Computation Offloading Optimization Based On Matching Theory In Fog Radio Access Networks

In recent years,computation-intensive and latency-sensitive applications,such as augmented reality(AR)and virtual reality(VR),are continuously emerging,which have stringent requirements on network latency and meanwhile bring a huge burden on user equipments(UEs).As a promising architecture,fog radio access networks(F-RANs)can reduce latency and UE energy consumption owing to the deployment of fog access points(F-APs),which allows UEs to offload computation tasks to the network edge.To improve system performance in computation offloading scenarios,the decision on to which FAP the task is offloaded and radio resource allocation for task uploading both play key roles.However,their joint optimization is challenging to deal with due to NP hardness,which currently draws a lot of attentions of researchers.Facing this issue,aiming at optimizing computation offloading performance in F-RANs with multi-UEs and multi-F-APs,this thesis studies matching theory based approaches to FAP selection and radio resource allocation that well balance the tradeoff between complexity and performance.The main contents and contributions are summarized below.Firstly,facing the issue that solving the FAP selection problem optimally in computation offloading scenarios with multiple UEs and FAPs possesses high complexity,this thesis develops a matching theory based approach,aiming to minimize total system energy consumption.Taking F-AP computation capability constraints into account,a joint optimization problem is formulated,whose solution identifies the location where each task is executed.Under any pre-fixed FAP selection,the decision on which tasks to go to the cloud is optimized for each FAP based on a greedy algorithm,while the remaining FAP selection problem is transformed into a matching problem with externalities that is efficiently solved by an iterative swap matching algorithm.The algorithm can converge to a stable matching and the complexity of each iteration has only a linear relationship with both the number of UEs and FAPs.Numerical results show that the proposed algorithms can achieve near-optimal performance but are much less complex.Secondly,facing the tight coupling of FAP selection and radio resource allocation and the non-convexity of their joint optimization,this thesis is based on matching theory to propose an efficient approach that intends to achieve low latency in computation offloading.Based on the setting of non-orthogonal multiple access that can help reduce latency,an optimization problem is formulated,which jointly optimizes resource-pare composed by FAPs and subchannels selection,UE transmission power in uplink and computation resource allocation at FAPs.To solve this mixed integer non-linear programming,the primal problem is first decoupled into a UE transmission power control sub-problem and a computation resource allocation sub-problem under pre-fixed resource-pare selection.After solving the two sub-problems,an iterative swap matching based algorithm with an inner loop and an outer loop is developed,where the outer loop is responsible for adjusting FAP selection based on a well-defined swap matching order.Simulation results verify the performance gain of the proposed algorithm.The above work provides theoretical support for F-RAN to support Augmented Reality and Virtual Reality,etc.and provides new ideas and methods for F-RAN's resource allocation theory research.