Reinforcement learning offers a way for robots to learn effective behavior, but it commonly lacks stability guarantees. Minjae Cho, the study’s lead author and a Ph.D. student, said that assurance is important for practical deployment because a robot that becomes unstable could, in a worst-case scenario, break and require costly repairs. CARL was designed to address both requirements: using reinforcement learning as the robot’s main controller while applying contraction theory to bound how large small tracking errors can become. The method does not claim that errors never occur; instead, it provides mathematical limits on their magnitude so that the robot remains stable while learning.

Cho initially simulated CARL before testing the algorithm with multiple physical robots. The experimental platforms were a TurboBot 3 Burger wheeled robot, a flapping aerial drone and a Jethexa six-legged robot. The team compared the progress of CARL with previous algorithms and reported that the performance achieved in simulation was consistently better in physical testing, in some cases by as much as 15%. This was an experimental result for the robots, tasks and comparisons tested, rather than a guarantee of the same improvement across every robot or application.

One challenge evaluated in the experiments was the sim-to-real gap, the difference between a simulated model and a physical robot. Cho illustrated the issue with mass: a model in the computer may have a mass of 1 kilogram, while the corresponding physical machine may weigh 1.2 kilograms. CARL was found to handle these uncertainties better than the existing methods used for comparison, allowing its simulated performance advantage to carry over into the flight-laboratory tests despite differences between the model and the real machines.

Cho identified wildfire management as one possible application. First responders could potentially deploy robots to help identify hotspots or deliver supplies, although those capabilities were not demonstrated in the reported experiments. The method has open-access availability, and its code is also provided through GitHub. The authors were Cho, Hiroyasu Tsukamoto and Huy T. Tran, with the study’s DOI given as 10.1109/TRO.2026.3727138 and its publication date listed as 24 August 2026. The reported results indicate that combining reinforcement learning with contraction theory can produce a stability-guaranteed controller that performs well across three distinct robot platforms, but the evidence presented concerns controlled simulation and laboratory experiments rather than field operation.

Subject of Research: Stability-guaranteed reinforcement-learning control tested on wheeled, aerial and six-legged robots.

Article Title: New method for guaranteed stability and improved performance for robots on the move

Article References: New method for guaranteed stability and improved performance for robots on the move. Original publication (Bibliographic details unavailable.)

Image Credits: AI Generated

DOI: Not provided

Keywords: robotics, reinforcement learning, control theory, CARL, contraction theory, robot stability, tracking performance, sim-to-real transfer, aerial robots, legged robots, method, guaranteed

The work was led by Alex Zhavoronkov, a member of the Aging editorial board who is affiliated with Insilico Medicine and the Buck Institute for Research on Aging. Its starting observation is blunt: established frameworks such as the Hallmarks of Aging and Mikhail Blagosklonny’s hyperfunction theory explain mechanisms of decline, but they do not prescribe how, when or whether those mechanisms should be targeted therapeutically. “Existing aging theories describe what changes with age but do not prescribe how to intervene,” the authors write. Control theory, the mathematical discipline concerned with directing dynamic systems toward desired states, offers precisely that prescriptive language.

At the heart of the framework lies the concept of a biological state space, a multidimensional representation of an organism’s condition spanning inflammation, mitochondrial function, epigenetic stability, immune competence and regenerative capacity. Rather than attempting to rewind every molecular feature to a youthful value, the framework prioritizes maintaining or recovering acceptable physiological function. The authors define a functional viability region, the set of biological states in which essential functions remain within acceptable limits, surrounded by a wider recovery region from which safe interventions could potentially restore function. As aging progresses, this recoverable territory may shrink, making functional restoration progressively harder and eventually impossible.

From this geometry follows a striking redefinition of two familiar concepts. Aging, in this framework, is a progressive loss of safe controllability, and biological age is interpreted not as a molecular signature but as the minimum safe cost required to restore or maintain function. This distinguishes the approach from conventional biological aging clocks, which generally estimate chronological age or health-related risk from molecular measurements. Here, the measure of age is practical: how difficult would it be, with the interventions actually available, to achieve meaningful functional improvement without pushing the system into a dangerous state?

Drugs enter the mathematics as vector fields, meaning mathematical descriptions of how an intervention changes a biological system’s trajectory through state space. Because a vector field depends on the current state, the same treatment can push one patient toward recovery and another toward harm. Treatment effects therefore depend on biological state, dose, duration and tissue conditions rather than being universally beneficial. The authors argue that this formulation could help explain why individuals of similar chronological age may respond so differently to the same intervention, a persistent puzzle in geriatric medicine and in trials of aging-related therapeutics.

The framework also delivers a formal account of why treatment order matters. Using mathematical objects known as Lie brackets, which capture how the composition of two influences depends on the sequence in which they are applied, the authors propose that administering two interventions in different sequences can produce different biological effects even when the same therapies are used. To demonstrate the principle, they built a five-variable computational model of an aged mouse liver incorporating senescent-cell burden, oxidative damage, epigenetic integrity, regenerative capacity and fibrosis, and compared three hypothetical strategies: senolytic treatment followed by partial cellular reprogramming, the reverse sequence, and simultaneous administration.

In the illustrative 56-day simulation, senolytic treatment before reprogramming produced the most favorable modeled outcome, with the proposed explanation that clearing senescent cells first creates conditions that improve subsequent reprogramming. The authors are careful about what this proves. These findings came from mathematical simulations rather than animal experiments, and the predicted benefits of treatment sequencing require biological validation before any clinical conclusions can be drawn. The value of the exercise lies in showing that sequencing effects, often treated as anecdotal, can be made explicit, testable predictions within a rigorous formal structure.

Safety occupies the center of the framework rather than its periphery. “Safety is not a secondary penalty appended after efficacy; it is the central constraint that defines the admissible control set,” the authors write. They distinguish between reversible interventions, such as many small-molecule treatments, and potentially durable approaches including gene editing, viral gene delivery and certain cell therapies. Under the framework, reversible interventions are prioritized when uncertainty is high, while longer-lasting therapies are considered only when their expected benefits and safety are sufficiently supported. The approach also acknowledges hard biological boundaries: extensive fibrosis or irreversible cell loss may mark points beyond which available treatments can no longer restore normal function, underscoring the potential importance of intervention timing and of preventive strategies over late-stage rescue attempts.

To explore clinical relevance, the researchers outlined three translational case studies covering immune aging, age-related muscle loss and ovarian aging, illustrating how intervention selection could account for tissue-specific changes, functional goals and safety constraints. A recurring theme is that molecular proxies can mislead: increasing muscle mass, for example, would not constitute successful treatment if strength, mobility and metabolic function remained impaired. The paper also applies an exploratory scoring framework to aging-related interventions across five biological stages, drawing on aging research databases, clinical trials and therapeutic-target resources, and evaluating treatments by biological coverage, safety, reversibility, durability and potential functional benefits. In this exercise, caloric restriction and aerobic exercise ranked highest, followed by the SGLT2 inhibitors dapagliflozin and empagliflozin, with GLP-1 receptor agonists including liraglutide and semaglutide also scoring relatively high. The authors stress that these scores depend on the framework’s assumptions and weighting criteria and do not demonstrate that any of these drugs extend human lifespan or should be prescribed specifically to treat aging.

Looking forward, the paper sketches a proposed computational system that would integrate molecular measurements, clinical data, mathematical modeling and repeated monitoring to estimate biological states and adjust interventions over time, an approach that could eventually support individualized therapeutic decisions, although a fully trained and experimentally validated implementation does not yet exist. The authors also present 20 falsifiable predictions for future investigation, including whether treatment sequencing influences functional restoration, whether biological states predict individual responses, and whether safety-constrained strategies outperform approaches based mainly on aging biomarkers or single molecular targets. The limitations are acknowledged plainly: biological states cannot be measured completely, the models rest on assumptions requiring experimental validation, the rankings depend on chosen scoring criteria, and the framework’s advantages have not been demonstrated in prospective studies. What the paper offers, ultimately, is a shift in attention for aging drug discovery, away from cataloguing age-related changes and toward determining how function might be restored safely, delivered as a set of testable principles that independent laboratories can now begin to challenge.

Subject of Research: A control theory framework for aging biology and safer gerotherapeutic drug discovery

Article Title: New control theory framework proposes a safer approach to aging drug discovery

Article References: New control theory framework proposes a safer approach to aging drug discovery. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: aging, control theory, gerotherapeutics, drug discovery, biological age, senolytics, cellular reprogramming, state space, vector fields, safety, caloric restriction, SGLT2 inhibitors

The work, published in Frontiers in Electronics by a team from the State Grid Chongqing Economic Research Institute and Chongqing University of Science and Technology, tackles one of the most stubborn problems in DC power distribution: the constant power load. When a downstream converter is tightly regulated to deliver a fixed power regardless of voltage, it behaves like a negative resistance on the bus. If the bus voltage dips, the load draws more current to keep its power constant, dragging the voltage down further. This positive-feedback-like interaction erodes the natural damping of the system and can amplify oscillations after sudden disturbances, sometimes to the point of outright instability.

Modern converters compound the problem because, unlike the rotating generators of the twentieth century, they possess no inherent electromechanical inertia. When a load steps abruptly or an input source fluctuates, the mismatch between incoming and outgoing energy has nowhere to go except into the small output capacitor. The bus voltage can then overshoot or undershoot severely, potentially tripping overvoltage or undervoltage protection and shutting the system down. The research team, led by Wenzuo Tang and Shuaicheng Hou, argued that a truly robust controller must do two things at once: counteract the negative damping of constant power loads and supply sufficient virtual inertia to soften transient excursions.

Their testbed is a two-phase interleaved Boost converter, a topology prized in medium- and high-power applications because the parallel phases share current stress and partially cancel input-current ripple. Interleaving means the two switching stages operate half a cycle out of phase, smoothing the current drawn from the source. The team derived an averaged mathematical model of the converter in continuous conduction mode, then recast it in Euler-Lagrange form, a framework that describes the system in terms of stored and dissipated energy. This energy-based description is the key that unlocks passivity-based control.

Passivity theory treats a converter as an energy-handling network: energy flows in from the source, accumulates in the inductor’s magnetic field and the capacitor’s electric field, and drains away through dissipation. If the dissipated energy always exceeds the difference between supplied and stored energy, the system is strictly passive and inherently stable. The researchers exploited this by defining an error-energy function measuring the gap between actual and desired inductor current, then injecting artificial damping to drive that error energy to zero. The convergence rate is set directly by the damping coefficient, giving the controller a clear physical tuning knob. Because the Boost converter is a non-minimum-phase system, direct voltage control would produce unstable internal dynamics, so the team selected the current-regulating equation as the duty-ratio law, guaranteeing global stability of the current loop.

An additional wrinkle arises in interleaved operation: the two phases must share current equally. The final control law incorporates a current-sharing term, so that when one phase’s inductor current exceeds its reference, the duty cycle shrinks, shortening the charging interval and pulling the current back. This creates a self-correcting, energy-dissipating feedback loop in hardware, not merely on paper. Yet the passivity-based inner loop alone cannot solve the voltage-side problem, because a conventional proportional-integral outer loop would still convert a sudden bus-voltage error into an aggressive current reference, slamming the converter through large transients.

Enter the virtual DC machine. The outer voltage loop deliberately emulates the external behavior of a DC motor, complete with an armature circuit, a virtual back electromotive force, a virtual moment of inertia, and a virtual damping coefficient. The measured bus-voltage error is processed through a PI stage to yield mechanical power, from which a virtual torque is computed. The mechanical equation of motion, with its inertia and damping terms, then filters the response, producing a virtual armature current that becomes the reference for the passivity-based inner loop. The result is that a sudden voltage error is no longer translated instantaneously into a current command; the virtual rotor must ‘spin up’ first, exactly as a physical machine would.

Small-signal analysis of this virtual machine reveals it behaves as a first-order inertial element whose time constant equals the virtual inertia. Bode plots show that increasing the inertia coefficient lowers the cutoff frequency and improves smoothing of high-frequency noise, at the cost of slower tracking, while raising the damping coefficient boosts low-frequency gain and improves phase margin, strengthening disturbance rejection. The trade-off is explicit and quantifiable: more inertia and damping buy stability but slightly delay recovery. The team’s frequency-domain study lets engineers choose those coefficients deliberately rather than by trial and error.

The experimental verification is where the approach becomes genuinely striking. The prototype, built with GaN power devices switching at 500 kilohertz, converts a 96-volt nominal input, spanning 80 to 120 volts, to a regulated 270-volt bus. Its volume of roughly 0.335 liters yields a power density of about 9 kilowatts per liter, a figure that underscores why gallium nitride matters: the higher switching speeds and lower losses of GaN allow the small inductors and capacitors that make such density possible. The converter was tested across no-load, half-load, and full-load conditions, holding output-voltage fluctuation within 1 percent across the entire power range.

The dynamic tests delivered the headline numbers. When input voltage stepped from 80 to 120 volts under full load, conventional PI plus PI control produced an 8.9 percent overshoot with a 9-millisecond settling time. Replacing only the current loop with passivity-based control halved the deviation to 3.7 percent. Adding the virtual DC machine on top cut the deviation to just 1.5 percent, an 83.1 percent reduction versus the conventional benchmark. Load steps proved even harsher, and the contrast starker: dropping the load from 3000 watts to 100 watts produced a 13.3 percent excursion under PI plus PI control, but only 1.1 percent under the composite scheme. In extreme simulations where load power suddenly tripled, the PI current loop descended into limit-cycle oscillation and lost stability, while the passivity-based loop still restored the rated voltage. The virtual-inertia approach did concede a millisecond or two of recovery time in some cases, but the authors argue this is a bargain: a modest slowdown in exchange for eliminating the voltage excursions that trigger protective shutdowns. For fuel-cell supplies, DC microgrids, and motor drives, where converters must survive wide operating ranges and violent load swings, the combination of energy-based damping and machine-like inertia may prove to be the stabilizing handshake that high-density power systems have been waiting for.

Subject of Research: Stability enhancement of interleaved boost converters with constant power loads using virtual DC machine and passivity-based control

Article Title: Stability enhancement method for interleaved boost converter based on virtual DC machine and passivity-based control

Article References: Tang, W., Li, B., Hou, S., Shao, X., & Yu, H. (2026). Stability enhancement method for interleaved boost converter based on virtual DC machine and passivity-based control. Frontiers in Electronics, 7, Article 1909575. https://doi.org/10.3389/felec.2026.1909575

Image Credits: AI Generated

DOI: 10.3389/felec.2026.1909575

Keywords: DC-DC converter, interleaved boost converter, constant power load, passivity-based control, virtual DC machine, gallium nitride, DC bus stability, power electronics, fuel cell power supply, control theory, power density, Euler-Lagrange model

The author, Lori Hogenkamp of the Center for Adaptive Stress in Cincinnati, frames the argument around a construct she calls Regulatory Bandwidth, which she previously introduced as the organism’s present coordinated capacity across interacting stress-responsive systems. Her central claim is stark: regulatory capacity is not an intrinsic property of an organism. It is a conditional relation among the architecture and current state of the regulatory system, the particular perturbations that reach it, the cost functional used to evaluate the response, the viability criterion, and the time horizon. Remove any one of those ingredients, and the question of which system has greater capacity has no answer.

The mathematical heart of the paper borrows from control theory, a field that engineers have used for decades to analyze feedback systems. Consider a biological system linearized around its operating state. The eigenvalues of the Jacobian matrix determine stability: if they all have negative real parts, perturbations eventually decay. But eigenvalues say nothing about what happens along the way. In non-normal systems, which are common in directed and hierarchical biological networks, non-orthogonal modes can combine to produce large transient excursions before recovery, so two systems with identical spectra can traverse wildly different paths and impose very different energetic or functional costs. Hogenkamp’s formal example drives this home with two three-dimensional models spanning autonomic, neuroendocrine, and inflammatory states. Both share the same eigenvalue spectrum, the same relaxation rate, and the same equilibrium, yet under one cost functional one system carries 14.7 times the other’s burden, while under a different metric the ranking flips entirely.

The tool that formalizes these reversals is the observability Gramian, a matrix that, in this setting, represents the accumulated weighted recovery burden of a trajectory starting from a given departure. For a declared quadratic cost, the Gramian solves a Lyapunov equation and quantifies exactly how much burden accrues given how perturbations propagate through the network’s coupling structure. Crucially, the paper separates three maps that routine biomarker composites silently conflate: the measurement map, which specifies what instruments can observe; the burden-evaluation map, which specifies which observed departures count as cost; and the latent state representation itself. A composite index inherits commitments from all three and then reports a single number as though it had recovered a property of the organism.

This is where the argument turns pointed. Standardization choices that researchers treat as neutral preprocessing are, in Hogenkamp’s framing, constitutive assumptions. Z-scoring cortisol and blood pressure together declares a one-standard-deviation departure in each to be commensurate, a geometric commitment with no biological justification. Changing the reference cohort can change a person’s standardized profile without any change in that person’s physiology. Cutoff methods, regression weights, and factor-analytic loadings each embed different commitments about what counts. And no amount of additional biomarkers or data-driven weighting solves the problem, because a model optimized to predict mortality estimates risk under a particular population and outcome definition; it does not thereby discover a context-free quantity called capacity. Prediction and construct identification, the paper insists, are different achievements.

The framework also draws sharp distinctions between concepts that stress science has long blurred. A resting snapshot indicates state, not dynamics: it cannot predict how the coupled system will respond to a challenge. Return rate is not recovery burden, because two systems may recover at similar asymptotic speeds while incurring very different transient costs. And recovery burden is not viability, which means remaining within declared physiological or functional constraints over a declared horizon. A trajectory can be costly yet remain viable, or it can cross a critical functional boundary despite a modest average-cost score. Most consequentially, viability is not adaptation. A response that preserves short-term function through costly compensation may produce delayed burden or narrow the range of perturbations manageable later; it may be adaptive for one function or horizon and maladaptively stabilized at another. Adaptation, in this account, is an indexed judgment tied to a specified level of organization, protected function, reference point, horizon, and cost criterion, not a synonym for stability or recovery.

What makes the framework scientifically serious rather than merely critical is that it generates falsifiable predictions. The primary comparison pits a flexible scalar-capacity model, which predicts one preserved ordering of systems across all declared perturbation classes, against a conditional-profile model, which permits reproducible system-by-perturbation rank reversals. Both models are fitted to the same data with the same covariates and evaluated by complexity-controlled held-out prediction on sessions, contexts, or perturbation classes they did not see. If the scalar model is non-inferior, the profile has no demonstrated value. If replicated crossing interactions survive adjustment for resting state, exposure history, and measurement error, capacity is better represented as conditional structure than as an intrinsic rank. The paper even specifies how measurement error in integrated burden outcomes grows with the observation horizon and must be estimated and removed before any rank comparison, because spurious crossings that replicate across sessions could otherwise masquerade as biology.

To show the framework is implementable, the paper walks through a prospective design specification using the orthostatic challenge, an established cardiovascular test in which a person transitions from lying down to standing or passive head-up tilt. Continuous beat-to-beat blood pressure and heart-rate recordings capture the multivariate response, while symptoms, task performance, and cerebral blood-flow velocity are recorded as functional outcomes. The reference trajectory, burden weights, viability endpoint, and time horizon are all declared before fitting. Notably, a retrospective tilt study cited in the paper found that cerebral blood-flow velocity and end-tidal carbon dioxide predicted lightheadedness while standard heart rate and blood pressure measures did not, illustrating exactly why the biological conclusion depends on which outputs are measured: preserved values in familiar cardiovascular channels do not establish preservation of cerebral perfusion or function.

The implications extend into contested territory. The framework offers a mathematical route to person-environment fit: two architectures can have similar average burden under one perturbation distribution and sharply different burden under another, so accommodations that change the input channel, timing, or predictability of demands can have large biological effects without changing a person’s underlying regulatory organization. Hogenkamp, an autistic scholar, is explicit that this framing is neurodiversity-affirming without insulating the theory from evidence. It does not deny impairment, energetic cost, or collapse; it changes the explanatory unit, locating costly outcomes in the interaction between a minority architecture and an environment concentrated along its fragility directions. At the same time, the paper is careful to note that rejecting the scalar model supports conditional structure but does not establish architectural heterogeneity, which requires a further comparison against a flexible shared-architecture model. Simple trajectory nonlinearity, the paper warns, is not evidence of architecture.

The author is candid about limits. The formal illustration is local, linear, and time invariant, while whole-organism physiology is nonlinear, stochastic, and nonstationary; the Gramian analysis does not characterize global regime transitions, which require separate nonlinear analyses. Identifying a biologically meaningful cost functional is an inverse problem, and viability boundaries are scale- and task-dependent. But the practical conclusion is direct and likely to unsettle a research field built on composite indices: perturb carefully, follow trajectories, justify the state and measurement maps, declare burden and viability criteria, model measurement error, and retain conditional response structure long enough to determine whether a scalar summary has earned the claims made for it. A number can still summarize, the paper concludes, but only after its conditions are fixed, and it can never convert a context-dependent relation into a universal rank.

Subject of Research: Measurement architecture for multisystem stress-regulatory capacity in network physiology

Article Title: Regulatory capacity is not a scalar: metrics, perturbations, and viability in network physiology

Article References: Hogenkamp, L. (2026). Regulatory capacity is not a scalar: metrics, perturbations, and viability in network physiology. Frontiers in Network Physiology, 6, Article 1961712. https://doi.org/10.3389/fnetp.2026.1961712

Image Credits: AI Generated

DOI: 10.3389/fnetp.2026.1961712

Keywords: network physiology, allostatic load, regulatory bandwidth, control theory, observability Gramian, stress physiology, viability theory, perturbation response, neurodiversity, allostasis, resilience, person-environment fit

What makes this gap so persistent is the nature of the characteristic equation itself. For an ordinary differential equation without delay, stability can be checked with the Routh–Hurwitz criterion, a finite algebraic test on the coefficients. Introduce even a single delay, however, and the characteristic equation becomes transcendental: the exponential term means there are infinitely many roots, and their collective behavior must be controlled as the coefficients vary. A root that sits safely in the left half of the complex plane at one parameter value can migrate across the imaginary axis at another, and because the roots form an infinite family, no finite checklist of crossings suffices in general. This is precisely why the literature has accumulated so many partial results—criteria valid under sign restrictions on the coefficients, or for special sub-cases obtained by setting certain parameters to zero—while the full four-parameter problem resisted a definitive treatment.

The approach taken by the CICESE researchers builds on a classical pillar of the field: Pontryagin’s theorem on quasi-polynomials. Quasi-polynomials are expressions that combine polynomial terms with exponentials, and they arise naturally as characteristic functions of delay differential equations. Pontryagin’s result provides conditions under which the zeros of such functions can be controlled, effectively reducing an infinite-dimensional spectral question to a finite geometric one. What the new work adds is a novel geometric layer: rather than examining each candidate crossing curve separately and counting roots, the authors define a family of parametric curves and take their supremum envelope. The envelope collapses the potentially complicated crossing geometry into a single computable object whose epigraph—everything above the curve—is exactly the stable region.

This constructive character is what separates the contribution from purely numerical alternatives. Methods such as pseudospectral discretization of the solution operator, or large-scale quasi-polynomial root mapping, approximate the rightmost characteristic roots efficiently and are invaluable for large systems, but they deliver answers at specific parameter points rather than a global picture. The cluster treatment of characteristic roots and frequency-sweeping tests similarly determine, for fixed coefficients, the set of admissible delays. The new result inverts the question: for a fixed delay, it describes the entire set of stabilizing coefficients at once. The two viewpoints are related by a scaling map—varying the delay traces a ray through parameter space, and stability holds exactly where that ray intersects the stable region—so delay margins can be recovered from the coefficient-space picture, but the inversion is what makes gain charts and robustness assessment natural.

The practical payoff of having an explicit boundary is considerable. Because the stable region is an epigraph with a computable boundary curve, robustness indices follow immediately as distances from a nominal operating point to that boundary. In control design, such margins quantify how much parameter drift—a change in gain, a shift in actuator latency, wear in a mechanical component—a closed-loop system can tolerate before losing stability. When the boundary must be found numerically, engineers typically sample and interpolate; here the boundary is available in closed parametric form, so the distance computation inherits the precision of the underlying formulas. The accompanying algorithm supports standard visualization workflows, meaning the stability landscape can be plotted, inspected, and used interactively during controller tuning.

The examples in the paper illustrate how the general characterization specializes. For structured sub-classes of the four-parameter family, the envelope description collapses to very simple closed-form stability criteria—inequalities that can be checked by hand. One such example concerns an internal combustion engine model, for which the authors derive a delay-independent stabilization condition: a set of coefficient relationships under which the system remains stable for every value of the delay, no matter how large. Delay-independent stability is a strong and desirable property in applications where the delay is uncertain or time-varying, since it removes the delay from the list of quantities that must be bounded or measured. That such a condition emerges as a corollary of the general theory, rather than requiring a separate ad hoc analysis, demonstrates the unifying power of the characterization.

It is worth appreciating which parts of the construction are classical and which are genuinely new. The parametrization of the imaginary-axis crossing locus, the invariance of the number of unstable roots on connected components of the complement of the crossing curve, and the computation of crossing directions via eigenvalue perturbation are all standard elements of the D-partition method, which dates back to mid-twentieth-century work on automatic control. The new ingredients are threefold: first, the identification of the stable component of the crossing geometry without having to perform a root count at a test point in every component; second, the closed-form description of that component as the epigraph of the supremum envelope; and third, a global geometric theorem showing that the crossing curve admits no simple counterclockwise loop, which allows a local instability observation to be propagated to every non-stable component and thereby delivers necessity of the conditions. Necessity is the hard part—sufficient conditions are comparatively easy to obtain, but proving that every point outside the described region is genuinely unstable is what elevates the result from a useful criterion to a complete characterization.

The broader context of delay-differential equations helps explain why second-order systems with a single delay occupy such an important niche. Retarded functional differential equations, in which the rate of change depends on past states, model phenomena ranging from population dynamics with gestation periods to cutting forces in machining that depend on the tool position one revolution earlier. In feedback control, delays arise from actuation latency, sensor processing, transport of material or information, and computation time. Even when the underlying plant is linear, the delay makes the dynamics infinite-dimensional, and phenomena such as stability lobe diagrams—familiar to machinists as the lobed boundaries separating stable from chattering cutting conditions—emerge from the interplay between the delay and the gains. The classical stability lobe analysis for the undamped sub-case of the second-order equation is one of the special cases that the new theorem subsumes, placing a graphical technique used in manufacturing for decades within a single rigorous framework.

The paper also situates itself within a decade of intense activity on delay-system stability. Time-domain methods based on the delay Lyapunov matrix have matured to the point of delivering necessary and sufficient tests, and frequency-domain criteria continue to be refined. A distinct recent line of research exploits multiplicity-induced dominancy—the observation that characteristic roots of maximal multiplicity in low-order retarded equations tend to be the dominant ones—to assign the spectral abscissa directly. In the PID-stabilization literature, generalized Hermite–Biehler arguments have characterized stabilizing gains for low-order plants, and pseudoconcavity arguments have yielded the exact delay margin achievable by PID control of second-order systems. These strands, however, typically optimize along the delay axis for a structured controller or treat restricted parameter sub-classes; the coefficient-space region for the full four-parameter class had remained open.

For practitioners, the significance lies in the transition from pointwise testing to global design. A gain chart produced by the method shows, for fixed delayed-channel parameters, the complete region of stabilizing local parameters and its boundary, rather than a scatter of tested points. This means a designer can see at a glance how much freedom remains after choosing an operating point, where the nearest instability lies, and how the region deforms as the delay or the delayed-feedback gains change. Because the region is connected—being an epigraph—there are no isolated stable islands to hunt for, which simplifies both computation and intuition. The authors suggest that the framework supports standard control-design workflows directly, from visualization to margin computation to tuning.

The mathematical community’s interest in complete low-dimensional characterizations is not merely academic. Low-order delay systems serve as canonical models in vibration suppression, human balance control, and networked control systems, and they are the building blocks from which intuition about higher-order and multi-delay problems is formed. A definitive answer in the single-delay second-order case provides a benchmark against which numerical methods can be validated and a template that may guide extensions to multiple delays or higher order, where the geometric complexity of the crossing curves grows substantially. The supremum-envelope technique, in particular, offers a way of organizing complicated boundary geometry into a single monotone object, and whether analogous envelopes exist for richer classes of quasi-polynomials is a natural question for further work.

Finally, the result underscores a recurring theme in control theory: that classical tools, when combined with the right geometric insight, can resolve problems that computational methods alone leave only approximately answered. Pontryagin’s theorem has been available since the 1940s, and the D-partition technique since the 1950s, yet the complete four-parameter characterization required identifying the correct envelope construction and proving the global loop-free property that makes necessity tractable. The outcome is a rare thing in the stability theory of delay equations—a closed-form, necessary and sufficient answer—and it arrives with the algorithmic machinery to make it usable in practice.

Subject of Research: Research on the complete stability region characterization for second-order single-delay linear systems.

Article Title: Complete characterization of the asymptotic stability region for single-delay second-order linear systems

Article References: Cuesta-García, J. R., & Márquez-Martínez, L. A. (2026). Complete characterization of the asymptotic stability region for single-delay second-order linear systems. Frontiers in Control Engineering, 7, Article 1885855. https://doi.org/10.3389/fcteg.2026.1885855

Image Credits: AI Generated

DOI: 10.3389/fcteg.2026.1885855

Keywords: time-delay systems, asymptotic stability, control theory, quasi-polynomials, stability region, retarded functional differential equations, Pontryagin theorem, robustness margins, Complete, characterization, asymptotic, stability

A team of researchers at the University of Electronic Science and Technology of China in Chengdu has now tackled one of the most stubborn demands in this field: guaranteeing not just that a hydraulic actuator eventually tracks its commanded trajectory, but that the tracking error shrinks to a small neighborhood of zero within a deadline chosen in advance by the designer. In a study published in Applied Intelligence, Chen Wang, Jinping Ye, Xin Chen, Zhenlei Chen, Dan Jiang, Qing Guo and colleagues present a predefined-time adaptive neural backstepping controller for electro-hydraulic actuator systems, or EHASs, that remains stable even when key hydraulic parameters are uncertain and unknown external loads push against the piston.

The distinction between eventually and on time matters more than it might first appear. Conventional adaptive and backstepping controllers, which have dominated the hydraulic control literature for years, typically guarantee asymptotic convergence: the error decays toward zero as time stretches toward infinity, with no firm promise about when it becomes small enough to matter. Finite-time and fixed-time controllers tightened that promise, guaranteeing convergence within a bound that depends on initial conditions or on system parameters. Predefined-time control goes further still. The settling bound is set directly by the designer, independent of where the system starts, so an engineer can specify, for example, that the actuator must lock onto its target within two seconds regardless of how far away it begins. That kind of hard deadline is exactly what safety-critical and high-throughput applications demand.

The central obstacle is uncertainty. In an electro-hydraulic actuator, the dynamics involve hydraulic natural frequencies, leakage coefficients, effective bulk moduli and friction characteristics that are difficult to measure precisely and that drift over the machine’s life. The Chengdu team’s approach begins by gathering these unknown parameters into two lumped nonlinear terms rather than trying to estimate each one separately. Those lumped terms are then approximated online by radial basis function neural networks, a class of neural networks well suited to control because their outputs are linear in their weights. This linearity lets the adaptation law adjust the network weights in real time using Lyapunov-based arguments, the standard mathematical machinery for proving that a controlled system’s energy-like quantity decreases and the state converges.

A subtler mathematical hazard lurks in predefined-time designs for hydraulic systems. The error surfaces used in these schemes often contain low-order fractional-power terms, expressions in which the tracking error is raised to a power less than one. Near zero error, the derivatives of such terms can blow up, producing singularities in the virtual controllers that the backstepping procedure computes at each step of the design. A controller that demands infinite effort at the very moment the system approaches its goal is obviously unusable. The researchers resolved this by introducing twice continuously differentiable segmented functions that regularize the fractional-power terms, smoothing the expressions near the origin so the virtual controllers remain finite and well defined while preserving the predefined-time convergence property away from the origin.

With these ingredients in place, the team carried out a rigorous theoretical analysis establishing what control theorists call practical predefined-time stability. In practical stability, the error is not promised to reach exactly zero but to enter an adjustable neighborhood of zero, a small residual region whose size the designer can tune. The analysis proves that both the tracking error and the errors in the adaptive parameter estimates converge into that tunable neighborhood within the designer-specified time. This is a meaningful guarantee: it means the deadline is not an empirical observation from simulations but a mathematical property of the closed-loop system, derived under explicit assumptions about the uncertainty bounds.

The practical stakes become clear when one considers where electro-hydraulic actuators must perform on schedule. In a leg exoskeleton driven by hydraulics, a joint that lags its commanded trajectory by even a fraction of a second can destabilize a walking gait. In flight-control actuation, deadline violations translate directly into degraded handling. In industrial automation, cycle-time guarantees determine throughput. Earlier work in the field, including adaptive backstepping sliding-mode schemes for hydraulic manipulators, high-gain extended-state-observer controllers for position tracking, and barrier Lyapunov function methods for prescribed performance, steadily improved transient and steady-state accuracy, but none of these approaches could guarantee convergence within a time fixed by the designer in the presence of the full combination of parameter uncertainty and unknown loads.

The fixed-time and predefined-time literature that inspired the new work has matured rapidly in other domains. Researchers have applied fixed-time adaptive fuzzy and neural control to spacecraft attitude tracking, vehicle suspensions, flexible-joint manipulators, robotic manipulator fault tolerance and multi-UAV formation control. Polyakov’s foundational 2012 work on fixed-time stabilization of linear systems opened the door, and subsequent designs extended the idea to nonlinear and uncertain systems. What the Chengdu team contributes is the careful translation of these ideas to the specific, messy setting of hydraulic actuation, where the nonlinear flow-pressure characteristics, unmatched uncertainties and singularity risks of fractional-power terms make the mathematics considerably harder than in the electrical and aerospace systems where predefined-time control was first demonstrated.

Validation came in two stages. Simulation studies first confirmed that the theoretical convergence bounds hold in the idealized model, with the tracking error and the neural network’s weight-estimation errors settling into their prescribed neighborhoods within the specified time. The team then ran comparative experiments against both finite-time controllers and conventional backstepping controllers on an electro-hydraulic actuator setup. The results showed measurably improved tracking accuracy over both benchmarks, with the new controller reaching its small-error neighborhood faster and holding it more tightly under the same uncertain conditions. The comparisons matter because they isolate the benefit of the predefined-time formulation: the same underlying plant, the same class of disturbances, but a controller whose convergence deadline is a design parameter rather than an emergent property.

The work, published as volume 56, article 488 of Applied Intelligence with support from China’s national Brain Science and Brain-like Intelligence Technology major project and provincial science programs, points toward hydraulic machines that can be certified against timing guarantees rather than merely tuned until they behave. Because the neural networks learn the lumped uncertainties online, the approach does not require an accurate hydraulic model to be measured in advance, and because the settling time is chosen by the designer, the same controller can be retargeted to faster or slower tasks by adjusting a parameter rather than redesigning the control law. For the factories, aircraft and legged robots that depend on fluid power, the study offers a step toward actuators whose punctuality is written into the mathematics rather than hoped for in the field.

Subject of Research: Predefined-time adaptive neural control of electro-hydraulic actuator systems with parameter uncertainties

Article Title: Predefined-time adaptive neural control for electro-hydraulic actuator systems with parameter uncertainties

Article References: Wang, C., Ye, J., Chen, X., Chen, Z., Jiang, D., & Guo, Q. (2026). Predefined-time adaptive neural control for electro-hydraulic actuator systems with parameter uncertainties. Applied Intelligence, 56(15), Article 488. https://doi.org/10.1007/s10489-026-07535-0

Image Credits: AI Generated

DOI: 10.1007/s10489-026-07535-0

Keywords: electro-hydraulic actuators, predefined-time control, adaptive control, neural networks, backstepping control, radial basis function neural networks, parameter uncertainties, tracking error convergence, practical stability, industrial automation, hydraulic systems, control theory

Flutter is the nightmare scenario of aeroelasticity. As an elastic wing moves through air, unsteady aerodynamic forces couple with the structure’s natural vibration modes, typically bending and torsion. Below a critical speed the oscillations decay; at the flutter boundary the wing sustains neutral harmonic oscillation; beyond it, vibrations grow until structural failure. Conventional aircraft avoid the problem by stiffening the structure, paying a penalty in mass. Active flutter suppression offers an alternative: feedback control that augments the damping of the critical modes, pushing the flutter boundary beyond its open-loop limit and enabling the slender, high-aspect-ratio wings envisioned for next-generation aircraft.

The challenge is dimensionality. Future wings will carry many distributed control surfaces and sensors, creating a high-dimensional multi-input multi-output plant. A full multivariable controller can in principle use everything optimally, but demands accurate models and hard-to-formulate objectives. At the other extreme, independent feedback loops on collocated sensor-actuator pairs are simple and robust, as demonstrated historically on the XB-70, the B-1, the B-52, NASA’s DAST program, the Active Flexible Wing, and the X-56A, but they do not target specific modal dynamics. Modal blending occupies the middle ground: static blending vectors map the large sensor and actuator arrays onto a small number of virtual inputs and outputs aimed at the critical flutter mode, and a low-order dynamic controller operates in that reduced space.

Previous work, notably the H2-optimal blending framework of Pusch and colleagues, established how to choose blending vectors that maximize the modal energy transferred through the blended channel, and even exploited the transmission zero that blending introduces for prescribed pole placement. What remained unclear was the deeper geometry: which modal subspace to project onto, how many virtual channels to retain, and what the blending-induced zero costs when the controller must stabilize an already unstable mode. The new paper develops a unified framework connecting subspace geometry, transmission-zero placement, and modal observability and controllability, validated on a generic rectangular-wing model with eight control surfaces and eight accelerometers.

The first central result is an equivalence theorem with a clean geometric reading. Closed-loop performance depends only on the blending subspace, not on the particular basis chosen within it. Three different representations of the same flutter subspace, a raw modal basis, an orthonormalized basis, and a pseudoinverse-based construction, produce identical closed-loop transfer functions, verified numerically to principal angles below one ten-trillionth of a degree. Formally, the design variable is not a matrix but a point on the Grassmann manifold of subspaces, and basis choice is a gauge freedom. Conversely, corrupting the subspace by angular displacement degrades performance measurably, with output-side corruption proving far more damaging than input-side corruption at equal angles.

To judge which subspaces the hardware can actually support, the study introduces an energy-normalized minimum singular value metric that quantifies whether all critical modes can be observed and controlled independently and simultaneously. Because the raw state mixes displacements, velocities, aerodynamic lag variables, and actuator states in incompatible units, the author first maps everything into energy coordinates where every component carries the unit of the square root of a joule. The metric then has a direct physical meaning: it bounds how strongly sensor noise corrupts the reconstruction of modal coordinates, and it reveals whether any direction in the target subspace is effectively unactuable.

Applied across a velocity sweep from 40 to 160 meters per second, the metric delivers a sobering structural verdict. The flutter subspace and the neighboring residual subspace are each individually well observable and controllable, but the combined four-dimensional subspace is neither, at any airspeed, with the eight-accelerometer, eight-surface layout. The consequence is that geometric mode isolation, projecting the residual pair out of the measurements before blending, is impractical on this wing: at the synthesis point of 130 meters per second, the flutter and residual subspaces, separated by nearly 38 degrees in the full state space, become collinear to within a sixth of a degree once mapped through the sensor array. No choice of blending vectors can separate directions the measurement map renders collinear. Remedies must come from sensor diversity, placement optimization, or observer-based modal filtering.

The second central result concerns what happens when a two-dimensional oscillatory mode is squeezed into a single virtual channel. A flutter mode is inherently a complex conjugate pair spanning a two-dimensional real plane, so rank-one blending restricts instantaneous measurement and actuation to one direction in that plane, discarding the quadrature component. The paper derives a closed-form map showing that the resulting transmission zero is governed by a single quadrature mismatch angle between the input and output blending directions: the zero sits at the real part of the modal pole minus the modal frequency times the tangent of the mismatch angle. For an unstable mode, small mismatch angles place the zero in the right half-plane, precisely where Bode’s sensitivity integral makes it fundamentally costly. Worse, the H2-optimal direction selection is nearly blind to this variable: for a lightly damped mode the energy objective is almost flat in the mismatch angle, and its residual preference actually lands the zero at the modal frequency in the right half-plane for an unstable mode.

The practical consequences are quantified in a head-to-head comparison under an identical structured H-infinity synthesis framework, with a fourth-order dynamic core, bandpass-weighted control-effort objectives, and hard disk-margin constraints of at least 6 decibels gain and 45 degrees phase. On the wing model at 130 meters per second, where the open-loop flutter poles are already unstable, the rank-two multi-input multi-output blending spanning the full flutter subspace meets all soft performance goals with a worst-case value of 0.645, while every tested rank-one direction misses them. The H2-optimal single-channel design reaches 1.149, and a mismatch sweep of a restricted rank-one family varies by a factor of 3252 between its best and worst feasible angles, confirming that the quadrature mechanism is no artifact of simplified models. A zero-aware design that pins the channel zero at the origin at a stable extraction point fails even more dramatically: as flight conditions change and the mode’s eigenvectors rotate, the zero drifts into the right half-plane, rendering the design margin-infeasible at the synthesis point. A random subspace control confirms the geometry, not the channel count, is decisive.

The study’s takeaway reshapes how engineers should think about modal control of oscillatory aeroelastic systems. The blending subspace must match the dimension of the target mode: for the unstable flutter mode studied here, two virtual channels spanning the full modal plane remove the projection-induced zero entirely, at the modest price of a second channel and a larger tunable parameterization. Rank-one blending remains a legitimate low-complexity option for damping augmentation of stable modes, but its performance is governed by quadrature mismatch rather than energy transfer, a distinction the H2 objective cannot capture. The author proposes orthonormal multi-channel blending built from energy-normalized pole vectors as a canonical choice, together with the minimum singular value metric as a principled test of whether a given sensor and actuator layout can support independent modal control at all. Extensions to gust load alleviation, swept planforms, sensor placement optimization, and gain-scheduled linear parameter-varying frameworks are outlined as next steps, and the wing model and analysis code are being released openly, inviting the aeroelasticity community to build on a result that turns an intuitive design art into a matter of measurable geometry.

Subject of Research: Subspace geometry and transmission-zero effects in modal blending for active flutter suppression of aeroelastic wings

Article Title: Subspace geometry and performance tradeoffs of modal blending for active flutter suppression

Article References: Eichelsdörfer, J. (2026). Subspace geometry and performance tradeoffs of modal blending for active flutter suppression. Aerospace Systems. https://doi.org/10.1007/s42401-026-00552-4

Image Credits: AI Generated

DOI: 10.1007/s42401-026-00552-4

Keywords: aeroelasticity, active flutter suppression, modal blending, control theory, transmission zeros, Grassmann manifold, observability, controllability, H-infinity synthesis, flexible wings, aeroservoelasticity, German Aerospace Center