PaperlayerView exact revision on arXiv

Source-linked AI summary

Pneumatic neurons for soft robots enable inflate-and-fire networks for rhythmic motion

Dongting Li, Michael Tolley, Nick Gravish

arXiv:2609.12258v1cs.RO

TL;DR

Robotics commonly centralizes control, while decentralized approaches offer modularity and parallelization. This work introduces low-boiling-point-fluid-driven Pneu-ron modules and reports oscillatory function after severing, alongside environment-dependent traveling-wave speeds.

  • Problem

    Centralized control is common in robotics, while decentralized controllers offer modularity and parallelization.

    Source: §2 · p. 26

  • Method

    Pneu-rons use low-boiling-point-fluid inflation for pressure generation, actuation, and internal signaling, with an integrate-and-fire-inspired model predicting network oscillations.

    Source: §2 · p. 24

  • Results

    After severing, both robot halves maintained oscillatory function; traveling-wave speed was sensitive to module mechanical and thermal properties and the external environment.

    Source: §2 · p. 25

  • Takeaways & Limitations

    Embedding actuation, logic, and timing in compliant modules offers a candidate substrate-agnostic framework for designing and analyzing self-oscillating systems.

    Source: §2 · pp. 26–27

    The heat-sensitive TPU pouches limited maximum heater temperature and therefore maximum inflation speed.

    Source: §2 · p. 26

Abstract

from arXiv · show

Animals coordinate their movements through distributed neural circuits, but soft robots still typically depend on external, centralized electronics for control. Building soft robots that operate without centralized electronic controllers while remaining responsive to their environment remains a frontier challenge in soft robotics. In this work we introduce a soft-robot control architecture inspired by leaky integrate-and-fire models of biological neural circuits. The Pneumatic neuron (Pneu-ron) is a soft actuator that unifies energy conversion, logic, and actuation in one component. Each module combines a low-boiling-point fluid (LBF), a heater, and a mechanical switch into a self-excitable unit. Boiling the LBF inflates the module and triggers excitation and inhibition of adjacent modules in a process we call "inflate-and-fire". When interconnected into excitatory-inhibitory rings, Pneu-rons generate stable, sequential oscillations whose frequency emerges from the material dynamics and environmental conditions. By harnessing the inflation of Pneu-rons for actuation these networks can drive oscillatory locomotion of soft robots. Pneu-ron networks sustain oscillation under mechanical load and thermal variations, adapting through material physics rather than computation. Dynamical modeling of these networks reveals a dimensionless bifurcation diagram that dictates the network's oscillatory behavior. Encoding logic and actuation into material-level modules presents a new avenue for adaptive, electronics controller-free, soft robots.

1 Results

Modeling each module as a leaky integrate-and-fire pressure unit, this section finds that sustained traveling-wave oscillation in a ring requires excitation, inhibition, and self-latching together.

1.1 Minimum signaling architecture for sustained network oscillation

  • Each module is modeled as a first-order RC unit whose normalized pressure rises while its heater is powered and decays otherwise, firing at a threshold.A single module is excitable but not bistable; after triggering, its pressure returns to rest.Source: §1.1 · p. 6
  • Forward excitation alone lets a triggered chain wave fade as modules reset, while ring closure instead drives every module into ON saturation.The ring's returning wave re-energizes the first module.Source: §1.1 · pp. 6–7
  • Adding inhibition without latching over-corrects: signal amplitude decays at each chain handoff, and the ring collapses to OFF.Source: §1.1 · p. 7
  • The same three-pathway architecture in a chain produces shift-register-like propagation that ends at a latched terminal module.Excitation provides forward signaling, inhibition provides upstream release, and latching provides module-level memory.Source: §1.1 · p. 7
  • With excitation, inhibition, and self-latching, rings sustain traveling-wave oscillation with a two-module “2-hot” wavefront.A module rises toward its excitatory threshold as its predecessor cools past its inhibitory threshold.Source: §1.1 · p. 7, Figure 2 · p. 8

gion

The model predicts a bounded oscillatory region between extinction and saturation, set by regeneration and clearance conditions. Two cross-sections show how threshold separation and thermal timing constrain that region.

gion

  • Regeneration requires the upstream module to remain active until its successor reaches the excitatory threshold and self-latches; failure to meet this condition prevents steady-state oscillation in chains and rings.The condition is expressed as propagation gain G, comparing excitation duration t_e with rise time t_r.Source: gion · p. 9
  • Ring clearance requires each module to recover below the excitatory threshold before the wave returns, with available recovery time set by (N−1)t_r and required recovery time by t_i.The clearance condition applies when all three signaling pathways are present.Source: gion · pp. 9–10
  • The two-parameter bifurcation diagram maps a bounded oscillatory region between extinction and saturation across representative cross-sections of the five-dimensional parameter space.Figure 3 presents threshold-adjustment and environmental cross-sections.Source: gion · pp. 10–11, Figure 3 · p. 10
  • A guard window that is too narrow drives saturation, while one that is too wide makes recovery exceed the ring’s transit time.The threshold-adjustment cross-section varies threshold separation while holding thermal time constants and network size fixed.Source: gion · p. 11
  • Fast cooling relative to heating can cause extinction, whereas slow cooling can cause saturation; the oscillatory region lies between these failure modes.The environmental cross-section varies heating and cooling time constants while fixing thresholds and network size.Source: gion · p. 11

1.3 Pneu-ron as an LIF-inspired excitable module

The Pneu-ron integrates a fluid-filled pouch actuator with a buckling-beam switch. Its dual-threshold contacts implement excitation and inhibition, while local electrical routing provides self-latching.

1.3 Pneu-ron as an LIF-inspired excitable module

  • The module combines a low-boiling-point-fluid pouch with a two-part buckling-beam switch, coupling inflation to electro-mechanical switching.Heating drives a liquid-vapor transition and pressure-driven pouch inflation; strain-limiting connectors direct the deformation into vertical bulging and linear contraction.Source: §1.3 · p. 12
  • The switch’s normally closed contact provides inhibition, while its normally open contact conducts at the excitatory threshold to activate the successor.Between the inhibitory and excitatory thresholds, both contacts are disengaged, creating a guard window.Source: §1.3 · pp. 12, 14
  • Local routing connects the module’s normally open contact in parallel with the upstream contact, so the module’s own heater remains powered after it reaches the excitatory threshold.The switch itself returns to rest when pressure is released, so the electrical routing supplies the required logical memory.Source: §1.3 · p. 14
  • The connected-module circuit assigns neighboring normally open contacts to excitation and self-latching, and the next module’s normally closed contact to inhibition.The equivalent logic representation shows all three pathways around each heater.Source: Figure 4 · p. 13, §1.3 · p. 14

1.4 Characterization of a single Pneu-ron.

Single-module tests characterize the Pneu-ron’s pressure response, actuator motions, and voltage-dependent mechanical output. At 9 V, it produces approximately 0.5 N contraction force and 4 mm stroke, with approximately 4 N expansion force.

1.4 Characterization of a single Pneu-ron.

  • Activation produced longitudinal contraction and perpendicular lateral expansion, while the pressure response was second-order and under-damped.The compliant switch also exhibited hysteresis during the activation-deactivation cycle.Source: §1.4 · p. 15
  • At an input voltage of 9 V, the module achieved a maximum contraction force of approximately 0.5 N and a stroke of 4 mm.Source: §1.4 · p. 16
  • Settling time decreased from approximately 600 s at 4.5 V to 300 s at 9 V.The expansion force driving the internal switch also increased with voltage.Source: §1.4 · p. 16

1.5 Three-segment CPG rings sustain oscillation

A three-segment Pneu-ron ring sustained traveling-wave oscillation at room temperature. Switch outputs and thermal images showed the predicted two-heater handoff pattern.

1.5 Three-segment CPG rings sustain oscillation

  • A three-segment ring produced sustained traveling-wave oscillation at room temperature.The experiments tested the bifurcation diagram’s predictions using assembled modules in a ring.Source: §1.5 · pp. 16–17
  • The ring showed the predicted “2-hot” handoff pattern, with two heaters conducting simultaneously: one module above the excitatory threshold and its successor rising toward it.Switch outputs and thermal images confirmed the pattern.Source: §1.5 · p. 17

1.6 Demonstration of network reconfiguration

A six-segment Pneu-ron ring was cut into two three-segment sub-rings during wave propagation; both resumed independent oscillation after one required manual excitation. The demonstration supports reconfiguration for ideal, symmetric cuts, while arbitrary cut locations remain future work.

1.6 Demonstration of network reconfiguration

  • At disconnection, M4 had not received enough excitation to self-latch, so it lost excitation; M1–M3 continued its cycle and re-excited M1.The two sub-rings were formed by separating the ring at the M3/M4 and M6/M1 junctions during active propagation.Source: §1.6 · pp. 17–18
  • Both three-segment sub-rings resumed independent oscillation within one cycle and continued autonomously for 13 minutes after M4 was manually restarted.The result demonstrates simultaneous oscillation of N=6 and N=3 networks within the predicted oscillatory region.Source: §1.6 · p. 18
  • The three-segment rings oscillated more slowly per module than the original six-segment ring because each handoff needed more time to clear the previous wave.No reprogramming or parameter adjustment was needed beyond manually exciting M4; arbitrary cut locations are proposed as future work.Source: §1.6 · p. 18

1.7 Response to mechanical perturbations

Mechanical loads shifted module switching thresholds, but the loaded four-segment ring continued oscillating with a slower rhythm. Localized clamping instead disrupted wave order, producing a reversible collision-induced conduction block.

1.7 Response to mechanical perturbations

  • The four-segment ring continued oscillating under 500 g loads on M1 and M3, while its period increased ∼34% (632 to 848 s).The figure documents two 500 g weights and continuing wave propagation; the period increase reflects slower release under load.Source: §1.7 · pp. 18–19, Figure 8 · p. 19, Figure 8A
  • Compression lowered the switching threshold and shortened the guard window Δth, reducing loaded-module rising time ∼40% (M1: ∼25 to ∼15 s).The shift remained within the oscillatory region of the bifurcation diagram.Source: Figure 8A
  • Clamping M2 forced its excitatory pass, allowing two waves to coexist and collide; the four-segment ring then locked with M1 and M3 ON and M2 and M4 OFF.For N=4 at R=2, the clearance condition allows only about 1.3 waves, so the second wave produces a deadlock.Source: §1.7 · p. 20
  • Forced-air cooling reset M1 below its inhibitory threshold, released the deadlock, and restored the normal single-wave pattern.The conduction block persisted after the external load was removed until thermal reset broke the latching.Source: §1.7 · p. 20

1.8 Thermal sensitivity and perturbations

Thermal and voltage perturbations move the network toward extinction or saturation. Cold and reduced voltage extinguished traveling waves, whereas ambient heat above the fluid’s boiling point produced spontaneous activation and a static saturated state.

1.8 Thermal sensitivity and perturbations

  • At −20 °C, rapid heat loss prevented modules from reaching excitatory threshold and extinguished the traveling wave.Cooling increased τon and decreased τoff, shifting the operating point toward the extinction zone.Source: §1.8 · p. 21
  • Reducing supply voltage from 9.0 V to 7.5 V caused signal amplitude to decay over approximately one additional cycle before the wave extinguished.The voltage drop slowed internal heating, increasing τon while leaving τoff nearly unchanged.Source: §1.8 · p. 21
  • On a 60 °C hotplate, M1 and M3 inflated spontaneously and formed a static arch that lifted M2 away from the heat source.The 34 °C boiling point places ambient temperatures above that value in spontaneous-activation conditions; the authors treat the hotplate case qualitatively because it lies outside model validity.Source: §1.8 · pp. 21–22, Figure 8C

1.9 Locomotion by mechanical coupling of the traveling wave

A three-segment Pneu-ron ring coupled its traveling activation wave to state-dependent-friction legs, producing peristaltic crawling without a separate timing signal. Over six cycles, the robot advanced 25 mm, with cycle time set by the chosen fluid and low-power heaters.

1.9 Locomotion by mechanical coupling of the traveling wave

  • The three-segment ring’s body wave consists of one module inflating as its predecessor deflates, and this sequence couples to the robot’s legs for crawling.The assembled worm-inspired robot was tested on a flat acrylic surface.Source: §1.9 · p. 22
  • Inflated legs engage the ground with a frictional claw, while deflated legs slide on the longer, low-friction arm.The state-dependent friction converts vertical inflation into directional ground engagement.Source: §1.9 · p. 23
  • Sequential activation makes each inflating segment push its released neighbor opposite the wave direction, advancing the ring one increment per cycle.The resulting gait follows reverse peristalsis; the same inflation event actuates a segment and triggers its successor, so no gait generator or actuator timing is required.Source: §1.9 · pp. 23–24
  • Over six cycles, the 180-mm robot advanced 25 mm, or 0.0231 body lengths per cycle.Its approximately 15-minute cycle period reflects the working fluid’s phase-change dynamics and the selected low-power heaters; speed is an implementation property, not an architectural one.Source: Figure 9 · p. 23, §1.9 · p. 24

2 Discussion

The discussion presents Pneu-ron networks as self-contained systems that sustain oscillation after damage and adapt actuation to environmental conditions. It also identifies material constraints and potential applications beyond locomotion.

Damage robustness

  • Severing a worm-like robot in half leaves both halves oscillating, effectively producing two robots through the self-contained, pump-free module design.The authors identify the modular architecture as the basis for this damage robustness.Source: §2 · p. 25

Environmental adaptation

  • Actuation-wave speed varies with module mechanics, thermal properties, external load, ambient temperature, and whether the robot is submerged or in air, without sensors or digital computation.The authors describe this as distributed morphological control rather than a learned or computed policy.Source: §2 · p. 25

Material constraints

  • TPU's heat sensitivity caps heater temperature and inflation speed; silicone pouches leaked low-boiling-point fluid through diffusion and swelling.The authors identify material selection and design as areas for future improvements in speed, displacement, fabrication, and reliability.Source: §2 · p. 26

Potential applications

  • The self-contained architecture could support underground, underwater, and space devices, as well as physical pacemakers or thermally driven end effectors.These are proposed applications, not demonstrated deployments.Source: §2 · p. 26

Decentralized control and framework

  • One DC supply and no controller, clock, or feedback sensor suffice for sequential peristaltic motion; the network template and bifurcation diagram offer a substrate-agnostic analysis framework.The modules embed actuation, logic, and timing in one compliant body.Source: §2 · pp. 26–27

3 Materials and Methods

The methods describe Pneu-ron construction, single-module characterization, network simulation, perturbation and reconfiguration tests, and a locomotion experiment. The protocols examine module behavior and ring dynamics under varied conditions.

3.1 Mechanical design of the pouch actuator

  • Each Pneu-ron combines a fluid-filled pouch that converts heat into pressure with a buckling switch that converts pressure into electrical contact changes.The switch geometry sets inhibitory and excitatory thresholds.Source: §3.1 · p. 27

3.2 Characterization methods

  • Single-module tests recorded pressure and switch behavior through heating-cooling cycles, fitted a second-order RC model, and measured mechanical output under varying voltage.Pressure thresholds were measured at NC and NO contact transitions; mechanical output was measured with a load cell.Source: §3.2 · pp. 27–28

3.3 Network simulation

  • Network simulations used a first-order RC model with inhibitory and excitatory thresholds to test signal propagation and map bifurcation cross-sections over thresholds, thermal time constants, and network size.The model simplified the physical system's second-order dynamics and hysteretic switching.Source: §3.3 · pp. 28–29

3.4 Experimental protocol of perturbation test

  • Perturbation tests applied 500 g weights, a clamp and fan-assisted recovery, freezer exposure at −20 °C, voltage reduction from 9.0 V to 7.5 V, and a 60 °C hotplate.The protocols tested mechanical loading, cooling, reduced voltage, and hot saturation.Source: §3.4 · p. 29

3.5 Experimental protocol of reconfiguration test

  • For reconfiguration tests, a six-module ring ran at 9.0 V for 14 minutes before being severed at two junctions during wave propagation; both sub-rings were then monitored for 13 minutes.Module M4 was manually restarted by brief mechanical compression.Source: §3.5 · p. 30

3.6 Experimental setup for locomotion experiment

  • The locomotion setup used a three-module ring with hinged L-shaped legs and gripping-tape claws, operated at 9.0 V for six cycles over 95 minutes.Displacement was measured after a further 10-minute cooling period using the distal feet on M3 as the reference point.Source: §3.6 · p. 30
Loading 2609.12258v1…