Novel energy-optimal solutions to the inverse kinematics problem of redundant manipulators

  • Alessandro Tringali

Student thesis: Doctoral Thesis

Abstract

Robotic manipulators are today used in many industrial and field applications, especially in the kinematically redundant ones, which offer the possibility to approach a specific end-effector pose in infinitive different ways. This freedom allows to optimise secondary task alongside the main goal to drive the end-effector through a specific trajectory, and such optimisation problem has motivated the work of many researchers in the last decades. Historically, the energy of the manipulator has been considered a particularly important optimisation cost function: it is relevant in industrial settings, where manipulators operations are a standing cost, and even more in field environments, such as space, where the available power is limited.;This thesis presents a study about inverse kinematics algorithms for redundant manipulators, aimed at optimising the energy required to perform manipulation tasks. First, a literature review surveying inverse kinematics and optimisation for both fixed-base and free-floating manipulators is presented. This presents the state of the art in the field and illustrates the motivation for this thesis. It also outlines the main challenges encountered in the development of optimisation algorithms for redundant manipulators.;After this, two algorithms are presented and discussed within the thesis, a global and a local one. Both are based on nonlinear optimisation techniques. The global problem is talked first, and a method is proposed that can optimise different cost functions related to either kinetic energy or torques, with linear and nonlinear constraints, such as torque, power, and periodic motion. Furthermore, the algorithm is able to individuate multiple optima when they are present, thus increasing the chances to find the best (global) optimum.;A local algorithm based on prediction of kinetic energy integral has also been developed. In order to illustrate related challenges, a workspace analysis is first presented that illustrates difficulties in providing reliable prediction of kinetic energy values along a specific end effector trajectory. Kinematic indexes are discussed through a qualitative and quantitative analysis aimed at assessing their correlation with kinetic energy, and results of a canonical correlation analysis is presented. Furthermore, it is illustrated that the a spaceborne robotic manipulator can be controlled concurrently with the Attitude and Orbit Control Systems of spacecraft, adding extra degrees of freedom.;Following this, a local algorithm based on a predictive estimation of kinetic energy integral along a specified trajectory is presented and discussed. This algorithm is based on a simplified optimisation problem that allows to assess the direction of motion that will cause the smallest increase in the kinetic energy integral. This produces solutions that are closer to the global optimum respect to traditional algorithms.;Simulations with a 3-DoF planar robot are used to validate the results. The global method is validated against a global algorithm existing in literature and shown to be able to solve a wider class of problems. The local algorithm is statistically compared against existing inverse kinematics methods, showing a reduction in kinetic energy up to 30%. The thesis is completed by a discussion about limits and further improvements of the work hereby presented.
Date of Award18 Feb 2021
Original languageEnglish
Awarding Institution
  • University Of Strathclyde
SponsorsUniversity of Strathclyde & Ministry of Science and Technology, P R China
SupervisorXiu Yan (Supervisor) & Erfu Yang (Supervisor)

Cite this

'