Source-linked AI summary
Tensegrity Continuum Robots Enable Task-Adaptive Morphologies for Cooperative Behaviors
Mahmud Hasan Saikot, Sydney Spiegel, Sudheera Akalanka Kariyawasam, Andrew Stefka, Josh Chrisler, Jianguo Zhao
TL;DR
Existing continuum robots provide compliance but generally cannot dock and reassemble into multi-robot configurations. This paper introduces TeCoBot, a tensegrity-based continuum robot with claw connections, and demonstrates task-adaptive morphologies for cooperative manipulation, locomotion, and loco-manipulation.
Problem
Existing continuum robots provide compliance but generally function as standalone systems rather than docking, detaching, and reassembling into multi-robot configurations.
Method
TeCoBot combines a compliant tensegrity continuum body that bends and collapses axially with claw-based mechanisms for modular multi-robot reconfiguration.
Results
TeCoBots self-reconfigure into task-adaptive morphologies that support cooperative manipulation, locomotion, and loco-manipulation across indoor and outdoor scenarios.
Takeaways & Limitations
The work demonstrates a robotic collective combining local compliance with global multi-robot reconfiguration for adaptable cooperative behaviors.
Takeaways & Limitations
Self-reconfiguration is teleoperated by human operators, although autonomous docking is demonstrated using pre-scripted motions or onboard vision and fiducial-marker feedback.
Abstract
from arXiv · showhide
Robots that can change their morphologies and behaviors for different tasks and environments hold great promise for adaptable, multifunctional systems. Modular reconfigurable robots (MRRs) can achieve such functionalities by docking and rearranging individual units, but most rely on rigid modules that lack structural compliance, resulting in limited capabilities. Continuum robots offer compliance through flexible backbones, yet they cannot self-reconfigure into task-adaptive multi-robot configurations. Here, we introduce an MRR that unifies the advantages of both architectures by combining a tensegrity-based compliant body with claw-based connection mechanisms. Each robot can manipulate and locomote independently, and multiple robots can self-reconfigure into different morphologies (e.g., chains, loops, branches) for cooperative manipulation and locomotion. We demonstrate the robots' capability across diverse tasks and environments, including coordinated object manipulation and transport, multimodal locomotion, and loco-manipulation in real-world scenarios. These results lay a foundation for adaptable and multifunctional robotic collectives, with broad potential applications in manufacturing, space exploration, and search-and-rescue operations.
INTRODUCTION
TeCoBot combines the compliance of continuum robots with the modular reconfigurability of MRRs in a self-reconfigurable tensegrity robot. Individual and connected robots support adaptive manipulation, locomotion, and cooperative loco-manipulation across diverse environments.
- Motivation: MRRs reconfigure through docking and rearrangement but generally use rigid modules, whereas continuum robots provide compliance for constrained and complex environments.The introduction contrasts adaptable modular architectures with the environmental advantages of flexible or extensible backbones.
- Robot concept: TeCoBot stacks multiple tensegrity modules into a compliant continuum body with claw-based connections, enabling axial collapse/extension and omnidirectional bending.Cable-driven actuation operates the stacked tensegrity modules, while modular claws support connection and manipulation.
- Individual capabilities: Each TeCoBot independently manipulates and locomotes, using body deformation and dual modular claws to operate within a 3D workspace.The overview identifies deformation-enabled manipulation and locomotion as single-robot capabilities.
- Cooperative loco-manipulation: Integrated multi-robot behaviors support complex loco-manipulation sequences such as climbing, object transport, and handover through dynamic reconfiguration.These behaviors are demonstrated across different environments using reconfigurable assemblies.
- Cooperative capabilities: Multiple TeCoBots self-reconfigure into task-adaptive morphologies, including chains, loops, and branches, for cooperative manipulation and locomotion.Examples include lifting a large ball, removing a container lid, rolling around pipes, and climbing ducts.
Characterization
Characterization shows that TeCoBot’s deformation-force trade-offs depend on tendon geometry and module count, while claw docking performance depends mainly on sleeve material and attachment type. A constant-curvature kinematic model predicts experimentally reconstructed shapes with tip errors of 1.9–6.4 mm.
- Deformation characterization: Increasing tendon thickness raises axial collapsing force, with diminishing gains beyond 2 × 2 mm; TeCoBot therefore uses 2 × 2 mm tendons.Peak force increases from 17.5 ± 1.0 N for 1 × 2 mm to 27.4 ± 0.8 N for 2 × 2 mm and 30.9 ± 1.0 N for 3 × 2 mm.
- Deformation characterization: Bending force increases linearly with angle and tendon thickness, while maximum bending angle increases with module count and peak force remains nearly constant at ~16–18 N.Peak bending forces are 11.1 ± 0.9 N, 19.0 ± 0.7 N, and 25.8 ± 0.5 N for 1 × 2 mm, 2 × 2 mm, and 3 × 2 mm tendons, respectively.
- Docking characterization: Claw-to-claw docking force is weakly affected by sleeve length but depends on material, with latex producing higher symmetric-docking force than silicone.Symmetric docking reaches 42.9 ± 4.5 N with latex and 38.9 ± 2.6 N with silicone.
- Kinematic modeling: The constant-curvature kinematic model predicts experimentally reconstructed backbone shapes with tip position errors of 1.9–6.4 mm, or approximately 0.8–3.0% of backbone length.Cable pulls combine into a resultant vector that determines the bending-plane direction.
- Docking characterization: Claw-to-body docking force increases with motor current, and latex sleeves with tendon-based attachment provide the highest forces.The characterization also evaluates claw-to-docking-station interactions, which show similar trends.
Manipulation and Locomotion with a Single TeCoBot
A single TeCoBot uses its compliant body and two claws to manipulate objects and locomote through diverse modes. Its axial extension–collapse, omnidirectional bending, and mode switching enable three-dimensional manipulation and navigation through complex environments.
- Manipulation: A single TeCoBot grasps objects using its compliant body and two claws, with axial extension–collapse and omnidirectional bending generating a three-dimensional workspace.The workspace was calculated from a model and demonstrated experimentally, including pick-and-place tasks.
- Locomotion: A single TeCoBot locomotes through axial extension–compression, inchworm, crawling, rolling, turning, and pipe-climbing modes.Timestamped image sequences illustrate each locomotion mode.
- Locomotion: TeCoBot switches between locomotion modes to navigate an obstacle course containing a pipe, ramp, and table.The obstacle course demonstrates navigation through complex environments.
Self-Reconfigurable Cooperative Manipulation with Multiple TeCoBots
Multiple TeCoBots self-reconfigure among four task-adaptive three-robot morphologies through sequential undocking, repositioning, and redocking. These configurations support cooperative manipulation of objects and tasks with different sizes and requirements.
- Reconfiguration: Each TeCoBot independently manipulates and locomotes while multiple robots self-reconfigure into task-adaptive morphologies for cooperative manipulation.The robots use claw-based connections and frame docking stations to anchor in different orientations.
- Morphologies: Four representative three-robot morphologies span vertical, mixed vertical-horizontal, and serial arrangements, enabled by docking stations along a cuboidal frame.M1 has three vertically docked robots; M2 has two vertical and one horizontal robot; M3 has one vertical and two horizontal robots; M4 connects three robots in series.
- Reconfiguration: TeCoBots transition between M1, M2, M3, and M4 through sequential claw detachment, repositioning, and redocking.M12, M23, and M34 illustrate the successive transitions, including robots undocking from the frame, landing, docking horizontally, and connecting sequentially to other robots.
- Cooperative manipulation: The morphologies enable cooperative manipulation across objects and tasks with different sizes and requirements.M1 lifts a 200 mm diameter ball, M2 picks up a test tube, removes its cork, and pours liquid into a beaker, while M3 stabilizes and tears a tissue roll.
Self-Reconfigurable Cooperative Locomotion with Multiple TeCoBots
Multiple TeCoBots self-reconfigure through docking and undocking into task- and environment-specific morphologies for cooperative locomotion. These assemblies support distinct locomotion mechanisms across flat ground, objects, ducts, gaps, and unified-body motion.
- Three-robot reconfiguration: Three-robot systems reconfigure into serial-chain, loop-with-trailer, H-shaped, and closed-loop morphologies through reversible claw-based docking pathways.A TeCoBot can detach, move, and reattach to another robot’s claw or body, including transitions through intermediate configurations.
- Three-robot locomotion: L1 locomotes through middle-robot extension and contraction while front and rear claws alternately anchor by friction.This mechanism enables anchored extension–contraction locomotion on flat ground.
- Three-robot locomotion: L2 encloses a large object with looped robots while its trailing robot advances using an inchworm gait.The configuration demonstrates object-assisted cooperative locomotion.
- Three-robot locomotion: L3 enables traversal of rectangular ducts, demonstrating that morphology-specific mechanisms support locomotion in constrained environments.The supplied passage identifies rectangular ducts as an environment for the H-shaped configuration.
- Beyond three robots: Two robots in a linear configuration roll across gaps impassable for one robot, while four robots enable lizard-like locomotion or collective forward translation as a unified body.Four-robot assemblies include H-shaped and square morphologies.
Self-Reconfigurable Cooperative Loco-Manipulation
Three TeCoBots self-reconfigure for multi-phase cooperative loco-manipulation in real-world indoor scenarios involving climbing, reaching, handover, and transport. Demonstrations include study-table cleanup and cabinet organization, with additional outdoor examples reported in supplementary materials.
- Demonstration overview: Three TeCoBots self-reconfigure to perform multi-phase loco-manipulation tasks involving climbing, reaching, handover, and transport in two real-world indoor scenarios.Two additional outdoor demonstrations are provided in Supplementary Note 9, Figs. S25–S26, and Movie S10.
- Task A: Study Table Cleanup: In Task A, three TeCoBots transition through standing, climbing, handover, and serial-chain formation to return a pen and dispose of scrap paper.The robots start from a ‘Y’ configuration on the floor, climb onto a 40 cm-high table, fetch the paper, and descend to discard it.
- Task A: Study Table Cleanup: Task A uses coordinated reconfiguration and manipulation: ground robots push a vertically reoriented robot onto the table, then form a serial chain for cooperative paper handover.The sequence also includes placing the pen into its holder and grasping the scrap paper.
- Task B: Cabinet organization: In Task B, the robots organize components between a supply bin and cabinet drawer using a wall-mounted docking station.Starting from a serial (---) ground configuration, one robot docks at the station while another opens the drawer; the docked robot retrieves a motor from the bin.
DISCUSSION
The work presents TeCoBot, a tensegrity-based continuum robot that combines compliant individual bodies with global multi-robot reconfiguration for task-adaptive cooperative behaviors. Demonstrations across indoor and outdoor loco-manipulation tasks show adaptability, while locomotion is partly automated and self-reconfiguration remains primarily teleoperated despite autonomous docking demonstrations.
- Design framework: TeCoBot combines local body compliance with global multi-robot reconfiguration to produce task-adaptive morphologies for cooperative behaviors.Its tensegrity body supports omnidirectional bending and axial collapse/extension, while dual claws connect multiple units.
- Demonstrations: Real-world demonstrations included workspace cleanup, cabinet organization, and outdoor tasks, highlighting adaptability across complex loco-manipulation scenarios.These applications complement the broader demonstrations of cooperative manipulation, transport, and locomotion described in the paper context.
- Design framework: Unlike rigid modular reconfigurable robots or continuum robots alone, TeCoBot integrates compliance from its tensegrity body with reconfiguration enabled by dual claws.The combined properties support embodied intelligence by offloading part of sensing and control to the compliant body.
- Autonomy and limitations: Self-reconfigurations are teleoperated, although autonomous docking was demonstrated with both pre-scripted open-loop motions and closed-loop onboard vision with fiducial-marker feedback.Some locomotion modes, including rolling and crawling, are automated.
METHODS · Robot Design and Fabrication
TeCoBot is designed as a compliant continuum robot built from vertically stacked icosahedral tensegrity modules. Its cable-driven actuation combines 3D-printed structural components, tendon networks, encoded motors, and durable routed cables.
- Robot Design and Fabrication: Each TeCoBot uses vertically stacked tensegrity modules to form a compliant continuum structure.The tensegrity architecture provides the robot’s overall body structure.
- Robot Design and Fabrication: Each module uses an icosahedral tensegrity geometry with six rigid carbon fiber rods.The rods are 58 mm long and 2 mm in diameter.
- Robot Design and Fabrication: A continuous planar tendon network made from thermoplastic polyurethane forms each module’s tensile structure.The tendon structure is fabricated using fused deposition modeling 3D printing and assembled into a 3D tensegrity form by inserting rods.
- Robot Design and Fabrication: The base, top module, and spacer disks are 3D-printed from polylactic acid and fastened using M3 screws and threaded heat-set inserts.Three spacer disks measure 110 mm in diameter and 2 mm in thickness.
- Robot Design and Fabrication: Three spacer disks guide internal actuation cables and maintain structural alignment between adjacent modules.The disks are placed between adjacent modules.
- Robot Design and Fabrication: The cable spool and motor cover use polycarbonate 3D printing for higher strength and durability.The actuation cable is a 40 lb. braided fishing line anchored at the top module with screws.
Control and Electronics · Claw and Roller Wheel Design
TeCoBots use wireless ESP32-based control with distributed motor electronics and separate batteries for control and motors. Interchangeable claws provide either friction-enhanced docking and locomotion or soft-object manipulation.
- Control and Electronics: Each TeCoBot is wirelessly controlled over Wi-Fi through ESP32-S3 microcontrollers.A central ESP32-S3 transmitter connected to a PC sends commands to an onboard ESP32-S3 receiver on each robot.
- Control and Electronics: The base module houses the ESP32-S3 controller and three Dual TB6612FNG motor drivers.The motor drivers are identified as SparkFun components.
- Control and Electronics: Two Lithium-Polymer batteries separately power the controller and motors.The base battery has 2 cells and 300 mAh; the top battery has 3 cells and 450 mAh.
- Claw and Roller Wheel Design: TeCoBots use two interchangeable claw designs selected according to task requirements.One claw targets locomotion and multi-point docking, while the other targets soft or delicate object manipulation.
- Claw and Roller Wheel Design: The docking-oriented claw uses 3D-printed PLA fingers with rubber tubular sleeves at their ends.The sleeves enhance friction for locomotion and multi-point docking.
- Claw and Roller Wheel Design: The manipulation-oriented claw combines a soft 3D-printed TPU pad with a PLA finger base.This design is optimized for soft or delicate objects but has limited connection capability.
- Claw and Roller Wheel Design: Each claw is actuated in one of two ways, including a smart servo.The smart servo is identified as Dynamixel XL330-M288-T.
Characterization Experiments · Locomotion and Manipulation Experiments
The study characterizes tensegrity-module mechanics, claw-based docking, and kinematic-model accuracy through controlled experiments. It also establishes an anchored aluminum-extrusion frame for cooperative manipulation and conducts indoor locomotion tests on a flat surface.
- Characterization Experiments: Collapse and bending tests record force and displacement during tensegrity-module actuation using a custom test rig.A force gauge applies vertical pulling force for collapse tests; bending tests use the same rig.
- Characterization Experiments: Docking characterization measures maximum claw-effector holding force across docking configurations and materials.Smart servo motors control gripping force, with claw-to-claw and claw-to-body docking types tested.
- Characterization Experiments: Kinematic-model validation compares experimentally constructed robot shapes with model-predicted backbones computed from cable displacements.Validation measures effective length, bending angle, and bending-plane direction under different cable actuation patterns.
- Locomotion and Manipulation Experiments: The manipulation setup uses an aluminum-extrusion frame equipped with customized 3D-printed PLA docking stations along its edges.The stations allow TeCoBots to anchor in various configurations for cooperative manipulation tasks.
- Locomotion and Manipulation Experiments: TeCoBots perform cooperative manipulation tasks by anchoring to docking stations arranged along the manipulation frame’s edges.The frame and docking-station arrangement supports multiple anchoring configurations.
- Locomotion and Manipulation Experiments: Indoor locomotion experiments evaluate TeCoBots on a flat table surface covered with background paper for visual contrast.The visual-contrast sheet is used during indoor locomotion testing.
Cost of Transport · Locomotion Speed Improvement Analysis · Autonomous Docking
The sections quantify locomotion efficiency, show a 2.4× speed increase through a motor replacement, and evaluate open- and closed-loop autonomous docking. The analyses emphasize baseline measurement conditions, actuation limits, and added autonomy hardware.
- Cost of Transport: Cost of transport is quantified for axial extension–compression and rolling locomotion on smooth wood using average electrical power over one locomotion cycle.CoT = P/(mgv), where m is robot mass, g is gravitational acceleration, and v is measured average forward speed.
- Cost of Transport: Smooth wood minimizes slip and terrain-induced losses, so the resulting cost-of-transport values represent a lower bound.
- Locomotion Speed Improvement Analysis: ~12 mm/s versus ~5 mm/s increases locomotion speed 2.4× on smooth wood without changing the robot’s morphology or gait.The improvement uses a faster 30:1 motor instead of the claw’s 380:1 Pololu gearmotor, retaining the same locking-worm-gear design.
- Locomotion Speed Improvement Analysis: The prototype prioritizes validating compliant modular reconfiguration rather than optimizing locomotion speed.Its axial extension–compression cycle is dominated by slow claw actuation lasting approximately 7–10 s per open/close.
- Autonomous Docking: Autonomous docking is evaluated beyond teleoperation in both open-loop and closed-loop modes.
- Autonomous Docking: In open-loop docking, a mobile TeCoBot completes pre-scripted rolling, alignment, and claw-actuation commands for 3-to-3 claw-to-claw docking with a stationary robot without visual feedback.
- Autonomous Docking: Closed-loop docking adds a Raspberry Pi 5, Pi camera, ESP32 transmitter, and Dynamixel smart-servo front claw inside the mobile robot’s top module.