arXiv subjects

Sebastian Escobar

Publications and source records attributed to Sebastian Escobar.

4 recordsLinked to original sources

arXiv2026

Conditional Timed Partial Orders: An Expressive and Interpretable Framework for Robot Task Specification and Planning

Timed Partial Orders (TPOs), originally proposed for workflows, provide an interpretable framework for robot task specification with planning algorithms based on mixed-integer linear programming (MILP). However, TPOs are limited in expressivity, capturing only partial-order events with simple timing constraints. In this paper, we introduce Conditional TPOs (cTPOs), which extend TPOs with richer relative-timing constraints and conditional event activations based on environmental conditions. We show that planning for cTPOs also reduces to an MILP problem; however, the added expressivity results in significantly larger MILPs that can become computationally intractable. To address this challenge, we propose a decomposition algorithm that partitions a cTPO into smaller sub-TPOs, yielding a sequence of smaller MILP problems. We prove that this decomposition is complete and preserves plan optimality while improving the interpretability of complex tasks. Experimental results demonstrate the effectiveness of cTPOs as a task specification framework and the efficiency of our decomposition approach, achieving up to four orders of magnitude speedup over the monolithic MILP.

cs.RO↗

arXiv2026

Predicted Capabilities of the SPRITE SmallSat for a Low-Redshift Lyman Continuum Emission Survey

Ionizing Lyman continuum (LyC; $\lambda < 912~\rm{\mathring{A}}$) radiation from low-redshift ($z \sim 0.3$) galaxies provides crucial insight into the processes that contributed to cosmic reionization. While the \textit{James Webb Space Telescope} has observed galaxies at redshifts as high as $z \sim 14$, detecting LyC beyond $z \sim 3$ is challenging due to absorption by neutral hydrogen in the intergalactic medium (IGM). Low-redshift LyC emitters (LCEs), therefore, act as proxies for their high-redshift counterparts, enabling direct measurements of LyC escape fractions with reduced IGM interference. These observations allow detailed ancillary studies of galaxy properties and the mechanisms driving ionizing photon escape, which cannot be directly observed at the Epoch of Reionization. This paper examines the capabilities of the Supernova remnants and Proxies for Re-Ionization Testbed Experiment (SPRITE) SmallSat, designed to study LyC emission from star-forming galaxies at $0.16 < z < 0.4$. SPRITE uses advanced mirror coatings and a highly sensitive far-ultraviolet imaging spectrograph, enabling it to probe LyC from galaxies that have been difficult to study with prior and existing instruments. To assess SPRITE's predicted performance in LyC studies, we select eight previously confirmed LCEs from the Low-redshift Lyman Continuum Survey as commissioning targets. Observations of these commissioning LCEs will validate SPRITE's LyC sensitivity and characterize its detection limits. This will enable the broader SPRITE low-redshift LCE survey, which will provide new constraints on the physics of LyC escape and help bridge the gap between low- and high-redshift LyC studies. SPRITE will also inform the design and scientific potential of future Lyman-UV missions, including the Habitable Worlds Observatory.

astro-ph.GA↗

arXiv2026

Near UV Stellar Activity and Brightness Fluctuations of the Alpha Centauri AB Star System from Weeks to Decades -- Inputs for Reflected Light Spectroscopy with HWO

We present the most comprehensive near-ultraviolet (NUV: 2550-3255 Angstrom) activity record to date for the Alpha Centauri AB system, combining archival IUE and HST observations spanning nearly five decades with new high-cadence CUTE measurements. We show that Alpha Centauri A exhibits predominantly quiescent NUV behavior, with the majority of observations remaining within 1 sigma of the median flux and only rare chromospheric flaring events (1 flare every 12 years), consistent with its weak chromospheric activity and 19-year stellar cycle inferred from X-ray and FUV observations. In contrast, Alpha Centauri B displays a broader variability envelope, characterized by more frequent and higher-amplitude chromospheric excursions that track its well-established 8-year magnetic activity cycle. Using Lomb-Scargle analysis on the Mg II index derived from CUTE observations, we estimate the rotational period of Alpha Centauri A to be on timescales of 15-20 days. We also confirm the coherence of the stellar activity cycle of Alpha Centauri B in the NUV with its X-ray activity cycle. These data establish a critical reference framework for interpreting reflected-light observations of terrestrial exoplanets and for assessing the detectability of ozone and other biosignature-related features at NUV wavelengths with future facilities such as the Habitable Worlds Observatory. These results indicate that HWO observations of terrestrial exoplanets in reflected light photometry and spectroscopy around magnetically inactive early G-type stars and early K-type stars may be expected to show 10-20 percent and 30-40 percent temporal flux variability, respectively, over the course of months to years from the changing stellar inputs alone.

astro-ph.SR↗

arXiv2026

Training with Hard Constraints: Learning Neural Certificates and Controllers for SDEs

Due to their expressive power, neural networks (NNs) are promising templates for functional optimization problems, particularly for reach-avoid certificate generation for systems governed by stochastic differential equations (SDEs). However, ensuring hard-constraint satisfaction remains a major challenge. In this work, we propose two constraint-driven training frameworks with guarantees for supermartingale-based neural certificate construction and controller synthesis for SDEs. The first approach enforces certificate inequalities via domain discretization and a bound-based loss, guaranteeing global validity once the loss reaches zero. We show that this method also enables joint NN controller-certificate synthesis with hard guarantees. For high-dimensional systems where discretization becomes prohibitive, we introduce a partition-free, scenario-based training method that provides arbitrarily tight PAC guarantees for certificate constraint satisfaction. Benchmarks demonstrate scalability of the bound-based method up to 5D, outperforming the state of the art, and scalability of the scenario-based approach to at least 10D with high-confidence guarantees.

eess.SY↗