The requirement to support an always increasing number of networking technologies and services to cope with context uncertainties in heterogeneous network scenarios leads to an increase of operational and management complexity of the Internet. Autonomous communication protocol tuning is then crucial in defining and managing the performance of the Internet. This chapter presents an evolutionary roadmap of communication protocols towards cognitive Internet in which the introduction of self-aware adaptive techniques combined with reasoning and learning mechanisms aims to tackle inefficiency and guarantee satisfactory performance even in complex and dynamic scenarios.In this survey, we overview and compare existing adaptive protocol stack solutions, review the principles of cross-layer design as well as the agent-based and AI based self-configuration solutions. The fundamental principles of cognitive protocols, such as adaptation, learning, and goal optimization, are presented along with implementation examples. Finally, the chapter discusses future research on the topic.