SOW.pdf

Statement of Work Adaptive Envelope Protection (AEP) Technology Maturation, Flight Integration, and Demonstration on NASA RAVEN SWIFT and NASA RAVEN eVTOL Aircraft

Collaborating Organization:

Massachusetts Institute of Technology (MIT) Reliable Autonomous Systems Laboratory (REALM)

Period of Performance:

12 Months

Government's Minimum Need

Summary

The objective of this NASA-funded effort is to support a one-year collaborative research and technology maturation effort between the NASA Transformational Tools and Technologies (TTT) project and the Massachusetts Institute of Technology (MIT) Reliable Autonomous Systems Laboratory (REALM) to further develop, implement, integrate, validate, and demonstrate the Adaptive Envelope Protection (AEP) framework on representative electric Vertical Takeoff and Landing (eVTOL) research aircraft.

This effort builds directly upon the successful completion of the initial Adaptive Envelope Protection research collaboration, during which a computational framework was established for learning runtime safety assurance capabilities using online and offline data-driven learning methods. The initial effort demonstrated the feasibility of integrating neural-network system identification, learning-based nominal control policies, policy neural control barrier function (PNCBF) safety filters, robust safety filtering concepts, and learning-based control allocation into a unified Adaptive Envelope Protection architecture within a high-fidelity eVTOL simulation environment. It also established a complementary offline learning methodology capable of learning the principal AEP components from simulation-generated datasets when direct simulator access is limited.

The purpose of this second effort is to mature these capabilities into a practical runtime safety assurance framework suitable for integration into NASA flight research vehicles and the associated flight-test workflow. Rather than focusing primarily on identifying candidate algorithms, this effort emphasizes technology maturation, software implementation, validation, progressive flight demonstration, and continual improvement through simulation, hardware integration, and flight-test data.

The research will support the NASA Transformational Tools and Technologies (TTT) objectives by developing a reusable runtime safety assurance capability that accelerates flight envelope expansion through independent, formally grounded safety guarantees that are largely independent of the nominal flight control and system identification methods under evaluation.

The effort will be conducted in two sequential phases.

Part I will focus on implementing, integrating, evaluating, and maturing the Adaptive Envelope Protection framework using the NASA RAVEN SWIFT sub-scale eVTOL aircraft (or an equivalent representative sub-scale vehicle if RAVEN SWIFT is unavailable). This phase will emphasize software integration, simulation validation, flight implementation, data collection, algorithm refinement, and evaluation across hover, transition, forward-flight, and landing operations.

Part II will transition the matured AEP framework to the full-scale NASA RAVEN eVTOL aircraft (or an alternative representative full-scale flight research vehicle if RAVEN is unavailable). Lessons learned during Part I will be incorporated into improved AEP algorithms and software workflows. Following implementation and evaluation on the full-scale vehicle, the updated framework will be regression tested on the RAVEN SWIFT platform to verify that improvements remain robust and transferable across multiple vehicle scales.

The long-term objective of this effort is to mature Adaptive Envelope Protection into a reusable runtime safety assurance capability that supports safe, efficient, and increasingly autonomous flight-test envelope expansion for present and future NASA eVTOL research vehicles.

Background The emergence of advanced eVTOL aircraft incorporating distributed electric propulsion, highly redundant control effectors, and vehicle configurations that change during flight presents significant challenges for conventional flight-test methodologies. Unlike traditional aircraft, many emerging eVTOL configurations including tilt-wing, lift-plus-cruise, tiltrotor, and other morphing concepts operate over broad flight envelopes in which aerodynamic characteristics, control effectiveness, and stability properties vary significantly between flight regimes.

These characteristics increase both the complexity and the cost of developing accurate aerodynamic models and high-confidence flight control systems. During early flight testing, uncertainties associated with aerodynamic modeling, actuator effectiveness, environmental disturbances, and vehicle configuration changes can substantially increase the engineering effort required to safely expand the operational flight envelope.

One objective of NASA's TTT project is to reduce these barriers by enabling more rapid aerodynamic model development and validation through increased use of online system identification, learning-based control methods, and data-driven flight-test techniques. Adaptive Envelope Protection directly supports this objective by providing an independent runtime safety assurance capability that operates alongside experimental controllers and system identification algorithms without requiring significant modifications to those systems.

The first phase of the NASA-MIT collaboration established the computational foundations of Adaptive Envelope Protection. Initial research focused on learning runtime safety assurance through direct interaction with a Python/JAX implementation of the NASA lift-plus-cruise simulator. This enabled implementation and evaluation of neural-network-based online system identification, Epigraph Form Proximal Policy Optimization (EFPPO) for learning stabilizing nominal control policies, Policy Neural Control Barrier Functions (PNCBF) for runtime safety filtering, learning-based control allocation, and robust extensions to safety filtering within a unified computational framework.

Subsequently, recognizing that direct executable access to high-fidelity simulations may not always be available, the collaboration developed a complementary offline learning methodology in which the principal AEP components are learned from simulation-generated datasets. This architecture allows training of nominal controllers, safety filters, robust safety filters, and future fault-tolerant policies using simulation or flight-test datasets without requiring continuous simulator access. This capability significantly improves flexibility and enables learning from simulation campaigns conducted by NASA as well as from experimental flight data.

The completion of these foundational activities provides the basis for transitioning Adaptive Envelope Protection from an algorithm research effort toward an integrated runtime capability suitable for implementation on actual flight research vehicles.

The second phase of this collaboration therefore focuses on maturing the AEP framework through software integration, simulation validation, hardware implementation, flight demonstration, continual refinement, and regression testing using both sub-scale and full-scale NASA eVTOL research platforms.

Technical Objectives The overall objective of this effort is to mature the Adaptive Envelope Protection framework into a practical, reusable runtime safety assurance capability suitable for implementation and evaluation on NASA eVTOL flight research vehicles while maintaining flexibility to incorporate future advances in learning-based flight control and safety assurance.

Specific technical objectives include:

1. Mature the existing Adaptive Envelope Protection software architecture into an integrated, modular runtime framework suitable for implementation on representative eVTOL flight research vehicles.

2. Continue development of data-driven methods for learning stabilizing nominal control policies, runtime safety filters, robust safety filters, online system identification, and learning-based control allocation.

3. Mature both simulator-in-the-loop and dataset-driven learning workflows to support varying levels of access to NASA simulation environments and flight-test datasets.

4. Implement and evaluate the AEP framework within representative NASA RAVEN SWIFT and NASA RAVEN simulation environments prior to flight implementation.

5. Integrate the AEP framework with the onboard software architecture, state estimation, control allocation, and flight-test instrumentation of the NASA RAVEN SWIFT and NASA RAVEN aircraft, as available.

6. Demonstrate Adaptive Envelope Protection during representative hover, transition, forward-flight, descent, and landing operations while progressively expanding the validated operational envelope.

7. Evaluate AEP performance under representative model uncertainty, measurement uncertainty, external disturbances, actuator limitations, and selected subsystem failures representative of flight-test conditions.

8. Extend the AEP framework to incorporate additional runtime safety capabilities, including robust safety filtering, fault-tolerant operation, subsystem operational constraints (e.g. power source health), airspace boundary constraints, and non-cooperative multi-vehicle safety concepts.

9. Establish an end-to-end AEP development and validation workflow spanning simulation, dataset generation, offline learning, runtime implementation, flight testing, post-flight analysis, continual algorithm improvement, and regression validation.

10. Produce documented software, collaborative technical reports, datasets, performance evaluations, and recommendations supporting future transition of Adaptive Envelope Protection into NASA flight research and other advanced aerospace applications.

Requirements The Adaptive Envelope Protection (AEP) research conducted during the initial NASA–MIT collaboration established the computational and algorithmic foundation for a data-driven runtime safety assurance framework for advanced eVTOL aircraft. Building upon these accomplishments, the objective of this effort is to mature the AEP framework into an integrated capability suitable for implementation, evaluation, and progressive demonstration on representative NASA flight research vehicles.

The research shall continue to emphasize model-light and data-driven methods that maximize the utilization of simulation and flight-test data while minimizing dependence on highly accurate a priori aerodynamic models. The resulting AEP framework shall operate independently of the nominal flight control system and any experimental system identification or adaptive control algorithms, intervening only when necessary to preserve safe operation or prevent loss of controllability.

The work performed under this effort shall be guided by the following requirements.

2.1 Runtime Safety Assurance

The primary purpose of Adaptive Envelope Protection shall remain the provision of an independent runtime safety assurance capability capable of monitoring vehicle behavior and preventing loss-of-control events during flight testing and progressive flight-envelope expansion.

The AEP framework shall operate in parallel with pilot commands, conventional flight control systems, adaptive controllers, learning controllers, or experimental system identification algorithms while remaining largely independent of their internal implementation.

The framework shall seek to intervene only when necessary to maintain formally defined safety guarantees or prevent departure from a dynamically determined safe operating envelope.

2.2 Data-Driven Adaptive Learning

The framework shall continue to emphasize learning directly from measured or estimated vehicle states, simulation-generated datasets, and experimental flight-test data while minimizing reliance on high-fidelity aerodynamic models.

The framework shall support multiple development paradigms, including:

direct simulator-in-the-loop learning when executable simulator access is available;

offline learning from simulation-generated datasets and flight-test datasets;

continual refinement using newly collected simulation or flight-test data.

The framework shall maintain compatibility with future advances in machine learning, reinforcement learning, system identification, optimal control, and runtime safety assurance methods.

2.3 Integrated Runtime Architecture

The Adaptive Envelope Protection framework shall continue to evolve as an integrated runtime architecture composed of interoperable software modules rather than a collection of independent algorithms.

The architecture should support, as appropriate, online or offline system identification;

local or global dynamic model learning;

nominal policy learning;

runtime safety filtering;

robust safety filtering;

learning-based control allocation;

adaptive control reallocation;

uncertainty estimation;

flight-test monitoring;

runtime envelope estimation;

post-flight learning and refinement.

The modular architecture shall facilitate future incorporation of additional capabilities without requiring substantial redesign of the overall framework.

2.4 Representative Flight-Test Conditions

The framework shall be developed and evaluated under conditions representative of actual flight testing of advanced eVTOL aircraft.

Representative evaluation scenarios should include, as practical, hover;

vertical takeoff;

climb;

forward-flight transition;

cruise;

approach;

descent;

hover landing;

transitions between flight modes;

changing vehicle configurations.

Whenever practical, demonstrations shall emphasize realistic flight-test scenarios rather than idealized control problems.

2.5 Practical Flight-Test Constraints

The Adaptive Envelope Protection framework shall account for practical aspects of real flight operations including, where appropriate, constrained control inputs;

actuator dynamics;

actuator saturation;

measurement uncertainty;

state-estimation uncertainty;

sensor bias;

sensor noise;

computation latency;

model uncertainty;

aerodynamic uncertainty;

mass-property uncertainty;

external disturbances;

rotor interactions;

control-surface uncertainty;

changing control effectiveness;

changing vehicle configuration.

The objective is not necessarily to solve every source of uncertainty but to ensure that the framework can accommodate these effects as the technology matures.

2.6 Robustness and Failure Tolerance

The framework shall continue development of methods capable of maintaining safe operation in the presence of representative failures and degraded operating conditions.

Examples include propulsion failures;

control-surface failures;

degraded actuator effectiveness;

subsystem operational limits;

energy-management constraints;

sensor failures;

control-allocation changes;

unexpected disturbances.

Where complete fault accommodation is not feasible, the framework shall provide graceful degradation of capability while preserving safety whenever practical.

2.7 Extensible Safety Framework

Although the primary emphasis of this effort remains vehicle-level Adaptive Envelope Protection, the software architecture shall be designed to accommodate future safety capabilities including, as appropriate, airspace boundary constraints;

geofencing;

mission constraints;

energy-awareness;

collision avoidance;

non-cooperative multi-vehicle operations;

dynamic operational rules.

These capabilities are intended to build upon the core runtime safety assurance framework established during this effort.

Technical Approach The technical approach emphasizes progressive technology maturation through iterative development, implementation, validation, experimentation, and continual improvement. Rather than pursuing isolated algorithm development activities, the effort shall mature the Adaptive Envelope Protection framework as an integrated runtime capability supporting the complete flight-test lifecycle.

Development shall proceed through repeated cycles of software implementation, simulation evaluation, hardware integration, flight demonstration, post-flight analysis, algorithm refinement, and regression validation.

The research shall remain sufficiently flexible to support multiple development workflows depending upon the level of access available to NASA simulation environments and flight hardware.

Simulator-in-the-Loop Development

When executable simulation environments are available, the AEP framework may continue to utilize direct interaction with the simulation for development of nominal control policies, safety filters, online system identification, and other adaptive learning components.

Simulator-in-the-loop development provides an efficient mechanism for rapidly evaluating candidate algorithms, generating representative disturbances, exploring broad operating envelopes, and conducting repeatable safety evaluations prior to hardware implementation.

Dataset-Driven Development

When direct simulator access is limited, the framework shall support learning from simulation-generated datasets or experimental flight-test datasets.

The dataset-driven workflow developed during the initial collaboration shall continue to mature by enabling learning of multiple AEP components from common datasets, thereby reducing training effort while improving consistency between the various learned modules.

Simulation-generated datasets, hardware-in-the-loop testing, and experimental flight data shall all be considered appropriate sources for continual improvement of the framework.

Progressive Flight Implementation

Implementation shall proceed incrementally through progressively higher levels of technical maturity.

Representative stages include

1. software implementation;

2. software simulation;

3. integrated vehicle simulation;

4. hardware integration;

5. laboratory validation;

6. sub-scale flight demonstration;

7. post-flight refinement;

8. larger-scale vehicle implementation;

9. regression validation.

Progression between stages should be guided by technical readiness, successful completion of milestone demonstrations, and evaluation of performance obtained during preceding stages.

Continual Learning and Regression Validation

A fundamental objective of this effort is to establish a continual improvement process in which every significant simulation campaign, laboratory experiment, and flight test contributes to improving the Adaptive Envelope Protection framework.

Collected datasets shall be used, as appropriate, to improve learned models;

refine nominal control policies;

improve runtime safety filters;

evaluate robustness;

quantify uncertainty;

validate software updates;

perform regression testing against previously demonstrated capabilities.

Regression testing shall be incorporated throughout the effort to ensure that improvements made for one vehicle configuration or operating condition do not degrade previously demonstrated capabilities.

End-to-End Development Workflow

One of the principal outcomes of this effort shall be the establishment of a documented end-to-end Adaptive Envelope Protection development workflow suitable for continued application to future NASA flight research programs.

The workflow should encompass simulation planning;

dataset generation;

data management;

system identification;

controller learning;

runtime safety-filter learning;

software integration;

simulation evaluation;

hardware implementation;

flight-test execution;

post-flight analysis;

algorithm refinement;

regression validation;

software release.

The resulting workflow should provide a repeatable methodology for maturing future Adaptive Envelope Protection capabilities from initial concept through operational flight demonstration.

Part I – RAVEN SWIFT Technology Maturation

Overview

The objective of Part I is to mature the Adaptive Envelope Protection (AEP) framework from a software research capability into an integrated runtime safety assurance system suitable for implementation and evaluation on a representative sub-scale eVTOL flight research vehicle. The primary flight research platform for this effort shall be the NASA RAVEN SWIFT aircraft.

Should the RAVEN SWIFT vehicle be unavailable, another sub-scale eVTOL aircraft that reasonably represents the RAVEN SWIFT configuration and operational objectives may be used upon mutual agreement between NASA and MIT.

Part I will emphasize implementation, integration, validation, experimentation, and continual improvement of the AEP framework using representative simulation environments, laboratory testing, and sub-scale flight demonstrations. The effort will also establish the software architecture, development workflow, and experimental procedures that will support transition of the framework to the full-scale NASA RAVEN vehicle during Part II.

A principal objective of Part I is to establish a repeatable development cycle in which simulation, hardware testing, and flight experiments continually improve the AEP framework through iterative learning, software refinement, and regression validation.

Task 1 – Project Planning and Technology Roadmap MIT shall prepare a detailed technical roadmap describing the planned development, implementation, validation, and demonstration activities for Part I.

The roadmap shall identify technical objectives;

software development activities;

expected simulation activities;

anticipated flight demonstrations;

required datasets;

anticipated software releases;

expected technical risks;

proposed mitigation strategies;

expected milestones.

The roadmap shall be reviewed jointly with NASA and updated throughout the project as significant technical findings are obtained.

Task 2 – AEP Software Framework Maturation MIT shall continue maturing the Adaptive Envelope Protection software architecture into a modular runtime framework suitable for application to representative eVTOL research vehicles.

Activities may include further development of online and offline system identification;

local and global model learning;

nominal policy learning;

runtime safety filtering;

robust safety filtering;

learning-based control allocation;

adaptive control reallocation;

uncertainty estimation;

runtime envelope estimation;

flight-test monitoring;

software modularization;

software documentation.

Software shall be developed to facilitate integration with representative NASA simulation environments and future onboard flight software.

Task 3 – Dataset Generation and Learning Workflow MIT shall further mature the dataset-driven AEP learning workflow developed during the first phase of the collaboration.

This task shall include continued development of methods for generating, organizing, processing, and utilizing simulation-generated and flight-generated datasets for learning nominal control policies;

runtime safety filters;

robust safety filters;

system identification models;

control allocation methods;

future fault-tolerant control policies.

Whenever executable simulation access is available, simulator-in-the-loop learning methods may continue to be utilized alongside offline learning methods.

The objective of this task is to establish a flexible learning workflow capable of accommodating both executable simulations and fixed datasets depending upon data availability.

Task 4 – Simulation Integration MIT shall integrate the AEP framework into representative NASA RAVEN SWIFT simulation environments as available.

Simulation activities should include representative flight-test scenarios covering hover;

vertical takeoff;

climb;

hover-to-forward transition;

forward flight;

approach;

descent;

hover landing;

repeated transitions.

Simulation campaigns should also evaluate representative aerodynamic uncertainty;

model uncertainty;

measurement uncertainty;

sensor noise;

actuator limits;

external disturbances;

changing control allocation;

degraded operating conditions.

Simulation results shall be documented and utilized to improve subsequent software releases.

Task 5 – Flight Software Integration

MIT shall support integration of the Adaptive Envelope Protection framework with the NASA RAVEN SWIFT flight software architecture, as practical.

Integration activities may include interfaces to state estimation;

flight control systems;

telemetry;

onboard computing;

control allocation;

flight-test instrumentation;

data logging;

safety monitoring.

The software architecture should minimize coupling between AEP and the nominal flight control system so that AEP remains an independent runtime safety assurance capability.

Task 6 – Sub-scale Flight Demonstration MIT shall support implementation and evaluation of Adaptive Envelope Protection during representative sub-scale flight experiments.

The demonstrations should progressively evaluate hover operations;

takeoff;

climb;

transition;

forward flight;

approach;

landing;

multiple transition cycles.

Where practical, testing should also evaluate representative modeling uncertainty;

environmental disturbances;

sensor uncertainty;

changing control effectiveness;

actuator saturation;

degraded operating conditions.

The objective of these demonstrations is not to explore the entire flight envelope but to progressively increase confidence in the runtime safety assurance capability while maintaining safe flight-test practices.

Task 7 – Flight-Test Data Analysis Following each simulation campaign or flight-test activity, MIT shall evaluate the collected data to assess controller performance;

safety-filter performance;

intervention frequency;

runtime computational performance;

model prediction accuracy;

uncertainty estimates;

flight envelope utilization;

robustness.

The resulting observations shall be incorporated into subsequent software improvements.

Task 8 – Continual Improvement and Regression Validation A fundamental objective of this effort is continual improvement of the Adaptive Envelope Protection framework.

Following significant software updates, MIT shall, where practical, retrain learned models;

regenerate representative datasets;

repeat selected simulation cases;

replay representative flight datasets;

evaluate software changes against previous results;

document improvements and regressions.

Regression validation shall seek to ensure that improvements introduced for one operating condition do not degrade previously demonstrated capabilities.

Task 9 – Part I Technical Assessment Near completion of Part I, MIT shall conduct an overall technical assessment summarizing accomplishments;

software maturity;

remaining technical challenges;

lessons learned;

recommendations for Part II.

The assessment shall identify improvements recommended before transition to the full-scale vehicle;

software modules requiring additional maturation;

proposed changes to the development workflow;

recommended future research activities.

The assessment shall also provide recommendations regarding readiness for implementation on the NASA RAVEN full-scale aircraft.

Part I Milestones The following milestones are intended to measure progress toward the objectives of Part I.

Milestone 1 — Project Initiation (Month 1)

Initial technical roadmap completed.

Initial software development plan completed.

Initial coordination meeting conducted.

Milestone 2 — Software Maturation (Month 2)

Updated modular AEP software architecture completed.

Initial software integration completed.

Initial learning workflow documented.

Milestone 3 — Simulation Demonstration (Month 3)

Representative simulation scenarios executed.

Initial performance evaluation completed.

Initial software refinements documented.

Milestone 4 — Flight Software Integration (Month 4)

Integration interfaces completed.

Initial implementation on the RAVEN SWIFT software architecture demonstrated, as practical.

Milestone 5 — Flight Demonstration (Month 5)

Representative sub-scale flight demonstrations completed.

Flight data collected.

Initial post-flight assessment completed.

Milestone 6 — Part I Completion (Month 6)

Updated AEP software released.

Regression validation completed.

Technical assessment completed.

Recommendations for Part II documented.

Part I Deliverables MIT shall provide the following deliverables during Part I.

Monthly

Monthly technical progress report.

Updated project roadmap, as appropriate.

Summary of software development activities.

Technical Deliverables

Updated modular AEP software.

Documented source code.

Updated software architecture documentation.

Updated end-to-end AEP workflow documentation.

Simulation scenarios and evaluation procedures.

Simulation datasets, as practical.

Flight-test data analysis reports, as practical.

Regression validation reports.

Software release notes.

Final Part I Deliverables

Part I technical assessment report.

Summary of lessons learned.

Recommendations for Part II.

NASA–MIT collaborative technical report suitable for publication following NASA approval.

Presentation summarizing Part I accomplishments and recommendations.

Part II – Full-Scale RAVEN Technology Demonstration and Capability Maturation

Overview

The objective of Part II is to transition the Adaptive Envelope Protection (AEP) framework matured during Part I from the NASA RAVEN SWIFT sub-scale flight research platform to the NASA RAVEN full-scale eVTOL aircraft. This phase emphasizes integration of lessons learned during Part I into an improved runtime safety assurance capability suitable for progressively more representative flight research operations.

The NASA RAVEN aircraft shall serve as the primary full-scale implementation platform. If the RAVEN aircraft is unavailable during the period of performance, NASA and MIT shall jointly identify an alternative flight research platform that reasonably supports the technical objectives of this effort.

Part II shall focus on refining the AEP framework using knowledge gained during Part I, extending the framework with additional runtime capabilities, evaluating performance within representative full-scale simulation and flight-test environments, and validating that improvements remain transferable across multiple vehicle scales through continued regression testing on the RAVEN SWIFT platform.

The long-term objective of Part II is to demonstrate a mature, reusable runtime safety assurance capability suitable for supporting future NASA flight research involving advanced eVTOL aircraft and increasingly autonomous flight-test operations.

Task 10 – Technology Transition Planning MIT shall develop a technology transition plan describing the migration of the Adaptive Envelope Protection framework from the sub-scale RAVEN SWIFT implementation to the full-scale NASA RAVEN aircraft.

The transition plan shall include lessons learned during Part I;

recommended software improvements;

anticipated integration activities;

required simulation updates;

anticipated flight-test support;

technical risks associated with scaling;

proposed mitigation strategies.

The plan shall also identify software modules requiring modification to accommodate differences between the sub-scale and full-scale aircraft.

Task 11 – Algorithm and Software Maturation Based upon results obtained during Part I, MIT shall continue maturing the Adaptive Envelope Protection framework.

Activities may include continued development of nominal policy learning;

runtime safety filtering;

robust safety filtering;

learning-based control allocation;

adaptive control reallocation;

online and offline system identification;

uncertainty estimation;

runtime envelope estimation;

software modularization;

computational performance improvements.

Additional capabilities may also be investigated, as appropriate, including subsystem operational constraints;

propulsion-energy constraints;

airspace boundary constraints;

geofencing;

mission-level operational constraints;

cooperative multi-vehicle safety concepts;

fault-tolerant policy learning.

The emphasis shall remain on preserving the modular architecture established during Part I while extending the operational capability of the overall framework.

Task 12 – Full-Scale Simulation Integration MIT shall integrate the updated Adaptive Envelope Protection framework into representative NASA RAVEN simulation environments as available.

Simulation campaigns should evaluate representative mission scenarios including vertical takeoff;

hover;

climb;

hover-to-forward transition;

forward flight;

maneuvering flight;

descent;

landing;

repeated flight-mode transitions.

Simulation studies should also investigate representative operational uncertainties including aerodynamic uncertainty;

mass-property uncertainty;

environmental disturbances;

degraded actuator effectiveness;

subsystem operational limits;

measurement uncertainty;

estimator uncertainty;

changing vehicle configurations.

Simulation results shall be utilized to refine subsequent software releases and guide implementation on the full-scale vehicle.

Task 13 – Full-Scale Flight Software Integration MIT shall support implementation of the Adaptive Envelope Protection framework within the NASA RAVEN software architecture, as practical.

Integration activities may include interfaces with state estimation;

flight control systems;

onboard computing;

telemetry;

data acquisition;

flight-test instrumentation;

control allocation;

safety monitoring.

The software architecture shall continue to preserve the independence of Adaptive Envelope Protection from the nominal flight-control system while enabling efficient runtime intervention whenever required to preserve safe operation.

Task 14 – Full-Scale Flight Demonstration MIT shall support implementation and evaluation of Adaptive Envelope Protection during representative full-scale flight-test activities.

Representative demonstrations should progressively evaluate hover operations;

takeoff;

climb;

transition;

forward flight;

descent;

landing;

representative flight-test maneuvers.

Where practical, demonstrations should evaluate representative model uncertainty;

environmental disturbances;

changing control allocation;

degraded actuator effectiveness;

subsystem operational constraints;

representative failure scenarios;

runtime computational performance.

The objective shall be progressive maturation of the runtime safety assurance capability rather than demonstration of any specific flight envelope.

Task 15 – Cross-Scale Regression Validation A distinguishing objective of Part II shall be validation that improvements developed for the full-scale aircraft remain applicable across multiple vehicle scales.

Following significant software updates implemented during Part II, MIT shall, where practical, repeat representative RAVEN SWIFT simulation scenarios;

evaluate updated algorithms using previously collected RAVEN SWIFT datasets;

repeat selected sub-scale flight demonstrations;

compare performance metrics obtained during Parts I and II;

document improvements and identify regressions.

This cross-scale validation activity shall provide confidence that the Adaptive Envelope Protection framework represents a reusable capability rather than a vehicle-specific implementation.

Task 16 – Flight-Test Data Analysis and Framework Refinement Simulation campaigns and flight demonstrations conducted during Part II shall continue to support iterative refinement of the Adaptive Envelope Protection framework.

Collected datasets shall be evaluated to improve learned dynamic models;

nominal policies;

runtime safety filters;

robust safety filters;

uncertainty estimation;

fault tolerance;

software robustness;

computational efficiency.

The objective is to establish continual improvement of the framework throughout the period of performance.

Task 17 – Final Technical Assessment Near completion of the effort, MIT shall conduct a comprehensive technical assessment summarizing the accomplishments of both Parts I and II.

The assessment shall include maturity of the Adaptive Envelope Protection framework;

demonstrated runtime capabilities;

lessons learned during simulation and flight implementation;

remaining technical limitations;

recommendations for future NASA research;

recommendations for future flight-test applications;

opportunities for transition into operational research programs.

The assessment shall also identify future research directions that could further improve runtime safety assurance for advanced aerospace vehicles.

Part II Milestones The following milestones are intended to measure successful completion of Part II.

Milestone 7 — Technology Transition (Month 7)

Technology transition plan completed.

Updated development roadmap approved.

Part I lessons incorporated into the software architecture.

Milestone 8 — Algorithm Maturation (Month 8)

Updated AEP software released.

Additional runtime capabilities implemented, as appropriate.

Updated learning workflow completed.

Milestone 9 — Full-Scale Simulation (Month 9)

Representative RAVEN simulation campaigns completed.

Performance assessment documented.

Updated software release generated.

Milestone 10 — Full-Scale Integration (Month 10)

Flight software integration completed, as practical.

Representative laboratory or integration testing completed.

Pre-flight technical assessment completed.

Milestone 11 — Full-Scale Demonstration (Month 11)

Representative flight demonstrations completed, as practical.

Flight data collected and evaluated.

Updated AEP framework generated.

Milestone 12 — Project Completion (Month 12)

Cross-scale regression validation completed.

Final software release completed.

Final technical assessment completed.

Final NASA–MIT collaborative report completed.

Recommendations for future research documented.

Part II Deliverables

MIT shall provide the following deliverables during Part II.

Monthly

Monthly technical progress report.

Updated project roadmap, as appropriate.

Summary of software development activities.

Summary of simulation and flight-test activities.

Technical Deliverables

Updated modular AEP software.

Updated documented source code.

Updated software architecture documentation.

Updated end-to-end development workflow.

Full-scale simulation evaluation reports.

Flight-test analysis reports, as practical.

Regression validation reports.

Performance assessment reports.

Software release documentation.

Final Project Deliverables

Final integrated Adaptive Envelope Protection software framework.

Final software architecture documentation.

Final end-to-end AEP development and validation workflow.

Cross-scale validation report.

NASA–MIT collaborative final technical report suitable for publication following NASA approval.

Final technical presentation.

Recommendations for future NASA Adaptive Envelope Protection research and technology maturation.

5. Program Milestones The following milestones establish measurable objectives for monitoring technical progress throughout the period of performance. Successful completion of each milestone shall be jointly evaluated by NASA and MIT based on the completion of the associated technical activities, software development, documentation, and demonstrations appropriate to the stage of the effort.

Milestone Target Time

Success Criteria

M1 – Project Initiation Month 1 Initial technical roadmap, software development plan, and project schedule completed. Initial coordination meeting conducted.

M2 – AEP Software Framework Maturation

Month 2 Updated modular AEP architecture completed. Initial software integration and learning workflow documented.

M3 – Simulation Validation

Month 3 Representative simulation campaigns completed. Initial software validation and performance evaluation documented.

M4 – RAVEN SWIFT

Integration

Month 4 AEP integrated with representative RAVEN SWIFT software architecture and simulation environment, as practical.

M5 – RAVEN SWIFT

Demonstration

Month 5 Representative sub-scale implementation and evaluation completed. Flight or hardware test data evaluated, as practical.

M6 – Part I Review Month 6 Part I technical assessment completed. Software updated.

Regression validation performed. Recommendations for Part II documented.

M7 – Technology Transition

Month 7 Transition plan completed. Part I improvements incorporated into the Part II development baseline.

M8 – Full-Scale Software Maturation

Month 8 Updated AEP software framework completed. Advanced runtime capabilities implemented, as appropriate.

M9 – Full-Scale Simulation

Month 9 Representative RAVEN simulation campaigns completed.

Software improvements validated.

M10 – Full-Scale Integration

Month

Full-scale integration activities completed, as practical.

Updated software release documented.

M11 – Full-Scale Demonstration

Month

Representative implementation and evaluation on the NASA RAVEN platform, or equivalent platform if necessary, completed as practical.

M12 – Final Program Review

Month

Final AEP framework, documentation, regression validation, technical report, and recommendations completed.

6. Deliverables The principal objective of this effort is to deliver a mature Adaptive Envelope Protection framework together with the supporting software, documentation, technical evaluations, and development methodology necessary to support continued maturation and application within future NASA flight research programs.

Unless otherwise agreed upon by NASA and MIT, the following deliverables shall be provided.

6.1 Monthly Deliverables

MIT shall provide:

Monthly technical progress reports summarizing accomplishments, technical challenges, planned activities, and updates to the project roadmap.

Summary of software development activities completed during the reporting period.

Summary of simulation activities, flight activities, and data analysis performed during the reporting period.

Updated risk assessment identifying significant technical risks and proposed mitigation strategies, as appropriate.

6.2 Technical Deliverables

Throughout the period of performance MIT shall provide, as appropriate:

Updated Adaptive Envelope Protection source code.

Software architecture documentation.

Documentation describing software installation, configuration, and execution.

End-to-end AEP software workflow documentation.

Simulation scenarios and representative evaluation procedures.

Representative simulation datasets, when distributable.

Flight-test data evaluation reports, when applicable.

Regression validation reports.

Software release documentation.

Technical memoranda describing significant algorithm improvements.

Recommendations for future software development.

6.3 Part I Deliverables

At completion of Part I MIT shall provide:

Part I Technical Assessment Report.

Updated software release.

Lessons learned summary.

Recommendations for transition to Part II.

Updated project roadmap.

6.4 Final Deliverables

At completion of the effort MIT shall provide:

Final Adaptive Envelope Protection software framework.

Final software architecture documentation.

Final end-to-end AEP development workflow.

Final regression validation report.

Final technology maturation assessment.

Final NASA–MIT collaborative technical report suitable for publication.

Final presentation summarizing technical accomplishments.

Recommendations for future NASA Adaptive Envelope Protection research, technology maturation, and flight implementation.

7. Program Management The success of this effort depends upon close technical collaboration between NASA and MIT throughout the period of performance.

The collaboration shall emphasize continual technical interaction, rapid identification of technical challenges, iterative refinement of research objectives, and frequent evaluation of progress toward the overall project objectives.

Technical Coordination

Regular technical meetings shall be conducted between the NASA Technical Lead and the MIT research team to review technical progress;

software development;

simulation activities;

flight-test support activities;

technical challenges;

project risks;

future work.

Twice weekly meetings are anticipated throughout the period of performance unless otherwise mutually agreed.

Technical Reviews

Formal technical reviews shall be conducted at the conclusion of

Part I (approximately Month 6) Part II (approximately Month 12)

These reviews shall summarize accomplishments;

remaining challenges;

software maturity;

recommendations for future work.

Additional technical reviews may be scheduled as mutually agreed.

Collaboration

NASA shall provide technical guidance regarding research priorities;

methods and implementation;

flight-test objectives;

representative simulation environments;

available datasets;

flight-test opportunities;

software integration priorities.

MIT shall perform the research, software development, technical analyses, documentation, and technical reporting necessary to accomplish the objectives of this Statement of Work.

Both organizations shall work collaboratively to maximize the technical value of the effort while maintaining sufficient flexibility to accommodate evolving research results, flight-test opportunities, and technology readiness.

8. Success Criteria Successful completion of this effort shall be demonstrated through the delivery of a useable prototype Adaptive Envelope Protection capability that includes:

A modular and documented runtime AEP software framework.

Demonstrated implementation within representative simulation environments.

Progressive implementation and evaluation on representative sub-scale and full-scale eVTOL flight research platforms, as practical.

A documented end-to-end workflow supporting simulation, learning, implementation, flight testing, post-flight analysis, and continual improvement.

Demonstrated regression validation across multiple vehicle scales.

Technical reports documenting the maturity, performance, limitations, and future research directions of the Adaptive Envelope Protection framework.

The intent of this effort is to demonstrate meaningful technology maturation rather than to achieve predetermined flight-test performance metrics or complete operational readiness.

9. Period of Performance The total period of performance shall be twelve (12) months, divided into two approximately six-month phases.

Part I – Technology Maturation on NASA RAVEN SWIFT

Duration:

Approximately six (6) months

Activities shall focus on software maturation, simulation integration, sub-scale implementation, representative demonstrations, data evaluation, and preparation for transition to the NASA RAVEN platform.

Part II – Technology Transition to NASA RAVEN

Duration:

Approximately six (6) months

Activities shall focus on full-scale implementation, technology maturation, advanced capability development, cross-scale regression validation, and completion of the Adaptive Envelope Protection framework.

The start of Part II shall be contingent upon successful completion of Part I and mutual agreement between NASA and MIT.