Reinforcement Learning for Routing in Cognitive Radio Ad Hoc Networks

Table 2

Reward representation for WCRQ-routing model embedded at SU node 𝑖.

Cost

𝑟^𝑖𝑡(𝑎^𝑖𝑡)=𝜔𝑟𝑡^𝑖,𝑗+(1−𝜔)𝑞^𝑗𝑡 where 𝑟𝑡^𝑖,𝑗 represents the number of retransmissions for a packet sent from SU node 𝑖 to SU neighbor node 𝑗 at time 𝑡, while 𝑞^𝑗𝑡 represents the number of packets in the queue of SU neighbor node 𝑗. Weight factor 𝜔=[0,1] is used to adjust the tradeoff between PUs’ and SUs’ network performance.