IROS 2017 论文列表

Robot teams require planning and adaptive capabilities in order to perform cooperative manipulation tasks in dynamic or unstructured environments. Since these capabilities are inherent to humans, it is suitable to consider human-robot team teleoperation for cooperative manipulation where a single human collaborates with the robot team. In this paper, we present a subtask-based control approach which enables a simultaneous execution of two subtasks by the robot team, interacting with the object: trajectory tracking and formation preservation. Control inputs for both subtasks are provided by the human operator. The commands are projected onto the spaces of subtasks using a command mapping strategy. Analogously, measured interacting forces are projected onto the space of feedback signals, provided to the human via wearable fingertip haptic devices through a feedback mapping strategy. Experimental results validate the proposed approach.

We present a novel force-limitation algorithm to protect fragile robot actuation and transmission systems (like tendon-driven systems) from early wear-out. Inspired by human muscle fatigue, we model artificial actuator fatigue by integrating applied forces over time, gradually limiting applicable forces when fatigue increases. The algorithm is applied to protect from long-term tendon wear-out on our Shadow Dexterous Hands as well as to restrict grasping forces for compliant grasping in unknown environments. In grasping experiments the efficiency of various grasp-force limitation approaches are compared to each other, including one exploiting tactile-based slip detection.

A method for real-time robot navigation with obstacle avoidance is presented. A two-stage approach is proposed: the use of online Simulated Annealing (SA) to optimize a convex operating region in configuration space is paired with Model Predictive Control (MPC) to determine a locally optimal motion plan. The method retains recursive feasibility guarantees from MPC and speed of solution of the convex optimal control problem. Meanwhile, the convexification performed by the SA enables the method to operate with unstructured environment representations, such as point clouds or line scans.

Occupancy grid maps are a popular method for representing the environment in the context of robot navigation tasks. However, occupancy grid maps can have a high memory demand that grows quadratically with the range of the sensor. In this paper, we introduce a memory-efficient map representation that is based on a constant set of individual scans. To make these scan-based maps suitable for autonomous robot navigation, we propose probabilistically sound methods for both mapping and localization. To solve the mapping problem, our approach incrementally selects scans based on the additional information they provide relative to the scans previously selected. Using these selected scans, we perform an Monte Carlo Localization (MCL) approach with a sensor model optimized for the scan-based representation of our map. We present extensive experiments in which we evaluate our approach using real world data recorded in a garage parking scenario with an autonomous car as well as a robot localization problem in an indoor environment. The results demonstrate that our approach can cope with high sensor noise and that it achieves comparable localization accuracy while at the same time consuming only a fraction of memory compared to regular occupancy grid maps.

Simultaneous localization and mapping (SLAM) is a core element of every autonomous mobile robot. The underlying engine of a SLAM system is its back-end, which aims at optimally estimating the trajectory and map of the environment based on sensor data abstractions. Over the past decade, SLAM solutions based on graph optimization approaches prevailed over the filtering based solutions, since they dominated in performance over a wider range of applications. In this paper we propose a novel filtering based SLAM back-end based on the exactly sparse delayed state filter (ESDSF) derived on Lie groups (LG-ESDSF). The proposed filter retains all the good characteristics of the classic ESDSF, but also respects the state space geometry by employing filtering equations directly on Lie groups. We have compared our SLAM system with two current state-of-the-art SLAM solutions, namely ORB-SLAM and LSD-SLAM, on the KITTI vision benchmark suite. Test results show that the proposed SLAM based on the LG-ESDSF back-end can achieve same level of accuracy as the methods based on the graph optimization techniques, while maintaining lower computation times.

In this work we present a new robotic system, Salto-1P, for exploring extreme jumping locomotion. Salto-1P weighs 0.098 kg, and has an active leg length of 14.4 cm. The robot is able to perform a standing vertical leap of 1.25 m, continuously hop to heights over 1 m, and jump over 2 m horizontally. Salto-1P uses aerodynamic thrusters and an inertial tail to control its attitude in the air. A linearized Raibert step controller was sufficient to enable unconstrained in-place hopping and forwards-backwards locomotion with external position feedback. We present studies of extreme jumping locomotion in which the robot spends just 7.7% of its time on the ground, experiencing accelerations of 14 times earth gravity in its stance phase. An experimentally collected dataset of 772 observed jumps was used to establish the range of achievable horizontal and vertical impulses for Salto-1P.

We present repetition sampling, a new adaptive strategy for sampling based planning, which extracts information from previous solutions to focus the search for a similar task on relevant configuration space. We show how to generate distributions for repetition sampling by learning Gaussian Mixture Models from prior solutions. We present how to bias a sampling based planner with the learned distribution to generate new paths for similar tasks. We illustrate our method in a simple maze which explains the generation of the distribution and how repetition sampling can generalize over different environments. We show how to apply repetition sampling to similar constrained manipulation tasks and present our results including significant speedup in execution time when compared to uniform sampling.

In this study, we develop an electromagnetic-based haptic interface to provide controlled magnetic forces to the operator through a wearable haptic device (an orthopedic finger splint with single dipole moment) without position feedback. First, we model the electromagnetic forces exerted on a single magnetic dipole attached to the wearable haptic device, and derive magnetic force-current mapping for the dipole moment. Second, this mapping is used as basis for parameter selection of the electromagnetic coils of the haptic interface, dipole moment of the wearable haptic device, and the operating workspace of the system. The electromagnetic-based haptic interface enables three-dimensional (3D) virtual object rendering in mid-air within a workspace of 150 mm × 150 mm × 20 mm, using magnetic forces in excess of 50 mN. Participants experimentally demonstrate a 61% success rate in distinguishing the geometry of 4 representative 3D virtual objects. However, our statistical analysis shows that the ability of the participants to distinguish between geometries is not statistically significant, for 95% confidence level.

In this paper, we solve the problem of relay robot placement in multi-robot missions to establish or enhance communication between a static operator and a number of remote units in an environment with known obstacles. We study the hardness of two different relay placement problems: 1) a chain formation of multiple relay robots to transmit information from an operator to a single unit; and 2) a spanning tree of relays connecting multiple remote units to the operator. We first build a communication map data structure from a layered graph that contains the positions of the relays as the unit moves. This structure is computed once and reused throughout the mission, significantly reducing plan re-computation time when compared to the best-known solution in the literature. Second, we create a min-arborescence tree that forms a connected component among the operator, relays, and units, and that has an optimal communication cost. Finally, we validate our ideas through software simulations, hardware experiments, and a comparison of our approach to state-of-the-art methods.

Reinforcement learning (RL) depends critically on the choice of reward functions used to capture the desired behavior and constraints of a robot. Usually, these are handcrafted by a expert designer and represent heuristics for relatively simple tasks. Real world applications typically involve more complex tasks with rich temporal and logical structure. In this paper we take advantage of the expressive power of temporal logic (TL) to specify complex rules the robot should follow, and incorporate domain knowledge into learning. We propose Truncated Linear Temporal Logic (TLTL) as a specification language, We propose Truncated Linear Temporal Logic (TLTL) as a specification language, that is arguably well suited for the robotics applications, We show in simulated trials that learning is faster and policies obtained using the proposed approach outperform the ones learned using heuristic rewards in terms of the robustness degree, i.e., how well the tasks are satisfied. Furthermore, we demonstrate the proposed RL approach in a toast-placing task learned by a Baxter robot.

This paper reports the electrostatic regulation of surface traction of a quadruped soft robot to improve its locomotion efficiency. The soft robot, containing five pneumatic channel networks (PneuNets) in different parts of its body, is actuated to achieve undulated locomotion. Electrostatic adhesion is applied to the bottom surface of each robot leg, by using a thin elastomeric adhesion pad embedded with interdigitated comb electrodes. The adhesion pad is fully compatible with the soft robot structure, and is able to adjust the level of surface traction on the robot leg during locomotion. We calibrate the adhesion force generated by the pad as a function of its size and the applied electrostatic voltage. We demonstrate the control of the moving direction and speed of the soft robot on horizontal surfaces with different frictional and electrical characteristics, by adjusting the level of electrostatic adhesion. With the electrostatic traction control, the robot can also climbing up an inclined metal surface with a low coefficient of friction, which cannot be achieved by the same robot without adhesion pads. This work illustrates the important role of surface friction on locomotion of the soft robot, and provides an efficient solution to surface traction control of soft robots.

This paper develops intelligent algorithms for robots to reorient objects. Given the initial and goal poses of an object, the proposed algorithms plan a sequence of robot poses and grasp configurations that reorient the object from its initial pose to the goal. While the topic has been studied extensively in previous work, this paper makes important improvements in grasp planning by using over-segmented meshes, in data storage by using relational database, and in regrasp planning by mixing real-world roadmaps. The improvements enable robots to do robust regrasp planning using 10,000s of grasps and their relationships in interactive time. The proposed algorithms are validated using various objects and robots.

In this research, we aim to implement the ability to navigate through unknown environments according to some given instructions on mobile robots. We proposed a Recursive Neural Network model, which takes the user instructions and data acquired by laser range finder as input, and outputs control velocities to make the mobile robot navigate in an unknown indoor environment. Instructions are using commonly verbal expressions, and instead of establishing models to represent the semantic meanings, we use the concept of word vectors to feed the information directly into the neural network models. Using human-controlled navigating records as training data, robots learn how to execute navigations according to various of different instructions. Experiments were conducted under both simulation and real world environments, and results show that the robot can successfully navigate to the goal positions without having any prior knowledge about the environments.

In this paper we provide a method for identifying and temporally localizing tactile force actions from measured force signals. Our key idea is to use the continuous wavelet transform (CWT) with the Complex Morlet wavelet to transform force signals into feature vectors amenable to machine learning algorithms. Our method uses these feature vectors to train a classifier that recognizes different actions. We demonstrate our approach in a system that records human activities with an instrumented set of tongs. Our system successfully identifies a wide range of actions based on a small set of labeled examples.

This study proposes an online foot position compensator (FPC) for improving the robustness of biped walking. The proposed FPC learns foot placement parameters through sensory feedback and preview control theory to re-plan the trajectories of the center of mass (CoM) and gait pattern; the re-plan is based on the online generation of a walking pattern using the policy gradient reinforcement learning method (PGRL) to reject external perturbations and recover the walking position. A method of upper-body posture control is also presented to further improve the ability of humanoid robots to suppress external disturbances. The proposed posture controller stabilizes the robot by utilizing hip joints to modulate the upper-body posture in real time. The effectiveness of the proposed methods is validated by simulations on a full-body NAO humanoid robot.

This paper explores a problem of reconfiguration of a large set of agents using global control signals: right, left, up and down. A field of obstacles guides the reconfiguration by limiting motion in the various directions. This paper extends work by Becker et al. ([4], [5] and [7]), and shows there exists a workspace a constant factor larger than the number of agents that enables complete rearrangement for a rectangle of agents.

Mobile microrobots can be used as on-chip remote micromanipulation tools to handle with delicate samples, such as cells and biological tissues. Many efforts have been made to improve microrobots actuation, controllability and power consumption. However, stiction phenomena occurring between the robot and the substrate on which it is moving are still a big issue to deal with. In this paper stiction problem is addressed with a new approach. To begin, the two contact surfaces are coated with low friction materials combination (Al2O3/Si3N4) and different roughnesses are taken into account, in order to find the best combination and to ensure uniform sample conditions. Tests have been made, proving that a rough bottom surface of microrobots moving on a Si3N4 coated glass substrate minimizes stiction events. Starting from this surface combination, the powering algorithm of the electromagnetic field generation system is changed: oscillations perpendicular to the robot trajectory are added to the standard actuation power to overcome static friction while parallel ones are used to avoid dynamic friction, which speeds down the robot. The new powering algorithm is adapted in realtime by visually tracked position and speed of the microrobots. Both the fabrication process flow and the powering algorithm are fully described in the paper. Coatings and adaptive powering show promising results: they manage to reach a long time closed-loop position control stability of the microrobot and they can be applied to find the smoothest path between two points after having performed a surface scanning to map the in-plane stiction distributions. Enhanced position control accuracy and the demonstrated robustness in long term motion control through the proposed approaches could be an important asset to on-chip microrobots toward their biological or micromanufacturing applications.

This paper proposes an approach to enlarge the impedance range of admittance-type haptic interfaces. Admittance-type haptic interfaces have advantages over impedance-type haptic interfaces in the interaction with high impedance virtual environments. However, the performance of admittance-type haptic interfaces is often limited by the lower boundary of the impedance that can be achieved without stability issue. Especially, it is well known that low value of inertia in an admittance model often causes unstable interaction. This paper extends recently proposed input-to-state stable approach [1] to further lower down the achievable impedance in admittance-type haptic interfaces with less conservative constraint compared with the passivity-based approaches. The primary challenge was identifying the nonlinear hysteresis components which are essential for the implementation of the input-to-state stable approach. Through experimental investigation and after separating and merging the admittance model and the position controller, the partial admittance model (from the measured human force to the desired velocity) and the velocity controller (from the velocity tracking error to the controller force) were found having counter-clockwise hysteresis nonlinear behavior. Therefore, it allows implementing the one-port input-to-state stable (ISS) approach for making both components dissipative and ISS. An additional advantage of the proposed ISS approach is the easiness of the implementation. No model information is required, and the network representation is not necessary, unlike the passivity-based approaches. Series of experiments verified the effectiveness of the proposed approach in term of significantly lowering the achievable impedance value compared with what the time-domain passivity approach can render.

In this paper, we present a real-time approach to local trajectory replanning for microaerial vehicles (MAVs). Current trajectory generation methods for multicopters achieve high success rates in cluttered environments, but assume that the environment is static and require prior knowledge of the map. In the presented study, we use the results of such planners and extend them with a local replanning algorithm that can handle unmodeled (possibly dynamic) obstacles while keeping the MAV close to the global trajectory. To ensure that the proposed approach is real-time capable, we maintain information about the environment around the MAV in an occupancy grid stored in a three-dimensional circular buffer, which moves together with a drone, and represent the trajectories by using uniform B-splines. This representation ensures that the trajectory is sufficiently smooth and simultaneously allows for efficient optimization.