Skip to content

Repository files navigation

HARP — Human-Autonomy Readiness Protocol

Release Status

v0.1 Portfolio Release — feature scope frozen.

Autonomy doesn't eliminate humans. It changes where humans enter the system. HARP measures whether they're ready when that moment arrives.

HARP is an independently developed operational architecture prototype exploring a human-readiness layer for safety-critical autonomous systems and Physical AI.


The 60-Second Version

Autonomous systems can reduce how often humans are needed without eliminating the moments when human judgment matters.

Traditional training systems are good at answering questions like:

Did this person complete training? Are they qualified?

HARP explores the next question:

What current evidence do we have that this person can still demonstrate the competencies required when human support is needed?

The prototype demonstrates a possible operational chain:

Scenario performance → competency evidence → individual readiness → team coverage → operational readiness signal


Live Prototype

Experience the HARP workflow from operator scenario performance through organizational readiness.

  1. Enter the Operations Center and work a Remote Assistance scenario.
  2. Receive deterministic after-action feedback and generate competency evidence.
  3. See how that evidence affects individual readiness.
  4. Open Organizational Readiness to see how individual competency evidence translates into team-level coverage.

HARP v0.1 uses entirely fictional and synthetic operational data.


The Core Idea

Qualification ≠ Readiness

An operator can remain formally qualified while the evidence supporting a rare or safety-critical competency becomes stale.

HARP keeps those concepts separate.

Qualification answers whether established requirements were completed.

Readiness asks how strong and recent the available evidence is that required competencies can currently be demonstrated.

HARP does not claim that a person's actual skill can be reduced to a precise percentage.

The readiness score in this prototype represents the strength and recency of available evidence of demonstrated proficiency.


How HARP Works

HARP follows a continuous operational readiness loop:

Train → Demonstrate → Qualify → Operate → Measure → Reinforce → Requalify

At the prototype level:

Scenario
   ↓
Operator Decision
   ↓
Deterministic Evaluation
   ↓
Competency Evidence
   ↓
Readiness Engine
   ↓
Individual Readiness
   ↓
Organizational Coverage
   ↓
Training / Reinforcement / Requalification

The simulator is therefore not the product by itself.

It is one mechanism for creating structured evidence that can feed a broader readiness-management system.


v0.1 Capabilities

  • 10 synthetic Remote Assistance scenarios
  • 8 competency areas
  • deterministic decision trees and critical-error paths
  • scenario-based after-action feedback
  • competency evidence ledger
  • performance, recency, and coverage-based readiness
  • explainable readiness calculations
  • qualification/readiness separation
  • synthetic 12-person operator roster
  • team competency heatmap
  • shift-level competency coverage
  • requalification and action signals
  • targeted training recommendations
  • organizational Mission Readiness concept

Competency Model

HARP v0.1 evaluates eight competency areas:

ID Competency
HARP-01 System Fundamentals
HARP-02 Situational Awareness
HARP-03 Remote Assistance
HARP-04 Escalation
HARP-05 Emergency Response
HARP-06 Communication
HARP-07 Incident Documentation
HARP-08 Human Factors

The model is designed to demonstrate how individual scenario performance could create evidence across multiple operational competencies.


Remote Assistance Scenario Set

HARP v0.1 uses a fictional Autonomous Operations Center and ten synthetic Remote Assistance scenarios:

Scenario Event
RA-001 Ambiguous Construction Flagger
RA-002 Emergency Vehicle Approach
RA-003 Sensor Obstruction
RA-004 Passenger Medical Event
RA-005 Network Degradation
RA-006 Conflicting Police Direction
RA-007 Blocked Roadway
RA-008 Post-Collision Response
RA-009 Vulnerable Road User
RA-010 Incomplete Information

RA-010 intentionally tests whether an operator recognizes that insufficient information can itself be an operational condition requiring restraint or escalation.


Evidence-Based Readiness

HARP treats demonstrated competency events as evidence.

The prototype Readiness Engine considers:

  • demonstrated performance
  • recency of evidence
  • competency coverage
  • scenario difficulty
  • critical errors
  • remediation
  • repeated demonstrations

Conceptually:

Readiness Evidence = Performance × Recency × Coverage × Criticality Adjustment

The exact formulas, thresholds, weights, and proficiency intervals used in v0.1 are illustrative.

They are not validated human-performance or safety metrics.


Individual Readiness

HARP maintains the distinction between:

Qualified

and

Ready based on current evidence

An operator might still hold a formal qualification while HARP identifies that evidence supporting a particular competency is becoming old or insufficient.

For example:

Operator: Qualified

Situational Awareness      READY
Remote Assistance          READY
Communication              READY
Emergency Response         WATCH

The WATCH condition does not mean the operator has forgotten the competency.

It means the organization has weaker or older evidence supporting current demonstrated proficiency in that area.


Organizational Readiness

Individual qualification records do not necessarily answer whether a team has the right competency coverage for an operation.

HARP therefore rolls individual evidence into an organizational view.

The prototype explores questions such as:

  • Do we have enough qualified operators?
  • Do we have sufficient Emergency Response readiness?
  • Do we have sufficient Escalation competency coverage?
  • Which competencies are becoming stale across the team?
  • Who should receive reinforcement or requalification?
  • What should the next training block prioritize?

This creates an important distinction:

Headcount ≠ Capability Coverage

A shift can contain enough formally qualified people while still having weak current evidence in a critical competency.


Human-Autonomy Exchange (HAX)

HARP addresses whether current evidence supports that a human is ready to perform a required competency.

That creates a second operational question:

When an autonomous system requires human support, how should the system determine which human to engage, what authority that person should receive, and when the system should return to autonomous operation?

HARP refers to this proposed interaction layer as the Human-Autonomy Exchange (HAX).

HAX extends the readiness concept by connecting an autonomous-system exception to the competency, readiness, availability, workload, and authority of the humans capable of responding.

Conceptually:

flowchart TD
    A["Autonomous System"] --> B{"Human Support Required?"}

    B -->|No| A
    B -->|Yes| C["Classify Operational Exception"]

    C --> D["Determine Required Competency"]
    D --> E["HARP Readiness Check"]

    E --> F{"Ready Operator Available?"}

    F -->|No| G["Fallback / Safe-State Logic"]
    F -->|Yes| H["Select Appropriate Operator"]

    H --> I["Check Workload and Intervention Capacity"]

    I --> J{"Capacity Available?"}

    J -->|No| G
    J -->|Yes| K["Define Human Authority Envelope"]

    K --> L["Human Support / Intervention"]
    L --> M["Validate Outcome"]
    M --> N["Return to Autonomous Operation"]
    N --> O["Capture Performance Evidence"]
    O --> E
Loading

HAX Decision Model

The proposed HAX layer explores six related decisions:

Function Operational Question
Exception Classification What happened, and does it require human support?
Competency Matching What competency is required to address the event?
Readiness Verification Is there current evidence that an available operator can perform that competency?
Operator Selection Which ready operator is best positioned to receive the event?
Authority Management What actions should that operator be permitted to take?
Return-to-Autonomy When and under what conditions should autonomous operation resume?

Human Intervention Capacity

HAX also introduces the concept of Human Intervention Capacity.

An autonomous operation may have enough formally qualified personnel while still lacking sufficient human capacity to safely absorb multiple simultaneous exceptions.

This creates another distinction:

Qualified staffing ≠ Available intervention capacity

For example, an operator may be qualified and supported by strong readiness evidence while already managing another high-demand event. HAX would therefore treat readiness as one input into intervention assignment rather than assuming that every ready operator is immediately available.

Conceptually:

Autonomous Exception → Required Competency → Ready Human → Available Capacity → Appropriate Authority → Intervention → Evidence

HAX is currently an architectural concept within HARP, not an implemented capability of the v0.1 prototype. Future development could explore event routing, operator workload, intervention capacity, authority boundaries, and multi-system supervision using entirely synthetic scenarios and data.


Architecture

                 HUMAN-AUTONOMY READINESS PROTOCOL
                               |
        +-----------------------+-----------------------+
        |                       |                       |
  SCENARIO ENGINE        READINESS ENGINE        QUALIFICATION
        |                       |                       |
  Decision Trees          Evidence Ledger          Lifecycle
  Critical Errors         Performance              Status
  AAR Feedback            Recency                  Requirements
        |                 Coverage
        +-----------+-----------+-----------------------+
                    |
               OPERATOR PROFILE
                    |
           +---------+---------+
           |                   |
    INDIVIDUAL READINESS   ORGANIZATIONAL READINESS
                              |
                         Competency Heatmap
                         Coverage Requirements
                         Requalification Queue
                         Training Priorities
                         Mission Readiness

More detail is available in docs/architecture.md.


Repository Map

HARP/
├── .github/
│   └── workflows/       GitHub Pages deployment
├── data/                Synthetic readiness and operator data
├── docs/                Architecture, concept paper, and documentation
├── organizational/      Organizational readiness prototype source
├── prototype/           Operator simulator source
├── scenarios/           Structured Remote Assistance scenarios
├── site/                Published GitHub Pages experience
│
├── DISCLAIMER.md
├── INDEPENDENCE.md
├── LICENSE
├── README.md
└── RELEASE_NOTES_v0.1.md

Development history, validation records, and release-preparation materials are retained under docs/development/.


Why This Project Exists

HARP explores how principles from safety-critical qualification, proficiency management, operational readiness, structured training, abnormal-event response, and human oversight might apply as autonomous systems become increasingly capable.

The central premise is that increasing autonomy does not necessarily remove humans from the safety architecture.

It can change where, when, and why human judgment is required.

That creates a readiness problem worth exploring:

The less frequently humans are required to intervene, the more important it becomes to know they are ready when intervention is necessary.


Potential Applications

HARP v0.1 focuses on autonomous-vehicle Remote Assistance, but the architecture is intentionally broader.

Potential future domains could include:

  • autonomous trucking
  • warehouse robotics
  • delivery robotics
  • industrial robotics
  • humanoid systems
  • unmanned aircraft
  • autonomous maritime systems
  • advanced manufacturing
  • other Physical AI environments

The underlying question remains the same:

How does an organization maintain evidence that the humans supporting increasingly autonomous systems are ready for the moments when human judgment still matters?


Design Principles

HARP is built around several principles:

Human-centered
Technology should support operational decision-making rather than obscure it.

Evidence-based
Readiness should be connected to demonstrated performance rather than training completion alone.

Explainable
Operators and leaders should be able to understand why a readiness signal changed.

Deterministic where safety matters
The v0.1 scenario engine uses predetermined decision logic rather than allowing generative AI to independently determine whether a safety-critical response is correct.

Operationally focused
The system is designed around decisions, competencies, readiness, and organizational capability rather than software features for their own sake.


What HARP Is Not

HARP is not:

  • autonomous-driving software
  • a remote-driving platform
  • a vehicle-control system
  • a production learning-management system
  • a regulatory framework
  • a certification standard
  • a validated human-performance model
  • a validated safety model
  • a replacement for an ADS safety case
  • a representation of any company's internal procedures

HARP is an operational architecture and portfolio prototype.


Public-Release Boundary

HARP is independently developed from general principles of safety-critical operations, training, qualification, proficiency, readiness management, and human oversight.

It does not contain or represent proprietary procedures, operational data, software, systems, controlled information, confidential information, or intellectual property belonging to any current or former employer, government organization, customer, or technology company.

All people, fleets, scores, thresholds, staffing requirements, scenarios, procedures, and operational data shown in the prototype are fictional or synthetic.

HARP is not sponsored by, endorsed by, affiliated with, or developed on behalf of any employer, government organization, autonomous-vehicle company, robotics company, or regulatory body.

See:


Methodology Boundary

The readiness equation, proficiency intervals, competency weights, thresholds, staffing requirements, scenario logic, and synthetic data used by HARP v0.1 are illustrative.

HARP does not claim that a person's true skill or operational capability can be represented by an exact percentage.

The prototype's readiness score represents the strength and recency of available evidence of demonstrated proficiency.

No HARP output should be used for real-world safety, qualification, staffing, certification, or operational decision-making.


Documentation

Additional project documentation:


Project Origin

HARP grew from a simple operational question:

How do we know someone is ready, rather than simply qualified?

The project applies transferable concepts from safety-critical operations, qualification systems, proficiency management, training, readiness reporting, and structured decision-making to the emerging human layer surrounding autonomous systems.

The goal is not to predict exactly how any particular autonomous-vehicle or robotics company should operate.

The goal is to demonstrate a framework for thinking about the problem.


North Star

The less frequently humans are required to intervene, the more important it becomes to know they are ready when intervention is necessary.


HARP v0.1 — Human-Autonomy Readiness Protocol

Launch the Live Prototype →

About

Human-Autonomy Readiness Protocol : an operational architecture prototype for measuring human readiness in safety-critical autonomous systems.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages