Application Task Allocation in Cognitive IoT: a Reward-Driven Game Theoretical Approach

In this study we consider the scenario of sensors belonging to different platforms and owned by different owners that join the efforts in an opportunistic way to improve the overall sensing capabilities in a given geographical area by forming clusters of nodes. The considered nodes have cognitive radio and exploit device-to-device communications. A solution is proposed which relies on a Cluster Head (CH) that guides the whole task allocation strategy. The addressed challenges are the following: i) collaborative spectrum sensing for effective communications within the cluster; ii) assignment of each request of sensing tasks to a single node in the cluster. The first challenge is addressed by proposing a collaborative sensing procedure where each node communicates to the CH the received signal energy of licensed users so that the latter makes a decision on the availability of the band by fusing the received information towards a minimisation of the uncertainty in detecting the free spectrum. The second challenge is addressed by proposing a non-cooperative Game theory based approach in which cluster nodes make effort to selfishly increase utility by winning the task. Each node takes part to the competition by considering two elements: the gain that is won for its contribution to sensing and for the execution of the task (in case it wins the competition); the cost in terms of energy to be consumed in case the task is executed. A Nash Equilibrium Point (NEP) is found for the aforementioned game in which each object has no incentive to deviate uni-laterally from the NEP. Extensive simulations are performed to evaluate the impact of probability of false alarm, utility function weighting factors and presence of licensed users on the cumulative system utility.