A grasp manipulation selection chart to pick-up objects lying on hard surfaces Vincent Babin Department of Mechanical Engineering Universit´ e Laval Qu´ ebec, Canada, G1V 0A6 vincent.babin.1@ulaval.ca Cl´ ement Gosselin Department of Mechanical Engineering Universit´ e Laval Qu´ ebec, Canada, G1V 0A6 gosselin@gmc.ulaval.ca Abstract—As robotic systems have to solve more and more complex problems, engineers are attempting to create systems that can perform the largest possible set of tasks. To evaluate the needs of the system to be implemented, both the task to be accomplished and the capabilities of the robot must be considered. One critical aspect of the studies is the link between the robot and the environment/object which is the end-effector. Specialised end-effectors are the ones designed with a well defined task in mind and are task specific, as opposed to general purpose end-effectors, which aim to be robust but execute most tasks rather poorly. General purpose grippers are designed to hold a variety of objects but are increasingly required to pick objects in real-world environments in a safe manner. With this in mind, this paper studies grasping methods found in the literature to compare them and emphasize the importance of compliant/soft end-effector compared to the use of force control techniques when interacting with rigid real-world environments. Index Terms—Grasping, Manipulation, Parallel Grasps, Grip- per, Scooping, Workspace, Constrained Environments. I. I NTRODUCTION The study of object grasping is a very diverse research topic, where the goal of the robotic system is to perform a task. The engineer must determine the needed applied forces on the environment or through a tool held by a robot. The set of forces and torques that can be applied is called the wrench space and is an ongoing research topic [1]–[3]. Because of the complexity of the interaction between an end-effector and an object, simply holding an object with a gripper/robot hand is a complex problem when in presence of external forces like gravity, uncertainties and positioning imperfections [4]. An interesting aspect of object grasping consists of the study of mechanisms that have stable grasp properties [5], meaning that the grasp is robust to external forces to a certain degree. The study of underactuation, which relies on extra degrees of freedom in the hand to mechanically adapt to the shape of objects to obtain form closure [6], [7], is also an ongoing research topic which alleviates the reliance on many motors and complex control. While previous studies provide an understanding of the quality of the pose of objects within an end-effector, another important research issue is the manipulation processes that can This work was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) and by Robotiq. lead to said poses. Some of the references in this area propose tools to quantify the ability of an end-effector to manipulate objects [8]. This leads to the concept of intrinsic [9] and extrinsic dexterity [10]–[12]. In this paper and according to [10], the difference between the methods is that the latter is allowed to use resources external to the gripper, like gravity or a surface, while the former is not. Studies have analysed the human hand, the grasping processes performed by humans [13], and the optimal use of anthropomorphic hands [14]. These all take advantage of extrinsic dexterity to a certain degree. Because of the fact that most objects are resting on a surface under the action of gravity, environment properties can be leveraged into effective grasp methods. The goal of this paper is first to analyse the grasping methods found in the literature for objects resting on hard surfaces and enumerate the features and algorithms necessary to perform them. Second, the need to interact with the en- vironment for performing grasping in future designs is made clear. First, the motivation of this paper and the scope of the grasping scenario are presented in Section II. This section aims at clearly stating which cases are considered and why this paper is a proposed starting point for future general purpose grasping algorithms and not a definitive step by step guideline. The methods and the features needed for each of them to work are then summarised in Section III. Section IV then presents a chart outlining each method’s steps, a suggested selection guide considering object properties, and examples objects that can be picked up. Finally the crucial features that make each method work and how each can be implemented are discussed in Section V. II. MOTIVATION As mentioned above, whether it is in industrial settings or in day to day life, most objects rest on surfaces ranging from the very hard to the very soft. Assuming that enough sensors or vision apparatus are available to detect the pose and orientation of an object, it is not always possible for a robot arm to pick said object. One possible cause for this is simply for the object to be too far away from the end-effector to even be touched. Those cases are not considered in this paper because solutions like using a mobile robot or, for example, a mobile shelve can