Papers
Topics
Authors
Recent
Search
2000 character limit reached

Rhombot: Modular Reconfigurable Robot

Updated 3 February 2026
  • Rhombot is a planar lattice modular self-reconfigurable robot that leverages a unique morphpivoting motion primitive to maintain global connectivity during continuous configuration changes.
  • The system uses a single degree-of-freedom actuation via a centrally mounted servo-driven cable and integrates fixed docking electromagnets to achieve sub-centimeter precision.
  • Experimental results demonstrate robust, repeatable reconfiguration cycles with RMSE around 4.8 mm in x and 15 mm in y, confirming the system’s high positional and docking accuracy.

Rhombot is a planar lattice modular self-reconfigurable robot (MSRR) constructed from rhombus-shaped modules designed for robust, medium-independent reconfiguration. Each module is a deformable parallelogram actuated through a single centrally mounted servo-driven cable. The system achieves morphing, docking, and locomotion with low control complexity. Rhombot's central contribution is the morphpivoting motion primitive, enabling continuous reconfiguration that maintains global structural connectivity. Experimental results demonstrate sub-centimeter positional and docking accuracy, as well as reliable operation under repeated reconfiguration cycles (Gu et al., 27 Jan 2026).

1. Rhombus Module Design

Each Rhombot module consists of a planar rhombus skeleton of rigid carbon-fiber links with side length $2a$, where a=140a = 140 mm. Opposite edges remain parallel through actuation, while a central parallelogram linkage enforces a single folding degree of freedom (θ\theta) defined as the interior angle between adjacent edges. The skeleton actively constrains its two diagonals such that 45θ13545^\circ \leq \theta \leq 135^\circ, preventing undesired planar rotation.

A single micro-servo motor drives a cable winch (radius r=5r = 5 mm), using a bevel-gear reduction (Z1=12Z_1=12, Z2=24Z_2=24 teeth) to increase delivered torque. The cable, looped beneath the skeleton and routed through guide pulleys, symmetrically pulls a pair of opposite vertices together or allows them apart, folding the rhombus along its long diagonal. This actuation both morphs the shape and realizes the morphpivoting motion.

Key module mechanical parameters include:

Parameter Value Description
aa 140 mm Half edge length
bb 70 mm Distance: module center to electromagnet
θ\theta 4513545^\circ-135^\circ Folding angle range
TT (servo stall torque) 4.5 kg·cm Torque capability
rr (winch radius) 5 mm
FeF_e (electromagnet) \approx25 N Docking holding force

The actuation moment between two adjacent edges for folding angle θ\theta is:

Md=TZ1rZ2  2asin(θ2)M_d = \frac{T\,Z_1}{r\,Z_2}\;2a\,\sin\Bigl(\frac{\theta}{2}\Bigr)

This compact mechanical package minimizes vertical profile, supports a single-degree-of-freedom (DoF) actuation, and embeds four docking electromagnets at fixed intervals for module connectivity.

2. Kinematic and Dynamic Modeling

Individual modules are modeled as quadrilaterals MiM_i with edges Ei0E_{i0}Ei3E_{i3}, where Ei0E_{i0} is typically oriented "downward" (grounded or parental), and the others serve as inter-module interfaces. Each module's local coordinate frame OiO_i is defined at the Ei0E_{i0} midpoint, with yiy_i directed toward the center and ziz_i perpendicular to the module plane.

The single DoF is parameterized by σi\sigma_i (σiθi\sigma_i \equiv \theta_i in the undeformed configuration). Homogeneous transforms from Ei0E_{i0} to midpoints of adjacent edges are:

  • T10(σ)T^0_1(\sigma):

(Rz(σ+π)[a+acosσ,  asinσ,  0]T 0  0  01)\begin{pmatrix} R_z(\sigma+\pi) & [\,a + a\cos\sigma,\;a\sin\sigma,\;0\,]^T \ 0\;0\;0 & 1 \end{pmatrix}

  • T20(σ)T^0_2(\sigma):

(I3[2acosσ,  2asinσ,  0]T 0  0  01)\begin{pmatrix} I_3 & [\,2a\cos\sigma,\;2a\sin\sigma,\;0\,]^T \ 0\;0\;0 & 1 \end{pmatrix}

  • T30(σ)T^0_3(\sigma):

(Rz(σ)[aacosσ,  asinσ,  0]T 0  0  01)\begin{pmatrix} R_z(\sigma) & [\,a - a\cos\sigma,\;a\sin\sigma,\;0\,]^T \ 0\;0\;0 & 1 \end{pmatrix}

Where Rz()R_z(\cdot) is a 3×33 \times 3 rotation about zz.

The forward kinematics for a chain of connected modules, forming a tree or loop (triangulated by constraint), is:

bTe=j=0eT(kj)(σkj){}^bT_e = \prod_{j=0}^e T^{(k_j)}(\sigma_{k_j})

Actuator requirements for connection/disconnection are set by the frictional resisting torque of the electromagnetic connector:

Mf=Fe(2ab)+εM_f = F_e\,(2a - b) + \varepsilon

with ε0.84\varepsilon \approx 0.84 kg·cm. To guarantee single-sided release during docking, Md>MfM_d > M_f is always enforced.

3. Morphpivoting Motion Primitive

Morphpivoting constitutes the atomic reconfiguration operation, defined as pivoting a single module about a still-connected edge by exploiting the module’s ability to fold and unfold. Execution occurs over four sequential phases:

  1. Pre-pivot morphing: Both pivoting and adjacent modules adjust θ\theta to an intermediate σpre\sigma_{\rm pre} to create mechanical clearance.
  2. Connection: The electromagnet on a previously unconnected adjacent edge-pair is activated.
  3. Disconnection: The former connector is deactivated, so only the new axis forms the pivot.
  4. Post-pivot morphing: Both modules restore angles to σpost\sigma_{\rm post}, finalizing the new configuration.

The transformation for the connection shift, considering module MaM_a with neighbors MbM_b (old) and McM_c (new), involves computing the net transformation between relevant edge frames before and after:

FnewXMa=[TpreFnew(σpre)][TpreFold(σpre)]1FoldXMa{}^{F_{\rm new}}X_{M_a} = \Bigl[T^{F_{\rm new}}_{\rm pre}(\sigma_{\rm pre})\Bigr]\, \Bigl[T^{F_{\rm old}}_{\rm pre}(\sigma_{\rm pre})\Bigr]^{-1}\,{}^{F_{\rm old}}X_{M_a}

This formalism supports issuing four servo-angle commands per module to perform each morphpivot.

4. Reconfiguration Planning and Control

Reconfiguration proceeds via a planning algorithm that takes as input a list of desired (Connect, Disconnect) edge-pairs. The current kinematic structure is maintained as a tree (KTree), which is updated after each primitive by re-establishing parent/child relationships.

For each requested edge-pair exchange, the system:

  • Identifies all modules and edges implicated
  • Invokes the Morphpivoting primitive with appropriate inputs
  • Updates the KTree and recomputes all σi\sigma_i mappings

Control is decentralized and minimal: each module contains a single actuator, a single angular sensor (AS5600 magnetic encoder), and independent on/off control of its four docking electromagnets. No external vision or global localization is required; all docking frames are defined from the local kinematic tree.

Throughout all steps, the morphpivoting approach ensures modules never fully detach, maintaining at least one connection and retaining global topological stability at all times.

5. Structural Stability and Precision Docking

During every morphpivoting action, the pivoting module keeps at least one edge connection, enabling uninterrupted global connectivity of the MSRR. The robot maintains static stability by keeping each module's center-of-mass within its support polygon, leveraging omnidirectional ball casters for low-friction ground support.

The docking system utilizes hermaphroditic connectors comprised of a normally‐on electromagnet and an iron plate. The nominal state is locked (powered off); power is applied only for intentional disconnection. The relationship Md>MfM_d > M_f ensures controlled, single-sided release.

Empirical positional accuracy is evidenced by kinematic-chain tests (4 modules), with end-effector root mean squared error (RMSE) 4.8\approx 4.8 mm in xx and $15$ mm in yy (the increased yy error is due to friction). During seven sequential morphpivotings, all connector offsets remained within the mechanical tolerance range of the electromagnet (a few millimeters), validating reliable and repeatable sub-centimeter docking precision.

6. Experimental Results, Limitations, and Future Work

The hardware deployment comprises multiple Rhombot modules on a ball-caster chassis atop a printed grid, powered by a wired 12 V Li-ion battery and coordinated through wireless nRF24L01 modules. Each module is controlled via Arduino Nano microcontrollers, servo drivers, and AS5600 angle sensors.

Demonstrated reconfigurations include three-module morphing (e.g., forming a 3-loop structure, reconnecting, and restoring shape) and simulated seven-module transitions (e.g., “μ” → “F”). Single servos in each module are responsible for all morphing and docking; robust operation was observed over many morphpivot cycles with no uncommanded disconnections.

Quantitative findings are as follows:

Metric Value
Kinematic-chain RMSE $4.77$ mm (xx), $14.96$ mm (yy)
Docking accuracy Within sub-centimeter mechanical tolerance
Robustness No spontaneous disconnections (dozens of cycles)

Limitations include cable slack-induced hysteresis of approximately 1500 encoder counts near end stops (Δθ131\Delta\theta \approx 131^\circ), mitigated by hardware tensioners. Anticipated improvements cite the use of an irregular-profile winch and development of an autonomous high-level sequencer for flexible, automatic (Connect, Disconnect) pair selection (Gu et al., 27 Jan 2026).

Rhombot demonstrates that single-DoF, rhombus-shaped modules with centrally mounted actuation yield stable, medium-independent self-reconfiguration. The morphpivoting primitive, in conjunction with robust local kinematic trees and electromagnetic docking, achieves low control complexity alongside high positional and docking accuracy.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (1)

Topic to Video (Beta)

No one has generated a video about this topic yet.

Whiteboard

No one has generated a whiteboard explanation for this topic yet.

Follow Topic

Get notified by email when new papers are published related to Rhombot.