Shuo Liu

Fast grasp quality evaluation with partial convex hull computation

We present Partial Quick Hull (PQH), an algorithm to efficiently compute one of the most commonly used grasp quality metrics. The metric relies on the computation of the convex hull of a set of points in a six dimensional space. Built on top of widely used QuickHull algorithm, PQH exploits the relationship between the convex hull and the grasp quality metric to avoid computing the whole convex hull. PQH determines at run time when the computation can be ended because the grasp quality metric can be already determined from a partially computed convex hull - hence the name of the algorithm. This improvement greatly accelerates grasp quality evaluation for force closure grasps. When the grasp is not force closure, the partial computation does not apply and PQH then behaves exactly like QuickHull. A large set of experimental tests show how PQH largely outperforms QuickHull and better scales with the size of the input.

Grasping the Performance: Facilitating Replicable Performance Measures via Benchmarking and Standardized Methodologies

It is comparatively easy to make computers exhibit adult-level performance on intelligence tests or playing checkers, and difficult to give them the skills of a one-year-old when it comes to perception and dexterity. More than 15 years after it was first stated, Moravecs paradox still holds true today. Fueled by vigorous research in machine learning, the gap has consistently narrowed on the perception side. However, most of the fine manual motor skills displayed by a toddler are, to date, far beyond what robots can do. It is true that many valuable tasks involving physical interaction with objects can be solved by contemporary robots as indicated by a thriving industrial robotics sector. However, in the future, robots are expected to work side by side with humans in unstructured environments, and the ability to reliably grasp and manipulate objects used in everyday activities will be an unavoidable requirement. Todays robots are far from being ready for this challenge.

A fast algorithm for grasp quality evaluation using the object wrench space

Grasp quality evaluation is an important problem when multifingered robotic hands are used to restrain objects in automation and manufacturing. In this paper we present a new algorithm to greatly expedite the evaluation of grasp quality considering the so-called object wrench space, i.e., the set of all disturbance wrenches that may be generated by a disturbance force acting on the object being grasped. While this metric has been known since a while, its practical use inside grasp planners has been limited because its exact computation is time consuming and thus prevents its use when many grasps have to be repeatedly evaluated during the planning process. Building on some geometric insights related to convex hulls, our algorithm determines on the fly which subset of the input data needs to be processed, and stops the computation as soon as it is known that the exact value of the metric has been determined. We show that our new algorithm significantly decreases the time needed to compute the grasping quality measure thus enabling grasp planners to use this metric to search the space of possible grasps.

Global grasp planning using triangular meshes

In this paper we present an algorithm to determine the location of contact points to obtain force closure grasps on tree dimensional objects. The shape of the object is assumed to be given by a triangle mesh - a format widely used in CAD software. Our algorithm can handle an arbitrary number of contact points and does nor require any prior information about their initial locations. Through an iterative process, contact point locations are updated aiming at improving a commonly used grasp quality metric. The process is global in the sense that during the process the whole surface of the object can be explored, and contact point locations can cross sharp edges that usually represent a problem for optimization algorithms relying on smooth surface representations. Extensive simulation results illustrate the performance of the proposed method, outlining strengths and directions for further research.

Partial convex hull algorithms for efficient grasp quality evaluation

We present two algorithms to efficiently determine the value of two grasp quality metrics formerly proposed in literature. The first one is heavily used in practice but has some drawbacks, e.g.,~it is not scale invariant and it does not focus on the disturbance forces that will occur in practice when the robot grasps an object. The second one overcomes these limitations, but is rarely used because it is computationally too demanding. The two algorithms we propose are based on the common intuition that both metrics can be efficiently computed through a modified version of the QuickHull algorithm that is commonly used to compute convex hulls. In both cases it is possible to establish when enough information has been generated to determine the desired value, and then stop the construction of a suitably defined convex hull. Extensive numerical evaluations demonstrate that our algorithms provide substantial computational gains when compared with the state of the art. The speedup provides an immediate benefit to planners using grasp quality metrics to guide the search through the space of possible grasps. Two algorithms to compute grasp quality evaluation metrics are presented.Algorithms are based on computational geometry insights derived from convex hull computation.Extensive simulations substantiate substantial performance gains over the state of the arton.Code and datasets are made freely available to the research community.

Grasp quality evaluation and planning for objects with negative curvature

We consider the problem of grasping concave objects, i.e., objects whose surface includes regions with negative curvature. When a multifingered hand is used to restrain these objects, these areas can be advantageously used to determine grasps capable of more robustly resisting to external disturbance wrenches. We propose a new grasp quality metric specifically suited for this case, and we use it to inform a grasp planner searching the space of possible grasps. Our findings are validated both in simulation and on a real robot system executing a bin picking task. Experimental validation shows that our method is more effective than those not explicitly considering negative curvature.

Kinematic noise propagation and grasp quality evaluation

To determine a force-closure grasp, current grasp synthesis algorithms either assume a deterministic model to compute a desired finger placement without noise, or model the end-effector position as a free-floating rigid body whose noise in pose is independent of the kinematic chain formed by the robot arm. In this work we instead explore a probabilistic approach that explicitly models noise in joint-angles. By sampling additive noise that is applied to a pre-grasp configuration and studying the resulting probability of force-closure when the robot fingers are closed, we observe in experiments that joint-angle positions can have a remarkable effect on the probability of successfully restraining an object. We systematically study the grasp quality value as a random variable and investigate the convergence of sampling based estimators for the mean, covariance and moments up to third order of this quantity by means of Montecarlo Sampling. We study illustrative examples of the impact of initial joint-configurations on the likelihood of force closure on a seven degree of freedom simulated Kuka lightweight robot arm.

Design and Application of Dorabot-hand2 System

We present Dorabot-hand2, a dexterous robot hand and its design principles. The goal of designing this hand is to gain capability of handling everyday tasks. The hand is tendon-driven and is based on modular design. We focus on certain aspects of the design, including strength, friction, cost and maintainability. We conclude with a description of the hand’s performance when competing in the Robotic Grasping and Manipulation Competition at IROS 2016.

A Robotic System for Autonomous Grasping and Manipulation

A robotic system that consists of only a gripper can be utilized for certain applications such as supporting disabled people. However, with a robot manipulator introduced into the system, it can achieve far more tasks such as automation of manufacturing and logistics processes. The autonomous track of the IROS2016 Robotic Grasping and Manipulation Competition was designed to bring a robotic system into ordinary everyday tasks involving grasping and manipulation. The main objective of this paper is the evaluation of the autonomous robotic system by comparing the performance against manual human-interacted system in terms of intelligence and robustness.