Q-Learning-based Routing Algorithm for 3D Network-on-Chips

Abstract

New communication methods have become inevitable due to the continuous increase in the number of components on the integrated circuits. Network-on-Chip (NoC) meets this need with its scalability and parallelism features. Furthermore, 3D-NoC architecture has been developed due to more speed and less power consumption demands. However, the routing problem for 3D becomes more complicated. Since deterministic algorithms frequently encounter congestion problems, adaptive algorithms give better results considering the system’s traffic load. Motivated by the effectiveness of learning algorithms on this type of problems, we present a Q-Learning based routing algorithm for the 3D-NoC routing problem. In our algorithm, each router node maintains a Q-Table and updates it by receiving the traffic information from neighboring routers. We select the output port of the packets coming to the node according to this table. We compared our method with the deterministic XYZ algorithm with different traffic models. The results show that our method achieved 8% performance improvement.