Self-organized Recovery of Coordinated Locomotion in Crickets via Prosthetic Limb Integration

Technical Summary: Self-organized Recovery of Coordinated Locomotion in Crickets via Prosthetic Limb Integration

Problem Statement
While distributed sensorimotor interactions enable coordinated multi-legged locomotion in insects without centralized control, the mechanisms allowing this coordination to re-emerge following limb loss remain poorly understood. A critical unresolved question is whether prosthetic-mediated recovery occurs uniformly across the temporal (phase coupling) and spatial (foot placement) dimensions of gait, or if these dimensions dissociate due to distinct functional properties of their underlying mechanosensory pathways. Previous studies, such as those on cockroaches with "peg legs," suggested that load-mediated ground contact could restore coordination, but they did not quantitatively determine if temporal and spatial recovery occurred to comparable extents or identify the specific mechanosensory pathways responsible. This study addresses the need to quantify inter-ganglion coordination across multiple leg combinations to understand how morphological interventions influence embodied locomotor control.

Methodology
The study utilized the cricket Gryllus bimaculatus as a model system, leveraging its three thoracic ganglia which control bilaterally symmetric leg pairs, allowing for independent quantification of fore-, mid-, and hindlimb coordination. The experimental design employed a within-subject approach across five conditions:

1. Intact: Normal walking.
2. Amp 1: Amputation of the right middle leg (femur-tibia joint).
3. Proth 1: Attachment of a passive, deformable metal wire prosthetic leg to the right middle leg stump.
4. Amp 2: Amputation of both left and right middle legs.
5. Proth 2: Attachment of prosthetic legs to both middle leg stumps.

Experiments were conducted on a spherical treadmill to record walking velocity and body rotation. High-speed video (120 fps) and optical flow sensors captured kinematic data. A deep-learning-based markerless tracking algorithm (DeepLabCut) was used to track 21 anatomical landmarks, enabling the quantification of:

• Temporal Coordination: Inter-leg phase differences computed via Hilbert transform of leg movement phases, with touchdown (TD) events as phase origins.
• Spatial Coordination: 2D foot trajectories (touchdown and liftoff positions) normalized by body length.
• Velocity: Forward and angular velocities.

Statistical analysis included circular statistics (Watson–Williams test) for phase distributions, Mann–Whitney U tests for velocity, and Kolmogorov–Smirnov tests for spatial distributions. Crucially, a Distribution Overlap (DO) metric was calculated to quantify the similarity between manipulated conditions and the intact baseline, using a split-half DO (SH-DO) from intact data as a reference for natural variability.

Key Contributions

• Spatiotemporal Dissociation: The study provides the first quantitative demonstration that prosthetic integration selectively restores temporal inter-leg phase coordination while leaving spatial foot-placement accuracy significantly impaired.
• Mechanistic Insight: It establishes a hierarchical architecture of distributed locomotion control, linking temporal recovery to the restoration of load-mediated signals (campaniform sensilla) and the persistence of spatial deficits to the absence of joint-angle proprioception (femoral chordotonal organs).
• Biohybrid Framework: It demonstrates that re-introducing mechanically relevant sensory constraints (ground contact and load) is sufficient to re-engage distributed coordination networks, even without anatomical integrity or active sensory replication.

Results

• Temporal Recovery: Leg amputation disrupted inter-leg phase coupling, broadening phase distributions and shifting preferred phases. The integration of passive prosthetic legs selectively reinstated coherent temporal coordination. For several leg pairs, the phase distribution overlap (DO) with the intact condition reached or exceeded the natural within-condition variability (SH-DO), and mean phase differences were statistically equivalent to the intact state. This recovery propagated through the distributed Central Pattern Generator (CPG) network, affecting even intact–intact leg pairs not directly involved with the prosthesis.
• Spatial Deficits: While the anteroposterior (y-axis) range of prosthetic leg trajectories was comparable to intact walking, the lateral (x-axis) placement remained significantly altered. The prosthetic legs failed to restore the precise spatial foot-placement trajectories of the intact condition. The DO values for spatial distributions remained substantially below the SH-DO reference, indicating a persistent spatial deficit.
• Velocity: Forward velocity was significantly reduced in all manipulated conditions compared to the intact state. While prosthetic integration led to a modest recovery in velocity distribution, it did not fully restore intact speeds, likely due to the elevated inertial load of the prosthetic limb (approx. 8x the mass of the biological leg). Angular velocity distributions showed asymmetry in unilateral conditions but recovered symmetry in bilateral prosthetic conditions.

Significance and Claims
The paper claims that coordinated locomotion can self-organize and recover following morphological intervention if task-relevant, load-dependent sensory feedback is reinstated, even in the absence of anatomical completeness. The selective recovery of temporal over spatial coordination reveals a hierarchical control structure where:

1. Temporal coordination relies on load-sensitive mechanoreceptors (campaniform sensilla) that detect ground contact forces; these are restored by the prosthetic limb contacting the substrate.
2. Spatial coordination relies on joint-angle proprioception (femoral chordotonal organs) that encodes limb position during the swing phase; this is absent because the passive prosthesis lacks articulated joints.

The authors conclude that load-mediated sensory signals are a key driver of distributed coordination in biohybrid systems. This finding elucidates an embodied principle where sensory–mechanical feedback facilitates the self-organization of resilient locomotion. The study suggests that for biohybrid prosthetics and sensorimotor rehabilitation, prioritizing the restoration of load-onset and load-release signals may be more critical for re-establishing gait timing than replicating precise morphological kinematics. The results provide a generalizable framework for understanding locomotor resilience and offer design principles for adaptive prosthetic engineering, emphasizing that functional recovery emerges from embodied sensory–mechanical interactions rather than strict anatomical replication.