Autonomous Flight Safety Levels for Certifiable AI/ML UAS
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The integration of AI/ML algorithms in unmanned aircraft systems (UAS) is hindered by the requirement for airworthiness certification, which is cost-prohibitive and challenging due to non-deterministic software behavior, making it difficult to certify autonomous flight in the National Airspace System (NAS).
Innovation Solution
The Intelligent Multi-level Safe Autonomous Flight Ecosystem (IMSAFE) architecture introduces a multi-level safety framework with defined safety levels (1-5) that allows collaboration between systems to ensure safe operation, enabling AI/ML usage in UAS by promoting or demoting safety levels based on individual and ecosystem capabilities, and leveraging existing certified software to reduce certification costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If airworthiness certification is required for UAS with AI/ML algorithms, then safety and reliability are improved, but certification cost and complexity increase significantly
Solution Approach 1:
The patent segments the safety certification process into hierarchical levels (Level 1: Basic safety, Level 2: Enhanced safety, Level 3: Advanced safety, Level 4: Highest safety). This allows UAS to be certified at appropriate levels without requiring full Level 4 certification for all applications, reducing overall certification complexity while maintaining necessary safety standards.
Solution Approach 2:
The patent introduces configurable safety parameters including safety_level (1-4), risk_tolerance thresholds, and operational constraints that can be adjusted based on mission requirements. This parameter-based approach allows flexible certification where lower safety levels are acceptable for certain applications, reducing certification burden while maintaining safety where needed.
2Reliability
If redundant equipment is added to UAS to meet certification requirements, then safety and reliability are improved, but vehicle weight increases
Solution Approach 1:
The patent implements multi-functional safety components that serve multiple purposes. For example, the safety management system integrates collision avoidance, terrain avoidance, and operational constraint enforcement into a single unified system rather than requiring separate redundant systems for each function, reducing overall weight while maintaining safety.
Solution Approach 2:
The patent allows dynamic adjustment of safety parameters and redundancy levels based on operational context. Critical safety functions are maintained, but non-critical redundant systems can be deactivated or reduced when not needed for the current mission, reducing weight while preserving essential safety capabilities.
3Reliability
If deterministic software is used to meet certification requirements, then airworthiness certification is achieved, but AI/ML algorithm performance and adaptability are reduced
Solution Approach 1:
The patent segments software into deterministic components (flight control, safety management, certification-compliant functions) and non-deterministic AI/ML components (target recognition, path planning, anomaly detection). The deterministic segment ensures certification eligibility while the AI/ML segment provides adaptability and intelligence, allowing both requirements to coexist.
Solution Approach 2:
The patent introduces a safety management system as an intermediary layer between deterministic certification requirements and non-deterministic AI/ML algorithms. This intermediary monitors, validates, and constrains AI/ML outputs to ensure they meet safety requirements, enabling AI/ML performance without compromising certification eligibility.
4Adaptability or versatility
If small UAS platforms are used for commercial and military applications, then utility and customization are improved, but certification cost becomes prohibitive
Solution Approach 1:
The patent introduces configurable safety parameters and operational constraints that can be adjusted based on mission requirements and certification needs. This allows small UAS to be certified for specific applications at appropriate safety levels, reducing certification costs compared to requiring maximum safety levels for all small UAS regardless of application.
Solution Approach 2:
The patent implements dynamic safety level adjustment where the certification and operational safety requirements can change based on the mission, environment, and risk assessment. This dynamic approach allows small UAS to operate with lower certification requirements in low-risk scenarios, reducing certification costs while maintaining high safety standards when needed.
Data Source
AI summary
A communications ecosystem and related methods of operation include an ecosystem safety level (ESL), a plurality of vehicles or systems, each vehicle or system having an individual safety level (ISL), a control system operable to determine whether the ISL for each vehicle or system meets or exceeds the ESL, and one or more communications links between any first vehicle or system having a first ISL that does not meet or exceed the ESL and a second vehicle or system with a second ISL that does meet or exceed the ESL such that the first vehicle or system operates at the second ISL.