Intelligence is the most important characteristic for cognitive wireless networks. A cognitive engine built on reconfigurable wireless networks is the key to implementing this characteristic. The design and implementation of a cognitive engine is important in research on the theory of initiative cognition for cognitive wireless networks. This paper first discusses research on cognitive loops, then investigates cognitive functions in the loop through the design of a universal cognitive engine functional architecture, and finally verifies the architecture on the platform of a cognitive engine prototype system. cognitive radio, cognitive wireless network, cognitive loop, cognitive engine, prototype system "Cognition" is the scientific term for the process of thinking. Mitola et al.[1] introduced "cognition" to wireless communication and coined the phrase "Cognitive Radio" in 1999.From then on, cognitive radio and cognitive networks have attracted increased attention from researchers and research institutions [2][3][4], and have become a hotspot in the communication domain [5]. A cognitive network is able to sense information from the outside world and its internal state, and can then adjust the wireless network (including working frequency, air access, data protocols, and so on) by parsing and orienting the sensing information to adapt to changes in the environment. Moreover, a cognitive network can still learn to form new knowledge. Thus, a cognitive network is an intelligent network that can sense, make decisions, and learn like humans. Implementing such a wireless network requires a functional architecture as the brain to control the whole cognitive process, that is, a cognitive engine. The cognitive engine integrates various cognitive functions (such as sensing, learning, and reasoning) so that it can intelligently control a wireless network to implement a cognitive cycle. Rieser [6], a researcher at Virginia Tech, first modeled the biologically inspired cognitive engine called BioCR. Later, his colleague Rondeau [7] further developed the theory of cognitive engines. He evolved cognitive engine modeling as a multi-objective optimization through genetic algorithms and built a cognitive engine architecture composed of a cognitive controller, sensors, a decision maker, optimizers, and interfaces. However, as this cognitive engine was purely designed for waveform optimization, it has limitations. In this paper, we set out to design a universal cognitive engine functional architecture by studying the cognitive loop.