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 act 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 object-specific values of physical parameters related to cutting. 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 resembling those performed by a human hand. Note to Practitioners—Automation of kitchen skills is an important step in the development of home robots, which are expected to relieve us from daily chores and help us care for the elderly and people with disabilities. The motivation of this research is to enable a robotic arm to cut natural foods with knife movements that bear the smoothness and efficiency of those executed by a human hand. Existing methods on robotic cutting have focused on force control to ensure material separation but not on execution of natural knife movements. This paper dissects a cutting action into three phases, as inspired from the human hand execution, and realizes them via different control policies. These policies are based on sensing and modeling the forces experienced by the knife through its interactions with the material and the cutting board. Preliminary experimental results have demonstrated cutting of various food items with speed and smoothness. In future research, we will address cutting of deformable objects and explore issues including energy efficiency, cutting by the knife held in a robotic hand, and food stabilization and manipulation by a second arm/hand.

Vegetable Peeling: A Case Study in Constrained Dexterous Manipulation

Dexterous Manoeuvre Through Touch in a Cluttered Scene

Visual Dexterity: In-Hand Reorientation of Novel and Complex Object Shapes

MORPHeus: a Multimodal One-armed Robot-assisted Peeling System with Human Users In-the-loop

Learning to use chopsticks in diverse gripping styles

Human-inspired Grasping Strategies of Fresh Fruits and Vegetables Applied to Robotic Manipulation

Tactile-Driven Non-Prehensile Object Manipulation via Extrinsic Contact Mode Control

Object-Centric Dexterous Manipulation from Human Motion Data

ForceMimic: Force-Centric Imitation Learning with Force-Motion Capture System for Contact-Rich Manipulation

Dexterous Manipulation from Images: Autonomous Real-World RL via Substep Guidance

A High-Efficient Reinforcement Learning Approach for Dexterous Manipulation

ReKep: Spatio-Temporal Reasoning of Relational Keypoint Constraints for Robotic Manipulation

Robotic in-hand manipulation with relaxed optimization

Constrained Reinforcement Learning for Dexterous Manipulation

Rotating without Seeing: Towards In-hand Dexterity through Touch

Dexterous Robotic Cutting Based on Fracture Mechanics and Force Control

Deep Dynamics Models for Learning Dexterous Manipulation

TacRot: A Parallel-Jaw Gripper with Rotatable Tactile Sensors for In-Hand Manipulation

REBOOT: Reuse Data for Bootstrapping Efficient Real-World Dexterous Manipulation

Learning Extrinsic Dexterity with Parameterized Manipulation Primitives

Learning thin deformable object manipulation with a multi-sensory integrated soft hand