Research

Lineage and scope

Progressive autonomy is not novel. PAA builds a vendor-neutral implementation architecture around mechanisms already visible in several automation domains.

Vehicle automation

SAE J3016

Source: SAE J3016_202104

SAE J3016 establishes the familiar lineage of discrete automation levels defined by the allocation of responsibility between a person and an automated driving system. Its levels are specific to driving and to an operational design domain.

PAA applies the level-based frame to bounded tasks and records changes in oversight.

SAE J3016 mapping mapping table
Source establishes PAA construct
levels of driving automation autonomy positions
responsibility allocation by level human oversight at each position
operational design domain scoped authority

Safety-critical facility control

Osprey

Source: Hellert, Montenegro & Sulc (2026)

Hellert, Montenegro, and Sulc describe plan-first agentic control at the Advanced Light Source. The system produces a human-readable execution plan before touching hardware; human review and policy checks gate execution.

Osprey demonstrates a typed boundary with blocking authorization. It does not define a general progression and regression lifecycle for reducing that oversight.

Osprey mapping mapping table
Source establishes PAA construct
plan-first orchestration pre-execution task boundary
human review before hardware action blocking evaluator
policy rejection and approval invariant check and escalation

Self-driving laboratories

Safe-SDL

Source: Zhang et al. (2026)

Zhang and colleagues describe advancement from human-on-the-loop operation to bounded autonomy after statistically sufficient incident-free operation. An incident or constraint violation causes regression, followed by investigation and corrective action before advancement resumes.

Safe-SDL applies progression and regression to laboratory safety. PAA generalizes that transition pattern beyond the laboratory and represents it with reusable contracts.

Safe-SDL mapping mapping table
Source establishes PAA construct
demonstrated competence promotion evidence
statistical validation window promotion eligibility
regression after an incident demotion and fallback
constrained autonomy domain scoped authority

Network operations

Malik

Source: Malik (2026)

Malik applies progressive autonomy with safety boundaries and closed-loop verification to hyperscale network operations. It does not define the evaluator and evidence contracts.

Scope of the claim

What PAA claims

PAA does not claim to have invented progressive autonomy. It specifies reusable contracts for governed tasks, evidence, authority decisions, and reversible transitions.

Schema