KHAWLA capsule platform for targeted gastrointestinal delivery with uncertainty aware release control

Abstract

Author preprint. Computational feasibility study and localization fault analysis. Not peer reviewed. Localized delivery by an ingestible capsule requires a release decision that accounts for localization error and stale measurements. This study presents KHAWLA, a proposed gastrointestinal capsule architecture, and evaluates a release supervisor in a reproducible two dimensional stochastic model. The supervisor combines estimated target distance, a covariance based margin, measurement freshness, commanded speed, and an eight sample dwell requirement. A total of 11000 simulation runs comprise 6000 primary comparisons, 2000 noise sensitivity runs, and 3000 comparisons with simpler release rules. With 1000 runs per primary condition, release occurred within the true 3 mm target radius in 1000 of 1000 nominal runs and 1000 of 1000 runs assigned temporary sensor outages. A distance threshold alone achieved 447 and 405 successful releases, respectively. Median release time increased from 14.7 to 16.2 s in the nominal condition and from 14.7 to 16.7 s in the outage condition. Simpler margin and dwell rules also achieved complete nominal success in their 500 run groups. Every tested rule failed under a hidden 4 mm localization bias. The results support further investigation of release inhibition during uncertain localization, while showing that covariance alone cannot establish physical targeting accuracy. The proposed hardware allocations and geometric deposition proxy remain unvalidated; no capsule prototype, biological experiments, or therapeutic efficacy data are reported. Data and code: The attached scientific supplement contains simulation source, pinned dependencies, saved primary and simpler-rule records, sensitivity summaries, figures, and checksums. Individual sensitivity records can be regenerated using the disclosed seeds. All numerical results are model generated. No human, animal, or patient experiments are reported. Project page: https://khawla-capsule-research.jovialorbit32.chatgpt.site AI assistance: OpenAI ChatGPT and Codex assisted with literature organization, drafting, simulation code, analysis, and figures. The named author retains responsibility. KHAWLA is named in memory of the author's grandmother, Khawla.