Abstract: This paper presents an ATR design paradigm that self configures and adapts to the diverse scenarios encountered during a mission. Today's ATR is constructed via inefficient and sub-optimal system configuration and training, whose process is very labor intensive, subjective and inaccurate. The resulting ATR is only capable of a limited amount of adaptation to changes in the environment. Moreover, the operation of such ATR systems require a user with expert algorithmic knowledge. Addressing the above-mentioned problems, the Honeywell effort is producing a self-adaptive ATR system. The system employs a Genetic Algorithm to autonomously and optimally perform configuration and training; the system also includes a specific knowledge capture mechanism, the Context Capture tool, which ties the context of the mission with an optimal configuration. Lastly, the system employs Case Based Reasoning to dynamically configure and control the ATR system based on the changing context during an ATR mission.!7
展开▼
