Physically accessible decision-making (PADM) and a measurement-access risk frontier are formulated: the Bayes-optimal target risk minimized over records realizable under cost, bandwidth, latency, disturbance, memory and actuation constraints.
Abstract
Rapidly scaling autonomous science is limited not only by algorithms, compute or data volume, but by which physical records a platform exposes before action. We formulate physically accessible decision-making (PADM) and a measurement-access risk frontier: the Bayes-optimal target risk minimized over records realizable under cost, bandwidth, latency, disturbance, memory and actuation constraints. The frontier gives a no-free-autonomy limit: automation cannot collapse decision uncertainty by computation alone; an optimal controller cannot remove target components absent from its record, and closing that gap requires expanded access, auditing, tolerated disturbance, slower operation or restricted deployment. In monitored feedback, displacement-only control remains exposed to a hidden switching force, whereas a finite-bandwidth cue recovers part of the missing projection before action. A chemistry-aware candidate-ranking audit with a 1000-target stress panel, Gaussian sensing, hidden-regime decisions and cost-aware/thermodynamic channel selection provide reproducible checks. PADM identifies target-specific audit value and residual oracle gaps before deployment.
CausalNav, a controller built around a signed, action-conditioned transition graph over identified state coordinates, is studied with CausalNav, a controller built around a signed, action-conditioned transition graph over identified state coordinates, which attains the best average rank.
Yi-Yao Zhang, Diksha Goel, Hussain Ahmad et al.· 0 citations
We study counterfactual protocol evaluation: whether data collected under a realised observation protocol determine the predictive value of alternatives that were never deployed. Protocol value is the population $R^2$ of the Bayes-optimal predictor of a fixed trajectory-level target from the measurements an alternative would collect. We show that even infinite benchmark data need not determine this value: distinct latent covariance structures can induce the same benchmark measurement--target law while assigning different values to the same alternative. We develop a value-specific identification theory in which only latent ambiguity that changes the alternative's value matters. For linear targets, invisible covariance directions certify non-identification, while targeted measurements can restore identification without recovering the full latent covariance; an exact permutation construction extends the result to nonlinear aggregate targets. With finite dense calibration data, uniform error bounds control protocol-selection regret and distinguishable value gaps. Exact marginal gains then support cost-constrained, target-aware observation design. Simulations and retrospective analyses of Sleep-EDF and Long-Term AF show that broad temporal-layout differences can be more reliably distinguished than fine placements selected from finite data. Together, these results connect identification, calibration resolution and observation design for undeployed protocols.
Simultaneous input-gain variation and large additive disturbance create a control problem in which a fixed observer or nominal controller may be unable to reproduce the performance of a regime-aware design. We study a training--deployment asymmetry: during simulation or commissioning, an expert controller is allowed to use the known gain and disturbance, whereas the deployed controller can use only the reference and measured states. Directly imitating expert actions is generally unsafe because the same instantaneous student observation may correspond to different privileged regimes and hence different expert actions. We propose a mechanism-guided transfer route rather than a new neural architecture. An exact sampled-data identity removes the additive disturbance from the expert law and reduces learning to a task-relevant inverse input gain inferred from causal state history. The latent target is reconstructed from expert actions and deployment-visible trajectories, so the true plant parameter is not required as a student label. A common-quadratic certificate is derived for the actual augmented sampled recursion, followed by explicit residual, coverage, switching, noise, and saturation qualifications. A parameter-regime scan shows that the nominal observer's error grows sharply as $a$ decreases and that the next gain above the best non-failing tuning diverges for every tested $a<1$. Direct action networks also fail in closed loop despite moderate offline error, whereas the structured student remains close to the privileged expert and reduces tracking RMSE by about 69\% relative to the tuned observer in unseen 60-s trials. The contribution is an interpretable design perspective for turning privileged multi-regime control knowledge into a deployable adaptive controller, together with conditions under which the transfer is meaningful.
Xitong Niu, Peifeng Hui, Zheyong Jiang et al.· 0 citations
This work studies the synthesis of active perception policies for predictive safety monitoring in partially observable stochastic systems. Operating under sensing and communication budgets, the proposed monitor dynamically schedules sensor queries to maximize information gain about the safety of future states. The underlying stochastic dynamics are captured by a labeled hidden Markov model, with safety requirements defined by a deterministic finite automaton. To enable active information acquisition, we introduce minimizing <inline-formula> <tex-math notation="LaTeX">$k$ </tex-math></inline-formula>-step Shannon conditional entropy of the safety of future states as a planning objective, under the constraint of a limited sensor query budget. Using observable operators, we derive an efficient algorithm to compute the <inline-formula> <tex-math notation="LaTeX">$k$ </tex-math></inline-formula>-step conditional entropy and analyze key properties of the conditional entropy gradient with respect to policy parameters. We validate the effectiveness of the method for predictive safety monitoring through a dynamic congestion game example.
S. Udupa, Jie Fu· IEEE Control Systems Letters· 0 citations
Artificial intelligence (AI) increasingly produces uncertainty outputs for sensing and measurement tasks, but the evidence supporting these outputs may not maintain a traceable correspondence with the relevant real-world conditions. This study conducted a structured 15-dimensional coding review of 566 studies, of which 556 formed stable dominant AI uncertainty claim units and entered the common analytic set. Each final code was linked to row-level audit evidence. The unidimensional distributions first showed breakpoint states with nonzero frequencies at the relevant nodes of the claim–condition–test–uncertainty-response chain, thereby confirming observable evidence discontinuities in the current corpus. The studies were then stratified using three sequential, non-compensatory evidence questions. Among the 556 studies, 347 (62.4%) were classified as Weak, 133 (23.9%) as Medium, and 76 (13.7%) as Strong. Descriptive cross-dimensional comparisons showed that specific OOD/drift risk or Error/quality estimation claims, Multiple uncertainty entry points, Multiple physical information types, and Sampling/ensemble approximation more often co-occurred with higher evidence traceability; Prediction reliability/confidence claims, a standalone Uncertainty proxy/score, and Latency/real-time inference constraints more often co-occurred with lower evidence traceability. This study summarizes the above evidence discontinuity as the physical fidelity gap (PFG), which refers to incomplete or unverifiable evidential links between AI uncertainty claims publicly reported in the literature and the relevant physical, measurement, or operational conditions. PFG provides a scope-bounded reference for locating links in the evidence chain that may need strengthening. The breakpoints observable in the current corpus indicate that the public evidence still has room for improvement in forming continuous, verifiable claim–condition–test–response correspondences; the tiered comparison indicates that subsequent work can strengthen the evidence chain by clarifying claim conditions, incorporating the relevant conditions into empirical testing, and reporting identifiable uncertainty responses and extended corroboration.
Lin Guo, Ai-Wen Ma, Heng Zhou et al.· Italian National Conference...· 0 citations
A calibrated stochastic world model can reveal how uncertain a future is without revealing why it branches. The same conditional future law can arise because an observation aliases physical states or because dynamics remain random after the declared full state is fixed. We prove that ordinary transitions cannot identify these two sources, even for a perfect probabilistic predictor. ClosurePairs makes them identifiable by crossing compatible microstates with repeated exogenous disturbances and estimating state, noise, and state-noise interaction variance. The central consequence is operational: under finite hierarchical sampling, forecast difficulty governs the useful compute scale, while the alias/process composition provides complementary information about its direction-resolving the current state or sampling future randomness. ClosurePairs recovers source attribution at unchanged likelihood, reduces equal-budget decomposition error in a nonlinear interaction benchmark, and supports observation-only routing. On exact-marginal MetaWorld twins, an output-only allocator is at chance while a Closure-supervised probe on frozen JEPA-WM features routes 89.8-100%. In an independent ManiSkill PushCube confirmation, a stochastic RSSM's outputs and latents remain at chance, whereas an RGB-only Closure probe routes 100% under both ID and geometry/camera OOD over five seeds, matching direct allocation rather than exceeding it. Across five unseen allocation menus, the same Closure probe routes 92.5%/90.4% ID/OOD with no new oracle labels, versus 37.9%/32.9% for a frozen direct allocator. ClosurePairs is therefore an identifiable, reusable mechanism target that cannot be recovered from forecast quality alone.
Yi-Xin Dong· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
