From Experiment Execution to Verifiable Capability: An Assurance Architecture for Self-Driving Laboratories

Abstract

Self-driving laboratories increasingly select experiments, orchestrate resources, execute protocols and update scientific models. Yet existing architectures often under-specify a critical runtime question: what evidence shows that the current physical system can satisfy a particular task under its operating conditions? Device availability, command completion and internal feedback do not establish that delivered volume, end-effector pose or sample temperature meets task-specific limits. Periodic calibration provides essential baseline evidence, but cannot support every operation through a single static pass label. We propose a three-layer architecture comprising task planning and scheduling, task-level capability assurance, and execution and physical operations. The assurance layer translates formal tasks into machine-readable capability requirements, represents measurement evidence through versioned, uncertainty-aware records with explicit applicability and validity, and uses these records to admit, constrain, reject, reassign or trigger verification of tasks. Maintenance, collisions, environmental excursions and version changes dynamically invalidate affected records. Parameter correction is treated as an optional bounded recovery process requiring authorization, readback, independent re-verification and commit or rollback. Examples in liquid handling, robotic positioning and temperature control illustrate the framework. Trustworthy self-driving laboratories must therefore maintain machine-actionable evidence that they are currently qualified to perform each task.