IROS 2016 论文列表

A water beetle is skillful at drag-powered swimming by using its oar-like legs. Inspired by this mechanism, a miniature robot, whose mobility is obtained by a pair of legs attached with swimming appendages, was studied in this work. The robot in this paper has optimally designed linkage structures to maximize the sweep angle, which is actuated by a single DC motor with series of gears and a spring. A simplified swimming appendage model was proposed to calculate deflection by drag force applied. Also, each of swimming appendage was optimized individually by considering the attached locations on the legs using the two fitness functions, and three different configurations were selected. In the experiments, the performance of the proposed robots was verified using a high-speed camera, and motion capture cameras. The robot with the proposed configuration showed the fastest movement comparing with other robots.

In this paper, we present the dynamical equations of a nonlinear adaptive natural oscillator (NANO) in order to exploit the natural dynamics in robotic systems. The presented oscillator tries to minimize an energy-based cost function by adapting the shape and frequency of the reference trajectory. Stability, convergence, and optimality of this oscillator are guaranteed analytically. Moreover, the performance of this oscillator is investigated by applying it to three different types of robotic models; i.e., the pendulum, the adaptive-toy, and the hopper-leg.

This paper presents a multilink-articulated robot with omni and hemispherical wheels (AIRo-II) for inspecting and exploring winding pipes. To quickly adapt to winding pipes, holonomic rolling movement without moving forward and backward is more useful. However, this requires the replacement of driving actuators with rolling actuators at the expense of the driving force. In this paper, we investigate the possibility of high maneuverability of multilink-articulated robots in winding pipes by using less actuators and by designing spring joints. We further validate this by experimental verification.

In this paper, the design, modeling and experimental verification of a large workspace, parallel aerial manipulator is presented. The proposed manipulator has 3 Degrees of Freedom (DoFs) and enables physical interaction on the sides, as well as below the aerial robot. The design parameters of the manipulator are chosen such that it achieves large and singularity-free workspace in combination with high dexterity. A Global Conditioning Index (GCI) is defined and used as a performance index for the manipulator over its complete workspace. Given the manipulator design parameters, a holistic model of the redundant aerial manipulation system is derived capturing the coupled dynamics of the aerial vehicle and the manipulator. Free flight experimental studies are utilized to validate the system, demonstrate its redundant DoFs, and evaluate its performance.

In this paper, we propose a new method for implementing a rotational nonlinear spring with user defined profile, based on the combination of a linear spring with a nonlinear transmission mechanism. The proposed structure consists of a non-circular cam, a roller which moves along outer circumference of the cam, and a stretched translational linear spring which is connected between center of the cam and center of the roller. We obtain a set of differential equations to design shape of the cam for any given torque-angle profile. Also, it will be shown that profiles with both positive and negative values can be implemented by the proposed method. At last, the cam of some popular nonlinear springs are designed, including constant, cubic, hyperbolic tangent and sinusoidal springs.

In this paper, a concentric wire-driven mechanism (CWM) is presented. The CWM comprises of two nested wire-driven mechanisms (WDM). Each WDM contains a flexible backbone and a set of wires. The backbone bending is controlled by pulling the wires. The stiffness of the outer WDM is controllable and dominants that of the overlapped section. Therefore, in the overlapped section the inner WDM conforms to the shape of the outer WDM and the tip of the outer WDM serves as the base of the distal separate section of the inner WDM. Compared with conventional flexible mechanisms, i.e., tendon/wire/cable-driven mechanism and concentric tube mechanism, the CWM has a much wider workspace. This advantage is demonstrated by simulations based on kinematic modeling. A prototype is developed and the preliminary experimental results validate the design concept.

This study demonstrates soft, epidermal resistive strain gauges capable of tracking finger joint angle during dexterous manipulation tasks. Intrinsically stretchable, biphasic, gallium-based metal films embedded in an elastomeric substrate allow for extremely thin (<50 µm) and skin-conforming wearable sensors with outstanding robustness. The sensors sustain repeated cycling to 50% strain and are insensitive to normal pressure up to 100 kPa. Following a calibration phase, we recorded flexions of a human finger using the soft sensors and compared their joint angle estimation to that of a commercial marker-based visual motion tracking system. The accuracy of our system (defined as the mean angular deviation between our sensors' output and the reference system) was below 9° over a set of 11 different grasping and motion tasks. We demonstrate the scalability and wearability of our technology with a three-finger sensing system used to track the fine movements of a pianist hand. Our soft technology is a promising candidate for implementation of truly wearable proprioceptive sensing systems.

A novel structure of a 2 DoF telescopic soft robot using a tri-layer Polypyrrole (PPy) soft micro-actuator with deployment is presented in this paper. The kinematic model is introduced and the Position Based Visual Servo (PBVS) control with path-planning and obstacle avoidance algorithms is developed. A prototype is presented and the control schemes are validated experimentally. A satisfactory accuracy with a submillimetric positioning error is obtained namely 287.6 microns for a circular path and 210 microns for obstacle avoidance.

We present the EP-Face connector, a novel connector for hybrid chain-lattice type modular robots that is high-strength (88.4N), compact, fast, power efficient, and robust to position errors.

This article presents the design, implementation and test of a joint mechanical device created to improve the safety and performance of robots interacting physically with human beings. The mechanism consists of a series of linear springs, preloaded and arranged in such a way to take advantage of the benefits of series elastic actuators and to limit the maximum applicable torque. This is a passive mechanism because it does not have a second actuator system on the same joint. However, it involves a joint system of variable stiffness given the three distinct performance phases it has, each one having a different stiffness value, elastic torque-dependent occurring in the joint.

We present a delta DLP 3D printer with large size in this paper. Compared with traditional DLP 3D printers that use a low-cost off-the-shelf consumer projector and a single vertical carriage, the platform of our delta DLP 3D printer can also move horizontally in the plane. We show that this structure allows the printer to have a larger printing area than the projection area of a projector. Our system can print 3D models much larger than traditional DLP 3D printers. The major challenge to realize delta 3D printing with large size comes from how to partition an arbitrary planar polygonal shape (possibly with holes or multiple disjoint polygons) into a minimum number of rectangles with fixed size, which is NP-hard. We propose a simple yet efficient approximation algorithm to solve this problem. The time complexity of our algorithm is O(n3 log n), where n is the number of edges in the polygonal shape. A physical prototype system is built and several large 3D models with complex geometric structures have been printed as examples to demonstrate the effectiveness of our approach.

In this study, a deformation control with spatial decoupling properties is designed and implemented. This control strategy treats the shift in position and posture attributable to an external force as the deformation of the robot. The deformation dynamics are constructed from the Maxwell and Voigt models, which describe plastic and elastic deformation, respectively. Next, a control method is proposed to passively achieve different deformation characteristics in each direction. Two physical simulations with a robotic arm are executed to validate the proposed control law.

Robots often operate in built environments containing underlying structure that can be exploited to help predict future observations. In this work, we present a deep learning based approach to predict exit locations of buildings. This technique exploits the inherent structure of buildings to create a model. A convolutional neural network is trained using a database of building blueprints and used to guide a search within a building. This technique is compared to standard frontier exploration and a traditional image processing approach of extracting features through histogram of gradients (HOG) and training a support vector machine (SVM). After validation through simulation, we show that the proposed deep learning technique reduces the amount of building exploration required to find the goal by 36%.

This paper presents a new method for parallel-jaw grasping of isolated objects from depth images, under large gripper pose uncertainty. Whilst most approaches aim to predict the single best grasp pose from an image, our method first predicts a score for every possible grasp pose, which we denote the grasp function. With this, it is possible to achieve grasping robust to the gripper's pose uncertainty, by smoothing the grasp function with the pose uncertainty function. Therefore, if the single best pose is adjacent to a region of poor grasp quality, that pose will no longer be chosen, and instead a pose will be chosen which is surrounded by a region of high grasp quality. To learn this function, we train a Convolutional Neural Network which takes as input a single depth image of an object, and outputs a score for each grasp pose across the image. Training data for this is generated by use of physics simulation and depth image simulation with 3D object meshes, to enable acquisition of sufficient data without requiring exhaustive real-world experiments. We evaluate with both synthetic and real experiments, and show that the learned grasp score is more robust to gripper pose uncertainty than when this uncertainty is not accounted for.

Graph-based SLAM approaches have had success recently despite suffering from ever-increasing computational costs due to the need of optimizing over the entire robot trajectory. To address this issue, in this paper, we advocate the decoupling of marginalization (node removal) and sparsification (edge reduction) to allow for short-term retention of dense factors induced by marginalization while enabling us to spread the computation of these two operations over time. In particular, we analytically show that during marginalization, the correct choice of linearization points in constructing dense marginal factors is to use the relative (local), instead of global, state estimates in the Markov blanket of the marginalized node, which has lacked a general consensus in the literature. Furthermore, during sparsification, we determine an online sparse topology through sparsity-regularized convex optimization, which guides us to construct consistent sparse factors to best approximate the original dense factors across the Markov blanket. The proposed approach is tested extensively on both 2D and 3D public datasets and shown to perform competitively to the state-of-the-art algorithms.

In this paper, we model an aerial vehicle, specifically a quadrotor, and a load attached to each other by a rigid link. We assume a torque input at the joint between the aerial vehicle and the rigid link is available. After modeling, we decouple the system dynamics in two separate subsystems, one concerning the position of the center of mass, which we control independently from the chosen torque input; and a second subsystem, concerning the attitude of the rigid link, which we control by appropriately designing a torque control law. Differential flatness is used to show that controlling these two separate systems is equivalent to controlling the complete system. We design control laws for the quadrotor thrust, the quadrotor angular velocity and the torque input, and provide convergence proofs that guarantee that the quadrotor follows asymptotically a desired position trajectory while the manipulator follows a desired orientation. Simulation and experimental works are presented which validate the proposed algorithms.

This article introduces a framework to compute decision making policies (mission planning) for a UAV-based delivery system serving impatient customers. Customers arrive on a finite number of locations L separated by arbitrary but fixed distances and eventually leave if not served. Policies seek to minimize the average net cost (maximize the average net revenue), i.e. loss from customers' abandonment deprived of the revenue from successful services. We introduce a novel model for the stochastic and dynamic pickup and delivery problem based on semi-Markov decision processes, and show the dependence of the optimal average net cost on the minimum distance between locations δ. Furthermore, we propose a feature-based state aggregation method to overcome the curse of dimensionality due to the exact modeling. The selection of relevant features is based on their correlation with the optimal performance. We show that the distance to the nearest pickup location dominates on other features for almost all δ. Based on this observation, we introduce the policy nearest neighbor or none, a policy with computational complexity O(L3) in many cases of interest. We show that this policy performs considerably better that the nearest neighbor (greedy) policy, and reaches the optimum for some δ.

Regrasping is the process of adjusting the position and orientation of an object in one's hand. The study of robotic regrasping has generally been limited to use of theoretical analytical models and cases with little uncertainty. Analytical models and simulations have so far proven unable to capture the complexity of the real world. Empirical statistical models are more promising, but collecting good data is difficult. In this paper, we collect data from 3300 robot regrasps, and use this data to learn two probability functions: 1) The probability that the object is still in the robot's hand after a regrasp action; and 2) The probability distribution of the object pose after the regrasp given that the object is still grasped. Both of these functions are learned using kernel density estimation with a similarity metric over object pose. We show that our data-driven models achieve comparable accuracy to a geometric model and an off-the-shelf simulator in classification and prediction tasks, while also enabling us to predict probability distributions.

Autonomous multi-robot teams can be used in complex coordinated exploration tasks to improve exploration performance in terms of both speed and effectiveness. However, use of multi-robot systems presents additional challenges. Specifically, in domains where the robots' actions are tightly coupled, coordinating multiple robots to achieve cooperative behavior at the group level is difficult. In this paper, we demonstrate that reward shaping can greatly benefit learning in multi-robot exploration tasks. We propose a novel reward framework based on the idea of counterfactuals to tackle the coordination problem in tightly coupled domains. We show that the proposed algorithm provides superior performance (166% performance improvement and a quadruple convergence speed up) compared to policies learned using either the global reward or the difference reward [1].