AUV Sonar Classification With Mission-Adaptive Program Selection

The patent implements a lightweight suitability calculation mechanism that can be executed with limited computational resources on the AUV. Instead of requiring complex offline analysis or extensive training data, the system uses simple polynomial chaos expansion and item response theory calculations that can be performed rapidly with minimal computational overhead, enabling the AUV to evaluate and switch between classification programs dynamically during missions.

The AUV autonomously performs suitability calculations for different classification programs based on current mission parameters and environmental conditions. The system self-evaluates which classification program is most appropriate without requiring external intervention or complex decision-making infrastructure, thereby reducing operational complexity while maintaining high adaptability.

The patent pre-calculates suitability metrics using polynomial chaos expansion and item response theory during system setup or previous missions. These pre-computed suitability values and their relationships to mission parameters are stored and can be rapidly retrieved and applied during actual operations, eliminating the need for time-consuming real-time training or extensive data processing during critical missions.

The system transforms the classification problem from requiring extensive training data to using parametric models (polynomial chaos expansion and item response theory) that describe the relationship between mission parameters and classification program performance. This parametric approach allows accurate predictions with minimal data while significantly reducing processing time.

The patent implements a dynamic program selection mechanism where the AUV evaluates the suitability of different classification programs based on current mission parameters and environmental conditions. Instead of statically carrying all possible programs with equal weight, the system dynamically determines which programs are most appropriate for the current situation, maintaining reliability while reducing the effective complexity and knowledge exposure.

The suitability calculation mechanism acts as an intermediary between the available classification programs and the mission requirements. This intermediary evaluates and ranks programs based on polynomial chaos expansion and item response theory, selecting the most suitable program without requiring the AUV to deeply process or expose all underlying program knowledge, thereby maintaining reliability while reducing complexity and security risks.