Observer Construction Under Observability Constraints: Conditional Requirements for Contamination Detection in Adaptive Multi-Agent Systems

Abstract

RT3 examines which measurement arrangements preserve task-relevant distinctions when adaptive systems can change their evidence and evaluation rules. It studies directional coverage, evidence retention through aggregation, measurement scope, and timely resource access. Conditional examples expose null-space blind spots, information loss under aggregation, and missed deadlines despite adequate average resources. A calibrated Gaussian-sensor benchmark demonstrates both coverage gains and sensitivity losses, including weak detection despite full rank. These results motivate observer design questions rather than a sufficient construction theorem. Adaptive co-drift, commitment benefits, resource inversion, and complete observer construction remain unresolved.