We present a unified framework for grasp planning and in-hand adaptation using visual, tactile, proprioceptive feedback. The main objective of the proposed is to enable fingertip grasping by addressing problems changed weight object, slippage, external disturbances. For this purpose we introduce Hierarchical Fingertip Space as representation enabling optimization both efficient synthesis online finger gaiting. Grasp followed step that consists force through impedance control...

A modularized and actuated landing gear framework allows a UAV to stably perch rest on wide range of different structures.

Constraining contacts to remain fixed on an object during manipulation limits the potential workspace size, as motion is subject hand's kinematic topology. Finger gaiting one way alleviate such restraints. It allows be freely broken and remade so operate different manifolds. This capability, however, has traditionally been difficult or impossible practically realize. A finger system must simultaneously plan for control forces while maintaining stability contact switching. letter alleviates...

Rearranging objects on a tabletop surface by means of nonprehensile manipulation is task which requires skillful interaction with the physical world. Usually, this achieved precisely modeling properties objects, robot, and environment for explicit planning. In contrast, as explicitly not always feasible involves various uncertainties, we learn rearrangement strategy deep reinforcement learning based only visual feedback. For this, model rewards train Q-network. Our potential field-based...

We address the problem of pregrasp sliding manipulation, which is an essential skill when a thin object cannot be directly grasped from flat surface. Leveraged on passive reconfigurability soft, compliant, or underactuated robotic hands, we formulate this as integrated motion and grasp planning problem, plan manipulation in robot configuration space. Rather than explicitly precomputing pair valid start goal configurations, then separate step path to connect them, our planner actively samples...

In this work, we address a planar non-prehensile sorting task. Here, robot needs to push many densely packed objects belonging different classes into configuration where these are clearly separated from each other. To achieve this, propose employ Monte Carlo tree search equipped with task-specific heuristic function. We evaluate the algorithm on various simulated and real-world tasks. observe that is capable of reliably large numbers convex non-convex objects, as well in presence immovable obstacles.

In this paper, we propose a solution to the problem of herding by caging: given set mobile robots (called herders) and group moving agents sheep), move latter some predefined location in such way that they cannot escape from while moving.We model interaction between herders sheep assuming former exert virtual "repulsive forces" pushing away them.These forces induce potential field, which does not increase their potential.This enables partially control motion sheep.We formalize behavior...

Numerous grasp planning algorithms have been proposed since the 1980s. The grasping literature has expanded rapidly in recent years, building on greatly improved vision systems and computing power. Methods to plan stable grasps known objects (exact 3D model is available), familiar (e.g. exploiting a-priori for different of same category), or novel object shapes observed during task execution. Few these methods ever compared a systematic way, objective performance evaluation such complex...

The purpose of this benchmark is to evaluate the planning and control aspects robotic in-hand manipulation systems. goal assess system's ability change pose a hand-held object by either using fingers, environment or combination both. Given an surface mesh from YCB data-set, we provide examples initial states (i.e.\ static poses fingertip locations) for various tasks. We further propose metrics that measure error in reaching state specific state, which, when aggregated across all tasks, also...

Dexterous in-hand manipulation of objects benefits from the ability a robot system to generate precision grasps. In this paper, we propose concept Fingertip Space and its use for grasp synthesis. is representation that takes into account both local geometry object surface as well fingertip geometry. As such, it directly applicable point cloud data establishes basis search space. We model hierarchical encoding enables multilevel refinement efficient The proposed method works at contact level...

We consider the problem of finding grasp contacts that are optimal under a given quality function on arbitrary objects. Our approach formulates framework for contact-level grasping as path in space supercontact grasps. The initial contains all grasps and each step along removed. For this, we introduce formally characterize search structure cost functions which minimal paths correspond to formulation avoids expensive exhaustive reduces computational by several orders magnitude. present...

The process of modeling a series hand-object parameters is crucial for precise and controllable robotic in-hand manipulation because it enables the mapping from hand's actuation input to object's motion be obtained. Without assuming that most these model are known priori or can easily estimated by sensors, we focus on equipping robots with ability actively self-identify necessary using minimal sensing. Here, derive algorithms, basis concept virtual linkage-based representations (VLRs),...

We propose a new object descriptor for three dimensional data named the Global Structure Histogram (GSH). The GSH encodes structure of local feature response on coarse global scale, providing beneficial trade-off between generalization and discrimination. Encoding structural characteristics an allows us to retain low variations while keeping benefit representativeness. In extensive experimental evaluation, we applied framework category-based classification in realistic scenarios. show...

We propose the Dexterous Manipulation Graph as a tool to address in-hand manipulation and reposition an object inside robot's end-effector. This graph is used plan sequence of primitives so bring desired end pose. translated into motions robot move held by use dual arm with parallel grippers test our method on real system show successful planning execution manipulation.

In this work, we propose a pick-and-place benchmark to assess the manipulation capabilities of robotic system. The is based on Box and Blocks Test (BBT), task utilized for decades by rehabilitation community unilateral gross manual dexterity in humans. We three robot benchmarking protocols work that hold true spirit original clinical tests-the Modified-BBT, Targeted-BBT, Standard-BBT. These can be implemented greater robotics research community, as physical BBT setup has been widely...

We address the problem of generating force-closed point contact grasps on complex surfaces and model it as a combinatorial optimization problem. Using multilevel refinement metaheuristic, we maximize quality grasp subject to reachability constraint by recursively forming hierarchy increasingly coarser problems. A is initialized at top then locally refined until convergence each level. Our approach efficiently addresses high dimensional synthesizing stable while resulting in from arbitrary...

Moving a human body or large and bulky object may require the strength of whole arm manipulation (WAM). This type places load on robot's arms relies global properties interaction to succeed- rather than local contacts such as grasping non-prehensile pushing. In this paper, we learn generate motions that enable WAM for holding transporting humans in certain rescue patient care scenarios. We model task reinforcement learning problem order provide robot behavior can directly respond external...

We present a new approach for modelling grasping using an integrated space of grasps and shapes.In particular, we introduce infinite dimensional space, the Grasp Moduli Space, which represents shapes in continuous manner.We define metric on this allowing us to formalize 'nearby' grasp/shape configurations discuss deformations such configurations.We work particular with surfaces cylindrical coordinates analyse stability popular L 1 grasp quality measure Q l under grasps.We experimentally...

We address the problem of fingertip design by leveraging on fact that most grasp contacts share a few classes local geometries. In order to maximize contact areas for achieving more robust grasps, we first define concept Contact Primitive, which represents set similar Thereafter, propose uniform cost algorithm, is formulated as decision making process in tree structure, cluster example into finite Primitives. fingertips optimization match geometry each primitive, and then three-dimensionally...

In this work, we present an algorithm that simultaneously searches for a high quality fingertip grasp and collision-free path robot hand-arm system to achieve it. The combines bidirectional sampling-based motion planning approach with hierarchical contact optimization process. Rather than tackling these problems in decoupled manner, the is guided by proximity configurations explored planner. We implemented 13-DoF manipulator show it capable of efficiently reachable grasps cluttered...

We address the problem of planning placement a rigid object with dual-arm robot in cluttered environment. In this task, we need to locate collision-free pose for that a) facilitates stable object, b) is reachable by and c) optimizes user-given objective. addition, select which arm perform with. To solve propose an anytime algorithm integrates sampling-based motion novel hierarchical search suitable poses. Our incrementally produces approach motions poses, reaching placements better objective...

We consider the problem of in-hand dexterous manipulation with a focus on unknown or uncertain hand–object parameters, such as hand configuration, object pose within hand, and contact positions. In particular, in this work we formulate generic framework for configuration estimation using underactuated hands an example. Owing to passive reconfigurability lack encoders hand’s joints, it is challenging estimate, plan, actively control manipulation. By modeling grasp constraints, present...