Robotic cutting: mechanics and knife control Journal Title XX(X):1–17 c The Author(s) 2021 Reprints and permission: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/ToBeAssigned www.sagepub.com/ SAGE Xiaoqian Mu, Yuechuan Xue, and Yan-Bin Jia Abstract Skills of cutting natural foods are important for robots looking to play a bigger role in kitchen assistance. The basic objective of cutting is to achieve material fracture via smooth movements of a kitchen knife, which in the process performs work to overcome material toughness, acts against blade-material friction, and generates shape deformation. This paper investigates how a robotic arm drives the knife to cut through an object in a sequence of three moves: pressing, touching, and slicing. To cope with evolving contacts with the material and cutting board, position, force, and impedance controls are applied, either separately or jointly, assisted by force sensing, and/or based on fracture mechanics, so the knife follows a prescribed trajectory to split the object. Force data acquired during the phase of pressing are used for estimating the material’s fracture toughness and the coefficient of blade-material friction (together with pressure distribution), all specific to the food item being cut and varying with its freshness. These estimated values are promptly used for control purpose to execute the phase of slicing. Experiments over several types of fruits and vegetables have exhibited natural cutting movements like those performed by a human hand. Keywords Dexterous cutting, fracture mechanics, knife pressing, slicing, robot control 1 Introduction Automation of kitchen skills is an important step towards the advent of multipurpose home robots, which have long been a public fascination. Until today, robotic kitchen assistance has been limited to peripheral tasks such as washing and sorting dishes (Srinivasa et al., 2012), carrying food trays (Iwata and Sugeno, 2009), and making pancakes and noodles (Bleetz et al., 2011), burgers 1 , etc., from prepared raw materials in very structured settings 2 . Typically in food industry, robots are each capable of only one task, whether cutting meat, deboning, or butchering chicken. Industrial food cutting enjoys high efficiency that benefits from specially designed tools for slicing or holding foods 3 . In our life, not so coincidentally, specialized kitchen tools sold at stores or online are also for single operations such as slicing lettuce, peeling potatoes, chopping fruits and vegetables, and so on. To prepare a meal with vegetables, meats, and fruits with no hand involvement in cutting would require a large collection of such specialized machines. This would be unrealistic given the high cost and limited kitchen space. It would also be clumsy and inefficient to constantly switch between foods and corresponding cutting machines. Robots can play a bigger role in the kitchen by becoming more versatile — even within the domain of one kitchen skill. Food cutting, as an integral part of automatic meal preparation, stands out as one of the ultimate tests on human- level dexterity for robots, which today still lack basic skills such as chop, slice, and dice. There are multiple reasons for the slow technical advance in robotic cutting. Grasping and stabilization of soft and irregularly-shaped food items aside, one main technical challenge is how to plan and control a knife’s movement through a material while reacting to forces of different natures (fracture, friction, viscosity, and contact) exerted by the material and cutting board. A smooth knife movement needs to make use of some estimates of these forces as well as shape deformations, much like they are felt by a knife-holding human hand. Elasticity theories (Saada, 1993; Novozhilov, 1999) and fracture mechanics (Anderson, 2005) can be drawn upon for the purpose of force and deformation modeling. Dexterous robotic cutting needs to resolve a range of issues, from fracture modeling, to knife holding and control, to object stabilization and maneuver, and to coordination of two hands/arms. We intend to take up these challenges one at a time with increasing complexity in this endeavor for years to come. This paper investigates a cutting task in which the knife is rigidly mounted on a robotic arm. To concentrate on knife skill realization, we make the following four assumptions throughout the paper: (A1) The object being cut deforms negligibly. (A2) The object remains stable during cutting. (A3) The knife moves in a vertical plane. (A4) The knife’s blade has negligible thickness. The first assumption arises from that some vegetables and fruits such as potatoes, onions, and apples barely deform Department of Computer Science, Iowa State University, Ames, IA, USA Corresponding author: Yan-Bin Jia, Department of Computer Science, Iowa State University, 131 Atanasoff Hall, Ames, IA, 50011, USA. Email: jia@iastate.edu Prepared using sagej.cls [Version: 2017/01/17 v1.20]