Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Cooperative Assembly with Autonomous Mobile Manipulators in an Underwater Scenario

Davide Torielli

TL;DR

This work addresses cooperative peg-in-hole assembly by two autonomous underwater manipulators, focusing on kinematic control under limited communications. It combines Task Priority Inverse Kinematics (TPIK) with an Arm-Vehicle coordination scheme and a cooperative scheme to drive a shared tool while minimizing data exchange; it also introduces a Force-Torque objective and a Change Goal routine to reduce contact forces and compensate pose errors during insertion. A third robot handles vision to estimate the hole pose, using detection (Find Square or Template Matching) and markerless tracking (ViSP) across mono, stereo, and RGB-D setups. Simulation in UWSim demonstrates the approach under perfect and imperfect hole poses, with and without vision, showing that Force-Torque feedback and goal-frame adjustment improve insertion robustness and reduce excessive contact. The work contributes a modular, scalable framework for cooperative underwater manipulation, offering practical pathways toward more autonomous, reliable intervention missions and guiding future dynamic and communication-aware extensions.

Abstract

[...] Specifically, the problem addressed is an assembly one known as the peg-in-hole task. In this case, two autonomous manipulators must carry cooperatively (at kinematic level) a peg and must insert it into an hole fixed in the environment. Even if the peg-in-hole is a well-known problem, there are no specific studies related to the use of two different autonomous manipulators, especially in underwater scenarios. Among all the possible investigations towards the problem, this work focuses mainly on the kinematic control of the robots. The methods used are part of the Task Priority Inverse Kinematics (TPIK) approach, with a cooperation scheme that permits to exchange as less information as possible between the agents (that is really important being water a big impediment for communication). A force-torque sensor is exploited at kinematic level to help the insertion phase. The results show how the TPIK and the chosen cooperation scheme can be used for the stated problem. The simulated experiments done consider little errors in the hole's pose, that still permit to insert the peg but with a lot of frictions and possible stucks. It is shown how can be possible to improve (thanks to the data provided by the force-torque sensor) the insertion phase performed by the two manipulators in presence of these errors. [...]

Cooperative Assembly with Autonomous Mobile Manipulators in an Underwater Scenario

TL;DR

This work addresses cooperative peg-in-hole assembly by two autonomous underwater manipulators, focusing on kinematic control under limited communications. It combines Task Priority Inverse Kinematics (TPIK) with an Arm-Vehicle coordination scheme and a cooperative scheme to drive a shared tool while minimizing data exchange; it also introduces a Force-Torque objective and a Change Goal routine to reduce contact forces and compensate pose errors during insertion. A third robot handles vision to estimate the hole pose, using detection (Find Square or Template Matching) and markerless tracking (ViSP) across mono, stereo, and RGB-D setups. Simulation in UWSim demonstrates the approach under perfect and imperfect hole poses, with and without vision, showing that Force-Torque feedback and goal-frame adjustment improve insertion robustness and reduce excessive contact. The work contributes a modular, scalable framework for cooperative underwater manipulation, offering practical pathways toward more autonomous, reliable intervention missions and guiding future dynamic and communication-aware extensions.

Abstract

[...] Specifically, the problem addressed is an assembly one known as the peg-in-hole task. In this case, two autonomous manipulators must carry cooperatively (at kinematic level) a peg and must insert it into an hole fixed in the environment. Even if the peg-in-hole is a well-known problem, there are no specific studies related to the use of two different autonomous manipulators, especially in underwater scenarios. Among all the possible investigations towards the problem, this work focuses mainly on the kinematic control of the robots. The methods used are part of the Task Priority Inverse Kinematics (TPIK) approach, with a cooperation scheme that permits to exchange as less information as possible between the agents (that is really important being water a big impediment for communication). A force-torque sensor is exploited at kinematic level to help the insertion phase. The results show how the TPIK and the chosen cooperation scheme can be used for the stated problem. The simulated experiments done consider little errors in the hole's pose, that still permit to insert the peg but with a lot of frictions and possible stucks. It is shown how can be possible to improve (thanks to the data provided by the force-torque sensor) the insertion phase performed by the two manipulators in presence of these errors. [...]
Paper Structure (63 sections, 20 equations, 36 figures, 2 algorithms)

This paper contains 63 sections, 20 equations, 36 figures, 2 algorithms.

Table of Contents

  1. Introduction
  2. State of the Art
  3. Previous Works in Underwater Missions
  4. The Control Framework
  5. The Peg-in-Hole Assembly Problem
  6. Motivation and Rationale
  7. Control Framework
  8. Definitions
  9. Control Objectives
  10. Control Objectives Classification
  11. Equality and Inequality Control Objectives
  12. Reactive and non Reactive Control Task
  13. Inequality Control Objectives Activation and Deactivation
  14. Task Priority Inverse Kinematics
  15. Notes on Conflicting Objectives
  16. ...and 48 more sections

Figures (36)

  • Figure 1: A scheme showing the two Task Priority Inverse Kinematics blocks for the arm-vehicle coordination
  • Figure 2: The frames of the two cooperative vehicles carrying a common object
  • Figure 3: The cooperation scheme with its different steps. The Agent b block is equal to the Agent a one.
  • Figure 4: The Scenario of the experiment. The two twin robots are carrying the peg, while the third robot is watching the hole to estimate its pose.
  • Figure 5: Schematic recap of the simulation comparison taken from usvsim. $\times$ stands for no implemented feature; $\surd\,$ for a feature that is a discrete representation of the real one; $\surd\surd\,$ for a good feature reproduction of the real one. More details on how each feature is evaluated are available in the original paper.
  • ...and 31 more figures