On optimal multipath rate-adaptive routing

A centralized rate-adaptive routing algorithm is presented that, in contrast to the distributed ones available in the literature, achieves provable stability, optimalilty with respect to optional linear or quadratic objective functions, and feasibility in that it can accommodate any admissible traffic matrix in the network without violating link capacities. We recast the routing problem in the framework of constrained optimal control theory to obtain optimal state feedback routing controllers, and we present simulations confirming that our routing controllers are viable in small- and middle-sized networks.