A Reinforcement Learning Approach to Adaptive Forwarding in Named Data Networking

This paper addresses Information Centric Networks, and considers in-network caching for Named Data Networking (NDN) architectures. We depart from forwarding algorithms which primarily use links that have been selected by the routing protocol for probing and forwarding, and propose an adaptive forwarding strategy using reinforcement learning with the random neural network (NDNFS-RLRNN), to leverage the routing information and actively seek possible deliveries outside these paths in a controlled way. Our simulations show that NDNFS-RLRNN achieves more efficient delivery performance than a strategy that strictly follows the routing layer or a strategy that retrieves contents from the nearest caches by flooding requests.