Searcharxiv⌕ Search

arXiv subjects

Zhaobo Liu

Publications and source records attributed to Zhaobo Liu.

17 recordsLinked to original sources

Control and Identification in Least-Squares Self-Tuning Regulation with Unknown Input Gain

We study stability, minimum variance tracking and parameter identification for the Åstr{ö}m--Wittenmark self-tuning regulator with unknown input gain. Ordinary least squares (LS) estimates are used directly in certainty-equivalent control. Under the stated plant and martingale noise assumptions, we establish almost sure stability and optimal tracking for every bounded reference, provided that the gain estimates remain nonzero. When at least two parameters are estimated, all coefficient estimates are strongly consistent. The proof obtains a logarithmic information lower bound from an LS energy identity and uses it to establish gain consistency without assuming closed-loop stability. For at least two estimated parameters and every bounded reference, the squared error is almost surely $O(\log\log n/n)$ for the estimated ratios of the plant coefficients to the input gain and $O(\log\log\log n/\log n)$ for the full parameter estimate. A law of the iterated logarithm shows that the latter rate is sharp under zero reference. The bound for the full parameter estimate improves with reference energy, which also appears explicitly in the logarithm law for cumulative squared tracking residuals. These results show that optimal tracking can coexist with substantially slower estimation of the full parameter vector than of the coefficient ratios used in feedback. For at least two estimated parameters, a chi-square limit yields asymptotically exact confidence ellipsoids for every bounded reference.

math.OC↗

The Critical Time for Local Controllability of a Water Tank

We determine the critical time for local controllability near equilibrium of a water tank governed by the one-dimensional Saint-Venant equations and controlled by its horizontal acceleration. With $C^1$ fluid trajectories and continuous controls, local controllability holds exactly for $T>T_c$. In normalized units, $T_c=2+d_c\approx 2.43674$, where $d_c$ is determined by two transcendental equations. In particular, local controllability fails for $2 2$. For times just above $T_c$, we construct a control that cancels the linear and quadratic terms of the endpoint expansion and apply a fixed point argument. At $T_c$, a weighted comparison with a critical plane curve, combined with estimates along the nonlinear characteristics, gives the obstruction.

math.OC↗

Expected Minimax Regret Rates for Adaptive LQG Control on Fixed Parameter Sets

We study how a fixed model family determines expected minimax regret in adaptive LQG control. The family is compact and affine, with partial observation and unknown noise covariances. We quantify how information from residual covariances and input responses can vanish relative to differences between optimal control laws. This relation yields matching bounds that classify regret as bounded, logarithmic, sublinear polynomial or linear. Polynomial exponents are rational. For real algebraic model and cost data, a terminating procedure determines the regret class and the exact exponent of polynomial growth before data collection. A causal policy attains these rates using model estimation and bounded probes, reselecting model sets from observations without a supplied common stabilizer. For costs with fixed positive definite output and input weights, polynomial exponents are at most one half.

math.OC↗

The Complexity of Minimal Positive Realization

We study the complexity of minimal positive realization for discrete-time systems. We prove that, given a strictly proper transfer matrix with rational coefficients and an order bound encoded in unary, deciding whether a positive realization exists within that bound is complete for the existential theory of the reals. We further prove hardness for stable single-input single-output transfer functions that admit positive realizations, even when the bound equals the McMillan degree.

math.OC↗

Critical Growth for Stochastic Adaptive Stabilization with an Unknown Input Gain

This paper studies adaptive stabilization of a class of scalar discrete-time nonlinear systems with unknown drift coefficients and an unknown nonzero input gain. Under iid Gaussian noise, we establish a necessary and sufficient polynomial criterion for global stabilizability, characterizing the admissible combinations of growth exponents. When the criterion holds, we construct a feedback combining finite calibration with adaptive control. The time average of the squared output converges almost surely to the noise variance, attaining the smallest possible value. If the criterion fails, then for every Borel causal feedback there exist fixed parameter values and an initial state for which the output magnitude diverges with positive probability.

math.OC↗

Computational complexity of static output feedback stabilization

We study the computational complexity of static output feedback (SOF) stabilization for continuous-time linear systems. For rational plant data, we prove that this problem is complete for the existential theory of the reals: it is equivalent, under polynomial-time transformations, to deciding whether a system of polynomial equations and inequalities with integer coefficients has a real solution. This result holds even for controllable and observable integer plants with two inputs. We characterize the effective gain dimension as the minimum number of linear combinations of gain entries that determine the characteristic polynomial for every gain. When this dimension is fixed, both feasibility testing and the construction of a rational stabilizing gain, whenever one exists, can be performed in polynomial time. SOF stabilization also remains strongly NP-hard for a family of minimal single-input integer plants. Every feasible plant in this family admits a stabilizing gain with entries in $\{-1,1\}$. For these gains, stability is preserved under simultaneous perturbations of all entries, provided that each perturbation has magnitude at most a uniform positive constant independent of the plant dimensions.

math.OC↗

Controller Identifiability: The Boundary Between Logarithmic and Square-Root Local Minimax Regret in Structured Adaptive LQR

We study when logarithmic regret is attainable in structured adaptive linear-quadratic regulation. The state and input matrices depend affinely on a common unknown parameter. We consider neighborhoods of a known nominal model that is stabilizable and observable through the output associated with the state cost, with radii proportional to the inverse fourth root of the horizon. The key condition is controller identifiability, which requires the optimal gain derivative to vanish in every parameter direction that leaves the closed-loop dynamics unchanged under nominal optimal feedback. Under suitable regularity conditions on the noise density, this condition at the nominal parameter yields logarithmic local minimax regret when the gain derivative there is nonzero. Failure of the condition yields square-root local minimax regret. If the nominal gain derivative vanishes, local minimax regret remains bounded. We prove that controller identifiability is equivalent to the optimal gain remaining unchanged under sufficiently small invisible perturbations of the nominal model. Motivated by this invariance, we construct a certainty-equivalent policy that estimates only the visible parameter component and attains the logarithmic upper bound under controller identifiability. The policy requires neither the horizon nor the noise law, and its guarantee holds for independent, identically distributed noise with zero mean and finite positive definite covariance.

math.OC↗

Global Output Lifting and Surjectivity in Invertible Nonlinear MIMO Systems

Global inversion of nonlinear multivariable systems raises the question of whether every target value of the selected output derivatives can be realized by a plant state. We prove that, under globally regular modified inversion with fixed row selection and finite termination at full input rank, every fixed homogeneous linear output assignment yields a surjective output-derivative map with a smooth global section. The result requires a state bound on each finite interval in terms of the initial state and finitely many output derivatives, uniformly over inputs. This condition is strictly weaker than weak uniform zero-state detectability, making the separate surjectivity requirement in a global inversion approach to stabilization a consequence of detectability under the stated inversion hypotheses. The section selects a realizing state smoothly for every derivative target, including in the presence of internal dynamics, and is constructed through complete lifting of compatible smooth references. For known plants, the same lifting method yields a smooth dynamic controller using full state feedback and current reference derivatives, with initialization from the plant state. From any initial state, it achieves exact tracking of any smooth desired output after any prescribed positive time.

math.OC↗

Fundamental Limits of Adaptive Stabilization with an Unknown Growth Exponent

A basic question in adaptive control is how rapidly a discrete-time nonlinear plant with unknown parameters may grow while remaining stabilizable. When the nonlinear growth exponent is known and only a scalar coefficient is unknown, existing theory identifies 4 as the exact critical exponent for stabilizability. We determine how this critical exponent changes when the growth exponent is also unknown. For any positive disturbance bound, one feedback law stabilizes all plants in some neighborhood of a nominal parameter pair if and only if the nominal exponent is below $3\sqrt{3}/2$. At equality, every compact parameter set whose exponents do not exceed this value remains stabilizable. If the exponent belongs to a known finite set, the critical exponent remains 4 for every nondegenerate compact coefficient interval. Hence finite exponent sets and arbitrarily short exponent intervals can have different critical exponents. Uncertainty in the growth rate therefore changes the range of growth that feedback can stabilize, not merely the size of the uncertainty set. The critical exponent remains $3\sqrt{3}/2$ under a known positive state-dependent multiplier bounded above and away from zero. For a system whose coefficient and exponent are known functions of an unknown parameter, a compact parameter family is stabilizable when its exponents do not exceed this value. At an interior parameter point where the Jacobian of these functions has rank two and the exponent is at least this value, no compact neighborhood is stabilizable.

math.OC↗

A Game-Theoretic Characterization of Feedback Capability for Fully Coupled Vector-Valued Nonparametric Systems

We study feedback stabilization for the discrete-time system $x_{t+1}=f(x_t)+u_t+w_{t+1}$ in $\mathbb{R}^d$ with unknown $f$ and arbitrary bounded disturbances. For scalar plants, the sharp feedback capability threshold under generalized Lipschitz uncertainty is $3/2+\sqrt{2}$. We treat fully coupled vector-valued systems, where scalar order and interval recursion are unavailable and coupling precludes a coordinatewise reduction. We introduce a response-history escape game in which the adversary seeks a finite envelope and an unbounded state radius. Borel determinacy ensures that exactly one player has a winning strategy at each slope. We prove that the same player wins from every finite response history, and slope monotonicity gives an independently defined game value $Γ_d$. We prove that $Γ_d$ is finite and is the strict feedback capability threshold for the plant problem. If $L<Γ_d$, one causal feedback law stabilizes every plant in the uncertainty class against every bounded disturbance sequence. If $L>Γ_d$, for every causal feedback law there exist a plant in the same class and a bounded disturbance sequence such that the closed-loop state sequence is unbounded. The proof uses one controller for all subcritical slopes and a realization in a Hilbert space based on the Kirszbraun--Valentine extension theorem. An explicit nearest-neighbor law gives a lower bound above one in every finite dimension, including $Γ_2\ge 2/\sqrt{3}$. Dimension monotonicity gives $Γ_d\leΓ_1$, and comparison with the scalar theory yields $Γ_1=3/2+\sqrt{2}$.

math.OC↗

Genie Sim 3.0 : A High-Fidelity Comprehensive Simulation Platform for Humanoid Robot

The development of robust and generalizable robot learning models is critically contingent upon the availability of large-scale, diverse training data and reliable evaluation benchmarks. Collecting data in the physical world poses prohibitive costs and scalability challenges, and prevailing simulation benchmarks frequently suffer from fragmentation, narrow scope, or insufficient fidelity to enable effective sim-to-real transfer. To address these challenges, we introduce Genie Sim 3.0, a unified simulation platform for robotic manipulation. We present Genie Sim Generator, a large language model (LLM)-powered tool that constructs high-fidelity scenes from natural language instructions. Its principal strength resides in rapid and multi-dimensional generalization, facilitating the synthesis of diverse environments to support scalable data collection and robust policy evaluation. We introduce the first benchmark that pioneers the application of LLM for automated evaluation. It leverages LLM to mass-generate evaluation scenarios and employs Vision-Language Model (VLM) to establish an automated assessment pipeline. We also release an open-source dataset comprising more than 10,000 hours of synthetic data across over 200 tasks. Through systematic experimentation, we validate the robust zero-shot sim-to-real transfer capability of our open-source dataset, demonstrating that synthetic data can server as an effective substitute for real-world data under controlled conditions for scalable policy training. For code and dataset details, please refer to: https://github.com/AgibotTech/genie_sim.

cs.RO↗

Stability of MIMO PID With Backward Differences Under Fast Sampling: An Exact Spectral Criterion

Backward differences are a standard digital realization of derivative action in proportional-integral-derivative control. This note proves that, for multivariable state-space plants, this implementation can be unstable no matter how small the sampling period is. The exact lifted model reveals fast eigenvalues created by the stored previous output sample. Away from boundary spectra, stability is equivalent to ideal loop stability plus Schur stability of the product of the output, input, and derivative gain matrices.

math.OC↗

Stabilizability Theorems on Discrete-time Nonlinear Uncertain Systems

This paper derives two stabilizability theorems for a basic class of discrete-time nonlinear systems with multiple unknown parameters. First, we claim that a discrete-time multi-parameter system is stabilizable if its nonlinear growth rate is dominated by a polynomial rule. Later, we find that a stabilizable multi-parameter system in discrete time is possible to grow exponentially fast. Meanwhile, optimality and closed-loop identification are also discussed in this paper.

math.OC↗

Inverse Eigenvalue Problem For Mass-Spring-Inerter Systems

This paper has solved the inverse eigenvalue problem for "fixed-free" mass-chain systems with inerters. It is well known that for a spring-mass system wherein the adjacent masses are linked through a spring, the natural frequency assignment can be achieved by choosing appropriate masses and spring stiffnesses if and only if the given positive eigenvalues are distinct. However, when we involve inerters, multiple eigenvalues in the assignment are allowed. In fact, arbitrarily given a set of positive real numbers, we derive a necessary and sufficient condition on the multiplicities of these numbers, which are assigned as the natural frequencies of the concerned mass-spring-inerter system.

math.OC↗

Asymptotic Behavior of Least Squares Estimator for Nonlinear Autoregressive Models

This paper is concerned with the least squares estimator for a basic class of nonlinear autoregressive models, whose outputs are not necessarily to be ergodic. Several asymptotic properties of the least squares estimator have been established under mild conditions. These properties suggest the strong consistency of the least squares estimates in nonlinear autoregressive models which are not divergent.

math.PR↗

Is It Possible to Stabilize Disrete-time Parameterized Uncertain Systems Growing Exponentially Fast?

This paper derives a somewhat surprising but interesting enough result on the stabilizability of discrete-time parameterized uncertain systems. Contrary to an intuition, it shows that the growth rate of a discrete-time stabilizable system with linear parameterization is not necessarily to be small all the time. More specifically, to achieve the stabilizability, the system function $f(x)=O(|x|^b)$ with $b<4$ is only required for a very tiny fraction of $x$ in $\mathbb{R}$, even if it grows exponentially fast for the other $x$. The proportion of the mentioned set in $\mathbb{R}$, where the system fulfills the growth rate $ O(|x|^b)$ has also been computed, for both the stabilizable and unstabilizable cases. This proportion, as indicated herein, could be arbitrarily small, while the corresponding system is stabilizable.

math.OC↗

When Adaptive Diffusion Algorithm Converges to True Parameter?

We attempt to answer the question what data brings adaptive diffusion algorithms converging to true parameters. The discussion begins with the diffusion recursive least squares (RLS). When unknown parameters are scalar, the necessary and sufficient condition of the convergence for the diffusion RLS is established, in terms of the strong consistency and mean-square convergence both. However, for the general high dimensional parameter case, our results suggest that the diffusion RLS in a connected network might cause a diverging error, even if local data at every node could guarantee the individual RLS tending to true parameters. Due to the possible failure of the diffusion RLS, we prove that the diffusion Robbins-Monro (RM) algorithm could achieve the strong consistency and mean-square convergence simultaneously, under some cooperative information conditions. The convergence rates of the diffusion RM are derived explicitly.

math.OC↗