Adaptive robust interaction control for low-cost robotic grasping

Mahboubi Heydarabad, S 2018, Adaptive robust interaction control for low-cost robotic grasping , PhD thesis, University of Salford.

[img] PDF - Accepted Version
Restricted to Repository staff only until 28 September 2020.

Download (16MB)

Abstract

Robotic grasping is a challenging area in the field of robotics. When a gripper starts interacting with an object to perform a grasp, the mechanical properties of the object (stiffness and damping) will play an important role. A gripper which is stable in isolated conditions, can become unstable when coupled to an object. This can lead to the extreme condition where the gripper becomes unstable and generates excessive or insufficient grip force resulting in the grasped object either being crushed, or falling and breaking.

In addition to the stability issue, grasp maintenance is one of the most important requirements of any grasp where it guarantees a secure grasp in the presence of any unknown disturbance. The term grasp maintenance refers to the reaction of the controller in the presence of external disturbances, trying to prevent any undesired slippage. To do so, the controller continuously adjusts the grip force. This is a challenging task as it requires an accurate model of the friction and object’s weight to estimate a sufficient grip force to stop the object from slipping while incurring minimum deformation.

Unfortunately, in reality, there is no solution which is able to obtain the mechanical properties, frictional coefficient and weight of an object before establishing a mechanical interaction with it. External disturbance forces are also stochastic meaning they are impossible to predict.

This thesis addresses both of the problems mentioned above by:

  • Creating a novel variable stiffness gripper, capable of grasping unknown objects, mainly those found in agricultural or food manufacturing companies. In addition to the stabilisation effect of the introduced variable stiffness mechanism, a novel force control algorithm has been designed that passively controls the grip force in variable stiffness grippers. Due to the passive nature of the suggested controller, it completely eliminates the necessity for any force sensor. The combination of both the proposed variable stiffness gripper and the passivity based control provides a unique solution for the stable grasp and force control problem in tendon driven, angular grippers.
  • Introducing a novel active multi input-multi output slip prevention algorithm. The algorithm developed provides a robust control solution to endow direct drive parallel jaw grippers with the capability to stop held objects from slipping while incurring minimum deformation; this can be done without any prior knowledge of the object’s friction and weight. The large number of experiments provided in this thesis demonstrate the robustness of the proposed controller when controlling parallel jaw grippers in order to quickly grip, lift and place a broad range of objects firmly without dropping or crushing them. This is particularly useful for teleoperation and nuclear decommissioning tasks where there is often no accurate information available about the objects to be handled. This can mean that pre-programming of the gripper is required for each different object and for high numbers of objects this is impractical and overly time-consuming. A robust controller, which is able to compensate for any uncertainties regarding the object model and any unknown external disturbances during grasping, is implemented.

This work has advanced the state of the art in the following two main areas:
  •  Direct impedance modulation for stable grasping in tendon driven, angular grippers.
  •  Active MIMO slip prevention grasp control for direct drive parallel jaw grippers.

Item Type: Thesis (PhD)
Contributors: Davis, ST (Supervisor) and Nefti-Meziani, S (Supervisor)
Schools: Schools > School of Computing, Science and Engineering > Salford Innovation Research Centre
Funders: Marie Curie Actions
Depositing User: Saber Mahboubi heydarabad
Date Deposited: 03 Oct 2018 09:32
Last Modified: 03 Oct 2018 10:00
URI: http://usir.salford.ac.uk/id/eprint/48346

Actions (login required)

Edit record (repository staff only) Edit record (repository staff only)

Downloads

Downloads per month over past year