COSMOD5G

Despite the explosion in the number of distributed optimization algorithms in the literature, there are still many limitations. Most algorithms require many auxiliary variables and parameters for each agent and do not converge to the global optimum. Others require the sharing of gradients and Hessians of local cost functions. Nevertheless, in many applications, agent variables are subject to various constraints. For example, in wireless networks, signal transmission power is limited by receiver sensitivity. Consequently, it is necessary to consider the distributed optimization problem with constraints of the inequality or equality type. The introduction of constraints makes the extension of existing algorithms non-trivial. Thus, the problem of this project concerns distributed optimization under constraints, also considering implementation constraints such as quantization and minimization of the amount of information exchanged in the network, which are found in resource allocation problems in 5G and beyond.

In order to overcome these locks, we propose on the one hand to extend existing algorithms to the case with constraints, such as limited bandwidth, throughput and energy consumption, as well as guaranteed quality of service (QoS). In comparison with existing research work, the novelty of our proposal lies in the following points in particular:

• The consideration of constraints in the distributed optimization algorithm while guaranteeing convergence to the optimal state in a user-defined time so as to satisfy the constraints associated with distributed resource allocation for 5G telecommunications networks and beyond with the development of 6G,
• The development of event-driven triggering mechanisms so as to reduce energy consumption for 5G networks and beyond,
• The implementation within telecommunications protocols of 5G and beyond.