Source-linked AI summary

Residual Wrench Certification and Margin-Aware Control Synthesis for Aerial Physical Interaction

Abhimanyu Khadga, Abhinav Sinha, Shashi Ranjan Kumar

arXiv:2609.38707v1cs.ROcs.SYeess.SY

TL;DR

Aerial interaction requires corrective wrench authority after hover and contact loading, but existing approaches do not connect reserves over coupled task families to feedback synthesis and certified local tracking. This paper develops reserve certificates and margin-aware control, establishing local tracking guarantees while noting that physical contact persistence requires additional contact conditions.

  • Problem

    Existing aerial-interaction approaches do not connect a uniform reserve over coupled force–moment task families to joint feedback-gain synthesis and a certified local tracking region.

    Source: §I · p. 1
  • Method

    The framework certifies task-relative wrench reserves, uses actuator-interiority bounds and redundant allocation, and synthesizes structured gains with margin-preserving command projection.

    Source: §I · p. 1
  • Results

    Within the certified local ellipsoid, the framework establishes nominal exponential convergence, robust ultimate boundedness, and actuator-feasible wrench commands under its stated assumptions.

    Source: §III · p. 4
  • Takeaways & Limitations

    Task-dependent margins quantify corrective authority that rank and conditioning cannot certify, supporting task-specific morphology and control design.

    Source: §V · p. 8

    Actuator-side wrench feasibility during sustained loading does not by itself establish physical persistence, which also requires unilateral, frictional, and contact-moment conditions.

    Source: §I · p. 1

Abstract

from arXiv · show

We develop a task-relative framework for certifying residual wrench authority after hover and contact loading in multirotors with bounded actuators. Using convex geometry, we derive signed margins for prescribed convex reserves, including Euclidean balls and weighted ellipsoids. We obtain computable reserve certificates from actuator-interiority bounds to support slack-maximizing allocation. To preserve the required reserve, we propose command projection onto a tightened feasible set. We then connect the available reserve to structured gain synthesis and certify a local tracking region that respects actuator limits. Within this region, we establish nominal exponential convergence and robust ultimate boundedness. For the prescribed task and morphology family, we show that the optimized octarotor attains a larger margin than the optimized hexarotor at equal total thrust. We evaluate the proposed framework in closed-loop simulations of sustained rigid-wall contact using a fixed-geometry octarotor and a variable-tilt quadrotor with matched installed thrust. The quadrotor's local certificate admits a higher normalized push limit for the prescribed task family. In tests beyond the realizability boundaries, we observe the predicted loss of authority through rotor-thrust limits for the octarotor and tilt-servo limits for the quadrotor.

I. INTRODUCTION

Multirotors can hover compactly, but conventional planar layouts couple translation to attitude, limiting interaction that requires maintaining pose under load. Existing work does not connect uniform reserves over coupled force–moment task families to feedback-gain synthesis and a certified local tracking region.

I. INTRODUCTION

  • Force–attitude coupling in conventional planar multirotors can restrict pressing, bracing, inspection, or manipulation while preserving pose.Safe interaction also depends on retaining corrective authority under sustained loading.Source: §I · p. 1
  • Prior work addresses aerial interaction, constrained allocation, and wrench-capacity measures, but does not jointly certify task-family reserves and local tracking regions with synthesized gains.The cited approaches include tilted and omnidirectional platforms,, allocation methods,, and force-polytope methods for zero-moment force margins.Source: §I · p. 1
  • A feasible contact wrench may leave negligible pose-correction authority, while certification must cover a task family and distinguish reserve exhaustion from actual infeasibility.Redundant allocation makes actuator clearance allocation-dependent, and exact wrench geometry can require extensive facet enumeration.Source: §I · p. 1
  • The framework combines reserve certificates, slack-maximizing allocation, margin-preserving command projection, and structured gain synthesis for locally certified tracking under stated assumptions.It establishes nominal exponential convergence and robust ultimate boundedness, and uses contact simulations to distinguish reserve thresholds from loss of realizability.Source: §I · p. 1
  • “Contact-persistent” refers to actuator-side wrench feasibility during sustained loading; physical persistence additionally requires unilateral, frictional, and contact-moment conditions.Contact-admissible constraints may be represented by intersecting the actuator-generated wrench set with a contact-admissible set.Source: §I · pp. 1–2

II. MODELING AND PROBLEM FORMULATION

The model maps bounded thrust commands from mounted modules to body-frame force and moment, then normalizes wrench coordinates to handle their different units. The problem asks which morphologies can accommodate an entire task family with a prescribed reserve and support actuator-feasible local tracking.

II. MODELING AND PROBLEM FORMULATION

  • Each of N thrust modules contributes force and moment according to its nonnegative thrust, mounting location, orientation, and reaction-moment parameters.The body-frame actuation wrench is assembled from the module contributions.Source: §II · p. 2
  • For a selected morphology, the allocation map M(α) sends thrust vectors in the admissible box Λ to attainable wrenches.The attainable set is W(α) = {w : w = M(α)λ, λ ∈ Λ}.Source: §II · p. 2
  • Because force and moment have different units, wrench distances and reserves use fixed scaled, dimensionless wrench coordinates.Interiority and rank are preserved by the nonsingular scaling, and a normalized reserve guarantees feasibility for perturbations within the corresponding reserve ball.Source: §II · p. 2
  • Constrained full actuation at an operating wrench requires it to lie inside the attainable set, not merely that the allocation map have rank six.The distance to the boundary also depends on the finite actuator box and operating point.Source: §II · p. 2
  • The formulation seeks morphologies that realize every wrench in a compact task family with a prescribed convex reserve, then uses available Euclidean authority for feasible commands and structured gains.The tracking region must provide nominal exponential convergence and robust ultimate boundedness under quantified disturbance bounds.Source: §II · p. 2

III. MAIN RESULTS

The results characterize task-relative actuator reserve with signed margins for convex reserve sets, then use those margins to certify tracking under actuator limits. They also provide computable allocation-based bounds and a command projection that preserves a required reserve.

  • The actuator-interiority certificate uses thrust-bound clearance alone unless contact inequalities are included directly; quasi-static authority comparisons do not certify servo rates, transient singularities, or closed-loop stability.The morphology parameterization distinguishes fixed tilt from admissible time-varying rotor angles.Source: §III · pp. 2, 5
  • For a feasible task wrench, the largest Euclidean residual radius equals its distance to the feasible-set boundary; convex reserves yield signed margins whose sign distinguishes interior, boundary, and infeasible wrenches.For a general convex reserve, the margin is the minimum facet slack divided by the reserve support value; ellipsoids provide direction-dependent reserves.Source: §III · p. 3
  • An interior actuator allocation certifies a Euclidean reserve of at least its actuator clearance divided by the selected right-inverse norm.Slack-maximizing allocation maximizes the clearance lower bound; positive optimized clearance certifies an interior allocation.Source: §III · p. 3
  • A common positive Euclidean reserve for an entire task family exists exactly when its worst-case signed margin is positive; a negative worst-case margin means some task is infeasible.Clipping a negative margin to zero is only a numerical diagnostic, not a nonnegative reserve certificate.Source: §III · pp. 3–4
  • Structured gain synthesis maximizes a certified Lyapunov level subject to pole-abscissa and local-coordinate constraints, yielding nominal exponential convergence and, when χ★ < 1, robust invariance and an ultimate bound.The guarantee assumes the task remains in the prescribed family, the disturbance bound holds, and the local model remains valid; the combined commanded wrench remains actuator feasible in the certified ellipsoid.Source: §III · p. 4
  • A desired command can be projected onto a tightened feasible set so admissible feedback and additional commanded wrenches preserve the required reserve.The tightened intersection must be nonempty; the stated reserve guarantee concerns commanded wrench demand, while dynamic disturbances are treated separately.Source: §III · p. 5
  • The push margin withholds authority in all normalized wrench directions, and a finite direction set is only diagnostic without a covering-error bound.The computations use exact half-space models; fixed rotor directions have allocation rank at most four and cannot provide a six-dimensional residual neighborhood.Source: §III · p. 5

IV. NUMERICAL EVALUATION

The evaluation compares task-relative margins and certified push across multirotor morphologies, then tests predicted actuator limits in wall-contact simulations. Within the stated task and morphology family, optimization favors the capacity-matched octarotor on maximum task margin and certified push, while the variable-tilt quadrotor has a higher simulated certified push limit than the octarotor.

Morphology comparison

  • Within the stated task family and actuator limits, the optimized capacity-matched octarotor attains 30.3% larger maximum task margin and 13.3% larger maximum certified push than the optimized hexarotor.At 36° tilt, the capacity-matched octarotor retains 96.9% of the hexarotor’s task margin and 95.4% of its certified push.Source: §IV · p. 6, Figure 1 · p. 6, Table I · p. 6

Residual-authority certificates

  • At γ=36°, slack maximization raises O8’s allocation-interiority certificate from 0.072 to 0.105, a 45.3% increase, without enlarging the polytope.The exact point margins for H6, O8, and O8,eq are 0.076, 0.137, and 0.073, respectively.Source: §IV · p. 6, Figure 2 · p. 6

Task-margin maps

  • The task-margin maps compare geometry and capacity under equal per-rotor thrust limits across 17,061 force–torque cases per architecture.The sweep spans ρmax∈[0, 0.38] and μmax∈[0, 0.12], with common contours and color limits.Source: §IV · p. 6, Figure 3 · p. 6

Controller synthesis

  • At γ=36°, certified levels are c★_6=0.200, c★_8=1.012, and c★_8,eq=0.188; the structured search returns k_p=0.076 and k_d=0.4 for all cases.The synthesis connects reserve-based gains to local tracking guarantees.Source: §IV · p. 6

Closed-loop push sweeps

  • At equal installed thrust, VTQ4’s certified push limit is ρ★≈0.696 versus O8’s ρ★≈0.247; at μmax=0.04, the certified push ratio is 2.81.The O8 envelope is limited by a rotor-thrust facet, while VTQ4 is limited by its tilt cone.Source: §IV · p. 7, Figure 4 · p. 7

Beyond-boundary tests

  • Beyond the certified boundary, O8 becomes infeasible when rotor bounds activate, while VTQ4 reaches its 36° tilt limit with the predicted 0.096 N force residual.The tests use O8 at ρ=0.280 and VTQ4 at ρ=0.730.Source: §IV · p. 7, Figure 5 · p. 8

V. CONCLUSIONS

The paper turns task-relative residual wrench authority into actuator-feasible local tracking guarantees for sustained aerial interaction. Its conclusions connect reserve margins to control synthesis, morphology comparisons, and contact simulations, while limiting architectural comparisons to tested task families.

V. CONCLUSIONS

  • Certified reserves support structured gain synthesis with nominal exponential convergence and robust ultimate boundedness under the stated assumptions.The framework converts residual wrench authority into actuator-feasible local tracking guarantees.Source: §V · p. 8
  • Task-dependent margins quantify corrective authority that rank and conditioning cannot certify, while redundant allocation improves lower certificates.Equal-thrust comparisons link morphology to attainable reserves.Source: §V · p. 8
  • Contact simulations distinguish exhaustion of a prescribed reserve from actual infeasibility and reproduce predicted thrust and tilt limits.The results support task-specific morphology and control design.Source: §V · p. 8
  • Architectural comparisons are confined to the tested task families, and the variable-tilt certificate remains local.Source: §V · p. 8
Loading 2609.38707v1…