PaperlayerView exact revision on arXiv

Source-linked AI summary

Distributed Cascade Force Control of Soft-Tactile-Based Multi-robot System for Object Transportation

Duy Anh Nguyen, Nhat Minh Dinh Le, Nhan Huu Nguyen, Pham Duy Hung, Van Anh Ho, Trung Dung Ngo

arXiv:2610.04775v1cs.ROcs.SYeess.SY

TL;DR

Maintaining pushing contact while directing an object toward a moving target is a central challenge for multi-robot transport. The paper proposes communication-free distributed cascade force control using soft-tactile robots, and reports successful transport in simulation and real-world experiments. The system is reported to remain stable and converge despite real-world friction uncertainty.

  • Problem

    The task requires robots to maintain pushing contact and direct an object’s motion toward a target.

    Source: §III · p. 3

  • Method

    A distributed cascade force control system uses soft-tactile robots to maintain contact and exert pushing forces without inter-robot communication.

    Source: §I · p. 2

  • Results

    Simulation and real-world experiments validated convergence and stability, and the hardware results showed transport to a moving target despite friction uncertainty.

    Source: C. Discussion · p. 6

  • Takeaways & Limitations

    The two-loop system coordinated robot groups for object transportation with proven convergence and stability.

    Source: §VI · p. 6

    The offset force and number of robots are predefined to counteract static friction, and robots need sufficient space along the object boundary.

    Source: C. Discussion · p. 6

Abstract

from arXiv · show

In this paper, we present a distributed cascade force control system (DCFC) for multiple robots with the aim of pushing a rigid object towards a desired moving target without their inter-robot communication. These mobile robots are equipped with 360-degree vision-based soft tactile sensors utilized to determine contact location and resultant impact force. By investigating the dynamics of moving rigid objects on the flat, we proposed a distributed cascade control. The inner loop control incorporates contact force and positioning, ensuring the robots' pushing contact and application of the desired force to the object. The outer loop control coordinates the robots to push the object in a desired direction without inter-robot communication, regardless of the unknown object mass. The stability and convergence of the system are verified using the Lyapunov stability theory. We also conducted simulation and real-word experiments to validate the performance of the proposed control method, and the experimental results showcase the successful coordination of multiple robots in pushing an object towards a moving desired direction.

I. INTRODUCTION

Multi-robot pushing is challenging because object–surface dynamics matter, yet many strategies do not model them or maintain continuous contact. This paper introduces communication-free distributed cascade control for soft-tactile robots transporting rectangular objects.

I. INTRODUCTION

  • Collective object transportation is relevant to warehouses, construction sites, and search-and-rescue missions.Source: §I · p. 1
  • Pushing avoids the physical connection required for grasping, but depends critically on rigid-object dynamics.Source: §I · p. 1
  • Earlier behavior-based pushing strategies did not account for object–surface interaction or rigid-body dynamics, limiting their treatment of transportation across terrains or slopes.Uncontrolled robot–object forces may exceed safety limits and harm robots or fragile objects.Source: §I · p. 1
  • Continuous pushing remains challenging, and one prior two-robot method maintained formation and relative positioning through continuous intercommunication.Source: §I · p. 2
  • The proposed distributed cascade force control system coordinates soft-tactile robots to push rectangular objects without inter-robot communication.It maintains contact while robots exert pushing forces.Source: §I · p. 2

II. PRELIMINARIES

The paper models a rigid object translating and rotating in a two-dimensional frictional environment under multiple external pushing forces. Ground friction is represented through motion-dependent force and torque terms.

II. PRELIMINARIES

B. Visual-based Tactile Sensor Modeling

The tactile sensor estimates contact force and angle from deformation observed by a camera. The interaction model also specifies when sensor–object friction may be ignored.

B. Visual-based Tactile Sensor Modeling

C. Interaction force between robot and object surface

III. PROBLEM DESCRIPTION

The task is to transport a rectangular rigid object of indeterminate mass toward a specified direction while maintaining pushing contact. The objectives include directing motion and ensuring each robot supplies the necessary force.

III. PROBLEM DESCRIPTION

  • Robots measure their orientation relative to the target direction and estimate their pushing forces, which are assumed to act perpendicular to the object surface.The setup places two robot groups on the same edge, on opposite sides of the center-of-mass projection.Source: §III · p. 3
  • The central challenge is maintaining robot–object pushing contact while synchronizing object velocity with the targeted direction.Source: §III · p. 3
  • When object velocity is unavailable, the control objective instead aligns pushing forces with the target direction and makes their total orthogonal component zero.Under these conditions, object velocity can be adjusted toward the desired direction.Source: §III · p. 3
  • With positive translational friction parameters, cross-target acceleration converges to zero, while constant total pushing force yields a stable target-direction velocity magnitude.Source: §III · p. 3

IV. DISTRIBUTED CASCADE FORCE CONTROL FOR SOFT-TACTILE-BASED ROBOTS (DCFC)

The DCFC separates object-direction coordination from contact and force regulation. Its outer loop sets a distributed force policy, while its inner loop maintains pushing contact and applies the desired force.

IV. DISTRIBUTED CASCADE FORCE CONTROL FOR SOFT-TACTILE-BASED ROBOTS (DCFC)

  • The DCFC combines a distributed outer-loop force policy with an inner multiple-input, multiple-output loop for contact maintenance and desired-force application.The cascade graph depicts these as distinct control responsibilities.Source: §IV · p. 3

A. Outer loop control - Distributed force policy making force direction aligned with the desired direction

The outer loop assigns robots to left and right groups and distributes force policies to align pushing with the desired direction. Its stability analysis relates moment compensation to convergence of the robots’ angle errors.

A. Outer loop control - Distributed force policy making force direction aligned with the desired direction

  • The distributed outer-loop policy is defined for left and right robot groups, with f0, KPo, and KIo as its control parameters.The control graph distinguishes this outer loop from the inner contact-and-force loop.Source: A. Outer loop control - Distributed… · p. 3, Figure 4 · p. 4
  • For rectangular objects, left- and right-group robots produce moments with opposite signs, and the integral term compensates for constant moment when needed.Equal robot counts and uniform object mass make the group moment average approximately zero; otherwise, the integral part compensates for the constant moment.Source: A. Outer loop control - Distributed… · p. 4
  • When the coefficients are positive and b(f0,di) is approximately zero, Lyapunov analysis gives convergence of the angle error to zero at a rate adjustable by KPo and KDo.The stability argument uses the global invariant set theorem.Source: A. Outer loop control - Distributed… · p. 4

B. Inner loop control - Distributed contact maintenance and force control

The inner loop combines contact-angle correction with translational motion to maintain contact and exert the desired force. Under the stated constant-velocity assumption, force error converges to zero.

B. Inner loop control - Distributed contact maintenance and force control

  • The robot’s rolling motion is represented by contact angle θi, while translational motion maintains deformation distance for exerting the desired force.The paper presents a control law for this inner loop and then gives differential-drive wheel-velocity equations.Source: B. Inner loop control - Distributed… · p. 4
  • The inner-loop stability derivation models robot force magnitude and substitutes the control law and force error while assuming vAi is constant over a period.Source: B. Inner loop control - Distributed… · p. 4
  • The inner-loop force error converges to zero at a rate adjustable by KP,in and KI,in.The derivation assumes vAi reaches a constant value over a period of time.Source: B. Inner loop control - Distributed… · p. 4

V. EXPERIMENT RESULT AND DISCUSSION

The section identifies the object, environment, robot, and control parameters used in simulation and real-world experiments. It also sets an offset force to overcome static friction and points to simulation results showing convergence to zero.

V. EXPERIMENT RESULT AND DISCUSSION

  • The simulation and real-world experiments use object and environment parameters listed in Table I and robot and control parameters listed in Table II.The object mass and kinetic friction in the real-world experiment are approximated as in the simulation.Source: §V · p. 4, Table I · p. 5, Table II · p. 5
  • A 1.5 N offset force per robot is set to overcome static friction, with the stated combined force exceeding 1.64 N.The passage describes adaptive estimation of f0 at the beginning of pushing using integral control with velocity feedback as feasible.Source: §V · p. 4
  • The simulation-results figure is described as demonstrating convergence to zero.Source: Figure 5 · p. 5

A. Simulation and Results

The simulation validates convergence of the inner-loop force and contact objectives and the outer-loop motion objective, with the system settling in about six seconds.

A. Simulation and Results

  • The full system settles in approximately 6 seconds.The simulation also reports stable rotation, with object angular velocity approaching zero.Source: A. Simulation and Results · p. 5
  • The measured force consistently converges toward the desired force signal.This is the inner-loop force objective represented by Eq. 13.Source: A. Simulation and Results · p. 5
  • Both robots’ contact angles θ_i converge to zero, satisfying the inner-loop contact objective.The contact-angle objective is represented by Eq. 14.Source: A. Simulation and Results · p. 5
  • The object’s motion direction aligns with the desired direction as its angle error converges to zero.The object’s angular velocity also approaches zero.Source: A. Simulation and Results · p. 5

B. Real-world experiments and Results

Real-world tests in constant- and changing-direction scenarios validate the control objectives: robots maintain pushing contact while the object tracks the desired direction with little deviation.

B. Real-world experiments and Results

C. Discussion

The discussion reports stable, convergent simulation and hardware performance, including transportation to a moving target despite real-world friction uncertainty. The method nevertheless predefines some control settings and has a restricted object-shape scope.

C. Discussion

  • Simulation and hardware experiments validate reasonable convergence time and stability, and hardware tests report transportation to a moving target despite uncertain friction conditions.The surface conditions in real-world experiments do not remain as constant as in the model.Source: C. Discussion · p. 6
  • The policy handles unknown object mass and friction, but f0 and the number of robots are predefined to counteract static friction.Robot positions need not be exact if sufficient boundary space is available.Source: C. Discussion · p. 6
  • The system does not account for every object shape; robots must be distributable on two sides of the center of mass along a straight edge.Triangles and some polygons are given as examples; nonpolygonal geometries and non-uniform mass distributions remain targets for future development.Source: C. Discussion · p. 6

VI. CONCLUSIONS

  • The conclusion characterizes DCFC as enabling soft-tactile-based robots to push an unknown object in a desired direction without communication.The paper also reports desired-force application, maintained pushing contact, and convergence and stability for the two-loop system.Source: §VI · p. 6
Loading 2610.04775v1…