Overview
- Date:Starts 17 October 2024, 13:15Ends 17 October 2024, 16:00
- Location:SB-H3, Sven Hultins Gata 6, Chalmers.
- Opponent:Assoc. Prof. Christian Smith, Division of Robotics, Perception and Learning, KTH - Royal Institute of Technology
- ThesisRead thesis (Opens in new tab)
In-hand perception and object manipulation are key areas that can significantly extend the capabilities of robotic systems. By enabling robots to sense and adapt to their environment, they can perform manipulations with new or unknown objects. Applications range from safely handling delicate items to increasing the workable space of a manipulator and achieving more efficient motions. This thesis advances these capabilities from multiple angles, including the development of new friction models for simulating in-hand slippage, as well as new sensors and parallel gripper hardware for real-world experimentation.
The robotic gripper interacts with objects through the contact surface between its fingers and the object. In this work, we explore and model the friction that occurs at this interface. During planar motion, where both tangential and angular velocities are present, a coupling arises between the tangential and torsional friction forces. We propose planar friction models based on the LuGre model, which captures this coupling using limit surface theory. Two friction models are introduced: a distributed planar friction model that discretizes the contact surface as a baseline, and a faster, numerically efficient model that leverages a pre-computed limit surface.
Slip-aware in-hand manipulation has not yet reached the maturity required for commercialization and readily available hardware. To address this, we designed a custom parallel gripper specifically for fast, closed-loop force control. The gripper is equipped with force-torque sensors and custom relative velocity sensors based on optical mouse technology. This hardware combination enables slip-aware manipulation using only in-hand perception. We demonstrate friction and contact property estimation from an exploration phase, along with four distinct slip-aware controllers. The four slip controllers include trajectory-following for gravity-assisted linear and rotational slippage, hinge control, and slip avoidance.
The robotic gripper interacts with objects through the contact surface between its fingers and the object. In this work, we explore and model the friction that occurs at this interface. During planar motion, where both tangential and angular velocities are present, a coupling arises between the tangential and torsional friction forces. We propose planar friction models based on the LuGre model, which captures this coupling using limit surface theory. Two friction models are introduced: a distributed planar friction model that discretizes the contact surface as a baseline, and a faster, numerically efficient model that leverages a pre-computed limit surface.
Slip-aware in-hand manipulation has not yet reached the maturity required for commercialization and readily available hardware. To address this, we designed a custom parallel gripper specifically for fast, closed-loop force control. The gripper is equipped with force-torque sensors and custom relative velocity sensors based on optical mouse technology. This hardware combination enables slip-aware manipulation using only in-hand perception. We demonstrate friction and contact property estimation from an exploration phase, along with four distinct slip-aware controllers. The four slip controllers include trajectory-following for gravity-assisted linear and rotational slippage, hinge control, and slip avoidance.