Slip-Aware In-Hand Object Manipulation with Parallel Grippers
Doktorsavhandling, 2026
Allowing controlled in-hand slip, rather than imposing rigid grasps, can substantially expand the manipulation capabilities of robotic systems equipped with parallel grippers. Such capabilities include more efficient object reorientation and increased effective reach. This thesis establishes the foundations for slip-aware in-hand manipulation with parallel grippers and demonstrates that such capabilities can be achieved using only a single-degree-of-freedom gripper and tactile sensing. To this end, the thesis contributes to advances in planar friction modeling and estimation, tactile sensing, gripper hardware design, and slip-aware control.
Robotic grippers interact with objects through the contact surface between their fingers and the object. We explore and model the friction at this interface, focusing on planar motion where translation and rotation occur simultaneously. This creates coupled tangential and torsional friction, which we model by extending the LuGre model using limit surface theory.
To experimentally realize slip-aware manipulation, we develop two generations of tactile sensors and gripper hardware. The first combines a high-performance force-controlled parallel gripper with commercial force/torque sensors and custom relative-velocity sensors based on optical mouse technology. This enables slip-aware manipulation using only in-hand perception. We use this system to estimate friction and contact properties and demonstrate several manipulation controllers, including gravity-assisted linear and rotational trajectory following, hinge control, and slip-avoidance. This framework is further extended with online estimation of contact properties, eliminating the need for a separate exploration phase.
The second-generation tactile sensor introduces a deformable sensing pad that measures sliding velocity, forces, torques, and pressure distribution. Using these sensors and an updated gripper, we demonstrate a variety of in-hand manipulation primitives by combining in-hand control with end-effector motion. Together, the extended friction model and contact-property estimation, tactile sensing and gripper hardware, and slip-aware manipulation controllers establish a unified basis for controlled slip and demonstrate versatile dexterous manipulation with simple parallel grippers.
Robotic grippers interact with objects through the contact surface between their fingers and the object. We explore and model the friction at this interface, focusing on planar motion where translation and rotation occur simultaneously. This creates coupled tangential and torsional friction, which we model by extending the LuGre model using limit surface theory.
To experimentally realize slip-aware manipulation, we develop two generations of tactile sensors and gripper hardware. The first combines a high-performance force-controlled parallel gripper with commercial force/torque sensors and custom relative-velocity sensors based on optical mouse technology. This enables slip-aware manipulation using only in-hand perception. We use this system to estimate friction and contact properties and demonstrate several manipulation controllers, including gravity-assisted linear and rotational trajectory following, hinge control, and slip-avoidance. This framework is further extended with online estimation of contact properties, eliminating the need for a separate exploration phase.
The second-generation tactile sensor introduces a deformable sensing pad that measures sliding velocity, forces, torques, and pressure distribution. Using these sensors and an updated gripper, we demonstrate a variety of in-hand manipulation primitives by combining in-hand control with end-effector motion. Together, the extended friction model and contact-property estimation, tactile sensing and gripper hardware, and slip-aware manipulation controllers establish a unified basis for controlled slip and demonstrate versatile dexterous manipulation with simple parallel grippers.
Grasp Control
Hardware
Tactile Sensing
Perception
In-Hand Manipulation
Friction Modeling
Friction Estimation
Robotics
EA, Hörsalsvägen 11
Opponent: Prof. Roberto Calandra, Head of the Learning, Adaptive Systems and Robotics (LASR) lab, Technische Universität Dresden, Germany
Författare
Gabriel Arslan Waltersson
Chalmers, Elektroteknik, System- och reglerteknik
Planar Friction Modelling with LuGre Dynamics and Limit Surfaces
IEEE Transactions on Robotics,;Vol. 40(2024)p. 3166-3180
Artikel i vetenskaplig tidskrift
Perception, control, and hardware for in-hand slip-aware object manipulation with parallel grippers
International Journal of Robotics Research,;Vol. 45(2026)p. 1497-1519
Artikel i vetenskaplig tidskrift
Gabriel Arslan Waltersson, Marko Guberina, Yiannis Karayiannidis. A Vocabulary of Slip-Aware In-Hand Motion Primitives
What happens when you make an assumption, like that robot hands are supposed to firmly grasp an object, but instead allow it to slip?
To demonstrate what I have been researching, I often take out my phone and move it around in my hand—sliding and reorienting it without dropping it. Then, I do the same thing without looking. At this point, the crowd is usually not very impressed. For us humans, using our hands is easy—almost instinctive. It is something we rarely need to think about. For robots, however, achieving this kind of dexterity is far from simple. The challenges range from material science to perception, planning, and control. In this thesis, I focus on allowing objects to slide within simple robotic grippers. These grippers can only open and close, yet through tactile sensing and fast actuation, we can precisely control how an object moves within the hand.
By exploiting this controlled motion instead of repeatedly repositioning the object to achieve a new grasp, we can add dexterity to simple grippers, allowing robots to solve tasks faster, extend their reach, and operate more energy efficiently. At the same time, this controlled motion gives us a unique opportunity to study how the object and gripper interact as they slide against each other. Rather than trying to avoid slip, we use it to better understand and control the interaction between the robot and the object. I view my research as consisting of two phases. First, we theoretically study the mechanics of sliding motion when the gripper is in contact with the object. After all, before we can control this motion, we first need to understand the forces that produce it. With these models in hand, the second phase focuses on giving the robot the ability to sense the slipping motion of the object and respond to it in real time. This requires dedicated sensing and actuation hardware, together with perception and control algorithms, to bring these ideas to life.
To demonstrate what I have been researching, I often take out my phone and move it around in my hand—sliding and reorienting it without dropping it. Then, I do the same thing without looking. At this point, the crowd is usually not very impressed. For us humans, using our hands is easy—almost instinctive. It is something we rarely need to think about. For robots, however, achieving this kind of dexterity is far from simple. The challenges range from material science to perception, planning, and control. In this thesis, I focus on allowing objects to slide within simple robotic grippers. These grippers can only open and close, yet through tactile sensing and fast actuation, we can precisely control how an object moves within the hand.
By exploiting this controlled motion instead of repeatedly repositioning the object to achieve a new grasp, we can add dexterity to simple grippers, allowing robots to solve tasks faster, extend their reach, and operate more energy efficiently. At the same time, this controlled motion gives us a unique opportunity to study how the object and gripper interact as they slide against each other. Rather than trying to avoid slip, we use it to better understand and control the interaction between the robot and the object. I view my research as consisting of two phases. First, we theoretically study the mechanics of sliding motion when the gripper is in contact with the object. After all, before we can control this motion, we first need to understand the forces that produce it. With these models in hand, the second phase focuses on giving the robot the ability to sense the slipping motion of the object and respond to it in real time. This requires dedicated sensing and actuation hardware, together with perception and control algorithms, to bring these ideas to life.
Ämneskategorier (SSIF 2025)
Robotik och automation
DOI
10.63959/chalmers.dt/5936
ISBN
978-91-8103-479-0
Doktorsavhandlingar vid Chalmers tekniska högskola. Ny serie: 5936
Utgivare
Chalmers
EA, Hörsalsvägen 11
Opponent: Prof. Roberto Calandra, Head of the Learning, Adaptive Systems and Robotics (LASR) lab, Technische Universität Dresden, Germany