Cooperation Reliability Based on Reinforcement Learning for Cognitive Radio Networks

Abstract

-The primary objective of cooperation in Cognitive Radio (CR) networks is to increase the efficiency and improve the network performance. However, CR users may act destructively and decrease both their own and others' performances. This can be due to Byzantine adversaries or unintentional erroneous conduct in cooperation. This work presents an autonomous cooperation solution for each CR user, i.e., each CR user decides with whom to cooperate. The objective of the proposed solution is to increase the spectrum access in cooperative CR networks. To realize this, a Reinforcement Learning (RL) algorithm is utilized to determine the suitability of the available cooperators and select the appropriate set of cooperators. In addition, the proposed solution determines the most appropriate number of cooperators to achieve the highest efficiency for spectrum access. Accordingly, the control communication overhead is reduced. The simulation results demonstrate the learning capabilities of the proposed to achieve reliable behavior under highly unreliable conditions.

A reinforcement learning based solution for cognitive network cooperation between co-located, heterogeneous wireless sensor networks

“…Authors of [7] tackled a problem of an efficient spectrum sharing, using the REINFORCE [? ]] algorithm.…”

“…Such a behavior could lead to an enforcement of sub-optimal states, since the optimal ones will get a small chance to be examined, which can ultimately result in a very long convergence time. Since performances in states 1 and 3 are quite similar (see Figures 6,7,8,9), as long as the system resides in any of the two states, its performance can be considered optimal. Therefore, percentages that correspond to these states (see Table 3) can be joined into a single value.…”

Application of Reinforcement Learning in Cognitive Radio Networks: Models and Algorithms

“…Security enhancement scheme [ 12 ] aims to ameliorate the effects of attacks from malicious SUs. Vucevic et al [ 13 ] propose a security enhancement scheme to minimize the inaccurate sensing outcomes received from neighboring SUs in channel sensing (A2). A SU becomes malicious whenever it sends inaccurate sensing outcomes, intentionally (e.g., Byzantine attacks) or unintentionally (e.g., unreliable devices).…”

“…Vucevic et al [ 13 ] propose a collaborative channel sensing (A2) scheme, and it has been shown to minimize error detection probability in the presence of inaccurate sensing outcomes. The purpose of this scheme is that it selects neighboring SU agents that provide accurate channel sensing outcomes for security enhancement purpose (A3).…”

“…The critic keeps track of c t +1 i ( a t i ) = c t i ( a t i ) + β · Δ c t i ( a t i ), where Δ c t i ( a t i ) is the temporal difference and β is a constant (e.g., β = 0.01). In [ 13 ], Δ c t i ( a t i ) depends on the difference between the delayed reward r t +1 i ( a t i ) and the long-term delayed reward , the number of incorrect sensing outcomes, and the suitability value π t i ( a t i ). Next, the actor selects actions using c t +1 i ( a t i ) given by the critic.…”