not only shows strong invariances and descriptive ability but also achieves satisfactory results in both recognition accuracy and efficiency. 6-D rigid body motion trajectories are first transformed into the Multi-layer Self-similarity Matrices (MSM) representation. The MSM is the combination of the square similarity matrices in three layers, which captures both local and global spatiotemporal features of the trajectories. Next, the Histogram of Oriented Gradients (HOG) features extracted from the MSM representation are concatenated as the final MSM-HOG trajectory descriptor. Then we train the Support Vector Machine (SVM) classifier with the linear kernel for multicalss motion recognition. Finally, rigid body motion recognition experiments on two public datasets are conducted to verify the effectiveness and efficiency of the proposed method.

We consider a pursuit-evasion problem between humanoids in the presence of obstacles. In our scenario, the pursuer enters the safety area of the evader headed for collision, while the latter executes a fast evasive motion. Control schemes are designed for both the pursuer and the evader. They are structurally identical, although the objectives are different: the pursuer tries to align its direction of motion with the line-of-sight to the evader, whereas the evader tries to move in a direction orthogonal to the line-of-sight to the pursuer. At the core of the control architecture is a Model Predictive Control scheme for generating a stable gait. This allows for the inclusion of workspace obstacles, which we take into account at two levels: during the determination of the footsteps orientation and as an explicit MPC constraint. We illustrate the results with simulations on NAO humanoids.

This work addresses the semantic segmentation of images of street scenes for autonomous vehicles based on a new RGB-Thermal dataset, which is also introduced in this paper. An increasing interest in self-driving vehicles has brought the adaptation of semantic segmentation to self-driving systems. However, recent research relating to semantic segmentation is mainly based on RGB images acquired during times of poor visibility at night and under adverse weather conditions. Furthermore, most of these methods only focused on improving performance while ignoring time consumption. The aforementioned problems prompted us to propose a new convolutional neural network architecture for multi-spectral image segmentation that enables the segmentation accuracy to be retained during real-time operation. We benchmarked our method by creating an RGB-Thermal dataset in which thermal and RGB images are combined. We showed that the segmentation accuracy was significantly increased by adding thermal infrared information.

Fusing RGB and depth data is compelling in boosting performance for various robotic and computer vision tasks. Typically, the streams of RGB and depth information are merged into a single fusion point in an early or late stage to generate combined features or decisions. The single fusion point also means single fusion path, which is congested and inflexible to fuse all the information from different modalities. As a result, the fusion process is brute-force and consequently insufficient. To address this problem, we propose a multi-scale multi-path multi-modal fusion network (M3Net), in which the fusion path is scattered to diversify the contributions of each modality from global and local perspectives. Specially, the CNN streams of each modality are fused with a global understanding path and meanwhile a local capturing path. By filtering and regulating information flow in a multi-path way, the M3Net is equipped with more adaptive and flexible fusion mechanism, thus easing the gradient-based learning process, improving the directness and transparency of the fusion process and simultaneously facilitating the fusion process with multi-scale perspectives. Comprehensive experiments demonstrate the significant and consistent improvements of the proposed approach over state-of-the-art methods.

We present an approach for estimating the absolute poses of a swarm Micro Autonomous Underwater Vehicles (μAUVs) by decomposing the problem into few absolute position estimations and many relative pose estimations. As power constraints are critical to small mobile robots, we develop an extension of active marker pose estimation using color information to solve the marker correspondence problem, and show that this approach is more energy efficient than reflective estimation approaches. We show the feasibility of this approach by localizing a robot navigating in an underwater test tank environment. Detailed analysis is presented characterizing the noise and error properties when estimating robot poses from fixed on-board markers. Moreover, we provide comparisons in power and computational cost for other popular methods of underwater localization.

Local communication and sensing between individuals are frequently used for control in multi-robot systems. However, in current flying multi-robot systems, such as swarms of quadrotors, these abilities are often only emulated using global communication and global position sensing. This is mainly due to the complexity, cost, weight, or power requirements for such sensors. Here we present a system that can allow for more natural swarming behaviors by enabling direct bearing and elevation sensing, as well as communication between nearby rotor-craft. This system takes advantage of the existing motion of the vehicles propellers, is low power, and can be adapted to existing vehicles with only simple modifications. We describe the system, present a working prototype, show performance of this prototype, and conclude by describing future work integrating the system into a rotor-craft swarm.

A novel method to compute the contact force in forward dynamics simulation of rigid bodies based on contact volume is proposed. The conventional methods are based on contact vertices, where the normal and tangential directions are estimated with respect to each pair of facets or facet and vertex of colliding bodies. However, the normal direction is intrinsically determined based on the solid deformation. This paper proposes a novel method that combines the constraint-based method and the volumetric intersection computation. Since the method estimates a 6-axis resultant force directly, there are two problems that are how to predict the preferred deformation of the volume and how to confirm the friction limit. The idea is to apply the least-square method for the former and to check if the computed force can be resolved into each stress within the friction limit. The proposed method checks the condition with linear programming method through a pyramidal approximation.

Assessing terrain ahead of a robot when repeating previously driven safe paths can be accomplished by looking for geometric changes (e.g., due to the appearance of humans or other obstacles). Previous work has shown that the incorporation of data-driven learning and place-dependence are useful aspects of making terrain classification viable in challenging terrain. This paper presents a learning, place-dependent (LPD) terrain classifier that uses a probabilistic model of the terrain to improve detection of small obstacles in uncluttered terrain while avoiding false positives in more challenging environments. Specifically, a Gaussian mixture model is used to account for multi-height terrain cells that arise in heavily vegetated areas (where both a ground plane and overhanging vegetation can occupy the same cell). A variational Bayesian technique is used to automatically determine the number of components required for each cell using a Dirichlet prior on mixing proportions and a Normal-Inverse-Wishart prior on the means and covariances of the components. The probabilistic nature of the model allows for the detection of much smaller obstacles in regions that exhibit low variance in the terrain surface, whilst still avoiding false positives in regions where the terrain is highly cluttered (e.g., vegetation). The algorithm is tested on almost 10 km of autonomous traverse and is shown to be able to classify a wider range of obstacles than two baseline change-detection algorithms based on absolute geometric differences.