COMPREHENSIVE MULTI-ENVIRONMENT OPERATIONAL DESIGN DOMAIN TAXONOMY FOR INTELLIGENT SYSTEMS SAFETY DESIGN

Abstract

In designing and deploying safety-critical intelligent systems, understanding and defining the operational environment are pivotal for ensuring the system's reliability and robustness. This paper introduces a comprehensive framework for a meticulous Operational Design Domain taxonomy process tailored to Intelligent Systems (IS) design and verification, particularly those integrating machine learning components and dataset development. The process systematically identifies and categorises environmental conditions—from natural lighting and weather to infrastructural and cultural variations—influencing the system's operational efficacy. The taxonomy considers air and land environments (extendable for sea and space in the future). The methodology begins by enumerating environmental aspects that an intelligent system must recognise and adapt to, especially when operating in complex areas. Then, it includes manmade ecosystem features. Each environmental parameter is meticulously defined to cover the range of variability and incorporated into a structured grading system (Grade A to E), which classifies environments from ideal to extremely challenging based on criteria such as lighting, weather conditions, and geographical features.