AI in Hypersonic Flight Control (Adaptive RL for Stability & Safety)

About This Course

This 3-week intensive course on AI in hypersonic flight control focuses on applying adaptive reinforcement learning (RL) in a safety-first, non-weaponized context. Instead of controller design, it emphasizes hazard analysis, perception and state awareness, verification & validation (V&V) planning, governance, and human oversight—guiding participants to build practical artifacts like a requirements matrix, V&V plan, and assurance case for safe high-speed testing.

Aim

To provide a safety-first, non-weaponized framework for using AI and adaptive RL in hypersonic flight control—focusing on governance, assurance, and human oversight rather than controller design.

Course Objectives

• Frame AI/adaptive RL for hypersonic flight within a safety-first, non-weaponized context.

• Build and document hazards, oversight roles, abort criteria, and geofencing for high-speed test articles.

• Specify perception and state-awareness requirements without exposing sensitive control algorithms.

• Draft a V&V plan, test envelopes, and safety monitors for AI-enabled high-speed systems.

• Outline a safety case and operator SOPs aligned with governance and certification thinking.

Course Structure

Module 1 – Safety, Ethics & High-Speed Flight Basics (Non-Weaponized)

• Where AI fits in safety-critical aerospace: hazard analysis, “human-on-the-loop,” and oversight

• High-level aerothermodynamics concepts (no design details): why high speeds amplify uncertainty and sensing challenges

• Assurance artifacts: requirements traceability, safety cases, and model/system cards

• Hands-on: Build a policy-aware requirements matrix and hazard log for a benign high-speed test article (e.g., generic aero model), focusing on oversight, abort criteria, and geofencing

Module 2 – Robust Perception & State Awareness (Conceptual)

• Sensor integrity at high dynamic pressure: fault concepts, latency awareness, and graceful degradation (no algorithms)

• Observability at a glance: what it means to estimate states safely without revealing implementation

• Validation & test planning: scenario coverage, limits, and transparency for operators and regulators

• Hands-on: Draft a V&V plan—define test envelopes, safety monitors, and operator intervention thresholds for a conceptual high-speed vehicle

Module 3 – Governance, Certification & Human Oversight

• Standards and certification thinking for AI components (conceptual): documentation, audits, incident reporting

• Envelope thinking without controllers: defining stay-out zones, rate limiters, and conservative defaults

• Human-machine interfaces: alerting, explainability for operators, and abort workflows

• Hands-on: Assemble an assurance case outline (safety case) with roles, evidence to collect, and an operator SOP for safe testing and shutdowns

Who Should Enrol?

• Engineers & researchers in aerospace, mechanical, electrical, controls, or related fields

• AI/ML & RL practitioners interested in safety-critical aerospace applications

• Flight test, safety, governance, and certification professionals

• Senior students, faculty, and industry R&D teams working on hypersonic or high-speed systems