Segmented backoff algorithm based on weighted reinforcement learning

Abstract

The invention discloses a segmented backoff algorithm based on weighted reinforcement learning, which comprises the following steps: S1, establishing a binary exponential backoff algorithm model, analyzing the condition that the channel effective utilization rate and the data packet loss rate of a wireless sensor network change along with the increase of data traffic in the network, establishing a segmented backoff window, and setting different numbers of nodes to change the data flow condition in the network so as to obtain the channel effective utilization rate of each segment of backoff window under the condition of different numbers of nodes. According to the weighted reinforcement learning-based segmented backoff algorithm disclosed by the invention, the channel access mode of the network access control layer is adjusted through the weighted reinforcement learning model; and therefore, the effective utilization rate of the channel of the wireless sensor network is improved and the packet loss rate is reduced while the fairness of competitive access of the node to the channel is ensured.

Description

technical field [0001] The invention belongs to the technical field of media access control layers of wireless sensor networks, and in particular relates to a segmented back-off algorithm based on weighted reinforcement learning. Background technique [0002] In a wireless sensor network, each sensor node transmits its own data by competing for access to a shared channel. With the expansion of the network and the increase in data traffic, the number of nodes competing for the channel increases, and the number of channel access failures due to conflicts increases. ; In addition, under the framework of binary exponential backoff algorithm, as the data flow increases, the number of collisions of nodes in the network due to competition for access channels increases, and after each collision, the collided nodes will be in [0, 2^(minBE+X)-1] (minBE is the smallest backoff index size, x is the number of collisions) randomly generates a backoff time, but if the backoff time is too s...

AI Technical Summary

Problems solved by technology

[0002] In a wireless sensor network, each sensor node transmits its own data by competing for access to a shared channel. With the expansion of the network and the increase in data traffic, the number of nodes competing for the channel increases, and the number of channel access failures due to conflicts increases. ; In addition, under the framework of binary exponential backoff algorithm, as the data flow increases, the number of collisions of nodes in the network due to competition for access channels increases, and after each collision, the collided nodes will be in [0, 2^(minBE+X)-1] (minBE is the smallest backoff index size, x is the number of collisions) randomly generates a backoff time, but if the backoff time is too small, it will increase the probability of collision and increase the node access channel When collision occurs, the possibility of data transmission failure will affect the effective utilization of the channel

[0003] And the existing backoff algorithm in IEEE 802.15.4 often has a smaller backoff window for nodes that have just transmitted data when nodes compete for access to the channel, so there is a greater probability of continuing to seize the channel. For other nodes Competition for access to channels can be unfair

However, the previous reinforcement learning algorithms did not provide convergence and action bias for the algorithm in solving this problem.

Images