Towards Optimal Control and Algorithmic Structure of Decompression Schedules
Pith reviewed 2026-05-21 21:00 UTC · model grok-4.3
The pith
Decompression schedules are solved via optimal control yielding bang-bang vertical rates and discrete algorithms with error bounds.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
We formalise decompression planning as an optimal control problem with gas feasibility windows (ppO2, END), affine ceilings, and convex penalties in normalised oversaturation. The depth trajectory is constrained to be a monotone ascent. In this setting we prove relaxed existence, derive bang-bang structure for the vertical rate control, and obtain nonsmooth dwell time KKT conditions. For finite stop grids we give resource constrained dynamic programming and label setting formulations with explicit discretisation error bounds.
What carries the argument
Optimal control problem under monotone ascent constraint with convex oversaturation penalties, which supports derivation of bang-bang vertical rate control and KKT conditions for dwell times.
Load-bearing premise
The depth trajectory must be a monotone ascent, which is invoked to derive the bang-bang structure and the KKT conditions.
What would settle it
Finding a decompression schedule with non-monotone depth changes that achieves lower risk than any monotone schedule under the oversaturation penalty, or an optimal control with more than bang-bang switches in vertical rate.
read the original abstract
We formalise decompression planning as an optimal control problem with gas feasibility windows (ppO$_2$, END), affine ceilings, and convex penalties in normalised oversaturation. The depth trajectory is constrained to be a monotone ascent, matching operational decompression practice. In this setting we prove relaxed existence, derive bang-bang structure for the vertical rate control, and obtain nonsmooth dwell time KKT conditions. For finite stop grids we give resource constrained dynamic programming and label setting formulations with explicit discretisation error bounds, while also stating the tissue state quantisation or label growth assumptions needed for pseudo-polynomial complexity. The time risk attainable set is generally nonconvex because gas, stop, and switching choices are discrete. We also isolate the precise scope of the two segment exchange argument. It orders terminal tissue tension under monotone inert fraction ordering, but it does not prove that re-descents are dominated for the oversaturation only penalty used here.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript formalizes decompression planning as an optimal control problem incorporating gas feasibility windows (ppO₂, END), affine ceilings, and convex penalties in normalised oversaturation. The depth trajectory is constrained to monotone ascent to match operational practice. Under this setting the authors prove relaxed existence, derive bang-bang structure for vertical rate control, and obtain nonsmooth dwell-time KKT conditions. For finite stop grids they supply resource-constrained dynamic programming and label-setting formulations together with explicit discretisation error bounds, while stating the tissue quantisation and label-growth assumptions required for pseudo-polynomial complexity. They also note the general nonconvexity of the time-risk attainable set arising from discrete gas, stop and switching choices, and isolate the precise scope of the two-segment exchange argument.
Significance. If the central claims hold, the work supplies a mathematically rigorous framework for decompression schedule design that combines existence theory, structural results, and discretised algorithms with error guarantees. The explicit discretisation bounds and the careful scoping of the two-segment exchange argument are concrete strengths that facilitate both theoretical analysis and practical implementation. These contributions are relevant to optimal control applications in safety-critical physiological systems and could support the development of more efficient, verifiable decompression protocols.
major comments (1)
- [Section deriving bang-bang structure and KKT conditions] The monotone-ascent constraint on depth trajectories is invoked to obtain both the bang-bang structure and the nonsmooth KKT conditions; while the modelling choice is operationally motivated, its necessity for the structural results should be examined more explicitly (e.g., by indicating whether a non-monotone trajectory could violate the derived switching law or KKT stationarity).
minor comments (2)
- [Introduction / Problem formulation] Notation for the gas feasibility windows (ppO₂ and END) and the precise definition of the normalised oversaturation penalty should be introduced with a short table or equation block early in the manuscript to aid readability.
- [Dynamic programming and label-setting formulations] The statement of the tissue quantisation and label-growth assumptions required for pseudo-polynomial complexity would benefit from a brief remark on how these assumptions scale with the number of stops or tissue compartments.
Simulated Author's Rebuttal
We thank the referee for the careful reading, positive assessment, and constructive comment. We address the single major comment below and will incorporate the requested clarification in the revised manuscript.
read point-by-point responses
-
Referee: [Section deriving bang-bang structure and KKT conditions] The monotone-ascent constraint on depth trajectories is invoked to obtain both the bang-bang structure and the nonsmooth KKT conditions; while the modelling choice is operationally motivated, its necessity for the structural results should be examined more explicitly (e.g., by indicating whether a non-monotone trajectory could violate the derived switching law or KKT stationarity).
Authors: We agree that the dependence of the bang-bang structure and nonsmooth KKT conditions on the monotone-ascent constraint merits more explicit discussion. In the revised manuscript we will insert a short remark immediately after the statement of the structural results. The remark will note that the proof of the bang-bang property for vertical rate control proceeds by applying the Pontryagin maximum principle to the Hamiltonian under the monotonicity constraint; this unidirectional depth evolution precludes the possibility of re-descents that could otherwise produce additional switching points or violate the derived stationarity conditions. A non-monotone trajectory could indeed admit such re-descents, which would generally invalidate the switching law obtained under the current formulation. Because the constraint is both operationally required and essential for the convexity arguments used to obtain the KKT conditions, we retain it; the added remark will make this dependency transparent without changing any theorems or proofs. revision: yes
Circularity Check
No significant circularity detected
full rationale
The paper derives relaxed existence, bang-bang vertical rate structure, and nonsmooth dwell-time KKT conditions directly from standard optimal control theory under the explicitly stated monotone-ascent depth constraint. Resource-constrained DP and label-setting formulations with explicit discretisation bounds follow from discrete gas/stop/switch choices and stated tissue quantisation assumptions. The two-segment exchange argument is isolated to ordering terminal tissue tension under monotone inert-fraction ordering without claiming re-descent dominance. No derivation reduces by construction to fitted parameters, self-definitions, or load-bearing self-citations; all steps rely on external mathematical frameworks that remain independent of the target results.
Axiom & Free-Parameter Ledger
axioms (2)
- domain assumption Depth trajectory constrained to monotone ascent
- domain assumption Gas feasibility windows (ppO2, END) and convex penalties in normalised oversaturation
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.