Adaptive routing planning method for q-learning optical on-chip network based on dijkstra algorithm

Abstract

The present invention relates to a Q-learning optical on-chip network adaptive routing planning method based on Dijkstra algorithm, comprising: S1: constructing a network model, and defining network model parameters; S2: constructing each node to other nodes according to Dijkstra algorithm and network model The shortest path tree of the node, at the same time store several pieces of the node to the target node v in each node according to the preset value d The shortest path of , and get the source node v s to the target node v d The routing hops of the shortest path h(v s ,v d ); S3: According to the Q-learning algorithm, the link selection mechanism based on the ε-greedy strategy is used for path planning, and the source node v s to the target node v d Several planned paths, obtain the reward value of the planned path, and the number of route hops of the planned route does not exceed the route hop number of the shortest path h(v s ,v d ); S4: Get the best path according to the reward value of the planned path. The method of the invention overcomes the shortcoming that each target point of the Dijkstra algorithm can only generate one shortest path.

Description

technical field [0001] The invention belongs to the technical field of dynamic routing planning, and in particular relates to a Q-learning optical on-chip network adaptive routing planning method based on the Dijkstra algorithm. Background technique [0002] With the exponential growth of data traffic and the rapid development of smart devices, the network becomes more complex and more diverse, and more factors need to be considered, including stability, security, bandwidth, delay, load, etc. Now that chip multiprocessor capabilities are increasing, on-chip communication efficiency is critical to overall performance. During the entire information transmission process, the intermediate routers need to select the next hop router according to the current state. However, from an overall and long-term perspective, the lack of global information makes the selected next-hop forwarding node not always optimal, so people pay more attention to using reinforcement learning to solve re...

AI Technical Summary

Problems solved by technology

Dijkstra's algorithm is a well-known algorithm for finding the shortest path. It can quickly give the shortest path, but it can only provide one shortest path for each target point, and cannot provide other alternative shortest paths, and it is only applicable to non-negative weight planning

Compared with the Dijkstra algorithm, the Bellman-Ford algorithm supports the presence of negative weights, and the code implementation is relatively simple, but the Bellman-Ford algorithm has a high time complexity, the convergence speed is lower than the Dijkstra algorithm, and the Bellman-Ford algorithm requires a lot of information Passing, especially when negative weights are encountered, requires multiple iterations

Images