Frame-Coded Legged Locomotion over Noisy Terrain

Summary

Abstract Summary

Open-loop multilegged locomotion over rough terrain has been interpreted as matter transport over a noisy channel: leg-ground interactions are discrete basic active contacts, terrain deletes or perturbs those contacts, and spatial redundancy concentrates the resulting thrust and arrival time. That construction is repetition-like because every module carries the same scalar locomotion task. It consequently provides neither a positive task rate nor a decoder that changes with the surviving contact set. Here we formulate locomotion instead as a quantized finite-frame expansion with erasures. A d-dimensional body-level command is mapped into N>d heterogeneous local contact commands. Rough terrain erases or corrupts frame coefficients, while a contact-gated compliant morphology physically realizes the weighted active-subframe decoder. For a linear-Gaussian model, mechanical equilibrium is exactly the posterior mean, tangent stiffness is posterior precision, and mechanical compliance is posterior covariance. Equal-norm Parseval frames are shown to be minimax optimal against one missing contact, two-contact robustness is governed by frame coherence, and a harmonic frame gives a directly realizable gait family. For independently surviving contacts of probability q, random Gaussian gait frames admit exact reconstruction at every analog dimension rate R<q, with a binomial reliability exponent, whereas recovery of arbitrary commands is impossible for R>q. Residual contact noise yields an asymptotic per-mode amplification 1/(q-R) and a vanishing mechanical stiffness margin at the threshold. An information-locomotion inequality and an exact incremental-redundancy rule direct the next gait component toward the softest task-relevant unresolved mode. The resulting analog frame-coding theorem establishes a finite relative redundancy and converse as part of a fundamental limit theory of legged locomotion.

Key Points

Open-loop multilegged locomotion over rough terrain has been interpreted as matter transport over a noisy channel: leg-ground interactions are discrete basic active contacts, terrain deletes or perturbs those contacts, and spatial redundancy concentrates the resulting thrust and arrival time.
That construction is repetition-like because every module carries the same scalar locomotion task
It consequently provides neither a positive task rate nor a decoder that changes with the surviving contact set.
Here we formulate locomotion instead as a quantized finite-frame expansion with erasures
A d-dimensional body-level command is mapped into N>d heterogeneous local contact commands

Frame-Coded Locomotion

Here we introduce the basic approach. Consider one quasi-static gait epoch. Let be a desired locomotion command. Its coordinates may be generalized wrench components, body twist components, support-plane pose variables, backbone modes, or coefficients of a short trajectory segment, consistent with the template-and-anchor and reduced-coordinate traditions in locomotion modeling [ 7 , 8 ] .

The high-level gait generator knows $U$, but the physical body has to realize it through distributed local legs (after uncertain terrain interaction).

Frame-Coded Locomotion — Frame-Coded Legged Locomotion over Noisy Terrain — Fig. 2: Worst surviving lower frame bound for representative $d=3$, $N=8$ Parseval designs. The equal-norm harmonic construction equalizes single-contact loss and remains comparatively robust to several erasures; clustered columns create redundant contact roles and a weak surviving mode.

Morphological Active-Subframe Decoding

This section focuses on the embodied computation of decoding in the presence of noise. Let $q\in\mathbb{R}^{d}$ denote the realized body-level state. Suppose active contact $i$ contributes the elastic energy and the body has intrinsic energy For active set $S$, The active set need not be enumerated explicitly as a computation, the body does it naturally.

A slack tendon or unloaded series-elastic element simply removes its energy term.

Morphological Active-Subframe Decoding — Frame-Coded Legged Locomotion over Noisy Terrain — (a) Finite-length noninjectivity probability for contact survival $q=0.7$.

Finite-Erasure Gait Design

For this section, first take an uninformative prior and white residual noise: For a surviving subframe, define frame bounds Exact reconstruction requires $A_{S}>0$; stable reconstruction requires $A_{S}$ bounded away from zero. If $A_{S}>0$, then For isotropic damping $C=cI$, the decoding time constant also satisfies $\tau_{S}\leq c/A_{S}$.

The formula is the Goyal-Kovačević-Kelner reconstruction-error criterion expressed in locomotion coordinates; equal-norm tight-frame and optimal-frame refinements were developed further in [ 2 , 14 , 15 ] .

Finite-Erasure Gait Design — Frame-Coded Legged Locomotion over Noisy Terrain — (b) Below threshold, the MSE diverges and the mechanical stiffness margin vanishes as $R\uparrow q$.

A Random-Frame Coding Theorem

We now let the useful locomotion dimension grow with the number of attempted contacts. This distinguishes coded locomotion from scalar repetition [ 1 ] and connects with probabilistic erasure-robust frame analysis [ 18 ] . For each $N$, let $d_{N}/N\to R\in(0,1)$ and draw Let contacts survive independently: Define frame-decoding failure in the noiseless model as noninjectivity of $u\mapsto F_{S_{N}}^{\mathsf{T}}u$.

Under ( 38 )–( 39 ): Conditional on $M_{N}\geq d_{N}$ , $F_{S_{N}}$ has rank $d_{N}$ almost surely.

A Random-Frame Coding Theorem — Frame-Coded Legged Locomotion over Noisy Terrain — Fig. 4: Incremental information for a two-mode posterior $P_{S}=\operatorname{diag}(1,9)$ with unit precision. A contact direction aligned with the soft second mode supplies the largest conditional information and the largest reduction in uncertainty volume.

Information, MMSE, and Mechanical Compliance

The frame operator controls more than reconstruction. In the Gaussian model it is also an information matrix, so it yields an intriguing direct correspondence among information, estimation, mechanics, and locomotion. Here all information quantities are evaluated for known realized frames.

Under ( 9 ) and ( 11 ), If independent contact $i$ has continuously variable precision $\rho_{i}$, then The mutual information of a linear Gaussian channel is The determinant lemma and $P_{S}=J_{S}^{-1}$ yield ( 53 )–( 54 ).

Incremental Redundancy and Morphological HARQ

In locomotion, one can consider not just feedforward but also feedback-based control analogous to hybrid automatic repeat request (HARQ). Recent work describes passive force redistribution as mechanical intelligence and combines it with sensor-based control in a forward error correction (FEC)/ARQ analogy [ 24 , 25 ] . Open-loop frame coding is the forward-error-control component.

Contact sensing or a compliance threshold can add a reverse channel.

Filter-Bank Gaits and Colored Terrain

The finite-frame model treats a gait epoch as a block. Continuous locomotion requires streams of body-mode commands and contact actions. Let $u[t]\in\mathbb{R}^{m}$ and let an $N\times m$ oversampled analysis filter bank be Leg stream $i$ receives Spatial phase shifts, body waves, and contact waves provide physically feasible bases for $H(z)$.

For a fixed surviving stream set $S$, stable noiseless reconstruction requires the filter-bank frame condition for almost every $\omega$, with $A_{S}>0$ [ 4 , 29 ] .

Sources and Demonstrations

Figures are reproduced from Varshney See the full paper for experimental details and the project page, when the paper links one, for demonstrations and videos.