This work considers the problem of fronthaul offloading in a multi-hop fog network where the user devices (denoted as F-UEs) enhanced by learning capabilities, can communicate and collaborate with each other using network coded packets. The paper studies the achievable offloading in relation to different transmission deadlines and network densities comparing it to a single-hop communication. The F-UEs use Q-learning to find autonomously the optimal paths which are defined as the paths requiring the least amount of energy for a single transmission. However, due to the fact that the F-UEs are half-duplex devices and the multi-hop nature of the communication, transmission collisions are inevitable. Thus, the central processing unit is tasked not only with finding the best schedule of transmitting/receiving F-UEs and files in two transmission phases, but also with resolving any possible conflicts between the F-UEs. The central processing unit is able to detect those collisions and resolve them by adjusting the schedule while obeying the time deadlines and thus, maintaining high quality of service (QoS). Results obtained in simulations prove that allowing multi-hop communication achieves better fronthaul offloading than the single-hop communication.

This paper considers a device-based Fog Radio Access Network (F-RAN), where users’ smart devices (denoted by F-UEs) can be exploited to cache popular files so as to communicate them when requested by other F-UEs among them. This architecture restricts the involvement of the central baseband processing unit (BBU) to serving those F-UEs not immediately served by their peers, thus offloading the BBU's fronthaul spectrum and increasing the overall capacity of the network. This paper aims to further maximize fronthaul offloading in this architecture using network coding (NC). The latter exploits both the cached and previously received files by different F-UEs as side information to serve more requesting F-UEs in each transmission from their peers or the BBU so as to maximize fronthaul offloading. Being half-duplex devices (they can only either send or receive in any given time), the problem of BBU fronthaul offloading in this setting is formulated over two transmission phases on an NC graph. After showing that this problem is NP-hard, the paper proposes two novel heuristic algorithms to solve it in real-time. In addition, a lower-bound on the offloading gain of these algorithms is derived for a special file placement case, their asymptotic optimality is proven, and their complexities are analyzed. Simulation results both show that these proposed algorithms perform closely to the optimal solution and quantify the significant offloading gains achieved by them.

This paper studies the problem of offloading the fronthaul of Fog Radio Access Networks (F-RANs) with smart user equipments (denoted by F-UEs) that can cache files and cooperate with each other to retrieve their requested files. This cooperation between the F-UE devices will reduce the load on the fronthaul of the central baseband processing unit (BBU), thus increasing the overall capacity of the network. By smartly employing network coding (NC), this paper aims to minimize the number of transmissions required from the BBU given a realistic half-duplex transmission scenario. In this setting, F-UE devices can only send or receive at a time, and thus must receive their requested files in maximum two time-slots in order to achieve high quality of experience (QoE). The above problem is first formulated over an NC graph. Being NP-hard, two heuristic algorithms are proposed to solve the problem in real-time. Simulation results show that these proposed heuristics perform closely to the optimal solution. They also demonstrate the significant fronthaul offloading gains achieved by our proposed algorithms.